Geographic variation in Caluromys derbianus and Caluromys lanatus ( Didelphimorphia : Didelphidae )

We analyzed the geographic variations in the shape and size of the cranium and mandible of two woolly opossums, Caluromys derbianus and Caluromys lanatus. Using geometric morphometrics we analyzed 202 specimens of C. derbianus and 123 specimens of C. lanatus, grouped in 7 and 9 populations, respectively. We found sexual dimorphism in shape variables only in the dorsal view of the cranium of Caluromys derbianus, which is not associated with geographical origin. We detected geographic variation in the size of the mandible in two populations (Nicaragua and Northern Panama), but no geographic variation in shape. The size of the cranium of C. lanatus varies significantly, with clinal variation in peri-Amazon populations, with a break between two populations, Bolivia and Paraguay. Shape analyses also revealed some separation between the Paraná population and all other populations. Our results suggest that the available name, Caluromys derbianus, should be maintained for all individuals throughout the geographic range of the species. The same is true for Caluromys lanatus, which can be separated into two distinct morphologic units, Caluromys lanatus ochropus, from the Amazon and Cerrado, and Caluromys lanatus lanatus, from the Atlantic forest.

The morphology and/or physiology of organisms usually vary across their distribution range.This is particularly true for species that are distributed over different biomes or biogeographic provinces (THORPE 1987).Such variation in intraspecific characters throughout a species' range is known as geographic variation (MAYR 1977).The study of geographic variation is key for understanding speciation and the role that ecological and geographical features may play in shaping biodiversity (HAFFER 1969, GOULD 1972, EMMONS 1984).Furthermore, geographic variation has been a central theme in evolutionary biology, from the works of Darwin to modern analyses based on molecular approaches (HALLGRÍMSSON & HALL 2005).
The purpose of this study was to evaluate and to quantify the morphological variation in the size and shape of the cranium and mandible of Caluromys derbianus and Caluromys lanatus throughout their geographic range.We used geometric morphometric tools to evaluate whether the variation supports the taxonomic status of each species and their currently recognized subspecies.

MATERIAL AND METHODS
We obtained 2D images of the crania in three views (dorsal, ventral and lateral), and lateral images of the mandibles.Only complete adult specimens, i.e., specimens with all three premolars and four molars fully erupted and functional (TRIBE We digitized a total of 92 landmarks -28 in dorsal, 28 in ventral, 22 in lateral views of the cranium, and 14 landmarks on the mandible -using TPS Dig (ROHLF 2006) (Fig. 1, Appendix 1).All landmarks were tested for repeatability (FALCONER & MACKAY 1996), which was set at 85% for inclusion in subsequent analyses.

Geographic variation in
We applied a a Generalized Procrustes Analysis (GPA) to all landmark configurations (ROHLF & SLICE 1990), to remove the effects of isometric size, orientation and position.Conse-quently, only shape information was retained (ADAMS et al. 2004(ADAMS et al. , 2013)).We obtained two formally independent set of variables, used in the subsequent analyses.One set includes centroid size for all specimens.Centroid size is the univariate size variable resulting from the squared-root of sums of the squared distances between each landmark and the centroid of its configuration.This set was used in the analyses of geographic variation in the size of the studied structures (for more details see ZELDITCH et al. 2012).GPA also yields the partial warps and uniform components, a set of variables that retain all the information on the shape of the landmark configuration of the studied structures that were used in the analyses of geographic variation in shape.Further detail on the geometric morphometric procedures can be found in ZELDITCH et al. (2012).
We obtained the geographic coordinates of the collecting localities of each specimen from their skin tags.When coordinates were not in the tags, we used standard ornithological gazetteers (PAYNTER 1982(PAYNTER , 1989(PAYNTER , 1992(PAYNTER , 1993(PAYNTER , 1995(PAYNTER , 1997) ) to recover them.Specimens from different localities were grouped into populations based on the features of the ecoregions (OLSON et al. 2001) found in the distribution of both species (specimens from geographically close localities in the same ecoregion were pooled into populations).Next, to increase the sample size of populations resulting from the classification using ecoregions, we decided to pool the populations that were geographically closer to each other and which lacked morphometric divergence.
Literature information on the absence of sexual dimorphism in both species (ASTÚA 2010) was obtained from a smaller and geographically restricted dataset.With this in mind we reevaluated the existence of sexual size dimorphism through a ttest on centroid size, and the existence of sexual shape dimorphism through a Hotteling T 2 test on shape variables.Since several populations were represented by only a few specimens,we pooled all males into one group and all females into another regardless of their geographic origin, in order to increase sample size and to avoid a type I error.To evaluate geographic variation in size, we compared populations with ANOVAs on centroid sizes, followed by Tukey a posteriori tests.To evaluate geographic variation in shape we compared shape variables between populations using Canonical Variates Analyses (CVA), following WEBSTER & SHEETS (2010), given that our total sample size was much larger than [(2k -4) + (G -1)], where k is the number of variables and G is the amount of groups analyzed.For each view, this parameter ranged from 30 to 58 for Caluromys derbianus, and 32 to 60 for C. lanatus, indicating that running a CVA is appropriate.Because all analyses were repeated on four views of both species, we employed Bonferroni correction again, using a significant p-value of 0.0125 (0.05/4).

Sexual dimorphism
Neither species presented sexual dimorphism in size.Sexual dimorphism was observed only in the shape of the dorsal portion of the cranium of Caluromys derbianus (Hotteling T 2 = 0.858, F = 1.82; d.f.= 56, p < 0.01, 85 males, 90 females).In view of the absence sexual dimorphism in size and shape variables among individuals in all other views of both species, we decided to pool the sexes together within populations for subsequent analyses.This allowed us to include in the analyses specimens for which the sex was unknown.

Geographic variation in Caluromys derbianus
When analyzing size variation, we only found a statistically significant difference in mandible size, between the Nicaragua and Northern Panama populations (ANOVA F = 2.89, p < 0.01031, p < 0.002, post-hoc Tukey test).As for shape variation, the CVA scores overlapped considerably, indicating little morphometric divergence in size (Fig. 4).Given that the variation within each population was equal to or larger than the variation between populations, we concluded that the variation is not geographically structured and that the populations cannot be considered morphologically distinct.

Geographic variation in Caluromys lanatus
Under all views, size varied geographically (ANOVA, Dorsal: F = 11.02,p < 0.0001; Lateral: F = 3.66, p < 0.001; Ventral: F = 9.62, p < 0.0001; Mandible: F = 10.91,p < 0.0001), but no clear grouping was observed among populations.However, a north-south clinal variation in skull size can be inferred, with specimens increasing in size from Colombia (smallest) to Bolivia (largest), with Ecuadorian and Peruvian specimens presenting intermediate sizes.This trend is then interrupted in southern Bolivia, with specimens from southeastern Brazil, Paraguay and Argentina being smaller than their Bolivian counterparts (Fig. 5).
Caluromys lanatus has a conserved skull shape throughout its geographic range.CVA scores show a partial separation of the Paraná population from all others, due to a variation in the morphology of the occipital and posterior roots of the squamosal, which are larger in Paraná specimens than in other individuals.The morphology of the rostrum also varies, with short and narrow nasals and basicranium with short frontals and longitudinal elongation of parietals (visualized through displacement of landmarks at the postorbital constriction) in the dorsal view of the cranium (Fig. 6).Additionally, an increase in occipital width, a more horizontally aligned molar tooth row, and shorter and narrow rostrum are found in Paraná individuals, in lateral view of the cranium (Fig. 7).

DISCUSSION
Structured geographic variation in cranial size and shape was not detected in Caluromys derbianus.However, it was observed in Caluromys lanatus populations.Despite the fact that subspecies have been recognized for C. derbianus, its populations belong to a single morphologic unit, which is spread throughout the geographic distribution of the species.Our results also corroborate that Caluromys lanatus is one species, but with two distinct morphological groups, one in the Amazon-Cerrado and the other in the Atlantic forest.
The absence of sexual size dimorphism in the skull of these species was already discussed (ASTÚA 2010), although that analysis, unlike ours, detected significant sexual shape dimorphism in both species.

Geographic variation in Caluromys derbianus
We did not find any evidence of structured geographical variation in the size of the skull of Caluromys derbianus, despite its occurrence in the congeneric species C. lanatus (this study) and C. philander (OLIFIERS et al. 2004).We were also unable to detect a pattern in the geographic variation of the shape and size of the skull that would match the current taxonomic structures proposed for this species at the subspecific level.BUCHER & HOFFMANN (1980) and GARDNER (2008) divided Caluromys derbianus into seven subspecies and one trans-Andean "unspecified" population.These populations were based on morphological differences such as fur color.As we used only cranial quantitative data, it is possible that other characters, particularly in the external morphology, may be the reason for the high number of subspecies.In particular, pelage color, which was not assessed in this study, is well known to vary geographically in this and other marsupial genera (THOMAS 1913, GOODWIN 1942) and might explain the discrepancy between the existing classification and the one that results from our quantitative results from skull morphology, which failed to support a separation.
The distribution of Caluromys derbianus represents a continuum of populations on a N-S stripe, most of which are in Central America and the remaining populations in the Andes in South America.The absence of geographic variation in the size of the cranium and mandible shape of this species is noteworthy, since several geographic and ecological discontinuities found throughout its distribution range are believed to cause variation among populations of other taxa (SAVAGE 1987, PÉREZ-EMÁN 2005, CASTOE et al. 2009).

Geographic variation in Caluromys lanatus
Clinal variation occurs throughout the range of many mammals (STORZ et al. 2001, CARDINI et al. 2007).We found clinal variation in the size of the skull of Caluromys lanatus from Andean populations,to the Bolivian-Paraguayan border, coinciding with those populations that overlap less in shape analyses.Even though we have not analyzed molecular data, we believe that the large overlap of CVA scores among all Amazon populations can be associated with reduced genetic divergence in this species.The latter has been already noted for populations distributed in this area (PATTON et al. 2000, PATTON & COSTA 2003).
The divergence among the Paraná population and the others may correspond to the geographic differences between the Amazon and the Atlantic Rainforest.A similar variation pattern has also been observed to occur in Didelphis, Marmosa, Caluromys philander and Metachirus nudicaudatus (COSTA 2003, PATTON & COSTA 2003).Both morphological and genetic divergence were observed in these species.Similar results were also recorded for rodent genera such as Rhipidomys, Oecomys, Hylaeamys and Euryoryzomys   al. 2006), which is characterized by medium and large trees such as Bignoniaceae, Leguminosae and grass fields (PENNINGTON et al. 2000).The increase in the Araucaria cover in the early Holocene (LEDRU 1993, SALGADO-LABOURIAU et al. 1998) over open areas may have served as a bridge between the forested areas of the Atlantic forest and the Amazon (AB 'SABER 2000).This plant cover probably allowed the dispersion of Caluromys lanatus from the Amazon and Cerrado to the southern Atlantic Rainforest (COSTA 2003, PATTON & COSTA 2003), where these new populations were later isolated by open lands that arose between these areas (LEDRU et al. 1998, VAN DER HAMMEN & HOOGHIEMSTRA 2000, BEHLING 2002).This contact and subsequent isolation hypothesis is particularly likely for Caluromys lanatus, since this species is strictly arboreal.Environmental discontinuities that incur in canopy fragmentation may hinder population movements (PIRES et al. 2002), thus providing an effective ecological barrier like the one that has been associated with speciation of the congeneric Caluromys philander (LIRA et al. 2007).Morphologi-cal similarities between populations from Central Brazil and the Amazon may be explained by the fact that Cerrado vegetation may not be uniformly affected by climatic changes (SALGADO-LABOURIAU et al. 1997).At higher altitudes the plant composition was less altered even in the dry periods of the Pleistocene and may have extended to lower areas during cold periods (BUSH et al. 2004).Grassland vegetation may have replaced only low-altitude forests (SALGADO-LABOURIAU et al. 1997, 1998).Due to climatic and pluviometric oscillations, eventual expansions of gallery forests may have created ecological corridors that allowed faunal and floristic population flow among Cerrado, Llanos, Amazonia and even Gran-Sabana (CERQUEIRA 1982, LEDRU 2002, OLIVEIRA-FILHO & RATTER 1995apud DE OLIVEIRA et al. 2005).Gallery forests house twice as many forest-related species than the entire Cerrado latu sensu (JOHNSON et al. 1999).These forested areas may not have been totally affected by climatic changes and may have been used as a corridor that kept Amazonian and Cerrado populations in contact (CARDOSO & BATES 2002).

Potential implications for the taxonomic classification of Caluromys derbianus and Caluromys lanatus
The similar skull morphology shared by all populations of Caluromys derbianus suggest that the seven subspecies -C.derbianus aztecus, C. d. canutus, C. d. centralis, C. d. derbianus, C. d. fervidus, C. d. nauticus and C. d. parvidus -may be considered a unique species on morphometric grounds.Likewise, the lack of geographic variation among the Amazon and Cerrado populations of Caluromys lanatus suggest that three of the four subspecies recognized by CABRERA ( 1958 All these subspecies were described based on external morphological characters, such as body, facial, dorsal, caudal or feet color, characters that usually present geographic variation (THOMAS 1899, 1913, ALLEN 1904, HOLLISTER 1914, GOODWIN 1942).Caluromys species were first described based on morphological characters of a single or a few individuals; subspecies were generally described after comparing individual variation with the holotype.For this reason, it cannot ruled out that these subspecies were based on individual variation.In all cases, pending a proper extensive review of coat color or other morphological variation in Caluromys, our extensive and quantitative results do not support separation of these taxa.
However, because phenotype is mainly the expression of the underlying genotype, morphological divergence is often interpreted as evidence of specific status.In didelphids, for example, morphological evidence has been used to support splitting of the black-eared and the white-eared opossums (CERQUEIRA & LEMOS 2000, LEMOS & CERQUEIRA 2002) of the genus Didelphis, and Bolivian species of Marmosops (VOSS et al. 2004).As such, it it possible that the morphologic groups found here may represent distinct species (see, however LÓSS et al. 2011, for a situation where morphologic differentiation does not coincide with species limits).The recognition of southern and southeastern populations of South American didelphids as distinct species appears to be a recurrent pattern that emerges after a deeper analysis of the existing variation, such as in Phi-  2003).Such changes are actually the reflection of our still incomplete knowledge on the taxonomy and systematics of didelphids.
A proper and definite appraisal of the taxonomic status of both woolly opossums would require an integrative approach (including other phenotypical and genetic characters) to unveil their actual status.Especially among Caluromys lanatus populations, a molecular approach may be useful to assess if these divergent groups constitute distinct evolutionary lineages that would ultimately validate their status as distinct species.Pending this, we suggest that the available name Caluromys derbianus (Waterhouse, 1841) is maintained for all individuals across the geographic distribution of its populations.The name Caluromys lanatus (Olfers, 1818) should also be considered valid, with at least two distinct morphometric units, namely Caluromys lanatus ochropus, representing Amazon and Cerrado populations, and Caluromys lanatus lanatus, encompassing Atlantic forest individuals.

Figure 1 .
Figure 1.Landmarks used in the cranium and mandible.Smaller versions of each view include landmarks with links, as used deformation grids in subsequent figures.See Appendix 1 for detailed description of landmark locations.Scale bar: 1 cm.

Figure 2 .
Figure 2. Distribution of the localities of Caluromys derbianus with specimens included in this study.Localities were grouped in populations for subsequent analyses, and are labelled accordingly.Numbers indicate localities as listed in Appendix 2.

Figure 3 .
Figure 3. Distribution of the localities of Caluromys lanatus with specimens included in this study.Localities were grouped in populations for subsequent analyses, and are labelled accordingly.Numbers indicate localities as listed in Appendix 2.

Figure 4 .
Figure 4. Canonical Variates Analysis on shape variables (partial warps and uniform components) of the skull in dorsal view of Caluromys derbianus, using localities as grouping factors, and percentage of variance explained by the first two CVs.Only the convex hulls for each population are shown.Grids indicate deformation associated with the extremes of each CV, from a multivariate regression of shape variables onto CV scores.Overlap for all other views are very similar, therefore only the dorsal view of the cranium is shown.
(COSTA 2003).Populations from Paraná are ecologically separated from others by the Chaco -xerophytic plant cover, located in Argentina and Paraguay(MARCO & PÁEZW 2002, BOLETTA et

Figure 5 .
Figure 5. Clinal variation in size of the skull and mandible of populations of Caluromys lanatus along the east of the Andes, from Colombia to Bolivia, with a break between Bolivian and Northern Paraguay/Southern Brazil populations, indicated by the dashed line.Numbers in the map refer to the same points in the two graphs.

Figure 6 .
Figure 6.Canonical Variates Analysis on shape variables (partial warps and uniform components) of the cranium in dorsal view of Caluromys lanatus, using localities as grouping factors, and percentage of variance explained by the first two CVs.Only the convex hulls for each population are shown.Grids indicate deformation associated with the extremes of each CV, from a multivariate regression of shape variables onto CV scores.
) -C. lanatus cicur, C. lanatus ornatus and C. lanatus ochropus -and four of the six suggested by GARDNER (2008) -C.lanatus cicur, C. lanatus ornatus, C. lanatus ochropus and C. lanatus vitalinus can be lumped based on morphometrical data.The geographic variation found in skull morphometric data of individuals from the southern Atlantic Forest also suggest that two subspecies proposed by GARDNER (2008) -C.lanatus lanatus and C. lanatus vitalinus from southern Brazil can also be lumped.

Figure 7 .
Figure 7. Canonical Variates Analysis on shape variables (partial warps and uniform components) of the cranium in lateral view of Caluromys lanatus, using localities as grouping factors, and percentage of variance explained by the first two CVs.Only the convex hulls for each population are shown.Grids indicate deformation associated with the extremes of each CV, from a multivariate regression of shape variables onto CV scores.