Morphometric and morphological variation in South American populations of Myotis albescens ( Chiroptera : Vespertilionidae )

Myotis albescens (É. Geoffroy, 1806) occurs from Mexico to Uruguay and Argentina. Despite a large number of specimens in collections, its variability in South America has been underestimated, potentially leading to errors in identification. In order to clarify the taxonomic limits of M. albescens and to evaluate previous hypotheses of geographic variation in size we analyzed the type material and studied the variability in South American samples using multivariate exploratory and confirmatory procedures, as well as frequency analyses of discrete morphological data. The presence of a fringe of hairs along the trailing edge of the uropatagium, the long and silky pelage with frosted appearance on the dorsum, ear 9 to 14 mm long, broad interorbital and postorbital constrictions, and a globular braincase were identified as the most useful traits to distinguish M. albescens from its South American congeners. In agreement with Bergman’s rule, larger specimens were found in the South. Beyond the geographic component, Individual variation is an important factor affecting the variability in the size and shape of the skull and pelage color.

Myotis albescens is well represented in collections (LAVAL 1973).A set of traits outlined by previous authors (e.g., LAVAL 1973, BARQUEZ et al. 1999, LÓPEZ-GONZÁLEZ et al. 2001, LÓPEZ-GONZÁLEZ 2005), and summarized in WILSON (2008) and BRAUN et al. (2009), is currently used to identify individuals and to characterize the species.Nevertheless, based on the analyses of South American samples of M. albescens, we observed both quantitative and qualitative variability in the diagnostic traits, which limit their use in diagnosing the species.In addition, previous analyses of geographic variation in South American populations of M. albescens were inconclusive with regards to a trend of larger specimens from north to south (LAVAL 1973, MYERS & WETZEL 1983, LÓPEZ-GONZÁLEZ et al. 2001).
Herein, as part of an ongoing systematic and biogeographic review of South American species of Myotis, we analyzed the type material and studied the morphological and morphometric variation in available samples of M. albescens from South America.These approaches permitted us to redefine the taxonomic limits of the species and to test the hypothesis that specimens of M. albescens tend to be larger in Southern locations.

MATERIAL AND METHODS
To address the variation in qualitative and quantitative characters among South American population samples, 455 specimens (256 females and 199 males, Appendix) were examined.These specimens were identified as M. albescens based on a set of traits present in the neotype (AMNH 205195), as well as other characters highlighted by LAVAL (1973), LÓPEZ-GONZÁLEZ et al. (2001) and MORATELLI & WILSON (2011a, b): presence of a fringe of hairs on the trailing edge of the uropatagium; pelage frosted, long and silky, with light brown dorsal hairs, and yellow to whitish ventral hairs; broad interorbital constriction; globular braincase; and moderate overall size (ear length ca.10-14 mm, greatest length of skull ca.13.2-14.5mm).These characters occur individually in other South American species of Myotis, but together they are exclusive to M. albescens (MORATELLI & WILSON 2011b).Specimens that exceeded the dimensions provided above were assigned to M. albescens based on cranial morphology and pelage color.Fifteen cranial and four external dimensions were measured using a digital caliper accurate to 0.02 mm.The measurements, reported in millimeters (mm), and their abbreviations are defined as follows (lengths were measured from the anteriormost extreme of the first structure to the posteriormost extreme of the second structure mentioned below): greatest length of skull (GLS), from the premaxillae, including the incisors, to the occiput; condylo-canine length (CCL), from the occipital condyles to the upper canines; condylo-incisive length (CIL), from the occipital condyles to the upper incisors; basal length (BL), from the foramen magnum to the upper incisors; zygomatic breadth (ZB), greatest breadth across the outer edges of the zygomatic arches; mastoid breadth (MAB), greatest cranial breadth across the mastoid region; braincase breadth (BCB), greatest breadth of the globular part of the braincase; interorbital breadth (IOB), least breadth across orbital bulges; postorbital breadth (POB), least breadth across frontals posterior to the postorbital bulges; breadth across canines (BAC), greatest breadth across outer edges of the crowns of upper canines; breadth across molars (BAM), greatest breadth across outer edges of the crowns of upper molars; maxillary toothrow length (MTL), from the upper canine crown to the crown of M3; molariform toothrow length (M13), from the crown of M1 to the crown of M3; mandibular length (MAL), from the dentary, without incisors, to the angular process; mandibular toothrow length (MAN), from the lower canine to m3; forearm length (FA), from the elbow to the distal end of the forearm including carpals; third metacarpal length (3ML), from the distal end of the forearm to the distal end of the third metacarpal; and length of the dorsal (LDH) and ventral hairs (LVH), from the base to the tip of the hair, measured between scapulas.The weight, reported in grams, and ear length (EL), in millimeters, were obtained from the skin labels.Descriptive statistics were calculated for all dimensions above.

Morphometric and morphological variation in
To assess secondary sexual dimorphism of craniodental and external variables we performed a series of analyses of variance (ANOVAs) for locality samples with at least 10 individuals of each gender, as follows: São Gabriel da Cachoeira, Amazonas, Brazil (13 females, 17 males) and Paiçandu, Paraná, Brazil (10 females, 10 males).
Due to the limited number of specimens from some localities, the available specimens were grouped to form pooled samples for the geographic analyses.To define the optimum number of pooled samples and respective sample memberships, the matrix of geographic coordinates was submitted to a kmeans procedure (HARDY 1994).This procedure iteratively minimizes the least-squared geographic distances between each locality and centroids of all possible locality-groups using the C g criterion of KRZANOWSKI & LAI (1988).Based on this procedure, a total of 189 specimens recognized as M. albescens were indexed to six geographical samples (Fig. 1 Capibara, Amazonas, Venezuela (n = 2), Cerro Neblina Base Camp, Amazonas, Venezuela (n = 1), San Juan, Amazonas, Venezuela (n = 4), Pto.Paez, 38 km NW, Apure, Venezuela (n = 1), Rio Supamo, Bolivar, Venezuela (n = 1), Rio Chico, 7 km E, Miranda, Venezuela (n = 1), Valera, 23 km NW Trujillo, Venezuela (n = 1), and São Gabriel da Cachoeira, Amazonas, Brazil (n = 30).As multivariate procedures require complete datasets, missing values (2.03% of total dataset) were estimated from the existing raw data using the Expectation-maximization (E-M) algorithm (LITTLE & RUBIN 1987, STRAUSS et al. 2003).All measurements and estimated values were then log-transformed and the covariance matrices were computed considering all variables.Geographic variation in morphometric characters was accessed by Principal Components Analysis (PCA) to summarize general trends of size and shape variation within the total dataset treated as a unique sample, and Canonical Variate Analysis was used to assess craniometric characters that best discriminate the geographic samples (NEFF & MARCUS 1980, MANLY 1994, STRAUSS 2010).
All statistical procedures, including the k-means procedure to define the optimum number of pooled samples and respective sample memberships, the estimation of missing data using the E-M algorithm, as well as the univariate and multivariate analyses were performed in Matlab for Windows, version 4.2c (Mathworks 1994), using functions written by R.E.Strauss available at http://www.faculty.biol.ttu.edu/strauss/Matlab/Matlab.htm (as accessed in November 07, 2010).
A set of qualitative characters selected by previous authors (e.g., THOMAS 1901, 1902, MILLER & ALLEN 1928, HANDLEY 1960, LAVAL 1973, BAUD & MENU 1993, LÓPEZ-GONZÁLEZ et al. 2001, LÓPEZ-GONZÁLEZ 2005, WILSON 2008, MORATELLI & WILSON 2011a, b, MORATELLI et al. 2011a, b) was used to distinguish and characterize M. albescens: plagiopatagium attachment (attached at ankles; at toes by a narrow band of membrane; or at toes by a broad band of membrane); occurrence of a fringe of hairs along the trailing edge of the uropatagium (absent or present); position of P3 (aligned with other premolars or displaced to the lingual side, and visible or not visible when observed in labial view); occurrence and height of sagittal and occipital crests (absent or present, and height: very low, low, medium and high); shape of braincase roof (parietal inclined forward or straight); shape of occipital region (occipital flattened when observed in lateral view, not projected much beyond the limit of occipital condyles, or occipital rounded, projected beyond the limit of occipital condyles).In addition to these traits, pelage color was also described and compared to characterize variation in M. albescens.Capitalized color nomenclature standards follow RIDGWAY (1912).

Non-geographic variation
Quantitative characters: There is considerable variability in the overall size of adult individuals from the same locality.For example, in one sample from the Tocantins River, Pará, Brazil (n = 12), the forearm ranges from 32.2 to 36.7 mm in length, and the total length of skull ranges from 12.88 to 14.08 mm.In another sample from Paiçandu, Paraná, Brazil (n = 31), the length of the forearm ranges from 34.8 to 39.5 mm, and the total length of skull ranges from 13.44 to 14.74 mm.Other measurements contrasting within-group variability are in Tab.I.The analyses of variance did not reveal any significant sexually dimorphic character in the sample from São Gabriel da Cachoeira, Amazonas, and only three out of 11 craniodental characters in the sample from Paiçandu, Paraná, were sexually dimorphic (Tab.II).These results indicate that secondary sexual dimorphism in cranial characters does not have a relevant role in the within-group variation sampled from distinct parts of the geographic range of M. albescens.Based on this result, males and females were pooled in geographic samples.
Qualitative characters: The skull profile varies considerably (Figs 2-5 and Tab, III), even among specimens from the same locality.The rostrum and the braincase are the most variable skull regions.The rostral slope, from the posteriormost region of frontal bone to the anteriormost region of maxillary bone, varies from slight (Fig. 4) to well-marked (Fig. 2).The braincase varies from globular to laterally flattened, with the braincase roof inclined forward or straight (Tab.III).Other characters vary within groups in the same way as among groups (Tab.III).Regarding pelage color, most specimens have bicolored dorsal and ventral pelages, with Cinnamon-brown bases and Antiquebrown tips dorsally and Mummy-brown bases and almost white tips ventrally.However, some individuals are slightly lighter or darker, but differences in skin color seem to be due to individual variation rather than geographically structured.

Geographic variation
The first two principal components account for 64% of the total craniometric variation.The first principal component (PC1, 46%) corresponds to a size vector based on the positive and relatively high magnitudes of all character loadings (Figs 6-8, Tab.IV).PC2 (18%) presented loadings of opposite signs for correlations with characters, with those related with length of skull (MAN) and rostrum (MTL and M13) contrasting with width of rostrum (POB).
A wide superimposition of scores is revealed for most geographic samples, but PC1 centroids of the Uruguayan sample (group 1) are the largest, followed by the samples from Paraná (group 2), Peru (group 4), Bolivia (group 3), Venezuela (group 6) and Amazon (group 5) (Fig. 6).The 95% confidence interval of scores of the Uruguayan sample is almost non-overlapping with scores from northernmost samples (groups 5 and 6).PC2 is not informative to the distinction of samples.The sample from Bolivia (group 2) is very variable and encompasses centroids of most samples, except Uruguay.
Similarly to the PC1 x PC2 plot of scores, the first two discriminant functions (CV1 and CV2), which accounted for 90% of the total discriminatory variation, partially separated two sets of samples, albeit with a wide superimposition mainly due to the sample from Bolivia.Four characters representing length of skull (CIL), length of rostrum (MTL and M13) and width of rostrum (POB) were most associated with CV1 (Figs 7 and 9, Tab.IV).A vector plot of the character loadings on CV1 ZOOLOGIA 28 (6): 789-802, December, 2011 (x-axis) revealed that the Uruguayan sample (group 1) is distinguished from the Amazonian (group 5), Peruvian (group 4) and Venezuelan (group 6) samples mainly by greater values in GLS and M13, showing more elongated skulls, whereas Paraná (group 2) and Bolivian (group 3) samples showed intermediate scores.Based on the discontinuities revealed, Uruguayan speci- An analysis of the descriptive statistics of craniodental measurements (Tab.I) shows that southernmost samples are, on average, larger than northernmost samples, with specimens from Uruguay (group 1) and Paraná (group 2) being the largest ones, followed by specimens from Bolivia, intermediate in size, and specimens from Peru, Amazon and Venezuela being the smallest.The two external measurements (FA and 3ML) do not distinguish northern and southern samples.
It is interesting to note that the geographically intermediate sample from Bolivia also showed intermediate scores in CV1 with respect to the samples from Southern Brazil and Uruguay and those from the Brazilian Amazon, Peru and Venezuela.This result, together with the size trend revealed by PC1, is indicative that the craniodental variation in M. albescens is structured in a trend of increasing size from northern to southern latitudes, without a detectable break at the scale of the analysis allowed by our samples.
Type locality: The description of M. albescens, was based on the chauve souris douzième of AZARA (1801).Azara's specimen   Remarks: As a consequence of its wide distribution across South America, M. albescens occurs in sympatry with almost all South American congeners.It differs from most of them by the presence of a fringe of hairs along the trailing edge of the uropatagium and by the frosted appearance of the dorsal pelage.The presence of a fringe of hairs on the uropatagium is a character shared with M. levis (I.Geoffroy, 1824) (MORATELLI & WILSON 2011b), and some specimens of this latter species display frosted appearance in the dorsal pelage (LÓPEZ-GONZÁLEZ et al. 2001, WILSON 2008).When this combination occurs, M. albescens can be distinguished from M. levis by its globular braincase and smaller ear length (9-14 mm in M. albescens, and 14-18 mm in M. levis).Skulls of fluid preserved specimens, with a flattened braincase resembling M. levis, can be identified as M. albescens based on overall size.Regarding other species, M.

DISCUSSION
Despite the number of specimens in museum collections (LAVAL 1973), the morphometric variation in M. albescens has been underestimated in previous studies.In addition, previous hypotheses of clinal variation in body size were confirmed here.Comparing specimens from Amazon basin with specimens from Mato Grosso (Brazil), Paraguay and Uruguay, LAVAL (1973) reported a possible pattern of clinal variation for two measurements with an increase in size towards the south, except for Peruvian and Ecuadorian samples.MYERS & WETZEL (1983) reported significant geographical differences in external and cranial measurements for samples from Bolivia, Chaco Boreal and eastern Paraguay, with Bolivian specimens as the most distinct.Our PCA and CVA analyzes refuted the distinctness of Peruvian and Bolivian samples regarding size and form of skull.PCA and univariate analyses revealed a southward trend of clinal variation for skull measurements, with specimens from northern South America cranially smaller than southern specimens.Results of CVA showed that the Uruguayan specimens have more elongated skulls in comparison with northern specimens.Despite the absence of central and northeastern Brazilian samples, our data revealed a southward trend of size increase.Part of the variation in M. albescens complies with Bergmann's rule, in that populations living in colder climates are larger than populations living in warmer climate regions (RENSCH 1938, MAYR 1942, 1956).This pattern is a valid ecological generalization for mammals, and has been verified for other New World species of Myotis, as well as other New and Old World species of bats (MEIRE 2003).In addition to this geographic component affecting cranial size, individual variation seems to have an important role in pelage color, as previously proposed by LAVAL (1973), as well as in the size and shape of skull.

Figures
Figures 10-11.Dorsal (10) and ventral (11) views of the skin of the neotype of Myotis albescens (AMNH 205195).Forearm length 34.5 mm.The electronic version of this article includes this figure in color.
was not preserved(WILSON 2008) and the type locality was fixed as Yaguaron, Paraguarí, Paraguay by the neotype designation of LAVAL(1973).

Table I .
Descriptive statistics, with measurements given in millimeters, for geographic samples (groups 1-6) of Myotis albescens, with males and females pooled.See text for a description of measurement methods and localities included in each sample.Morphometric and morphological variation in South American populations of Myotis albescens mens have more elongated skulls in comparison with Amazonian, Peruvian and Venezuelan specimens.

Table II .
F and p values (ANOVA) for secondary sexual dimorphism in craniometric and external measurements of two samples (São Gabriel da Cachoeira, Amazonas, Brazil, and Paiçandu, Paraná, Brazil) of Myotis albescens.Numbers in bold indicate statistically significant dimorphism (p < 0.05).
rance to ventral pelage; wing and interfemoral membrane colors ranging from Cinnamon-brown to Mummy-brown; plagiopatagium attached to feet by a broad band of membrane

Table III .
Occurrence and distribution of selected qualitative characters for geographic samples (groups 1-6) of Myotis albescens.

Table IV .
Vector correlation coefficients ("loadings") between original variables and principal components (PC1 and PC2) and between original variables and canonical variates (CV1 and CV2) for South American samples of Myotis albescens.Numbers in bold indicate vector correlations with magnitudes larger than |0.29|.
Handley, 1960M.atacamensis(Lataste, 1891)in having bicolored dorsal fur and the dorsal surface of the uropatagium almost naked, whereas in the last the fur extends distally on the dorsal surface of the uropatagium to a level halfway between the knee and ankle, and the dorsal fur is soft and tricolored, with black bases(LAVAL 1973); from M. aelleniBaud, 1979, M.albescens differs by the slightly bicolored fur (as opposed to tricolored); from M. chiloensis (Waterhouse, 1840) by the smaller overall size and globular braincase; from M. diminutusMoratelli  & Wilson, 2011by its wider skull, globular braincase and larger overall size; from M. oxyotus (Peters, 1867) by the lighter dorsal fur coloration and smaller overall size; from M.nesopolus Miller,  1900by the paler dorsal pelage and larger overall size; from M. nigricans (Schinz, 1821), M. lavaliMoratelli et al., 2011, and M.  izecksohni Moratelli et al., 2011by its globular braincase; from M. ripariusHandley, 1960, M. keaysi J.A. Allen, 1914, and M. ruber  (É.Geoffroy, 1806)by the color of fur, absence of sagittal crest, globular braincase and shorter rostrum; from M. simus Thomas, 1901 by the plagiopatagium attached at toes by a broad band of membrane and silkier and longer dorsal fur (> 5 mm). albescens