Cranial morphometric study of four giant flying squirrels ( Petaurista ) (Rodentia: Sciuridae) from China

: The present study revisited the controversial taxonomic status of Petaurista yunanensis , P. philippensis , P. hainana , and P. petaurista by using a considerably extended set of morphometrical characters (26 cranial variables from 60 adult specimen skulls). The results revealed no sexual dimorphism in any of the four species but confirmed significant craniometric differences among the four species in both the principal components analysis (PCA) and discriminant function analysis (DFA), with the greatest distinction observed between P. petaurista and other Petaurista species. Both univariate and multivariate analysis indicated that the morphological differences between P. yunanensis and P. philippensis were less than that between P. philippensis and P. hainana . The morphometric results were concordant in geographic patterns with mtDNA data from previous studies and indicated that P. petaurista , P. hainana, P. philippensis , and P . yunanensis could be recognized as valid species.

Petaurista are widely distributed in China and more than ten distinct species are recognized (Corbet & Hill, 1992;Wang, 2003;Zhang et al, 1997). However, several of these species are referenced with very few specimens or based solely on skins with no corresponding skulls (Allen, 1940;Ellerman, 1940), and some are actually the synonyms or subspecies of either the P. petaurista complex or the P. philippensis complex due to intraspecific geographic variations across their distributions in Asia. Corbet & Hill (1992) treated P. albiventer in Pakistan and southwest China as the synonym of P. petaurista and recognized P. philippensis as a distinct species consisting of many forms formerly assigned to P. petaurista, including forms distributed in Taiwan (P. grandis), southwest Yunnan (P. yunanensis), and Hainan (P. hainana). After comparing the pelage and cranial characteristics of P. petaurista and P. hainana, Huang et al (1995) considered P. hainana to be a valid species, but Wang (2003) treated P. hainana as a subspecies of P. yunanensis. Thorington & Hoffmann (2005) treated all Petaurista forms as eight valid species instead of nine as suggested by Corbet & Hill (1992), but they accepted the specific validity of P. philippensis and the subspecies status of P. yunanensis and P. hainana. Patterns of genetic variations observed in the complex of P. philippensis based on cytochrome b genes indicated that P. hainana, P. albiventer, and P. yunanensis could be distinct species (Yu et al, 2006). Some forms included in P. philippensis warranted separate specific rank based on molecular data (Oshida et al, 2000a, b;Yu et al, 2006), but without further evidence from morphometric data, much remains to be done to ascertain conclusively these specific conclusions.
Most recent phylogenetic studies have focused on molecular data analysis, but tracing changes in morphological characters is also an important way to evaluate the distribution of the characters on which those taxonomic units are based. Morphometric data are important to understand biological phenomena and have been used to evaluate cranial, dental, and body measurements of many mammals (Muñoz-Muñoz & Perpinan, 2010;Slábová & Frynta, 2007;Zelditch et al, 2004). Quantitative analysis of intra-and inter-specific variations at the morphological level is useful for detecting patterns of geographic variations and delimiting intra-or inter-specific evolutionary units. To date, however, there are currently no published reports of quantitative analysis based on morphological characteristics that would allow the identification of the morphotypes in the complex of P. philippensis and P. petaurista.
To discuss the taxonomic relationships of P. philippensis, P. yunanensis, P. hainana, and P. petaurista and test previous taxonomic hypotheses, the present study conducted a comprehensive morphometric study on the above Chinese Petaurista species based on samples subsequently collected from southwest Yunnan and the Island of Hainan, China. Multivariate analyses were used to produce an overview of the associations between morphological variables and species patterns and discuss the taxonomic implications of these flying squirrels. Our morphometric study could be complementary to studies of variations of DNA sequences in flying squirrels.

Specimens and data collection
According to the taxonomic assignments of Allen (1940) and Zhang et al (1997), a total of 60 intact adult skull specimens of P. petaurista, P. yunanensis, P. hainana, and P. philippensis were examined for morphometric study (Append. I). These specimens are from the Kunming Institute of Zoology, Chinese Academy of Sciences (KIZ, CAS) (Kunming, China), the Institute of Zoology, Chinese Academy of Sciences (IOZ, CAS) (Beijing, China), and the Guangdong Entomological Institute (GDEI) (Guangzhou, China).

Data analysis
Statistical analyses were performed using SPSS version 11.0 (SPSS Inc., Chicago, IL, USA). All variables were transformed into logarithms to eliminate the bias effect of large measurements in multivariate analysis (D'Elía & Pardiñas, 2004). Statistical differences were considered significant at P<0.05.
In this study, all related data were subjected to oneway ANOVA for calculating mean±SD. T-test was used to assess the sexual dimorphism between male and female groups by comparing the group means of cranial measurements. Multiple comparisons between taxa were made for all 26 cranial measurements to evaluate variations between samples. Multivariate analyses, including principal components analysis (PCA) and discriminant function analysis (DFA), were carried out to evaluate the degree of similarity and dissimilarity in cranial structures between the putative species and to determine how the taxa were related when all cranial characters measured are considered simultaneously.
The PCA is based upon the variance-covariance matrix of the log-transformed variables. The eigenvector scores describing the relative significance of each variable to the principal components were used to compare the cranial morphological similarities and differences. The PCA scatter-plot visually represented the variation among different individuals of the samples. The DFA was performed to investigate the integrity of the pre-defined groups and to predict group membership of specimens with the linear models of variables. Based on the derived discriminant functions, each individual was allocated to the group with nearest centroid, and the proportion of individuals allocated to each group was calculated.

Results
Mean±SD of 4 external and 26 cranial variables for the four taxa are presented in Tab. 1.

Univariate analysis
Univariate comparison revealed that the means of all variables were significantly different and, in general, tended to become progressively larger from P. petaurista, P. hainana, P. yunanensis, to P. philippensis. The t-tests of Equality of Group Means on 54 (30 males, 24 females) out of 60 specimens indicated there was no sexual dimorphism in the 26 cranial variables in the four Petaurista groups (Tab. 2). Quantitative pairwise comparisons of all cranial variables between taxa indicated that P. yunanensis was morphologically similar to P. philippensis, with 11 cranial measurements showing no significant difference (P>0.05) (Tab. 3). Also, five cranial variables were not significantly different between P. yunanensis and P. hainana.

Multivariate analysis
In PCA, the eigenvalues for the first three principal components were 20.62, 1.64 and 0.85, respectively, accounting for 88.91% of the total variance (Tab. 4). Most characteristics with high positive loadings on the first principal component suggested that this component (79.32% of the total variance) represented size variation within the samples. All specimens on the first principal component were clustered as three groups, P. petaurista, P. hainana, and the group of P. yunanensis and P. philippensis with considerable overlaps. The second principal component (6.32% of the total variance) was strongly correlated with ROB, ABL, ABB, and BZP (loadings>0.50), and the third principal component (3.27% of the total variance) was correlated primarily Fig.1 with HO (loadings>0.50) (Tab. 4). The first two principal components separated all specimens as four distinct groups (Fig. 2). The DFA identified the major patterns of morphological divergence in the crania among the four Petaurista groups. The variation pattern reflected by the first two discriminant functions was consistent with the morphological variations observed in the PCA, and all samples were clearly clustered as four distinguishable groups based on the 1 st and 2 nd discriminant functions (Fig. 3). In specimen reclassification by DFA, all individuals were properly assigned to their original groups on the basis of the studied measurements. Fig. 4 is the geographic distributions of all Petaurista samples used in this study.

Discussion
One contentious issue regarding the Chinese Petaurista is the taxonomic status of P. yunanensis, P. philippensis, P. hainana, and the populations of P. petaurista in China, which have long been controversial (Corbet & Hill, 1992;Ellerman, 1940;Ellerman & Morrison-Scott, 1950;Hoffmann et al, 1993;Huang et al, Tab. 3   1995; Oshida et al, 2000aOshida et al, , 2000bThorington & Hoffmann, 2005;Wang, 2003;Yu et al, 2006). By using a considerably extended set of morphometrical characters (26 cranial variables) and applying multivariate morphometric analyses, results of the present study confirmed the significant craniological differences in P. petaurista, P. hainana, P. yunanensis, and P. philippensis, with P. petaurista having the most pronounced morphological variations, particularly in metrical components of cranial and body size. Pelage coloration had been applied for classification of flying squirrels and led to many taxonomical disagreements due to numerous color variations within Petaurista, even between different sexes (Allen, 1940;Ellerman & Morrison-Scott, 1950;Oshida et al, 2004b). A series of color variations in pelage were observed among the Petaurista forms including both sexes (Allen, 1940;Oshida et al, 2004a), but the pairwise comparison of each of the 26 cranial variables revealed no sexual dimorphism in any of the four groups, implying that the divergence of coloration patterns in forms was due to environmental or genetic fluctuations over time.
Both univariate and multivariate analysis revealed that the skull morphometric characters used in this study were effective for discriminating the four Petaurista groups. Our analyses demonstrated that the four morphotypes of Petaurista were distinguished by a number of cranial characteristics. The morphometric variables, which caused the major distinction between those groups, were specifically located in the occipital, supraocular, and rostral regions, as well as in the prooticsquamosal length. Patterns of molecular sequence variations from previous studies (Oshida et al, 2000a;Yu et al, 2006) and the cranial morphological differences observed in this study indicate that P. hainana, P. philippensis, and P. yunanensis could be recognized as three distinct species. These differences were clear and reinforced the existence of three morphotypes of the complex P. philippensis; although the degree and form of the morphological differences might be related to their geographical variations. Even though molecular data suggested that P. philippensis is closely related to P. hainana and significantly distinct from P. yunanensis (Yu et al, 2006), many characteristics beyond those related to external morphology and pelage coloration were observed to be held in common between P. yunanensis and P. philippensis. A good example is the similarity in the pattern of overall cranial structure in multiple comparison analysis. The degree of the morphological variations between P. yunanensis and P. philippensis was less than that observed between P. philippensis and P. hainana, with eleven cranial measurements showing no significant difference (P>0.05) (Tab. 3). Petaurista yunanensis occurs from extreme southwestern Yunnan into Myanmar and Indochina and is extensively sympatric with P. philippensis in southwestern China (Wang, 2003;Zhang et al, 1997). The sharing of morphological characteristics between P. philippensis and P. yunanensis is related to their similar living conditions.
Petaurista hainana was considered a valid species based on both molecular and morphological data (Huang et al, 1995;Yu et al, 2006). Our morphometric results were concordant with mtDNA data of previous research (Oshida et al, 2000a;Yu et al, 2006) and demonstrated the significant differences between P. hainana and P. yunnanensis/P. philippensis. In both PCA and DFA, P. hainana was clearly separated from other three groups (Fig. 2, 3), with 21/26 cranial variables being significantly different (P<0.05) (Tab. 3). Petaurista hainana is confined to tropical forests on Hainan Island of China and P. philippensis and P. yunanensis are distributed in mountainous coniferous, dry deciduous and evergreen forests at different elevations in western Yunnan of China. The phenotypic divergence of P. hainana in relation to P. philippensis and P. yunanensis is likely associated with their geographical distributions and living conditions and could be viewed as a reflection of adaptations to various ecological niches. The differences in skull morphology suggest that P. hainana is neither the synonym of P. philippensis nor a subspecies of P. yunanensis or P. petaurista (Corbet & Hill, 1992;Thorington & Hoffmann, 2005;Wang, 2003), but a valid species in its own right.
The greatest distinction observed was between P. petaurista and other Petaurista forms. Petaurista petaurista displayed a relatively high level of diversity in skull morphology, with 22/26 cranial variables significantly different from P. hainana, P. philippensis, and P. yunanensis at P<0.001 level (Tab. 3). Based on 26 morphological cranial variables, the specimens of P. petaurista formed a distinct aggregate in both PCA and DFA (Fig. 2,3), consistent with the results of Oshida et al. (2000a) and Yu et al (2006). It is obvious that P. petaurista, P. hainana, P. philippensis, and P. yunanensis are taxonomically distinct and distinct valid species.