Evolution of erythrocyte morphology in amphibians (Amphibia: Anura)

We compared the morphology of the erythrocytes of five anurans, two toad species – Bufo gargarizans (Cantor, 1842) and Duttaphrynus melanostictus (Schneider, 1799) and three frog species – Fejervarya limnocharis (Gravenhorst, 1829), Microhyla ornata (Duméril & Bibron, 1841), and Rana zhenhaiensis (Ye, Fei & Matsui, 1995). We then reconstructed the ancestral state of erythrocyte size (ES) and nuclear size (NS) in amphibians based on a molecular tree. Nine morphological traits of erythrocytes were all significantly different among the five species. The results of principal component analysis showed that the first component (49.1% of variance explained) had a high positive loading for erythrocyte length, nuclear length, NS and ratio of erythrocyte length/erythrocyte width; the second axis (28.5% of variance explained) mainly represented erythrocyte width and ES. Phylogenetic generalized least squares analysis showed that the relationship between NS and ES was not affected by phylogenetic relationships although there was a significant linear relationship between these two variables. These results suggested that (1) the nine morphological traits of erythrocytes in the five anuran species were species-specific; (2) in amphibians, larger erythrocytes generally had larger nuclei.

Prior to statistics, all variables were tested for normality and homogeneity.We used linear regression, one-way ANOVA, principal components analysis and Tukey's post hoc comparisons to analyze the data.Throughout this paper, values are presented as mean ± SE, and the significance level is set at ␣ = 0.05.All statistical analyses were performed with the Statistica software (version 6.0 for PC, Tulsa, OK, USA).
The tests detailed previously were carried out using the topology including all collected amphibian species from Gymnophiona, Caudata and Anura.This topology of species was based on proximate phylogenetic correlation assembled from PYRON & WIENS (2011).We drew the tree and reconstructed the evolutionary history of ES and NS of amphibians by parsimony ancestral states in the program Mesquite 2.75 (MADDISON & MADDISON 2011).Because branch lengths lacked divergence time and genetic distance and any other metric proportional to the expected variance for the evolution of each analyzed trait were unavailable, we arbitrarily set the initial branch length to 1, which is appropriate for a speciation model of evolution (MARTINS & GARLAND 1991).
We used ordinary least squares (OLS) and phylogenetic general least squares (PGLS) regressions to estimate the slope for all conventional analyses.These two analyses were implemented in R 2.15.3 (R Development Core Team 2013), using the RMS (HARRELL 2012) andCaper (ORME et al. 2012) packages.We used PGLS regression to examine the relationship between NS and ES in amphibians.The PGLS analyses incorporate phylogenetic information into generalized linear models.They offer a powerful method for analyzing continuous data, and have been applied to estimate the evolutionary model and the relationships among the traits of interest (BARROS et al. 2011, WARNE & CHARNOV 2008).In PGLS, the strength and type of the phylogenetic signal in the data matrix can be accounted for by adjusting branch length transformations, which show the degree of phylogenetic correlation in the data.In this study, we used from a maximum likelihood approach to evaluate the phylogenetic effects ( = 0 indicates no phylogenetic effect, and = l indicates the strongest phylogenetic effect equivalent to that expected under the Brownian motion model).We used the Akaike Information Criterion (AIC) to estimate merits and drawbacks of the models tested.The best model has the lowest AIC.The model with better ût can be determined by a maximum-likelihood ratio test in which twice the difference in the natural log of the maximum likelihoods (LnL) of OLS and PGLS models will be distributed approximately as a 2 with degrees of freedom equal to the difference in the number of parameters estimated in the two models (WARNE & CHARNOV 2008).

Morphological traits of erythrocyte
The erythrocytes of the five anuran species are oval, and their morphological traits are depicted in Table 1.The results of One-way ANOVA indicate that the nine variables of erythrocyte morphology were all significantly different among the five species (Table 1).We found that (1) the mean values of EL and ratio of EL/EW and NL/NW were largest in D. melanostictus and smallest in F. limnocharis, the mean value of EW was larger in B. gargarizans than in the other species, the mean value of ES was larger in B. gargarizans and D. melanostictus than in the other species; (2) the mean values of NL and NS were largest in D. melanostictus and smallest in F. limnocharis and M. ornata, the mean value of NW was largest in B. gargarizans and smallest in M. ornata; (3) the mean value of nucleo-cytoplasmic ratio was largest in D. melanostictus and R. zhenhaiensis and smallest in M. ornata (Table 1).The variable coefficient was significantly different in NW (F 4, 45 = 4.59, p < 0.01, Fig. 1), but not in other erythrocyte morphological traits among the five species (all p > 0.05).The variable coefficient of NW was significantly larger in D. melanostictus and R. zhenhaiensis than in B. gargarizans, with F. limnocharis and M. ornata in between (Fig. 1).
A principal component analysis resolved two components (eigenvalues у 1) from nine variables of erythrocyte morphology, accounting for 77.6% of the variation in the original data (Table 2).The first component (49.1% of variance explained) had high positive loading for EL, NL, NS and ratio of EL/EW.The second axis (28.5% of variance explained) mainly represented EW and ES.Erythrocyte morphology differed significantly among the five anuran species in their scores on the first axis (F 4, 45 = 45.95,p < 0.0001; BG b , DMª, FL c , MO c , RZ b , Tukey's test; a > b > c) and the second axis (F 4, 45 = 7.38, p < 0.001; BGª, DM b , FL b , MO b , RZ b , Tukey's test; a > b) (Fig. 2).

Variability of erythrocyte morphology in amphibians
We assembled published data with our own data on ES, NS for amphibians (Appendix 1).Data from 109 species of amphibians show that mean ES ranged from 119.4 µm 2 to 2649 µm 2 (N = 108) and the mean NS ranged from 18.1 µm 2 to 517 µm 2 (N = 71).Our reconstruction of evolutionary changes in these variables shows strong positive correlations between NS and ES in amphibians (Fig. 3).The ES and the NS were both significantly different among the three orders of Amphibia (Both p < 0.01).Both traits were greater in Caudata than in Gymnophiona and Anura (Fig. 4).Table 3 summarizes the relationships between NS and ES in amphibians according to OLS and PGLS analyses.Mean NS was positively correlated with mean ES in both the OLS and PGLS model (Fig. 5, Table 3).PGLS analysis showed that phylogenetic relationships did not affect NS and ES ( = 0) although there were significant linear relationship between NS and ES (Fig. 5, Table 3).

DISCUSSION
Hematological parameters vary significantly among amphibian species (ARIKAN et al. 2010, BARAQUET et al. 2013).For example, OLMO & MORESCALCH (1975) documented that interspecific variation is significant in the volume of erythrocytes and nuclei of seven Plethodontidae (Amphibia: Urodela) species.In our study, we found species-specificity in nine morphological traits of erythrocytes in the five anuran species.In general, variation in the morphological traits of erythrocytes in toads (B.gargarizans and D. melanostictus) was larger than in frogs (F.limnocharis, M. ornata, and R. zhenhaiensis).Furthermore, GÜL et al. (2011) found that the number of erythrocytes is also different in toads and frogs.The mean value of erythrocyte counts was greater in toads (Pseudepidalea viridis and Pelobates syriacus; n = 850530/µl; GÜL et al. 2011) than in frogs (Hyla arborea, Rana dalmatina and Pelophylax ridibundus; n = 741332/µl; GÜL et al. 2011).The morphological traits of erythrocytes were different between toads and frogs and this difference may be attributed to the following three reasons.First, the different habitats of toads and frogs may affect the variability of erythrocyte morphology (ROMANOVA & EGORIKHINA 2006).Toads mainly inhabit terrestrial environments, whereas frogs inhabit semi-aquatic or aquatic environments (GÜL et al. 2011).The terrestrial habitat has selected a series of adaptive structures and mechanisms in frogs that have enabled them to function under conditions of changeable humidity and partial oxygen pressure in terrestrial environments (BARAQUET et al. 2013, FOXON 1964, WOJTASZEK & ADAMOWICZ 2003).Second, erythrocyte size may be dependent on the level of metabolism in vertebrates (WOJTASZEK & ADAMOWICZ 2003).Through our field investigation, we found that two toad species (B.gargarizans and D. melanostictus) that crawl slowly and have lower metabolic rate consume less energy than the other three species that are agile in their jumping and swimming activity.Therefore, erythrocyte morphology may have evolved to adapt to various levels of activity in vertebrates.Finally, the body size of animals influences erythrocyte size (FRÝDLOVÁ et al. 2012).In our study, the means obtained for the snout-vent length of two toad species (B.gargarizans and D. melanostictus) were greater than the means of the other three frog species (F.limnocharis, M. ornata, and R. zhenhaiensis); this distinction was consistent with erythrocyte size.This finding is logical from a physiological point of view, since smaller erythrocytes have relatively larger surface areas, and therefore, exchange oxygen more efficiently.It is reasonable to expect that erythrocyte size is adjusted to the actual mass-specific metabolic rate that gradually decreases during ontogenetic growth (CLEMENTE et al. 2009, SMITH et al. 2008).
The morphological traits of erythrocytes are variable among individuals of a species.HOTA et al. (2013) found that the erythrocyte profile of M. ornata is variable during the larval and adult periods.The coefficient of variation (CV) indicated that the level of difference varied among individuals in the same species.Our results showed that the mean values of CV of NW in D. melanostictus and R. Zhenhaiensis were greater than in B. gargarizans (Fig. 1).These differences may be attributed to the different habitats (RUIZ et al. 1983, SALAMAT et al. 2013) and/or variable activity levels (ALLANDER & FRY 2008, SYKES & KLAPHAKE 2008).Moreover, erythrocyte morphology varies with geography in amphibian species.We pooled erythrocyte size data on B. gargarizans from previous studies and our current study, and found that the erythrocyte profile (EL and EW) differed among three populations from different sampling locations (GUO et al. 2002, ZHOU et al. 2011).The EL and EW of B. gargarizans in Lishui (28°27'N, 119°53'E) were greater than in Chongqing (29°81'N, 106°39'E, GUO et al. 2002), which were greater than in Shuicheng (26.58'N, 104°82 'E, ZHOU et al. 2011).However, erythrocyte shape (ratio of EL/EW) showed an op-   1.50; Shuicheng: 1.57).These geographic variations in erythrocyte morphological traits may be associated with differences in latitude, elevation, or environmental and climatic variables in different sampling locations (GOODMAN et al. 2013).Previous studies have found that morphological variation in the erythrocyte traits of amphibians was greater than that in mammals, birds and reptiles (DUELLMAN & TRUEB 1994, GREGORY 2001a, LI et al. 1989, SEVINÇ et al. 2004, WU et al. 1998).Erythrocyte size in animals is generally negatively correlated with the place where the species appears in an evolutionary tree (whether more basal or more apical, indicating a more recent divergence in time).
Howerver, within Amphibia, species of Gymnophiona have larger erythrocytes than the other species of Caudata and Anura (SZARSKI & CZOPEK 1966).Similar results were found in our study, indicating that the ES and NS in Aunra were the smallest among the three orders, but the ES and NS in Caudata were larger than in Gymnophiona (Fig. 4).This may be the result of insufficient data from a limited number of species (only two species in Gymnophiona) collected from previous reports.Likewise, we still could predict that erythrocyte size in Caudata and Gymnophiona evolved to be larger than that in Anura.PGLS analysis to recover phylogenetic relationships, showed that these did not affect NS and ES, although there were significant linear relationships between NS and ES (Fig. 5, Table 3).Similar results were found in 24 species of salamanders, which indicate that the more standard relationships between cell size and NS are similarly significant whether phylogenetically-corrected or not (GREGORY 2003).The increase in erythrocyte size may occur adaptively (e.g., to provide more efficient metabolism), and is correlated with an increase in genome size (GREGORY 2001b).MUELLER et al (2008) demonstrated that positive direct correlations between genome size and NS are significant in the salamander family Plethodontidae.Moreover, the "nucleoskeletal" theory emphasizes the need for a balanced ratio of nuclear and cytoplasmic volumes for the maintenance of cell growth and division, and the key importance of cell size to organismal fitness (GREGORY 2003).

ACKNOWLEDGMENTS
Our experimental procedures complied with the current laws on animal welfare and research in China.Funding for this work was supported by the National Science Foundation of China (31270443, 31500308 and 31500329) and the Natural Science Foundation of Zhejiang Province (LY13C030004, LQ15C040002 and LQ16C040001).We thank Rui-Yu Yang for helping to collect the animals.

Figure 2
Figure 2 Positions of five anuran species in the space defined by the first two axes of a principal component analysis based on nine variables of erythrocyte morphology.Enlarged symbols show the mean values of scores on the two axes.

Figure 3
Figure 3 Mirror trees of the evolutionary history reconstructions of erythrocyte size (left side) and nucleus size (right side) of (blank branch is lack of data), according to phylogenetic hypotheses of PYRON & WIENS (2011).

Figure 5
Figure 5 Ordinary least squares (OLS) regression of nucleus size on erythrocyte size in amphibians.Regression equation and coefficient are given in the figure.

Figure 4
Figure 4 The erythrocyte size and nucleus size of different orders in Amphibia.Different superscripts indicate significant difference (Tukey's post hoc test, ␣ = 0.05, a > b).

Table 1 .
Descriptive statistics, expressed as mean ± SE and range, for morphological traits of erythrocytes in five anuran species in Lishui, China, and results of one-way ANOVA for each variable of erythrocytes with species as the factor.

Table 2 .
Loading of the first two axes of a principal component analysis on nine variables of erythrocyte morphology.
* Variables with the main contribution to each factor.

Table 3 .
Regressions of nuclear sizes (NS) on erythrocyte size (ES) in amphibians based on ordinary least squares (OLS) regression and phylogenetic generalized least squares (PGLS) regression.Significant associations between variables are shown in bold.