Contributions to Zoology, 68 (3) 181-203 (1999) SPB Academic Publishing bv. The Hague Geographic variation and taxonomy of crested newts (Triturus cristatus superspecies): morphological and mitochondrial DNA data ² J.W. Arntzen¹ & Graham+P. Wallis 1 Department of Zoology and Anthropology, Faculty of Sciences, University of Porto, CECA/1CETA/UP, 2 Campus Agrdrio de Vairao, 4480 Vila do Conde, Portugal; Department of Zoology and Centrefor Gene Research, University ofOtago, P.O. Box 56, Dunedin, New Zealand crested Keywords: asymmetric hybridisation, biogeography, newt, geographic variation, morphology, Triturus mtDNA, taxonomy, cristatus Abstract Introduction Within the in the the newt genus Triturus, large-bodied species Studies of geographic variation within and among T. show cristatus (crested newt) superspecies an unusual degree closely related taxa have provided crucial infor- ofvariation in relative trunk length as a result ofamong-taxon mation on the nature of species and the process of variation in interlimb vertebral count. Here we examine the speciation (Mayr, 1963). Studies of spatial mor- systematic value of this feature as assessed by both exterior of phological heterogeneity are interest from the measurement (Wolterstorff Index) and direct radiographic count of view of adaptation since differences in ofrib-bearing vertebrae, with particular reference to a number point ofconfoundingfactors (sex differences,hybridisation, geographic morphology are likely to be functional or at least mtDNA variation, allometry, preservation effects). Using our have consequences on fitness and ecology. A proper which with haplotype data, are largely concordant geographic analysis of morphological variation requires inde- distributionof species, wefind that direct count ofthe rib-bearing pendent assessment of the phylogenetic relation- vertebrae performs more reliably (14% misclassification) than ships of morphotypes (e.g. Losos, 1990). A central external measurement (31% misclassification) as a species question to the systematist is: does geographically identifier. We therefore recommend this feature as a taxonomic differentiation reflect local tool, although (like external measurement) it breaks down near bounded, morphological for the observed which would be hybrid zones. To account biogeographicalpattern adaptation (in case convergence - scenario is and phenotype genotype discrepancies, a presented expected to be common), or is it a function of phylo- that combines movement ofthe contact zonebetween taxa the which geny (in case morphotypes should be mono- with This asymmetric hybridisation. scenario applies to spe- phyletic) (Harvey and Pagel, 1991; Bernatchez, cies interactions in eastern Yugoslavia and western France. 1995)? Adaptation and phylogeny are not, of course, mutually exclusive explanations of character dis- tributions. Contents Adaptive morphological change can in occur and be proliferated by cladogenesis the absence of reversal. To resolve this, the first Introduction 181 step Material and methods 182 is to determine whether different character sets Results 183 define the same taxa. Concordance oftaxon bound- Discussion 193 aries resolved different by means, e.g., morphol- Performance of WI and RBV for species diagnosis 193 ogy and genetics, is strong evidence for their real- Adaptation 194 ity (Hillis, 1987; Avise and Ball, 1990). Indeed, Taxonomy and phylogeny 194 often becomes Distribution and biogeography 195 morphological divergence only Acknowledgements 197 apparent after the taxa have been resolved geneti- References 197 cally (e.g., Arntzen and Garcia-Paris, 1995; Appendix 199 McDowall and Wallis, 1996). Here we look at 182 J. W. Arntzen & G. P. Wallis - Geographic variation in crested newt superspecies geographic variation in vertebral count in the Tri- of the five phenotypes was associated with one or turus cristatus superspecies (or Artenkreis sensu more characteristic mitochondrial genotypes (Wallis and in WI Rensch) and re-evaluate this character as a tool Arntzen, 1989). Significant differences for species identification, with particular reference were observed between groupscarrying these geno- be tool for to the several hybrid zones present in the group types, and therefore WI can used as a Limb interlimb dis- and our genetic analyses of these zones (Wallis species diagnosis. length and and Arntzen, 1989; Arntzen and Wallis, 1991; JW tance are sexually dimorphic characters, making it Arntzen, in prep.). necessary to calibrate WI for males and females Wolterstorff revised technical mtDNA data In his classic paper, (1923) separately. For reasons, were the taxonomy of the crested newt, Triturus cristatus available for (almost exclusively) femalenewts only index. WI for using a morphological He distinguished four (Wallis, 1987). However, we measured males taxa at the subspecies level [carnifex (Laurenti, from the same populations and because (with a few 1768), cristatus (Laurenti, 1768), dobrogicus (Kiri- exceptions) they belong to the same species, WI tzescu, 1903) and karelinii (Strauch, 1870)] and can be calibrated for males as well. The number of provided compelling phenotypic descriptions for rib-bearing vertebrae (RBV) varies in parallel with each of these. He provided a crucial discriminator the inverse of WI (Arntzen and Wallis, 1994; for taxon identification, which has become known Crnobrnja-Isailovic et ah, 1997). Similar patterns as the Wolterstorff Index (WI). WI is the ratio have been documented for lizards (Greer, 1987; between forelimb length (PaL) and interiimb dis- Griffith, 1990; Caputo, Lanza and Palmieri, 1995), * WI = PaL 100 tance (LiE) and is defined as / fish (Lindsey, 1975) and other salamanders (Jock- LiE. We have reviewed the available data on WI ush, 1997). and have discussed some of the practical and theo- Here we present additional data to calibrate WI and and document the existence retical pitfalls associated with its use (Arntzen and RBV of geographic Wallis, 1994). WI succinctly summarises the mor- variation across and within species. The results differentiation to much the confirm the limited value of WI for phological among taxa diagnostic RBV same extent as more sophisticated methods of purposes. The character on the other hand is multivariate morphometric analysis used in crested shown to be diagnostic and without much intraspe- cific variation. it to newts and other urodeles (Kalezic et ah, 1990; geographic Moreover, appears Arntzen and Wallis, 1994; Sket and Arntzen, 1994; be largely immune to observer bias, developmen- Arntzen and Sket, 1997; Cvetkovic, Kalezic and tal or sexual variation, or preservation artefacts. Dzukic, 1997). From a statistical standpoint, indi- Applying the character at the population level pro- undesirable duces coherent of variation ces may have certain properties (Atchley a pattern geographic that et ah, 1976), but the WI has become a popular tool coincides with that of the species, although external it for species identification of crested newts in the (like measurements) breaks down near field (for example Grillitsch et ah, 1983). How- hybrid zones. To account for the observed pheno- ever, to avoid circular reasoning, the calibration of type - genotype discrepancies and biogeographi- WI requires independent criteria, i.e., an identifi- cal patterns in eastern Yugoslavia and western is that combines the cation without reference to body shape. France, a scenario presented In and its the crested newt superspecies sister movement of the contact zone between taxa with taxon T. marmoratus (Latreille, 1800), WI increases asymmetric hybridisation. in the order: dobrogicus — cristatus — carnifex —— karelinii —marmoratus, describing a morphologi- cal series from slender and short-legged to stout Material and methods and The also character- long-legged. taxa possess istic colouration patterns of that are described on Forelimb length and interlimb distance weremeasured with plastic Plates I-II (for approximate species distributions callipers (0.1 mm precision) on preserved or live material, the representing 142 populations across total range. X-ray see Fig. I). We previously gathered mtDNA RFLP photographs were taken on preserved or sedated specimens data from female newts and concluded that each (representing 116 populations) with an ‘Elinax 90/20’ and an - Contributions to Zoology, 68 (3) 1999 183 Fig. I. Distribution of Triturus marmoratus and five taxa in the T. cristatus superspecies (after Arntzen, 1995). The crossed hatching covering Yugoslavia and adjacent regions refers to T. carnifex macedonicus, a taxon recognition resurrected from Karaman (1922). of0.7 seconds 4mA D10DW exposure at 38kV, on Agfa-Gevaert type (Table I). For reference, we include published and X-ray film. Additional data on WI (31 populations) RBV data from studies dealing with three or more taxa (38 populations) were obtained from the literature(Vallee, 1959; that were identified phenotypically (Table I). Statis- Kalezic et ah, 1990; Crnobrnja-Isailovic et ah, 1997). Analysis tically significant differences were found between ofvariance (ANOVA) and othertests followed Sokal and Rohlf all combinations for WI one-tailed t- (1981) and Siegel and Castellan (1988). Logistic equations were species (four determinedwith SPSS 1990), tests were P < (SPSS, Non-parametric tests between consecutive groups for each sex: performed with StatView (StatView, 1988), Association of 0.001 in all cases, except for males T. cristatus - intraspecific morphological variation and geographical