Performance of WI and RBV for species diagnosisnext section
In our previous work concerning the capacity of the Wolterstoff Index to discriminate females of the taxa, we identified several potential problems (Arntzen and Wallis, 1994): 1) statistical representation (the need for a mean value), 2) non-biological variation (preservation and measurement differences), 3) sexual variation (WI is higher for males than for females), allometric variation (WI decreases with size), 4) geographic variation (nearby animals may tend to be more similar within species), 5) hybridisation (hybrids between two species can have intermediate values typical of a third species), and 6) circular reasoning (the need for an independent character set to determine the significance of WI). All of these factors to some extent reduce the efficacy of WI and compromise classification made solely on this basis. The WI purports to capture information useful in taxonomy but in fact confounds the variables limb length, vertebral number, and possibly vertebral length.
Using mtDNA haplotype (Wallis and Arntzen, 1989) we showed that WI makes a good approximation to species classification, but can only be used with confidence for discriminating adult T. dobrogicus from the other species. In contrast, the number of rib-bearing vertebrae (RBV) as assessed by radiography eliminates all but one of these problems. Because RBV is a direct discrete meristic count, stable through the lifetime of the individual, and with limited intraspecific geographic differentiation its use in conjunction with a diagnostic genetic character leaves only the issue of hybridisation to be addressed. Variation caused by hybridisation near regions of parapatry is difficult to disentangle from intraspecific geographic variation on the basis of morphological data alone. Under both scenarios, variation will be most clearly expressed when samples from remote parts of the geographic range are compared. However, the observation that character state changes are consistently in the direction to that of the neighbouring species supports the hypothesis of hybridisation, rather than that of intrinsic geographic variation. For example, the significantly different values in T. carnifex macedonicus (high WI, low RBV compared to T. carnifex carnifex) may well be a result of introgression from T. karelinii, whose mtDNA prevails in some T. carnifex macedonicus populations (Wallis and Arntzen, 1989). Note, however, that the meristic count in hybrids is not necessarily intermediate to that of the parental species, as documented for salmonid fishes (Leary, Allendorf and Knudsen, 1985). If hybridisation between taxa of crested newts is a common phenomenon, it should be possible to find genetic markers for the species covarying with interspecific morphological variation. The observed breakdown of the diagnostic power of RBV in areas where taxa meet may reflect the true nature of characters in a contact zone.
It is possible that RBV is influenced early on in development. Indeed, vertebral count is often highly labile in fish and salamanders (e.g. McDowall, 1970; Jockush, 1997), with cooler conditions generally slowing development and increasing several meristic counts (Barlow, 1961 and references in Jockush, 1997). In Triturus vulgaris the average RBV count increases with the temperature at which the embryos are raised (Orska and Imiolek, 1962), while for other salamander species more complicated environmental effects were found (Lindsey, 1966; Peabody and Brodie, 1975). To address the question to what extent the variation in RBV is genetically determined and to gain insight into the relationship between embryonic development and adult morphology requires experimental work. Triturus dobrogicus might be the best species to work with because it is naturally polymorphic for RBV.
Elongation of the body and a reduction in the length of limbs in vertebrates generally indicates a more piscine locomotion by sinusoidal body undulation. Although this can be associated with some terrestrial habitats (as in sand-swimming skinks), it is more usually an adaptation to a more aquatic mode of life. Reduction of trunk size and the development of robust legs is more unequivocally associated with a terrestrial mode of life where the body requires more support (Young, 1950; Lande, 1978). The wide variation in body shape in the leg-less Gymnophiona, ranging from stout to thread-like, may also be associated with locomotory behaviour and ecological adaptation (Renous and Gasc, 1989). In the salamander genus Batrachoseps, extensive geographic variation in RBV, possibly related to fossoriality, has been observed, most of which was shown to be genetically determined (Jockush, 1979). Selection on female fecundity (correlated with interlimb length) and male sexual performance (stature at display correlated with leg length) may also play a role (Arntzen and Wallis, 1994).
Morphology predicts the aquatic period of T. marmoratus to be short, that of T. carnifex to be intermediate and that of T. cristatus - and T. dobrogicus in particular - to be long. While of course the phenology of breeding may be different from year to year, this prediction appears to be corroborated by field data. Triturus marmoratus spends annually approximately three months in the water, T. carnifex four months, and T. cristatus five months (Bouton, 1986; Griffiths and Mylotte, 1987; Andreone and Giacoma, 1989), while the aquatic phase of T. dobrogicus usually lasts six months (Karaman, 1948; Jehle et al., 1996). No data are available on the phenology of T. karelinii. This species is predicted to have a short aquatic phase, of intermediate length to that of T. carnifex and T. marmoratus. In terms of performance, the sinusoidal swimming ability should be best in the lowest WI / highest RBV count taxon, i.e., T. dobrogicus and relatively poor in T. marmoratus. Indeed, the ecological niche of T. dobrogicus is different from that of the other big-bodied species. It may co-exist with fish in oxbows, river margins and other non-temporary water bodies (Arntzen et al., 1997). The observed ventral aposematic colouration pattern in the T. cristatus superspecies versus the dorsal aposematic colouration of T. marmoratus (particularly evident in the entirely terrestrial juveniles) provides further support to our interpretations. Predictions about performance, such as in the gathering of food or predator avoidance behaviour in the aquatic versus terrestrial habitat could be tested experimentally.
Taxonomy and phylogeny
The depth of the differences among taxa, and the relative sharpness of the contact zones led us to follow earlier suggestions to raise the taxa to full species status (Bucci-Innocenti, Ragghianti and Mancino, 1983) as have others (Frost, 1985). The available data, unfortunately, do not support a single phylogenetic hypothesis for the four taxa comprising the T. cristatus superspecies. RBV is primitively 14 in the genus Triturus (B. Lanza et al., in prep.), rendering RBV of 13 an autapomorphic character state for T. marmoratus and RBV of 15 - 18 a synapomorphic character state series for T. carnifex - T. cristatus - T. dobrogicus. This character alone would suggest that T. karelinii represents the oldest extant crested newt lineage, followed by T. carnifex, T. cristatus and T. dobrogicus. This evolutionary classification is supported somewhat ambiguously by the phenetic analysis of protein electrophoretic data (Crnobrnja, Kalezi’c and D’zuki’c, 1989) but contradicted by another such study (Litvinchuk et al., 1994). The phylogenetic analysis of molecular data (mtDNA RFLP’s) suggests a different phylogeny. Looking at the most-parsimonious mtDNA tree (Wallis and Arntzen, 1989: Fig. 4), a tree that optimizes RBV character-state change [tree : character structure
involves moving only the ‘DOB’ branch (with terminal taxon number 11). [CAR? and KAR? refer to deeply differentiated haplotype lineages within T. carnifex and T. karelinii. We now recognize the first of these as belonging to T. carnifex macedonicus (see below) while the other will be subject to taxonomic description at the subspecific level (S. Litvinchuk et al., in. prep.)]. If DOB were placed with the T. cristatus (CRI) haplotypes, increased RBV becomes a derived character interior to the tree, with the more massive built newts basal. Although this haplotype tree has 70 steps as opposed to 67 in the published maximum parsimony tree (Wallis and Arntzen, 1989: Fig. 4), there is no bootstrap support above 50% for any of the crested newt species-level structure. That is to say, the relationship DOB(KAR(CAR,CRI)) is only defined by three synapomorphies in total and the tree could more conservatively be depicted as a four-way polychotomy at this level. This incomplete resolution is appreciated by Wallis and Arntzen (1989: 99) and emphasised by further analysis (Faith and Cranston, 1991; Faith, 1992). However, strong support is obtained from the ‘CAR’ and ‘CAR?’ mtDNA haplotypes for the sister taxon status of Italian and central Balkan crested newts. These groups of populations are also united by the synapomorphic character state RBV = 15. We therefore consider the crested newts from the central Balkan to belong to T. carnifex. The range of this species is disjunct (see below). Newts from both parts of the range are phenotypically distinct: T. carnifex from the western (Italian and Slovenian) part of the range typically have few, large, ill-defined black dots on the bellies, whereas T. carnifex from the eastern part of the range (F. R. Yugoslavia and Greece) have ventral coloration patterns with many sharp-edged spots, as Freytag (1988) observed, not unlike that of T. cristatus (Plates I-II). Crested newts of the eastern group were described as Molge karelinii var. macedonica Karaman, 1922. Considering the morphological and genetic differentiation between the forms, we propose raising this taxon to the subspecies level and, supported by phylogenetic arguments, classifying it as belonging to T. carnifex (not T. karelinii as suggested by Karaman, 1922). Therewith, Triturus carnifex var. albanicus Dely, 1959 is a junior synonym of T. carnifex macedonicus (Karaman, 1922). Following our correspondence with co-author J. Crnobrnja-Isailovi’c (J. W. Arntzen, in letter, 1996) this taxonomic solution is accepted by Kalezi’c et al. (1997), although the taxon is incorrectly referred to in feminine gender.
The two more massive newt species, T. karelinii and T. carnifex, show much greater restriction site variation than the other two species (Wallis and Arntzen, 1989). They also have slightly larger mitochondrial genomes and a greater tendency for insertions in the control region (Wallis, 1987). These factors suggest that the two northern species may have been subjected to long-term small population size during glaciations (Wallis and Arntzen, 1989), and it is conceivable that the evolutionary change in vertebral count is related to this population genetic feature.
Distribution and biogeography
Crested newts appear to be absent from the largest part of Bosnia-Hercegovina (see for example Schmidtler and Schmidtler, 1983; Kalezi’c, D’zuki’c and Tvrtkovi’c, 1990; Kalezi’c et al., 1997). The southeasternmost localities of T. carnifex to the northeast of the perceived gap in the species distribution are sites 21 - 23 [Belovar Moravce (Table 2); Plitvice (Fejervary-Langh, 1943) and Licki Osik (Kalezi’c et al, 1990). Further to the southeast T. carnifex is found at sites 24 - 27 [Donja Dubrava (Kalezi’c et al., 1990), Sarajevo (Bolkay, 1929; communicated by G. D’zuki’c), the Zelengora Mountain (Bolkay, 1928) and Dobrsko Selo (Kalezi’c and D’zuki’c, 1990). The easternmost recorded locality is site 29 at Dimitrovgrad (Radovanovi’c, 1964). Crested newts of unknown taxonomic affinity were recorded at the Dalmatian coast [site 28, situated in between Sebenico (= S’ibenik) and Spalato (= Split) (Werner, 1897, also mentioned by Buresh and Zonkov, 1941), but with no clearly independent confirmation for over a century we doubt the validity of this record. Dzuki’c (1993) considers the distribution of T. carnifex not to be interrupted but continuous, following a strip of land to the south of the Sava river, without, however, presenting data supporting this view. The area where crested newts are absent coincides with the core area of the karst (Sket, 1994), where most natural water bodies are ephemeral and do often not support the larval development of species with a prolonged larval phase, such as crested newts. The small-bodied newts such as T. alpestris and T. vulgaris in contrast are widespread and locally abundant. They may reproduce successfully in shallow and temporary ponds such a wheel ruts (Winkler and Brauns, 1990) and the dispersal rate for the small newt T. vulgaris is estimated to be higher than that for the big newt T. cristatus (Stensjö, 1998). While most contemporary newt ponds are man-made and rarely desiccate (i.e., watering holes for cattle), the puddles formed by fallen trees and springs may originally have been the typical breeding habitat for the small bodied species.
The distribution of the four crested newt species in F. R. Yugoslavia is complex (Fig. 3c). Triturus dobrogicus is found all over the Pannonian and Dobrogean Plains. Both parts of the range are probably connected by the Danube where flowing through the Iron Gate (Arntzen et al., 1997). Triturus cristatus has a wide European range, is widespread over Romania and reaches southwards over the Iron Gate into Yugoslavia. Triturus carnifex macedonicus is widespread over most of Yugoslavia, the Former Yugoslavian Republic of Macedonia, Albania, and northern Greece. Triturus karelinii is found immediately south and southeast of Belgrade. The available evidence suggests that the local distribution is in a small pocket - an enclave, geographically isolated from the main T. karelinii distribution in Bulgaria, Thrace, and Turkey (Fig. 3c). However, a link between the parts, along a narrow strip in northeastern Yugoslavia (as in Arntzen, 1995 and in Kalezi’c et al, 1997: Fig. 6), cannot be excluded. The further surveying of eastern Yugoslavia and northwestern Bulgaria is required to settle this issue.
On a gross geographic scale, phenotype distributions, and mtDNA haplotype distributions are concordant. However, in northern Yugoslavia the ‘KAR?’ mtDNA haplotype is more widespread than the T. karelinii phenotype distribution would suggest (Fig. 3) (Wallis and Arntzen, 1989). The ‘KAR?’ mtDNA haplotype is locally found in T. dobrogicus, T. cristatus, and T. carnifex macedonicus populations. The reverse situation, with a foreign haplotype in T. karelinii, has been observed once (the ‘DOB’ haplotype in population 15). Populations with foreign haplotypes possess either two haplotypes - the original plus an alien, such as at site 15 and 34 in T. karelinii and T. dobrogicus),or just the alien haplotype [‘KAR?’ in T. cristatus (site 30 and 37, N = 5) and in T. carnifex macedonicus (site 35 and 36, N = 16; Wallis and Arntzen, 1989). To account for these observations we suggest the following scenario. In former times T. karelinii was more widespread than at present, with a range approximately coinciding to the present day distribution of the ‘KAR?’ haplotype. By dispersing southwards and northwards, respectively T. cristatus and T. carnifex macedonicus superseded T. karelinii, in which process the range of T. karelinii south of Belgrade became isolated from the main stock (see the arrows in Fig. 3c). The genetic interactions between T. karelinii at one side and T. cristatus and T. carnifex macedonicus at the other where such that the formation of F1 hybrids was asymmetric, with hybrid offspring and the subsequent backcrosses possessing the (maternally inherited) T. karelinii mtDNA. This scenario is surprisingly similar to the one we described for T. cristatus - T. marmoratus interactions in western France (Arntzen and Wallis, 1991). In France, T. cristatus supersedes T. marmoratus, forming T. marmoratus enclaves in the process. Hybridisation between the species is strongly asymmetric, with F1 adults derived from matings of T. cristatus mothers and T. marmoratus fathers significantly outnumbering the reverse combination. The facts responsible for this phenomenon are largely unknown but may involve the genetic incompatibility of the nuclear and mtDNA genomes (J. W. Arntzen et al., in prep.). By comparing past and present distributions, the rate at which T. cristatus takes over from T. marmoratus has been estimated as averaging one km a year. The process may be triggered, or accelerated, by the removal of hedgerows, modifying a landscape with terrestrial features favourable to T. marmoratus, the most terrestrial among the big-bodied newt species, into one favourable to the more aquatic T. cristatus. The habitat preferences of the various crested newt species in eastern Europe, with the exception of T. dobrogicus, are poorly understood and it is unclear which ecological parameters affect their distribution or change in distribution. Another area of complexity is that around Vienna, where T. carnifex, T. cristatus and T. dobrogicus meet (Fig. 3b). At sites 12 and 36 newts were found with T. carnifex phenotype and the mtDNA haplotype typical for T. dobrogicus, matching a similar observation at Tasovice in the Czech Republic (48º49’ N, 16º09’ E; J. Pialek, V. Zavadil and J. W. Arntzen, unpubl.).
As noted by Crnobrnja-Isailovi’c et al. (1997), the remarkable variability in Balkan crested newts should provide valuable insights into the evolution of the group. Palaeontological and various molecular methods have provided some clues towards the timing of the radiation of the T. cristatus superspecies (reviewed in Oosterbroek and Arntzen, 1992). This period, which can be placed at 2-5 Ma, was one of great geographical and geological complexity in southeastern Europe (Crnobrnja-Isailovi’c et al., 1997 and references therein). However, our ability to associate the historical patterns of fragmentation, speciation and dispersal with palaeogeography is, as yet, hampered by the absence of a well-supported phylogeny.