Fossil Aphanius (Teleostei, Cyprinodontiformes) from southwestern Anatolia (Turkey): a contribution to the evolutionary history of a hotspot of freshwater biodiversity Tanja SCHULZ-MIRBACH Bettina REICHENBACHER Ludwig-Maximilians-University, Department of Earth and Environmental Sciences, Section Palaeontology, Richard-Wagner Str. 10, D-80333 Munich (Germany) t.schulz-mirbach@lrz.uni-muenchen.de b.reichenbacher@lrz.uni-muenchen.de Schulz-Mirbach T. & Reichenbacher B. 2008. — Fossil Aphanius (Teleostei, Cyprinodontiformes) from southwestern Anatolia (Turkey): a contribution to the evolutionary history of a hotspot of freshwater biodiversity. Geodiversitas 30 (3) : 577-592. KEY WORDS Teleostei, Cyprinodontiformes, otolith, Neogene, Lake Burdur, Aphanius, Anatolia, Fourier analysis, freshwater biodiversity. ABSTRACT Until recently, only a single fossil species of the cyprinodontiform genus Aphanius was known from Anatolia (Turkey), mainly based on fossil otoliths. As Anatolia is a diversity hotspot of this genus, it was of peculiar interest to investigate recently found fossil otoliths of ?upper Pliocene-lower Pleistocene age from the Yassigüme section located in the Burdur Basin in southwest Anatolia. We studied the morphological relations to otoliths of extant Aphanius anatoliae sureyanus inhabiting present-day Lake Burdur by conducting Fourier shape and statistical analyses (principal components analysis PCA, canonical discriminant analysis CDA). For further comparisons, we included a population of A. a. anatoliae (at Lake Salda) nearby Lake Burdur and one population of A. danfordii from the Kızılırmak River in northern central Anatolia. The contour of the fossil otoliths closely resembles that seen in the otoliths produced by the extant subspecies A. a. sureyanus from Lake Burdur. On the other hand, the fossil otoliths are distinctly different from those of the extant species A. danfordii. Furthermore, the fossil otoliths reveal more proximity to A. a. sureyanus than to A. a. anatoliae of that region, and both extant subspecies show a certain distance to one another in the CDA; thus we suggest that this might be explained by a diversification of the subspecies that had begun before the investigated fossil populations existed. Based on the strong similarity of the fossil otoliths with A. a. sureyanus, we conclude that they do not represent a new fossil species, and thus are preliminary denominated as †A. cf. anatoliae sureyanus. GEODIVERSITAS • 2008 • 30 (3) © Publications Scientifiques du Muséum national d’Histoire naturelle, Paris. www.geodiversitas.com 577 Schulz-Mirbach T. & Reichenbacher B. MOTS CLÉS Teleostei, Cyprinodontiformes, otolithe, Néogène, Lac Burdur, Aphanius, Anatolie, analyse de Fourier, biodiversité des faunes d’eau douce. RÉSUMÉ Des Aphanius fossiles (Teleostei, Cyprinodontiformes) du Bassin de Burdur au sudouest de l’Anatolie (Turquie). Une contribution à l’histoire évolutive d’un centre de diversification de biodiversité des faunes d’eau douce. Jusqu’à présent le genre Aphanius n’était connu en Anatolie (Turquie) que par une seule espèce fossile surtout représentée par quelques otolithes. L’Anatolie étant un important centre de diversification pour ce genre, il est intéressant d’étudier de nouveaux otolithes fossiles récemment récoltés dans le Pliocène (inférieur ou supérieur ?) à Yassigüme dans le bassin de Burdur (sud-ouest de l’Anatolie). Nous avons comparé les morphologies de ces otolithes fossiles avec celles des otolithes de l’espèce actuelle Aphanius anatoliae sureyanus qui vit dans le lac de Burdur. La morphométrie a été menée par analyse de Fourier des contours et les résultats ont été traités statistiquement par une analyse en composante principale (ACP) complétée par une analyse canonique discriminante (ACD). Pour enrichir le champ des comparaisons, nous avons également pris en compte une population d’A. a. anatoliae provenant du Lac Salda proche du lac de Burdur et une population d’A. danfordii provenant du fleuve Kızılırmak situé dans la partie septentrionale de l’Anatolie centrale. Les contours extérieurs des otolithes fossiles sont morphologiquement très proches de ceux des A. a. sureyanus qui vivent actuellement dans le lac de Burdur. Par contre les otolithes fossiles sont clairement distincts de ceux de l’espèce actuelle A. danfordii. En outre, la similitude morphologique est plus forte avec A. a. sureyanus qu’avec A. a. anatoliae, deux sous-espèces actuelles qui vivent à peu de distance l’une de l’autre mais qui, comme le montre l’ACD, n’occupent pas exactement la même partie de l’espace morphologique. Ces résultats nous conduisent à supposer que la différenciation sous-spécifique était déjà bien avancée lors de l’épisode de fossilisation du Pliocène. La forte similitude morphologique qui existe entre les otolithes fossiles et A. a. sureyanus ne permet pas de considérer les formes fossiles comme une nouvelle espèce. Nous préférons donc, au moins provisoirement, les désigner comme †A. cf. anatoliae sureyanus. INTRODUCTION Otoliths are composed primarily of aragonite and of some organic material. They lie in membranous sacs of the inner ear of teleostean fishes where they are involved in the senses of balance and hearing (cf. Popper et al. 2005). Three different otolith types on each side of the head can be distinguished: the utricular, the saccular, and the lagenar otolith (cf. Nolf 1985), which are named after the three types of otolithic endorgans (utriculus, sacculus, and lagena) in which they are located. Based on the symmetry of the inner ear in the head, a left and a right member of each otolith type can be identified. In most Teleostei, the saccular otolith is the largest or most 578 robust otolith, and thus usually has the potential of becoming fossilized. As the saccular otolith has a species-specific morphology, studies on its contour and characteristics have contributed considerably to our knowledge and understanding of fossil and extant teleost diversity (e.g., Koken 1884; Nolf 1995; Reichenbacher et al. 2007). In the following, the saccular otolith is referred to as “otolith”. ZOOGEOGRAPHY OF THE ANATOLIAN APHANIUS TAXA The Anatolian part of Turkey is one of the diversity hotspots of the cyprinodontiform genus Aphanius Nardo, 1827, which occurs in this area with four endemic species, i.e. Aphanius anatoliae GEODIVERSITAS • 2008 • 30 (3) Fossil Aphanius otoliths from Anatolia (Leidenfrost, 1912), A. danfordii (Boulenger, 1890), A. villwocki Hrbek & Wildekamp, 2003, and A. asquamatus (Sözer, 1942) (Wildekamp et al. 1999; Hrbek & Wildekamp 2003). Aphanius anatoliae is subdivided into four different subspecies: A. anatoliae anatoliae (Leidenfrost, 1912), whose populations live in the southwest and the western central part of Anatolia, A. a. splendens (Kosswig & Sözer, 1945), which is now restricted to Lake Salda (Fig. 1A), and A. a. sureyanus (Neu, 1937) and A. a. transgrediens (Ermin, 1946), which are endemic to Lake Burdur (Fig. 1A) and to Lake Acı, respectively (Wildekamp 1993; Wildekamp et al. 1999). Analysis of mitochondrial DNA indicates that Aphanius a. sureyanus, A. a. splendens, A. a. transgrediens, and several populations of A. a. anatoliae from the Lakes District (southwestern Anatolia) belong to a single clade (Hrbek et al. 2002; Hrbek & Meyer 2003). Based on crossbreeding experiments, Villwock (1964, 1982) had already demonstrated that three different population groups of A. anatoliae exist (Wildekamp et al. 1999), i.e. a southwestern group, which includes the first three of the above listed subspecies and several populations of A. a. anatoliae from the Lakes District, a second group that is composed of the western central populations of A. a. anatoliae, and a third group that links the first and second groups with one another. Hrbek et al. (2002) and Hrbek & Meyer (2003) hypothesized that the zoogeographic distribution of Aphanius is a result of the complex geological history of Turkey. Adding support to this hypothesis is the fact that the zoogeographic distribution pattern of the cyprinid genus Pseudophoxinus from Turkey is very similar to that observed for Aphanius (Hrbek et al. 2004). Based primarily on studies of A. anatoliae, however, Villwock (1964, 2004) postulated that the separation events occurred relatively recently, i.e. during the Plio-Pleistocene (about 1.8 Ma ago), or are of postglacial age, and resulted from climatic changes that led to a considerable reduction in size of the lakes. Conversely, Hrbek & Meyer (2003) suggested that the separation events took place at least several million years ago; molecular clock estimates indicate that the diversification of A. anatoliae occurred some 11.79 ± 0.52 Ma ago, and the diversification of GEODIVERSITAS • 2008 • 30 (3) the A. anatoliae populations in the Lakes District 7.48 ± 0.49 Ma ago. THE FOSSIL RECORD OF APHANIUS IN TURKEY Fossil remains of Aphanius from Anatolia can assist in answering the above raised uncertainties concerning the separation events. However, until recently only a single fossil species of this genus was known from Turkey, i.e. Aphanius kayai Reichenbacher & Rückert, 2002 from the upper Miocene-lower Pliocene of Manisa near Izmir (Rückert et al. 2002), which is known almost exclusively from otoliths. As a result, no suitable fossil record of the genus Aphanius existed from the Lakes District in southwest Anatolia. In addition, only a few studies have focused on Aphanius otoliths, e.g., Malz (1978), Reichenbacher & Sienknecht (2001), and Schulz-Mirbach et al. (2006). Therefore, it was of special interest to study fossil Aphanius otoliths from the ?upper Pliocene-lower Pleistocene of the Burdur Basin in the Lakes District. This study addresses the following questions: a) Do distinct differences exist in otolith contour between the fossil specimens and the extant A. anatoliae from the Lakes District and A. danfordii, respectively? b) Which parts of the otolith contour are most significant for distinguishing the Aphanius groups? and c) How similar or dissimilar are the otoliths from the two fossil samples and extant A. anatoliae subspecies to one another and what does this possibly imply with regard to the diversification models of the A. anatoliae populations in the Lakes District? MATERIAL AND METHODS SAMPLING AND PREPARATION Extant otoliths Otoliths from each of the populations of A. anatoliae anatoliae, A. a. sureyanus and A. danfordii were obtained from wild catches (bycatches) of the Süleyman Demirel University (Isparta, Turkey) and the University of Hamburg (Germany) (Fig. 1A; Table 1). Skulls of fishes were opened ventrally and left and right otoliths were removed. Otoliths were cleaned from organic residues by soaking in a 1% potassium hydroxide solution for four hours and subsequent rinsing with distilled water for 12 hours. 579 Schulz-Mirbach T. & Reichenbacher B. TABLE 1. — Overview of species (subspecies), sample number, number of otoliths, length range of otoliths (Lot) used in Fourier analysis (FA), locality, and age. The first numeral in brackets represents the number of left otoliths; the second numeral specifies the number of right otoliths and the third numeral after the semicolon the number of intact whole otoliths. Bold numerals in bold brackets represent the specimens used in Fourier analysis. Abbreviations: A., Aphanius. Taxon Sample no. No. of otoliths Lot [μm] for (Fig. 2) FA A. anatoliae anatoliae (Leidenfrost, 1912) A. anatoliae sureyanus (Neu, 1937) – A. danfordii (Boulenger, 1890) – – †A. cf. anatoliae sureyanus T01 259 †A. cf. anatoliae sureyanus T01 258 †A. cf. anatoliae sureyanus T01 257 †A. cf. anatoliae sureyanus T01 256 †A. cf. anatoliae sureyanus T97 242 †A. cf. anatoliae sureyanus T01 255 (29/29) (11/0) (45/45) (15/0) (36/36) (34/0) (20/12;23) – (1/3;4) – (2/3;5) – (15/13;25) (7/11) (15/14;22) (9/7) (6/5;9) – Fossil otoliths Otoliths come from sediments that crop out near Yassigüme (Fig. 1B), to the south of the presentday Lake Burdur (cf. Sen et al. pers. com.). The samples (Fig. 2) were screen-washed with peroxide, sieved, and microfossils were sorted under a stereomicroscope. A total of 108 otoliths of fossil Aphanius were obtained. Thirty-four otoliths from samples T97 242 and T01 256 were selected for Fourier shape and the statistical analyses (cf. Table 1). The otoliths from sample T01 259 were not included in the quantitative analysis because they mainly represent juvenile or sub-adult/adolescent specimens. The fossil otoliths did not show any deformation and exclusively intact otoliths were included in the analysis. Preparation Fossil and extant otoliths were stored dry in small plastic boxes (FEMA-cells). Otoliths are deposited in the Bavarian State Collection for Palaeontology and Geology in Munich (BSPG-2003 IV 167-260), Germany. 580 Locality (Fig. 1) Age extant 600-838 Nearby Lake Salda Lake Burdur extant 751-970 Karpuzatan extant – Burdur Basin early Pleistocene – Burdur Basin early Pleistocene – Burdur Basin early Pleistocene 636-879 Burdur Basin ?Plio-Pleistocene 545-798 Burdur Basin ?Plio-Pleistocene – Burdur Basin ?Plio-Pleistocene 623-888 SEM IMAGES For qualitative description of the otolith contours, SEM images were taken with a LEO 1430 VP at the Zoological State Collection in Munich (ZSM). The nomenclature of otolith features follows Chaine & Duvergier (1934) and Nolf (1985) (see Figure 3). DIGITISING OF CONTOURS Left otoliths of the extant taxa, and left and right otoliths of fossil Aphanius were positioned (with their outer face down) on plasticine, and digital images were taken with a magnification of 152×. Images were imported and measured with a Leica Image Software (IMAGIC 1000) via a CCD camera connected to a PC. Contours were processed in Adobe Photoshop CS2 with a final contrast of 100% (white object on black background). Digitising of the contours was applied in tpsDig vers. 2.0 (Rohlf 2004) with the tip of the rostrum used as the starting point and saving raw x-y values. Prior to the digitalization of the contours, images of right otoliths were mirrored. With the fossil specimens, it was not possible to use only left GEODIVERSITAS • 2008 • 30 (3) Fossil Aphanius otoliths from Anatolia N A Istanbul Ankara Izmir Karpuzatan Lake Burdur Kayseri B Lake Salda Antalya 200 km N B Lake Burdur (850 m) Burdur v v v v v v v v v v v v v v v Isparta v v v v v v v v v v v v v v v v v v v v v Lake v v v Gölcük v v v v v v v v PP ES Yassigüme Bey Dağları NA Site LY C IA N Igdelı Quaternary v v Gölcük Volcanics v (Plio-Quaternary) Burdur Formation (Plio-Quaternary) 10 km Pre-Pliocene Basement FIG. 1. — A, Map of Turkey showing the sample sites of the extant species; B, geological map of the northern part of the Burdur Basin with the sampling site at Yassigüme. GEODIVERSITAS • 2008 • 30 (3) 581 Schulz-Mirbach T. & Reichenbacher B. dorsal rim Recent posterodorsal edge posterior rim antirostrum excisura sulcus T01 259 Pleistocene rostrum T01 258 T01 257 posteroventral edge ventral rim FIG. 3. — Left saccular otolith of Aphanius anatoliae sureyanus (Neu, 1937) (BSPG-2003 IV 167), showing the most important morphological features of the contour and inner face. Scale bar: 100 μm. ?Pliocene T01 256 T97 242 pre-Pliocene basement T01 255 0.5 m conglomerate sandy marl limestone of approximately 600 to 900 μm (A. anatoliae and fossil Aphanius) and 700 to 1000 μm (A. danfordii) were selected for further analyses to minimize sizedependent effects (see also Table 1). These otolith length ranges correspond to total lengths of adult fish of the extant species between 27 and 50 mm. Juvenile or sub-adult specimens (TL < 27 mm) were not included in the shape analysis to avoid ontogenetic effects and otoliths of old and large individuals (TL > 50 mm) were omitted as well. Old and large specimens possess otoliths that tend to display (strongly) crenulated otolith rims whereas rims are smooth in otoliths of fish with TL values between 27 and 50 mm. marl with pebbles marl dolomite T01 255 position of sample mammal teeth FIG. 2. — Schematic overview of stratigraphy, lithology, and position of the samples from the cross section at Yassigüme near Lake Burdur. otoliths because otherwise the sample sizes would not have been sufficiently large enough for statistical analyses (< 10 specimens). Otoliths with a length 582 FOURIER ANALYSIS Images were processed in the Hshape Software of Crampton & Haines (1996) based on a Fast Fourier Transform (FFT) algorithm. This software consists of the three programs HANGLE, HMATCH, and HCURVE. In HANGLE, the Fourier functions are fitted to a function of the tangent angle dependent of the arc length that is connected to the x-y values. It results in two computationally independent Fourier descriptors per harmonic (Haines & Crampton 2000). Moreover, Fourier descriptors of higher number harmonics are not downweighted as in elliptic Fourier analysis. Therefore, Fourier descriptors of high number harmonics considerably contribute to the overall contour (Haines & Crampton 2000) GEODIVERSITAS • 2008 • 30 (3) Fossil Aphanius otoliths from Anatolia which is especially important with regard to the statistical analyses (see below). Normalization of size was performed automatically in HANGLE (harmonics: 20; smoothing iterations: 13), and normalization of orientation was attained by HMATCH for the entire sample set because the sum of eigenvalues of variance-covariance based PCA was lowest for fitting all contours by HMATCH (see also Haines & Crampton 2000 for quantitative determination of best fitting method). A number of at least 20 harmonics necessary for the analysis was indicated by the amplitude vs harmonic number plot and by visual comparison of back-calculated contours (HCURVE) with 5 to 30 harmonics and the original contour. Average contours: Fourier descriptors (FDs) of every group (A. a. anatoliae, A. a. sureyanus, A. danfordii, fossil samples T97 242 and T01 256) were averaged and 1024 x-y values were back-calculated in HCURVE. STATISTICAL ANALYSES A variance-covariance-based principal components analysis (PCA) was applied to the FDs in SPSS vers. 14.0 (SPSS Inc. 2005). In Hshape also FDs of higher number harmonics explain a partially important amount of the contour and therefore are of more or less “equal” scale. Hence, in this study, it did not seem to be reasonable to standardize original variables (FDs), i.e. to perform a correlation-based PCA (cf. Crampton 1995). The number of relevant PCs that explain more variability than would be explained by chance alone was determined based on the method outlined in Jackson (1993: 2207; Fig. 2). 95%-confidence ellipses for groups and for group means of the PC-plots were calculated based on the formulas provided in Sokal & Rohlf (1995). For visualization of morphospace of PC-plots, Haines & Crampton (2000) proposed calculation of synthetic model shapes that may represent real contours as well as extreme artificial shapes. Model shapes for all plots were calculated from -2 to +2 S.D. A canonical discriminant analysis (CDA) was conducted with the first three PCs for five groups (A. a. anatoliae, A. a. sureyanus, A. danfordii, and GEODIVERSITAS • 2008 • 30 (3) the fossil samples T97 242 and T01 256) and for three groups (the subspecies of A. anatoliae merged, A. danfordii, and both fossil samples merged). The first three principal components were used instead of the Fourier descriptors (number of FDs = 38) because the CDA requires considerably more specimens in the smallest groups than variables characterising these specimens (Ponton 2006). With the principal components as new variables this basic assumption was fulfilled: the number of individuals in the smallest group was 11 compared to three variables (PCs) used in the CDA. The assumption of homogeneity of the group covariance matrix was examined by the Box’s M test. ABBREVIATIONS BSPG CDA FA FD FFT Lot PC PCA S.D. TL ZSM Bayerische Staatssammlung für Paläontologie und Geologie (Bavarian State Collection for Palaeontology and Geology, Munich); Canonical discriminant analysis; Fourier analysis; Fourier descriptor; Fast Fourier Transform; Length range of otoliths; Principal component; Principal components analysis; Standard deviation; Total length; Zoologische Staatssammlung München (Bavarian State Collection for Zoology, Munich). RESULTS PRINCIPAL COMPONENTS ANALYSIS (PCA) The three first PCs explain more variability than would be explained by chance alone (Fig. 4). The three first PCs account for approximately 44.9% of the overall variance of the dataset with PC 1 covering 27%, PC 2 10.1% and PC 3 7.8%. 95%-confidence ellipses of the group means The PC 1 vs PC 2 (Fig. 5A) and PC 1 vs PC 3 (Fig. 5C) plots show that the two extant species A. danfordii and A. anatoliae (A. a. anatoliae and A. a. sureyanus) are clearly separated, whereas the minor overlap that occurs between A. danfordii and A. a. sureyanus is visible in the PC 2 vs PC 3 plot (Fig. 5E). The fossil samples do not display overlap with A. danfordii, with exception of the PC 1 vs 583 Schulz-Mirbach T. & Reichenbacher B. 30 Variance as % of total variance 25 observed data 20 15 10 5 0 theoretical random data 0 2 4 6 8 10 12 14 16 18 20 22 24 26 28 30 32 34 36 38 Eigenvector FIG. 4. — Eigenvector vs variance as % of total variance plot showing the position of the eigenvalues of the eigenvectors with respect to the theoretical eigenvalues expected by chance alone. Theoretical eigenvalues were determined based on the formula provided in Jackson (1993: 2207 “Broken Stick”). PC 3 plot (Fig. 5C) where sample T01 256 depicts a somewhat intermediate position between both extant species with the affinities tending towards A. danfordii. On the other hand, the fossil samples are characterized by large overlap between one another, and show overlap, or at least more proximity, to A. anatoliae, which is especially true of sample T97 242. 95%-confidence ellipses of the groups The overlap between A. danfordii and the remaining groups in PC 1 vs PC 2 plot is relatively small, with exception of the fossil sample T01 256 (Fig. 5A). Considerable overlap occurs between A. danfordii and the other four groups in the PC 1 vs PC 3 and PC 2 vs PC 3 plots. Outliers or data points near the boundary of the confidence ellipses of the groups contain references with regard to the later interpretation of the morphospace generated by the PC-plots. The PC 1 vs PC 2 plot (Fig. 5A) displays “extreme outlines”, one of which belongs to A. danfordii, and a second 584 to the fossil sample T01 256. Moreover, the latter falls into the 95%-confidence ellipse of the group mean of A. danfordii. Both otoliths reveal a distinct shape of the posteroventral edge, a relatively deep excisura and a prolonged rostrum. The five “extreme contours” of the PC 1 vs PC 3 plot (Fig. 5C) are peculiar in that they are located in-between their groups due to the shape of the antirostrum, deepness of the excisura and development of the posterodorsal edge. The PC 2 vs PC 3 plot (Fig. 5E) again indicates that the shape of the antirostrum and the posteroventral and posterodorsal edges are of importance for the extraordinary positions of the four “extreme outlines”. SYNTHETIC MODEL SHAPES The PC 1 vs PC 2 and PC 1 vs PC 3 plots (Fig. 5B, D) indicate that the more negative the PC 1 values are, the deeper the excisura and the longer the rostrum become. In addition, the more negative the values of PC 1 are, the more prolonged the tip of the rostrum becomes. More positive values along GEODIVERSITAS • 2008 • 30 (3) Fossil Aphanius otoliths from Anatolia B A 2 PC 2 (10.1%) 0.007 1 0.005 0 0.003 -1 0.001 -2 -0.001 -0.021 -0.017 -0.013 -0.009 -0.005 -0.001 C -2 -1 -2 -1 0 1 2 0 1 2 D 2 PC 3 (7.8%) 0.002 1 0 0 -0.002 -1 -0.004 -2 -0.006 -0.021 -0.017 -0.013 -0.009 -0.005 -0.001 PC 1 PC 1 (27%) E F 0.002 2 A. a. anatoliae A. a. sureyanus A. danfordii †A. cf. a. s. (T01 256) †A. cf. a. s. (T97 242) PC 3 (7.8%) 0 -0.002 1 0 -1 -0.004 -2 -0.006 -0.001 0.001 0.003 0.005 0.007 PC 2 (10.1%) -2 -1 0 1 2 PC 2 FIG. 5. — A, PC 1 vs PC 2 scatter plot; B, synthetic model shapes for the PC 1 vs PC 2 plot in units of S.D.; C, PC 1 vs PC 3 scatter plot; D, synthetic model shapes for the PC 1 vs PC 3 plot in units of S.D.; E, PC 2 vs PC 3 scatter plot; F, synthetic model shapes for the PC 2 vs PC 3 plot in units of S.D; A, C, E, open ellipses represent 95%-confidence ellipses of the groups, shaded ellipses show the 95%-confidence ellipses of the group means. “Extreme contours” are indicated by original outlines of the respective otoliths. Abbreviations: A., Aphanius; a., anatoliae; s., sureyanus. the PC 2 axis (Fig. 5B, F) account for a more ventrally bent antirostrum and a less angular and more pointed tip of the rostrum. Furthermore, the PC 2 GEODIVERSITAS • 2008 • 30 (3) explains the degree of curvature/angularity of the posteroventral edge, whereas the PC 3 (Fig. 5D, F) characterizes the development of the posterodorsal 585 Schulz-Mirbach T. & Reichenbacher B. TABLE 2. — Jackknifed classification matrix of the canonical discriminant analysis of all five investigated groups of Aphanius. The percentages in rows represent the classification into the groups given in columns; the corresponding number of specimens is given in brackets. The percentages of correctly classified individuals are in bold. Overall classification success is 63.8% (Wilks’λ = 0.169). Abbreviations: A., Aphanius; a., anatoliae; s., sureyanus. Groups A. a. anatoliae (Leidenfrost, 1912) A. a. sureyanus (Neu, 1937) A. danfordii (Boulenger, 1870) †A. cf. a. sureyanus (T01 256) †A. cf. a. sureyanus (T97 242) A. a. anatoliae A. a. sureyanus 72.7 (8) 26.7 (4) 0.0 (0) 5.6 (1) 12.5 (2) 18.2 (2) 46.7 (7) 0.0 (0) 16.7 (3) 18.8 (3) A. danfordii †A. cf. a. s. (T01 256) †A. cf. a. s. (T97 242) 0.0 (0) 0.0 (0) 94.1 (32) 11.1 (2) 0.0 (0) 9.1 (1) 13.3 (2) 5.9 (2) 38.9 (7) 31.3 (5) 0.0 (0) 13.3 (2) 0.0 (0) 27.8 (5) 37.5 (6) TABLE 3. — Jackknifed classification matrix of the canonical discriminant analysis of the three groups: Aphanius anatoliae (A. a. anatoliae and A. a. sureyanus), A. danfordii and †A. cf. a. sureyanus (T97 242 and T01 256). The percentages in rows represent the classification into the groups given in columns; the corresponding number of specimens is given in brackets. The percentages of correctly classified individuals are in bold. Overall classification success is 78.7% (Wilks’λ = 0.197). Abbreviations: A., Aphanius; a., anatoliae. Groups A. anatoliae (Leidenfrost, 1912) A. danfordii (Boulenger, 1870) †A. cf. a. sureyanus A. anatoliae A. danfordii 69.2 (18) 0.0 (0) 26.5 (9) 0.0 (0) 97.1 (33) 5.9 (2) †A. cf. a. sureyanus 30.8 (8) 2.9 (1) 67.6 (23) edge and posterior rim. The more negative the values Aphanius and 69.2% A. anatoliae) for the merged become along the PC 3 axis, the more dorsally bent groups (Table 3). the posteroventral edge is. This is particularly well AVERAGE CONTOURS (FIG. 7) visible in the PC 2 vs PC 3 plot (Fig. 5F). The development of the dorsal and posterior rims in the fossils more closely correspond to that seen in CANONICAL DISCRIMINANT ANALYSIS (CDA) The Box’s M test yields no significant result (p > 0.1) A. anatoliae than to that of A. danfordii (for comparison for both CDA plots, and thus the assumption of see also Fig. 8C, F vs H, I). The average contour of homogeneity of the group covariance matrix cannot A. danfordii is marked by a distinct posteroventral edge. be rejected. The plots of the CDA reveal both a dis- The fossil otoliths and A. danfordii are characterized by tinct separation of A. danfordii and the A. anatoliae a relatively deeply incised narrow excisura, while the subspecies and the fossil samples (Fig. 6). Accord- excisura of A. a. anatoliae and A. a. sureyanus is wider ing to the jackknifed classification matrices, there and flatter. The shape of the tip of the antirostrum is no misclassification between the extant species in the fossils resembles that seen in A. danfordii, but A. danfordii and A. anatoliae (Tables 2; 3). Minor the overall curvature, especially in sample T97 242, misclassification occurs between A. danfordii and parallels that of A. a. anatoliae and A. a. sureyanus the merged fossil samples (Table 3), and A. dan- (Fig. 8C, D, G vs H). The rostrum of the fossils is not fordii and the fossil sample T01 256 (Table 2). as round as in A. a. anatoliae and also not as broad In comparison to A. danfordii, which always has and rectangular as in A. danfordii, but very similar more than 90% correct classification, the indi- to A. a. sureyanus. The contours of the fossil samples vidual subspecies as well as the subspecies and the and A. a. sureyanus are similar in the region of the fossil samples show correct classification rates of dorsal and the posterior rim, with exception of the 37.5% (T97 242) to 72.7% (A. a. anatoliae) for posteroventral edge where the outlines of the extant all groups (Table 2) and about 70% (67.6% fossil subspecies show almost the same course. 586 GEODIVERSITAS • 2008 • 30 (3) Fossil Aphanius otoliths from Anatolia A. a. anatoliae A. a. sureyanus A. danfordii †A. cf. a. sureyanus (T01 256) †A. cf. a. sureyanus (T97 242) group centroids A 5 4 Discriminant axis 2 (6.8%) 3 2 1 0 -1 -2 -3 -4 -5 -5 -4 -3 -2 -1 0 1 2 3 4 5 Discriminant axis 1 (92.3%) B A. anatoliae 5 A. danfordii †A. cf. a. sureyanus group centroids 4 3 Discriminant axis 2 (7.7%) 2 1 0 -1 -2 -3 -4 -5 -5 -4 -3 -2 -1 0 1 2 3 4 5 Discriminant axis 1 (92.3%) FIG. 6. — Discriminant function scores for the first three principal components based on 38 Fourier descriptors generated by the program Hshape: A, for all five investigated groups; B, for three groups, the subspecies of Aphanius anatoliae (Leidenfrost, 1912) and the fossil samples merged, respectively. Abbreviations: A., Aphanius; a., anatoliae. GEODIVERSITAS • 2008 • 30 (3) 587 Schulz-Mirbach T. & Reichenbacher B. chapter “Material and methods”) for this purpose, but also takes indirectly into account that the sulcus of A. danfordii often runs obliquely compared to the median straight sulcus of the A. anatoliae subspecies and fossils (Fig. 8H vs A, B, E, G). A. a. anatoliae A. a. sureyanus A. danfordii †A. cf. a. sureyanus T97 242 †A. cf. a. sureyanus T01 256 FIG. 7. — Average contours based on the averaged 38 Fourier descriptors for every group, back-calculated with the program HCURVE (1024 x-y values for each contour) of A. a. anatoliae (Leidenfrost, 1912), A. a. sureyanus (Neu, 1937), A. danfordii (Boulenger, 1870), and the two fossil samples of †A. cf. a. sureyanus (T97 242 and T01 256). Abbreviations: A., Aphanius; a., anatoliae. DISCUSSION In this study, A. danfordii primarily served as “outgroup” for a more accurate similarity estimate between the fossil Aphanius and the extant A. anatoliae otoliths. In general, the results of the CDA (Fig. 6; Tables 2; 3) show that the fossil Aphanius und A. anatoliae are distinct from A. danfordii. The differences between the two extant species concur with the results of mtDNA analyses presented by Hrbek et al. (2002), and Hrbek & Meyer (2003), the crossbreeding studies conducted by Villwock (1964), and the qualitative analysis of otolith morphology by Schulz-Mirbach et al. (2006). The average contour of A. danfordii otoliths appears to be slightly rotated to the average contours of the other groups due to the fitting by the program HMATCH (Fig. 7). One might argue that this may be the reason for the differences between A. danfordii and the other groups. However, fitting the contours by HMATCH not only was the best method (see 588 GROUP-SPECIFIC TRAITS OF THE OTOLITH CONTOUR The morphological features of the otolith contour explained by the PC 2 are significant in distinguishing A. danfordii from A. anatoliae and the fossil Aphanius. These are: the shape of the antirostrum and curvature/angularity of the posteroventral edge. Moreover, the qualitative analysis of A. danfordii indicates that most of the otoliths are characterized by a tapering posteroventral edge and straighter and slightly dorsally pointed antirostrum (Figs 7; 8H). On the other hand, the otoliths of the fossil Aphanius and extant A. anatoliae populations are usually marked by a ventrally bent antirostrum and broad and angularly shaped posteroventral edge (Figs 7; 8C, D, F). Aphanius danfordii, the fossil Aphanius, and A. anatoliae are separated mainly along the PC 1 axis, which explains the deepness of the excisura and length of the rostrum. However, especially the depth and shape of the excisura vary considerably within the A. anatoliae populations as can be seen by comparing series of individual otoliths from each of the populations. As a result, this feature must be regarded less significant in the discrimination of the individual groups, and may also explain the somewhat unexpected proximity of the fossil samples (especially T01 256) to A. danfordii. There are several fossil otoliths in sample T01 256 that display a distinctly deepened excisura (see extreme original contours in Fig. 5C). Morphological features of the contour explained by the PC 3 (e.g., the development of the posterodorsal edge) seem to play only a subordinate role in separating the investigated groups from each other, which is expressed in the great overlap of all groups in the PC 2 vs PC 3 plot (Fig. 5E). DIVERSIFICATION OF A. ANATOLIAE LAKES DISTRICT If all three PCs are taken into consideration for the CDA, both extant species are well separated IN THE GEODIVERSITAS • 2008 • 30 (3) Fossil Aphanius otoliths from Anatolia A B C D E F G H I FIG. 8. — SEM pictures of: A, †Aphanius cf. anatoliae sureyanus (T97 242), right otolith (mirrored); B, †A. cf. a. sureyanus (T97 242); C, †A. cf. a. sureyanus (T01 256), right otolith (mirrored); D, A. a. anatoliae (Leidenfrost, 1912), Yeş ilova at Lake Salda, TL (total fish length) 45 mm, female; E, A. a. anatoliae, in-between Salda and Doğanbaba at Lake Salda, TL 38 mm, male; F, A. a. sureyanus, Lake Burdur, TL 31 mm, male; G, A. a. sureyanus (Neu, 1937), Lake Burdur, TL 37 mm, female; H, A. danfordii (Boulenger, 1870), Karpuzatan, TL 39 mm, male; I, A. danfordii, Karpuzatan, TL 39 mm, male; A, B, BSPG-2003 IV 245-246; C, BSPG-2003 IV 227; D, E, BSPG-2003 IV 182-183; F, G, BSPG-2003 IV 168-169; H, BSPG-2003 IV 193. If not mentioned otherwise, the pictures show left otoliths. Scale bar: 0.5 mm. (Fig. 6), and the fossil samples show great proximity to one another and also seem to have affinities with the extant A. a. sureyanus (Fig. 6A). The fossils are slightly more similar to A. a. sureyanus than to A. a. anatoliae (see for example Figs 6A; 7; Tables 2; 3), and these two extant subspecies of A. anatoliae reveal a certain distance in the CDA (Fig. 6A; Table 2). Based on these results, we suggest that A. a. anatoliae and A. a. sureyanus had begun to develop GEODIVERSITAS • 2008 • 30 (3) independently from each other already before the fossil populations established. However, we have to bear in mind that ecological influences may also play a certain role with regard to the formation of differences in otolith morphology. The distance between A. a. anatoliae and A. a. sureyanus could be also due to ecological differences in the respective habitats or differences in the life history. The sampled A. a. anatoliae population inhabits shallow vegetation-rich 589 Schulz-Mirbach T. & Reichenbacher B. freshwater at the shores of Lake Salda, while A. a. sureyanus displays a limnetic (open-water dwelling) phenotype (Hrbek & Meyer 2003) in the brackish Lake Burdur that does not contain large amounts of vegetation (Wildekamp et al. 1999). Volpedo & Echeverría (2003) demonstrated that in certain marine fishes, the substrate type, along with bottom oriented vs pelagic life, is reflected in the rostrum length to maximum otolith length ratio. Moreover, in some marine fishes, complexity of otolith outline and within-population variability was shown to decrease from deep to shallow water habitats (Gauldie & Crampton 2002). Although these studies focused on marine fishes and the investigated depth ranges encompass hundreds or thousands of metres, they demonstrated that certain ecological parameters are affecting otolith contour. Moreover, Gauldie & Crampton (2002) hypothesized that particularly high levels of symmetry between left and right otoliths may point to a largely genetically determined otolith morphology. Left and right A. a. anatoliae otoliths are morphologically quite different, indicated by a CDA with the raw FDs (data not shown). As environmental parameters always affect the entire animal, one would expect to find the left and right otoliths shaped in exactly the same way. As a result, it is unlikely that the asymmetry between the left and right A. a. anatoliae otoliths is caused by environmental influences. Rather, the asymmetry appears to be due to local differences in the concentration of intrinsic factors involved in the developmental processes (see Yoshioka et al. 2004) that are responsible for the formation of the otoliths. In future studies, populations of the second (e.g., Lake Eğirdir) and third population-groups (western central Anatolia) of A. anatoliae will be incorporated in the CDA to elaborate on the relevance of the difference between A. a. anatoliae and A. a. sureyanus presented in this study. TAXONOMICAL CONCLUSIONS Based on the fact that the fossil Aphanius and A. anatoliae are not distinctly separated, but rather show more or less similarity to each other in the CDA 590 plots (Fig. 6B; Table 3), and because the fossils and the extant populations display a certain degree of variability with regard to otolith contour indicated by qualitative analyses (for variability of extant populations of A. anatoliae, see SchulzMirbach et al. 2006), it does not seem reasonable to establish a new species for the fossil Aphanius otoliths. We suggest that the fossil otoliths can be assigned to A. anatoliae or the subspecies A. a. sureyanus. However, the nomenclatural situation within A. anatoliae is problematic insofar as the subspecies A. a. sureyanus, A. a. splendens, A. a. transgrediens, and populations of A. a. anatoliae of the Lakes District form a single clade according to Hrbek et al. (2002), and show total fertility between one another in the crossbreeding experiments of Villwock (1958, 1964), whereas other A. a. anatoliae populations in the western central part of Anatolia form a separate clade and are marked by different degrees of sterility to those populations of the Lakes District. Therefore, Villwock (2004) proposed to regard the various population-groups of A. anatoliae as “species in statu nascendi” and hence, to denominate them as A. anatoliae ssp. On the other hand, Hrbek & Meyer (2003) hypothesized that the identified A. anatoliae clades might represent new species. Based on the biospecies concept, the population-groups should only be regarded as distinct subspecies because they do not reveal total sterility among each other, which, however, is an essential criterion in the determination of true species. With regard to the classification of the fossils, there are two nomenclatural possibilities: they could be denominated as A. cf. anatoliae ssp. “Lakes District”, which would be impracticable, or as †A. cf. anatoliae sureyanus, which is preferred because of the close correspondences between the fossil otoliths and those of the extant subspecies A. a. sureyanus. Acknowledgements We wish to thank Dr A. Poisson, Sciences de la Terre, Université Paris-Sud, Orsay and Dr S. Sen, Muséum national d’Histoire naturelle, Paris, for providing samples and otoliths from the Yassigüme locality as well as information on the age and geological GEODIVERSITAS • 2008 • 30 (3) Fossil Aphanius otoliths from Anatolia situation. We thank Prof. W. Villwock, University of Hamburg, and Prof. M. Z. Yildirim, Süleyman Demirel University, Burdur, Dr M. A. Atalay, Ministry of Agriculture, Kütahya, Turkey, and Dr U. Sienknecht, Purdue University, USA, for making the extant fishes available. We acknowledge Mr. H. Meeus, Wommelgem, Belgium and J. L. Blanco, Museo Nacional de Ciencias Naturales, Madrid, for additional specimens of A. a. sureyanus. Our thanks go also to Dr D. A. Jackson, University of Toronto, Dr J. S. Crampton, Institute of Geological and Nuclear Sciences, Lower Hutt, New Zealand, and Dr H. Scherb, National Research Centre for Environment and Health (GSF), Neuherberg, Germany, for their helpful advice concerning the Fourier analysis and the PCA. In addition, we thank Prof. W. Villwock, University of Hamburg, Dr C. Stransky, Institute for Sea Fisheries (ISH), Hamburg, and Dr J.-L. Dommergues, Université de Bourgogne, Dijon, for their comments on an earlier version of the manuscript. We are grateful to Dr M. Krings, Ludwig-Maximilians-University (LMU), Munich, for fruitful discussion and improving the English. REFERENCES CHAINE J. & DUVERGIER J. 1934. — Recherches sur les otolithes des poissons – étude descriptive et comparative de la sagitta des téléostéens. Volume 1. Actes de la Société linnéenne de Bordeaux, Drouillard, Bordeaux, 524 p. CRAMPTON J. S. 1995. — Elliptic Fourier shape analysis of fossil bivalves: some practical considerations. Lethaia 28: 179-186. CRAMPTON J. S. & HAINES A. J. 1996. — User’s manual for programs HANGLE, HMATCH, and HCURVE for the Fourier shape analysis of two-dimensional outlines. Institute of Geological & Nuclear Sciences, Science Report 96 (37): 1-28. GAULDIE R. W. & CRAMPTON J. S. 2002. — An ecomorphological explanation of individual variability in the shape of the fish otolith: comparison of the otolith of Hoplostethus atlanticus with other species by depth. Journal of Fish Biology 60: 1204-1221. HAINES A. J. & CRAMPTON J. S. 2000. — Improvements to the method of Fourier shape analysis as applied in morphometric studies. Palaeontology 43 (4): 765-783. H RBEK T. & M EYER A. 2003. — Closing of the Tethys Sea and the phylogeny of Eurasian killifishes GEODIVERSITAS • 2008 • 30 (3) (Cyprinodontiformes: Cyprinodontidae). Journal of Evolutionary Biology 16: 7-26. HRBEK T. & WILDEKAMP R. H. 2003. — Aphanius villwocki, a new species from the Sakarya River basin of central Anatolian plain, Turkey (Teleostei: Cyprinodontiformes). Ichthyological Exploration of Freshwaters 14: 137-144. HRBEK T., KÜÇÜK F., FRICKEY T., STÖLTING K. N., WILDEKAMP R. H. & MEYER A. 2002. — Molecular phylogeny and historical biogeography of the Aphanius (Pisces, Cyprinodontiformes) species complex of central Anatolia, Turkey. Molecular Phylogenetics and Evolution 25: 125-137. HRBEK T., STÖLTING K. N., BARDAKCI F., KÜÇÜK F., WILDEKAMP R. H. & MEYER A. 2004. — Plate tectonics and biogeographical patterns of the Pseudophoxinus (Pisces: Cypriniformes) species complex of central Anatolia, Turkey. Molecular Phylogenetics and Evolution 32: 297-308. JACKSON D. A. 1993. — Stopping rules in principal components analysis: a comparison of heuristical and statistical approaches. Ecology 74 (8): 2204-2214. KOKEN E. 1884. — Ueber Fisch-Otolithen, insbesondere über diejenigen der norddeutschen OligocänAblagerungen. Zeitschrift der deutschen geologischen Gesellschaft 36: 500-565. MALZ H. 1978. — Vergleichend-morphologische Untersuchungen an aquitanen Fisch-Otolithen aus dem Untergrund von Frankfurt am Main. Senkenbergiana lethaea 59: 441-481. NOLF D. 1985. — Otolithi piscium. Handbook of Paleoichthyology 10: 1-145. NOLF D. 1995. — Studies on fossil otoliths – The state of the art, in SECOR D. H., DEAN J. M. & CAMPANA S. E. (eds), Extant Developments in Fish Otolith Research. The Belle W Baruch Library in Marine Science, Columbia 19: 513-544. PONTON D. 2006. — Is geometric morphometrics efficient for comparing otolith shape of different fish species? Journal of Morphology 267: 750-757. POPPER A. N., RAMCHARITAR J. & CAMPANA S. E. 2005. — Why otoliths? Insights from inner ear physiology and fisheries biology. Marine and Freshwater Research 56: 497-504. REICHENBACHER B. & SIENKNECHT U. 2001. — Allopatric divergence and genetic diversity of extant Aphanius iberus and fossil Prolebias meyeri (Teleostei, Cyprinodontidae) from Southwest and Western Europe as indicated by otoliths. Geobios 34: 69-83. REICHENBACHER B., SIENKNECHT U., KÜCHENHOFF H. & FENSKE N. 2007. — Combined otolith morphology and morphometry for assessing taxonomy and diversity in fossil and extant killifish (Aphanius, †Prolebias). Journal of Morphology 268: 898-915. 591 Schulz-Mirbach T. & Reichenbacher B. ROHLF F. J. 2004. — tpsDig, Digitize Landmarks and Outlines, version 2.0. Department of Ecology and Evolution, State University of New York at Stony Brook, New York, 20 p. RÜCKERT N., BÖHME M., REICHENBACHER B., HEISSIG K., WITT W. & BASSLER B. 2002. — Die Fossilführung des kontinentalen Neogens (Ober-Miozän/ Unter-Pliozän) von Develiköy (Manisa, Türkei). Mitteilungen der Bayerischen Staatssammlung für Paläontologie und historische Geologie 42: 51-74. SCHULZ-MIRBACH T., REICHENBACHER B., YILDIRIM M. Z. & ATALAY M. A. 2006. — Otolith characteristics of species, subspecies, and populations of Aphanius Nardo, 1827 (Teleostei, Cyprinodontiformes) from Anatolia (Turkey). Journal of Natural History 40 (27/28): 1687-1705. SOKAL R. R. & ROHLF F. J. 1995. — Biometry – The Principles and Practice of Statistics in Biological Research. Freeman and Company, New York, 887 p. SPSS INC. 2005. — SPSS 14.0 Base. SPSS Inc., Chicago, 845 p. VILLWOCK W. 1958. — Weitere genetische Untersuchungen zur Frage der Verwandtschaftsbeziehungen anatolischer Zahnkarpfen. Mitteilungen aus dem Hamburgischen Zoologischen Museum und Institut 56: 81-153. VILLWOCK W. 1964. — Genetische Untersuchungen an altweltlichen Zahnkarpfen der Tribus Aphaniini (Pisces: Cyprinodontidae) nach Gesichtspunkten der neuen Systematik. Journal of Zoological Systematics and Evolutionary Research 2: 267-382. VILLWOCK W. 1982. — Aphanius (Nardo, 1827) and Cyprinodon (Lac., 1803) (Pisces: Cyprinodontidae), an attempt for genetic interpretation of speciation. Zeitschrift für zoologische Systematik und Evolutionsforschung 20: 187-197. VILLWOCK W. 2004. — Synopsis of classic and molecular investigations of Old World cyprinodontids of the genus Aphanius Nardo, 1827 (Teleostei: Cyprinodontidae), with special concern of the Anatolian species, their speciation phenomena and their probable historic development. Mitteilungen aus dem Hamburgischen Zoologischen Museum und Institut 101: 35-46. VOLPEDO A. & ECHEVERRÍA D. D. 2003. — Ecomorphological patterns of the sagitta in fish on the continental shelf off Argentine. Fisheries Research 60: 551-560. WILDEKAMP R. H. 1993. — A World of Killies. Volume 1. 1st ed. American Killifish Association, Mishawaka, Indiana, 311 p. WILDEKAMP R. H., KÜÇÜK F., ÜNLÜSAYIN M. & NEER W. V. 1999. — Species and subspecies of the genus Aphanius Nardo 1827 (Pisces: Cyprinodontiformes) in Turkey. Turkish Journal of Zoology 23: 23-44. YOSHIOKA Y., IWATA H., OHSAWA R. & NINOMIYA S. 2004. — Analysis of petal shape variation of Primula sieboldii by elliptic Fourier descriptors and principal component analysis. Annals of Botany 94: 657-664. Submitted on 30 June 2007; accepted on 21 December 2007. 592 View publication stats GEODIVERSITAS • 2008 • 30 (3)