Natural History Museum of Crete, Knossou Avenue, PO Box 2208, Irakleio, 71409, Greece 2

Transcription

1 Blackwell Science, LtdOxford, UKBIJBiological Journal of the Linnean Society The Linnean Society of London, 2003? ? Original Article mtdna PHYLOGENY and MORPHOLOGICAL DIVERSITY IN MASTUS Biological Journal of the Linnean Society, 2003, 78, With 5 figures Mitochondrial DNA phylogeny and morphological diversity in the genus Mastus (Beck, 1837): a study in a recent (Holocene) island group (Koufonisi, south-east Crete) A. PARMAKELIS 1,3 *, E. SPANOS 2, G. PAPAGIANNAKIS 2, C. LOUIS 2,3 and M. MYLONAS 1,3 1 Natural History Museum of Crete, Knossou Avenue, PO Box 2208, Irakleio, 71409, Greece 2 I.M.B.B., FO.R.T.H GR, Vassilika Vouton, PO Box 1527, Irakleio, Crete, Greece 3 University of Crete, Department of Biology, Vassilika Vouton, Irakleio, Crete, Greece Three endemic Cretan land snail species of the genus Mastus (Beck, 1837) from the island group of Koufonisi (southeast Crete) and the eastern part of Crete, were studied by multivariate analysis of shell morphology and analysis of mtdna sequences. The phylogeny of the populations studied and the processes effecting the genetic and morphological diversity of the species were investigated. Extremely high mtdna sequence divergence was observed both within and between populations. The Cretan populations could not be distinguished morphologically, while the populations of the islets were more distinct. We argue that the active geological past of the area (including sea level changes) and the long-term presence of humans has produced a mixing up of Mastus populations leading to the accumulation of high divergence of mtdna sequences on a small spatial scale. The limited morphological diversity and the distinct shell identity of the islets populations can be attributed to the selective pressures of the island group.. ADDITIONAL KEYWORDS: human activities land snail multivariate analyses natural selection sequence divergence shell 16S rrna. A. PARMAKELIS ET AL. INTRODUCTION Island biota provide useful models to study many evolutionary hypotheses including those pertaining to repeated cycles of colonization and extinction, adaptation and drift-induced population differentiation (see Brown, Thorpe & Baez, 1993; Thorpe & Malhotra, 1996). The Aegean Archipelago is a relatively young formation as it attained its present form during the Late to Middle Pleistocene (Dermitzakis, 1990). Since then, and up to today, orogenic movements and sea level fluctuations have been changing the size and form of the islands comprising it, and have been continually altering the connections between, and isolation of, individual islands. The intense geotectonic history of the Aegean Archipelago and the long-term presence of humans in the area create a complex pattern, and the inference of evolutionary relationships of *Corresponding author. parmakel@nhmc.uoc.gr animal or plant taxa is not always a straightforward process (see Douris et al., 1998; Kasapidis, 2002). Depending on the island group examined, different historical processes can be considered as the factors responsible for the phylogenetic patterns observed. The island of Crete is surrounded by numerous island groups and each one of them has a completely different history regarding its past connection to Crete, the age of its formation and the extent of human exploitation of its resources. The island group examined in this study is that of Koufonisi (south-east of Crete) which was formed during the Holocene (Peters, 1985) and was never connected to the island of Crete. The flora and fauna of this island group was formed by dispersal over the sea barrier. The intense human activities of the Cretan civilizations are witnessed by archeological findings situated on the islets and, in some cases, a few metres below sea level. The isolation of the islets now comprising the island group took place during the historic ages, as evidenced by the archaeological and historical data. 383

2 384 A. PARMAKELIS ET AL. The animal taxon investigated is the land snail genus Mastus (Beck, 1837) which is one of those land snail genera along with Albinaria, Partula, Cerion and Cepaea that exhibit high levels of morphological diversity. Mastus is distributed all around the Mediterranean except for the eastern and south-eastern regions and comprises 32 currently recognized species (Heller, 1976; Maassen, 1995). Twenty-seven of these have been reported in the Aegean Archipelago, of which 16 are distributed in Crete and its surrounding island groups (Maassen, 1995). The study of its differentiation throughout the Aegean area is likely to prove informative about the processes of microevolution that have affected this genus. Three endemic Cretan land snail species of the genus Mastus from the island group of Koufonisi and the eastern part of Crete were studied through a multivariate analysis on shell morphology together with an analysis of mtdna sequences. The study had two main aims: (1) to determine the phylogenetic relationships between the populations, and (2) to examine the influence of historical and natural processes on the genetic and morphological diversity of the species. A combination of molecular data, such as mtdna sequences (which, hopefully, are minimally confounded by selection), and morphology can provide evidence for the driving forces of the morphological and genetic diversity of island ecosystems, as has been clearly indicated in the Canary island group by Thorpe & Malhotra (1996). Some systematic implications are also discussed, since the taxonomic classification of the Cretan Mastus species relies on the shell, the genitalia and the spermatophore morphology. In addition, this is the first report of mtdna sequences for the genus Mastus. LYBIAN SEA Greece I. Makroulo I. Marmara Crete Irakleio CRETE N CRETAN SEA I. Stroggylo I. Koufonisi Ferma Km Agia Fotia Itanos Atherinolakos Goudouras 1 Km MATERIAL AND METHODS COLLECTION LOCALITIES The study took place at the island group of Koufonisi that is located approximately 6.5 km south of Cape Goudouras, the south-eastern extremity of Crete. The group consists of one relatively large island, namely Koufonisi, and four smaller islets: Marmara, Makroulo and Stroggylo to the north, and Trachylos to the south (Fig. 1). The depths of the channels separating the islets from the island of Koufonisi are less than 5 m (Table 1). In the case of Marmara and Trachylos the depth is less than 2 m. The island group is separated from Crete by a narrow but quite deep (200 m) sea basin. On Koufonisi island, limestone, sand dunes and marls are present, while on the surrounding islets limestone dominates. The vegetation of the islets consists of phrygana and degraded maquis, in contrast to the island of Koufonisi where a variety of habitats is present (Table 1). I. Trachylos Figure 1. Map showing the localities from where specimens were collected. The historical presence of man on the islands is evidenced by archeological findings on the island of Koufonisi and the islets of Stroggylo, Marmara and Trachylos. The findings date back to the Minoan (6000 years before present, (b.p)) and the Roman ages (2000 years b.p). The malacofauna of the island group also reflects the large influence of man in the shaping of the island group fauna, since the portion of anthropochorus land snail species on the islands is very high compared to other island groups surrounding Crete (Welter-Schultes & Wiese, 1997). According to Peters

3 mtdna PHYLOGENY AND MORPHOLOGICAL DIVERSITY IN MASTUS 385 Table 1. Major aspects of the island group Koufonisi Marmara Makroulo Stroggylo Trachylos Area (km 2 ) Geological substrate Pleiocene Pleiocene Pleiocene Pleiocene Pleiocene limestones limestones limestones limestones limestones Highest point (m) Sediments Neogene and Quaternary Neogene and Quaternary Neogene and Quaternary Neogene and Quaternary Neogene and Quaternary Types of vegetation Maquis, phrygana, steppe, sand Maquis, phrygana Phrygana, maquis Phrygana, maquis Maquis, phrygana dunes Distance (km) from Goudouras (Crete) Distance (km) from Koufonisi Maximum depth (m) separating the islet from Koufonisi 2 <5 <5 <5 Table 2. List of collection localities, species names and sample sizes of the morphometric and molecular analyses. Numbers in parentheses indicate the sample size in the molecular analyses. The outgroups apply only to the molecular analyses. + indicates that the locality is the type locality of the species. UTM = Universal Transverse Mercator Grid Locality Species Type locality UTM Longitude (E) Latitude (N) Sample size Itanos Mastus itanosensis Maassen, MV (2) Ferma Mastus ierapetrana Maassen, LU (2) Atherinolakos Mastus sitiensis Maassen, 1995 MU (3) Goudouras Mastus sitiensis Maassen, 1995 MU (3) Agia Fotia Mastus sitiensis Maassen, MU (2) Koufonisi Mastus sitiensis Maassen, 1995 MU (3) Marmara Mastus sitiensis Maassen, 1995 MU (2) Trachylos Mastus sitiensis Maassen, 1995 MU (1) Stroggylo Mastus sitiensis Maassen, 1995 MU (2) Makroulo Mastus sitiensis Maassen, 1995 MU (1) Cyprus Paramastus cyprius Zilch, 1951 (1) (outgroup) Syria (outgroup) Buliminus labrosus (Olivier, 1801) (1) (1985), all five islands are built up of Neogene and Quaternary sediments. The geological and archeological findings indicate that the islets of Marmara and Trachylos were connected to the main island (Koufonisi) until the historic ages. There are no geological data indicating a connection of the island group with the opposite shores of Crete (see Peters, 1985). In sharp contrast to the areas situated on the coast of Crete directly opposite, the island of Koufonisi is cut by a dense pattern of normal faults (Peters, 1985). In addition to the populations of the island complex, we also studied populations on the island of Crete from areas that are either geographically close to the island complex or for which historical documents and archeological findings indicate interaction with the island complex through human activities. The vegetation of the localities sampled in Crete is mainly degraded maquis and phrygana with similar plants to each other and almost equal plant cover. The dominant plant species of these localities are Genista acanthoclada, Coridothymus capitatus and Sarcopoterium spinosum. In total we sequenced 22 individuals from ten localities (Table 2). In addition, 13 quantitative shell variables (see Table 5) were used in a multivariate analysis of the differentiation of the ten populations studied. Adult snails were collected from the islets of the Koufonisi island group (Koufonisi, Marmara,

4 386 A. PARMAKELIS ET AL. Makroulo, Trachylos and Stroggylo), from the areas of Ferma, Goudouras and Atherinolakos which are located on the south-east shore of Crete, and from the areas of Agia Fotia and Itanos in the north-east part of Crete (Fig. 1). In each locality the area sampled did not exceed 50 m 2. TAXONOMIC STATUS OF THE POPULATIONS STUDIED The collected material can be assigned to three different species according to the taxonomy proposed by Maassen (1995), while according to Vardinoyannis (1994) all populations belong to the endemic Cretan land snail species Mastus olivaceus (Pfeiffer, 1846). The populations of Ferma, Itanos and Agia Fotia are the type localities of the species involved in this study (Table 2), while the island specimens were assigned to the appropriate species after dissection and examination of the terminal genitalia. The taxonomic status proposed by Maassen (1995), which relies mainly on the spermatophore and genitalia morphology, is presented in Table 2. STATISTICAL ANALYSES OF SHELL MORPHOMETRIC CHARACTERS A digital image of the shell of each of the 235 individuals (ten populations) was obtained. The number of individual shells measured from each population varied from (Table 2). Thirteen variables were measured by digital image analysis on each individual shell. Relationships between individuals were first explored without a priori distinction of geographical entities with principal component analysis (PCA) on the correlation matrix of the log 10 -transformed values of the raw variables. The log transformation was performed in order to minimize deviations from normality and distortion effects caused by allometric relationships of the raw variables. The relations of the predefined geographical entities were analysed by discriminant analysis (DA). The DA, which maximizes the statistical separation of a priori defined groups, was performed in order to confirm or reject the cohesiveness of each geographical entity. For each population the coefficient of variation (CV: according to Sokal & Rohlf, 1995) was calculated for all shell variables. The mean coefficient of variation for each population (computed as the mean value of CV s obtained for each variable) is used as a measure of the relative morphological variation of each population. All analyses were performed using Statistica for Windows, version 5A. DNA EXTRACTION Total DNA was isolated from frozen (-80 C) or alcohol-preserved individuals. In order to overcome problems of polymerase chain reaction inhibition by mucopolysaccharides, we extracted DNA with CTAB (hexadecyl-trimethyl-ammonium bromide) by modifying the protocol of Winnepenninckx, Backeljau & De Wachter (1993). From each individual, a small piece of foot tissue (c. 25 mm 3 ) was sliced, ground and placed in 700 ml 2 CTAB extraction buffer, which contained 10 ml Proteinase-K (10 mg/ml). The mixture was incubated at 60 C until the tissue dissolved completely. After incubation, proteins were extracted once with phenol-chloroform and once with chloroform alone. DNA was precipitated with an equal volume of ice-cold isopropanol and stored at -20 C overnight. The DNA pellet was centrifuged and washed in 70% ice-cold ethanol twice. The pellet was dried overnight at room temperature and resuspended in ml of nanopure water. AMPLIFICATION AND SEQUENCING Amplification of a part of the mitochondrial large ribosomal gene (16S rrna) was carried out using the universal 16S primers 16Sbr-H [5 CCG GTC TGA ACT CAG ATC ACG T 3 ] and 16Sar-L [5 CGC CTG TTT ATC AAA AAC AT 3 ] (Palumbi et al., 1991). Each PCR was performed in a 10 ml volume, where 1 ml of template DNA was mixed with 0.2 mm dntps, 2.5 mm MgCl 2, 0.2 pmol of each primer, and 1 unit of Taq Polymerase (Gibco). Thermocycling was performed in a PTC-100 thermocycler (MJ-Research). The cycle program comprised of an initial denaturation at 94 C for 2 min, followed by 40 cycles of 1 min at 94 C, 1 min at 48 C, and 32 s at 72 C. The cycling was ended with 10 min sequence extension at 72 C. A set of internal primers for the 16S rrna gene was designed and used when the universal primers were not effective. These primers are the following: 16Siar-L [5 CTT TAA CGG CCG CAG TAC AYC T 3 ] and 16Sibr-H [5 TCT GAA CTC AGA TCA CGT AGG GT 3 ]. The PCR conditions with the internal primers differed at the annealing temperature, which was lowered to 47 C to increase the yield of the product. The PCR product was used directly for sequencing in a PE-ABI377 sequencer (using dye-terminator chemistry). Both strands of the PCR product were sequenced. The primers in the sequencing reactions were the same as in the amplification procedure. The number of individuals sequenced per sampling locality is presented in Table 2. MOLECULAR DATA ANALYSES Multiple sequence alignments were done using CLUSTALW (Thompson, Higgins & Gibson, 1994) of the ClustalW WWW Service at the European Bioinformatics Institute (

5 mtdna PHYLOGENY AND MORPHOLOGICAL DIVERSITY IN MASTUS 387 Alternative gap opening and gap extension penalties were used. Selection of the optimal parameters was performed according to Lecanidou, Douris & Rodakis (1994) and were those that gave the minimum number of required base changes when each alignment was used as input into the program DNAPARS of the Phylip 3.57c package (Felsenstein, 1995). The computer-generated alignment was then further adjusted based on published secondary structure models (e.g. Thollesson, 1999; Lydeard, Mulvey & Davis, 2000) of the 16S rrna mtdna gene. Nucleotide composition was computed from the entire data set. The number of transitions and transversions occurring among each pairwise combination of individual sequences, for the entire data set, and the respective ratio of TS : TV were plotted against pairwise (p) genetic distances to evaluate possible mutational saturation (following Lydeard et al., 1996). The correlation coefficient r (Sokal & Rohlf, 1995) was used to compare the pairwise numbers of transitions and transversions vs. pairwise genetic distances for the entire data set. Kimura (1980) two-parameter genetic distances were estimated for the entire data set, using MEGA (Kumar, Tamura & Nei, 1993) and were compared to the respective genetic distances from other studies of land snails (e.g. Thomaz, Guiller & Clarke, 1996; Douris et al., 1998; Chiba, 1999). Maximum parsimony trees (in PAUP, Swofford, 1998) and neighbour-joining (NJ, Saitou & Nei, 1987) genetic distance trees were used to test relationships among the individuals of the study. NJ trees (Saitou & Nei, 1987) were constructed from the Kimura (1980) two-parameter distances, using MEGA (Kumar et al., 1993). Two separate runs were made in the NJ analysis, one with pairwise deletion of gaps/missing data and another one with complete deletion of gaps/ missing data. Maximum parsimony analyses were conducted using the heuristic search algorithm in PAUP (Swofford, 1998) on the entire data set after the exclusion of phylogenetically uninformative characters. Two separate runs were made, one with the gaps of the aligned sequences treated as missing data and another one with gaps and missing data completely removed. No suitable outgroup 16S rrna sequence was available for Mastus in GenBank and thus we also sequenced one individual of Buliminus labrosus (Olivier, 1801) from Syria and one Paramastus cyprius Zilch, 1951 from Cyprus, in order to root the trees. The land snail species Albinaria coerulea (Rossmässler, 1835; GenBank Accession Number: X83390) was also used as an outgroup, although it belongs to a different land snail family than that of Mastus. Bootstrap analyses (Felsenstein, 1985; Swofford et al., 1996) with 1000 replications tested the support of the data set for the nodes of the maximum parsimony and genetic distance (neighbour-joining in MEGA, Kumar et al. 1993) trees. NUCLEOTIDE SEQUENCE ACCESSION NUMBERS Nucleotide sequences reported in this study have been assigned the GenBank Accession Numbers AF to AF RESULTS SEQUENCE VARIATION AND SECONDARY STRUCTURE OF THE 16S rrna FRAGMENT The amplified fragment corresponds to bases of the Albinaria coerulea mtdna (Hatzoglou, Rodakis & Lecanidou, 1995). The 16S rrna data set consisted of 474 aligned nucleotides. In total 215 (45.3%) bases are variable and 139 (29.3%) are parsimony-informative. Within the ingroup 133 (28%) are variable and 112 (23.6%) are parsimonyinformative. Kimura (1980) two-parameter distances from the entire data set are given in (Table 3). The genetic distances, d, ranged from 0 to with a mean value of within populations of the ingroup, from to between populations of the ingroup, and from to between ingroup and outgroup populations (Table 4). The genetic distances separating individuals from the same sampling locality ranged in the island group populations from d = to d = 0.048, and in the Cretan populations from d = to d = (Table 3). The maximum value of divergence of single sequences within the ingroup was 0.237, while the largest genetic distance between an ingroup and an outgroup sequence was (Table 3). The genetic distances separating ingroup species (sensu Maassen, 1995) ranged from d = to d = (Table 3). Nucleotide composition is clearly biased towards A- T. The mean values of A, T, G and C within the ingroup are 32.6%, 32.3%, 13.86% and 19.23%, respectively. Regression analysis of the pairwise numbers of TS and TV with pairwise genetic distances (p) showed a linear relationship (Fig. 2). The TS : TV ratio varied from 4 to 0, and in most pairwise comparisons it was lower than 1 (Fig. 2). The alignment of the sequences obtained in this study with those of Thecacera pennigera resulted in the identification of the homologous parts of the sequenced fragment of the 16S rrna gene. The most conserved fragments of the molecule as reported in Lydeard et al. (1996) are retained in our sequences too. There were some deletions and insertions of fragments though, but these occurred in parts of the sequences where molluscs exhibit a high degree of variation (see Lydeard et al., 2000).

6 388 A. PARMAKELIS ET AL. Table 3. Pairwise genetic distances of individual mtdna sequences computed with the Kimura two-parameter model. Numbers at top refer to individuals shown at left Koufonisi 1 2. Trachylos Marmara Makroulo Marmara Koufonisi Koufonisi Marmara Ferma Ferma Stroggylo Stroggylo Agia Fotia Agia Fotia Itanos Itanos Goudouras Goudouras Goudouras Atherinolakos Atherinolakos Atherinolakos B. labrosus P. cyprius A. coerulea

7 mtdna PHYLOGENY AND MORPHOLOGICAL DIVERSITY IN MASTUS 389 Table 4. Within and between localities average genetic distances based on the Kimura two-parameter model Between localities average genetic distance Agia Fotia Itanos Goudouras Atherinolakos Koufonisi Trachylos Marmara Makroulo Ferma Stroggylo Within localities average genetic distance Locality Koufonisi (N = 3) Trachylos (N = 1) Marmara (N = 3) Makroulo (N = 1) Ferma (N = 2) Stroggylo (N = 2) Agia Fotia (N = 2) Itanos (N = 2) Goudouras (N = 3) Atherinolakos (N = 3) Outgroups PARSIMONY AND GENETIC DISTANCE ANALYSES Parsimony analysis yielded three equally parsimonious trees (442 steps, CI: 0.64, RI: 0.366, after the exclusion of uninformative characters) when all characters were weighted equally and gaps were treated as missing data. The general topology of the most-parsimonious tree when gaps and missing data were completely removed from the data set was similar to the one obtained from the first analysis. The general topology of the NJ tree (both runs) was identical to the maximum-parsimony tree (50% majority-rule consensus tree; gaps treated as missing data; Fig. 3). PHYLOGENY OF INDIVIDUAL SEQUENCES The phylogenetic relationships among the individuals of the study based on the most-parsimonious tree are shown in Figure 3 (50% majority-rule consensus tree; gaps treated as missing data). Two major groups are supported statistically. The first group ( Group A ) comprises the individuals from Stroggylo islet and the individuals from the Cretan populations of Goudouras, Atherinolakos, Agia Fotia and Itanos, while the second group ( Group B ) comprises the individuals from the Cretan population of Ferma and all the individuals of the islets of the Koufonisi group except those of Stroggylo islet. In Group A, all the individuals of the Goudouras population cluster together (subgroup A1), which is also the case for most of the remaining populations of Group A where the individuals of a single population cluster together forming a monophyletic group (i.e. subgroup A2, A3 and A4). However, an unexpected clustering of individual sequences occurs in Group A relating to the individuals of Stroggylo islet, which are clustered apart from each other. Individual Stroggylo 1 clusters with the individuals of the Goudouras population (maximum parsimony bootstrap: 67%, NJ bootstrap 87%) while individual Stroggylo 2 is placed apart (NJ bootstrap 51%). In Group B the division into subgroups of individuals from a single population is evident (i.e. subgroups B1 and B2) and, as was the case in Group A, some individual sequences from a single population cluster separately from one another (i.e. individual Koufonisi 3 clusters apart from individuals Koufonisi 1 and Koufonisi 2 with a bootstrap of 100% in the NJ analysis), while other clades incorporate individuals from different islets (i.e. Koufonisi/Trachylos with maximum parsimony bootstrap 52%, NJ bootstrap 66%). In Group B what is quite impressive is the clustering of the individuals from Ferma population. Individual Ferma 1 falls within the clade combining all the individuals of the island group of Koufonisi, while individual Ferma 2 is basal to this clade (maximum parsimony bootstrap 94% and NJ bootstrap 98%).

8 390 A. PARMAKELIS ET AL. Number of substitutions TS/TV Ratio (a) TS TV (b) Genetic distance (p) Figure 2. (a) Regression analysis of the absolute number of transitions against percent sequence divergence for all pairwise comparisons (TS: r 2 = ; P < ) and transversions (TV: r 2 = ; P < ). (b) Ratio of transitions/transversions (TS : TV ratio) plotted against percent sequence divergence. Applying the contemporary taxonomy for the genus Mastus of Crete (Maassen, 1995) we can see that Mastus sitiensis appears to be polyphyletic with conspecific sequences falling into two different terminal groups (Group A and Group B) and the divergence between these two groups of sequences reaches up to a mean value of 19%. Mastus itanosensis is more closely related to Mastus sitiensis of the A Group, while Mastus ierapetrana is also polyphyletic and phylogenetically closer to Mastus sitiensis of the B Group. STATISTICAL ANALYSES OF SHELL MORPHOMETRIC CHARACTERS PCAs were computed with the data of 235 individuals, and the first two accounted for 78.4% of the total morphological variation. The loadings of the log 10 - transformed variables on the first two PCAs are given in Table 5. Principal component 1, which accounted for 70.2% of the total variance, was essentially the illustration of a size effect, i.e. a multivariate approximation of overall size. The second variable incorporated 8.20% of the remaining conchological variation. The loadings with absolute values above 0.4 (Table 5) of the second principal component indicated that it was mostly related to width variables (MD, LWD and WFWP). Thus, it is most likely that the second principal component was an expression of the intensity of aperture (or shell) expansion. The plot of the principal components scores of all the populations included in the analysis are presented in Figure 4. In the discriminant analysis all 13 variables were retained in the model (Wilks l = 4.94, P < ) and the mean value of the percent of correct classifications of all individuals in their predefined group was 58.3% (Table 6). In most Cretan populations the percentage of correct classifications was lower or equal to 50%, while only the populations of Itanos and Agia Fotia gave values well above 50% (56.67 and 60%, respectively). The majority of the individuals that were not grouped to their predefined group were assigned to another Cretan population. On the other hand, individuals of the populations of the island group when misclassified were mostly assigned to a Cretan population, but the percentage of correct classifications of the islets populations was in most cases well above 60% and only in the case of Makroulo was the value below 40%. In a separate discriminant analysis (results not presented here) including only the island group s populations, the correct classifications ranged from 63% to 100%. In the respective analysis including only the Cretan populations, the correct classifications ranged from 58% to 67%. The coefficients of variation of the shell variables for each population used in the study are presented in Table 7. DISCUSSION SEQUENCE DIVERGENCE The present analysis of the 16S rrna gene demonstrates that there is a high degree of genetic variation in the mitochondrial DNA of Mastus even within a population. The divergence values between populations of a species reached up to 21.4% (Table 3), while the divergence between individual sequences reached up to 23.66% (Table 3). Land snail populations exhibit high levels of mitochondrial diversity (Thomaz et al., 1996; Douris et al., 1998; Ross, 1999; Goodacre, 2001). Making the simplest assumptions, the data of these studies would suggest times of divergence as long as 20 million years between haplotypes now coexisting within a single population (Thomaz et al., 1996), or an evolutionary rate times faster than those found in other invertebrates (Chiba, 1999; Hayashi & Chiba, 2000; Watanabe & Chiba, 2001). There are several studies of land snail phylogenetics in which high values of mtdna sequence divergence have also been

9 S u b grou p A 1 S ub grou p A 2 S u b gro u p A 3 Su bg ro u p A 4 Sub g ro up B 1 S u b group B 2 Su b g rou p B 3 mtdna PHYLOGENY AND MORPHOLOGICAL DIVERSITY IN MASTUS 391 M astus sitiensis M astus itanosenis M astus ierapetrana Goudouras 1 Goudouras 2 67 Goudouras 3 Stroggylo Atherinolakos 1 Atherinolakos 2 Atherinolakos 3 Group A Stroggylo islet and Cretan populations Stroggylo Agia Fotia 1 Agia Fotia Itanos 1 Itanos 2 84 Koufonisi Koufonisi 2 Trachylos 1 75 Makroulo 1 Group B Marmara 1 Marmara 2 Marmara 3 Koufonisi 3 Ferma population and Koufonisi group (except Stroggylo islet) 96 F erma 1 F erma 2 Bulim inus labrosus Paramastus cyprius Albinaria coerulea OUTGROUP Figure 3. Evolutionary tree of mitochondrial 16S rrna sequences. The phylogenetic relationships of individuals from single populations are presented. Each number following the localities name corresponds to a different individual from the same locality. The tree was constructed using maximum parsimony. Values on the tree (50% majority rule-consensus tree) indicate the support (%) for the branches in 1000 bootstrap replicates. estimated, for example the genus Mandarina from the Bonin islands (Chiba, 1999) where the divergence between species reached up to 18.7% for the 16S rrna and 17.7% for the 12S rrna. Other similar cases are those of Helix aspersa (Thomaz et al., 1996; Guiller et al., 2001), Cepaea nemoralis (Thomaz et al., 1996), species of the genera Achatinella and Partulina (Thacker & Hadfield, 2000), and most recently the species Euhadra peliomphala (Hayashi & Chiba, 2000) and E. quaesita (Watanabe & Chiba, 2001). There is only one known case of even more extreme mtdna sequence divergence in land snails and that is in Partula from Society Island (Goodacre & Wade, 2001). In that study the mean value of genetic distances was 24% with a maximum value of 37% among cytochrome b sequences, which is generally accepted to evolve faster than the 16S rrna gene that was used in the current study. The high levels of divergence encountered in our study and other studies of land snails are comparable to those found for homologous sequences within taxa described as genera, families or even orders (see Thomaz et al., 1996).

10 392 A. PARMAKELIS ET AL. Table 5. List of morphometic variables measured on the shells and abbreviations used in the morphometric analyses. Component loadings of the shell variables on the first two components of the Principal Component Analysis, eigenvalues and percentage of variance explained by each component No. Variable Abbreviation Component 1 Component 2 1 Shell height SH Shell aperture height MH Shell aperture diameter MD Last whorl height LWH Last whorl diameter LWD Penultimate whorl height PLWH Penultimate whorl diameter PLWD Height of the third whorl after the protoconch HTWP Width of the first whorl after the protoconch WFWP Width of the second whorl after the protoconch WSWP Width of the third whorl after the protoconch WTWP Distance between shell apex and the point AL where the last spire connects to the shell aperture from the left side 13 Distance between shell apex and the point AS where the last spire connects to the shell aperture from the right side Eigenvalues Percentage of variance explained 70.19% 8.20% Four overlapping explanations of the high molecular divergence both within and among populations of land snails have been proposed (see Thomaz et al., 1996) but the most tenable one was that the population structure of pulmonates favours the persistence of ancient haplotypes, since pulmonate populations are almost invariably arranged in stepping-stone patterns, with notably limited migration between nearby demes (Lamotte, 1951; Murray & Clarke, 1984). Thus, land molluscs fulfill the conditions for the persistence of mitochondrial haplotypes in the very long-term (Thomaz et al., 1996; Guiller et al., 2001). High evolutionary rate of land snail mtdna has also been supported (Chiba, 1999; Hayashi & Chiba, 2000) to account for the high levels of intra- and interspecific mtdna divergence occurring in land snail taxa, but no detailed analysis investigating the rate of evolution in molluscs has been attempted. Each land snail species population could actually be composed of cryptic species subpopulations and this would effectively explain the high intrapopulation sequence divergence occurring in land snail populations. This possibility, according to Thomaz et al. (1996), has been considered and denied by successful crosses, producing fertile offspring, between snails from places known to have individuals with very different mitochondria. The heteroplasmy hypothesis, that could also produce large differences in the mtdna sequences, has been rejected in pulmonates (Thomaz et al., 1996; Davison, 2000) and consequently the possibility of mtdna recombination (Ladoukakis & Zouros, 2001) is also minimized. In the recent study of Watanabe & Chiba (2001), concerning the land snail Euhadra quaesita, it was concluded that historical processes such as vicariance and repeated mixing of populations can produce an exceptionally fine scale of geographical variation and may lead to the accumulation of high genetic diversity between and within populations of a species. SATURATION AND AUTHENTICITY OF mtdna SEQUENCES Transitional bias has been reported in most mtdna studies (Brown et al., 1982) and proportions of the TS : TV ratio may be used to interpret the relative degree of mutational saturation (Hillis, Mable & Moritz, 1996). Multiple substitutions accumulate progressively at given nucleotide sites with evolutionary time, resulting in progressively lower TS : TV ratios (Kocher & Carleton, 1997). In our data set, TS : TV ratios decreased with increasing level of taxonomic comparison (Fig. 2), suggesting increased saturation and phylogenetic noise. Additionally, in almost all pairwise comparisons transversions exceeded transitions (Fig. 2). On the other hand both TS and TV continue to accumulate linearly with genetic distance (p), suggesting that saturation did not obscure the phylogenetic signal (Hillis et al., 1996), and the transitional

11 mtdna PHYLOGENY AND MORPHOLOGICAL DIVERSITY IN MASTUS 393 Figure 4. Scatter-plot of the principal component analysis on the first two principal components accounting for 78.4% of the total shell variation. (a) Cretan populations (b) Island group populations. Percentage numbers indicate the portion of each sampling locality s individuals, included in the circled areas. bias reported for other taxa does not occur here as is the case in Doridacea as well (Thollesson, 1999). Typically, TS values are expected to level off while TV increases linearly with genetic distance (Mindell & Honeycutt, 1990). The linear increase of TS with genetic distance does not suggest substantial saturation (Lydeard et al., 1996). Data from vertebrate taxa showed that transitions in the 16S rrna gene did not exhibit saturation up to an estimated time of divergence of 300 million years. Desalle et al. (1987) and Desalle (1992) found similar results. They reported that the 16S rrna gene showed little if any saturation effect for approximately 200 million years after a comparison of the sequences from Drosophila species and the mosquito, Aedes albopictus. The authenticity of the mtdna sequences is confirmed by the nucleotide composition that is clearly biased towards A-T, as is the case in animal mitochondrial DNA (Thollesson, 1999). The most conserved fragments of the 16S rrna molecule of molluscs, as reported in Lydeard et al. (1996), are retained in our sequences and this minimizes the possibility of having sequenced nuclear copies of the 16S rrna mtdna gene. These nuclear copies have been reported to diverge under more relaxed evolutionary constraints (Bensasson, Zhang & Hewitt, 2001) and therefore

12 394 A. PARMAKELIS ET AL. Table 6. Assignment of individuals to populations according to their shell morphology and per cent of correct assignments of individuals to their predifined group. Rows are the observed assignments and columns are the assignments predicted by the Discriminant Analysis (Wilks l = ; approximate F (117,1605) = 4.94; P < ) Locality Per cent correct Itanos Ferma Atherinolakos Goudouras Agia Fotia Koufonisi Marmara Trachylos Stroggylo Makroulo Itanos Ferma Atherinolakos Goudouras Agia Fotia Koufonisi Marmara Trachylos Stroggylo Makroulo Total Table 7. Coefficients of variation (%) for each shell variable (see Table 5). The mean CV is used as a measure of the relative morphological variation of each population Itanos Ferma Atherinolakos Goudouras Agia Fotia Koufonisi Marmara Trachylos Stroggylo Makroulo SH MH MD LWH LWD PLWH PLWD HTWP WFWP WSWP WTWP AL AS Mean CV of population could explain the high levels of sequence divergence we have obtained. However, besides the general congruence of our sequences to the secondary structure of the 16S rrna molecule of molluscs, the use of muscle tissue for DNA extraction (which is rich in mtdna relative to nuclear DNA) also reduces the possibility of having amplified numts (nuclear copies of mtdna genes), as is concluded by Greenwood & Pääbo (1999). Additionally, in the chromatographs of our sequences we did not encounter any double peaks that are indicative of numts contamination of the PCR product (Kimura et al., 2002). MOLECULAR PHYLOGENY OF INDIVIDUAL SEQUENCES The clustering pattern of the individual sequences in the most-parsimonious tree (Fig. 3) is highly unexpected. In most cases individuals from single islets cluster apart from each other and do not form a monophyletic group (i.e. islets of Stroggylo, Koufonisi and Marmara) while at the same time the island group is divided into two major clades, that of Group A and Group B. The islet of Stroggylo clusters along with the Cretan populations of Goudouras, Atherinolakos, Agia Fotia and Itanos (Group A), while the remaining islets of the island group cluster together and are closely related to the Cretan population of Ferma (Group B). The clustering pattern of the individuals in this study is surprising in the sense that one would expect that all the individuals of the island group would form a monophyletic group, since all the islets were connected to each other until the historic ages. In addition, it would be expected that the clade of the

13 mtdna PHYLOGENY AND MORPHOLOGICAL DIVERSITY IN MASTUS 395 whole island group would be more closely related to some of the Cretan populations, depending on which one of those had served as the colonization source of the island group. However, the data are in almost complete contrast to this. The islets of the group seem to have been colonized from different Cretan populations, as indicated by the most-parsimonious tree, while individuals from different islets cluster together and individuals from the same islet cluster apart. This unexpected clustering of individual sequences can be explained by two different evolutionary scenarios. The first one would assume the complete mixing of all the populations, resulting in a mixing up of the mtdna lineages within an individual population. The second one would assume that the different mitochondrial variants of the populations of this study (that do not cluster according to population or islet) are derived from polymorphisms in an ancestor predating the colonization events themselves. These two scenarios are complementary rather than mutually exclusive, and both have been used to explain similar patterns obtained in land snails (Goodacre & Wade, 2001; Watanabe & Chiba, 2001). Both these scenarios could account for the high genetic diversity and the unexpected clustering of Mastus individuals in this study. Let us consider the results that support the mixing of populations scenario. The migration of individuals (mainly through human activities) from different areas could lead to the mixing up of separated populations. If this is true then it would be expected that in some individual populations the different haplotypes found would not fall as a monophyletic group. In addition, phylogenetically similar haplotypes should be found at separate sites at long distances from their main distribution areas. In our case there are individual populations that fit this description. For example on the Stroggylo islet the two individuals sequenced cluster apart from each other and diverge as much as 4.7%. The individuals from Koufonisi islet also behave the same way, although the genetic distance between them is only 2.14%. Finally, in the case of the Ferma population the two individuals sequenced differ as much as 13.09% and while one of the individuals of this population clusters with the individuals of the Koufonisi island group, the second individual is basal to this group. In the study of the land snail Euhadra peliomphala (Hayashi & Chiba, 2000; Shimizu & Ueshima, 2000) it was suggested that vicariance and mixing of populations on a small spatial scale might have occurred in the past, arising from sea level changes, but in those studies the levels of intrapopulation variation were not investigated. Watanabe & Chiba (2001) investigated both the intra- and interpopulation variation of the sequences of Euhadra quaesita from Kanto region in Japan, and concluded that the mixing up of populations that took place on a small spatial scale, due to regional sea level changes, have generated the complex geographical patterns of genetic subdivision encountered in the species. Our results are quite similar to those of Watanabe & Chiba (2001) and we argue that it is the active geological past (including sea level changes) of the area and the long-term presence of man that has produced this mixing of Mastus populations. The long-term and intensive presence of man in an area has been concluded before to influence the genetic structure of animal populations, for example in the Australian frog species Limnodynastes tasmaniensis and L. peronii (Schäble & Moritz, 2001), where two highly diverged (up to 10.24%) ND4 mtdna lineages were present in several single populations. Repeated cycles of isolation and secondary contacts of populations that diverged whilst separated, without reaching the point of reproductive isolation, play a major role in the formation of the genetic diversity appearing in Mastus populations in this study. This repetition of isolation and mixing results in an exceptionally fine scale of geographical variation and the accumulation of high genetic diversity within a single population, and consequently along the range of the species distribution. Thus, besides the demographic structure of land snail populations that allows the retention of ancient polymorphisms in the mtdna lineages, the mixing of diverged populations complicates the geographical distribution of genetic diversity. Genetically distinct individuals that are able to reproduce are brought together and restricted gene flow, which occurs between specific demes due to the highly structured populations of pulmonates, allows the maintenance of the mixed-up mtdna lineages through stochastic processes (Thomaz et al., 1996; Guiller et al., 2001). On the other hand, in the face of adequate gene flow, natural selection could favour specific associations between mtdna and nuclear genes (associations developed whilst populations are separated) and that would radically slow the homogenization process of the mtdna lineages (Thomaz et al., 1996; Goodacre, 2001), thus allowing certain mtdna haplotypes to be maintained. Besides natural selection, the very mode of land snails dispersal could be responsible for the maintenance of the various mtdna lineages in a population. Snails are believed to be leptokurtically dispersed. Most individuals do not migrate very far, but there are occasional longdistance migrants. Leptokurtic dispersal has been shown to increase genetic patchiness, since initial rates of inbreeding in demes formed by long-distance migrants are high (Ibrahim, Nichols & Hewitt, 1996). The variation between the populations founded by a few long-distance dispersers is governed by genetic drift alone and need not necessarily involve selection of haplotypes (Goodacre, 2001).

14 396 A. PARMAKELIS ET AL. It is interesting to note that the area of Ferma (which exhibits the highest level of intrapopulation divergence observed up to now in the literature for 16S rrna mtdna sequences) and the island complex which also exhibits high levels of intrapopulation divergence (Fig. 5) are areas which have become available for land snails to colonize relatively recently, since the sediments present there are mostly Pliocenic and Holocenic (Peters, 1985). This tends to support the mixing of populations hypothesis, leading as it does to the high level of genetic diversity of land snails populations on a small spatial scale. The molecular analyses in the present study are subject to the restriction of small sample sizes. Despite this our data highlight that there are localities where more detailed investigation of population genetic structuring would be rewarding and should produce insights into general biogeographical and historical processes governing the genetic and morphological diversity of land snail species in Crete. STATISTICAL ANALYSES OF SHELL MORPHOMETRIC CHARACTERS In the principal component analysis of the shell variables, the plot of the factor scores of the Cretan populations (Fig. 4) showed that there is a high degree of overlap between the individuals from Agia Fotia and Itanos. The populations with the most variable shell morphology are those of Goudouras and Ferma, whose individuals are scattered throughout the morphological variation of all populations together in the PCA scatter-plot. These two populations greatly overlap each other as well. On the other hand, the population of Atherinolakos seems to be less variable and is the most distinct among the others. In contrast to the Cretan populations, those of the island group do not overlap. All the islets seem to have a more or less distinct shell morphology with a very small degree of interconnection (Fig. 4) and it is mainly the overall size that discriminates the islets populations. Combining both scatter-plots of Figure 4, it can be seen that the populations from the islets are interconnected with certain Cretan populations. In the discriminant analysis it was indicated that the majority of the individuals of the Cretan populations that were not grouped to their predefined group were assigned to another Cretan population. On the other hand, when misclassified, the individuals of the populations of the island group were mostly assigned to Cretan populations, but the percentage of correct classifications of the islets populations was in most cases well above 60%. The pattern obtained with the PCA and DA is confirmed by the calculations of the mean coefficients of variation for each population (Table 7). From the values of the coefficients of variation the rough conclusion that all Cretan populations exhibit more or less equal levels of morphological variation can be deduced. The least variable Cretan population is that of Ferma (Table 7), while the most diverged is that of Goudouras. In addition, it can be seen that in the island group populations a significant decrease in the shell variation is observed compared to that of the Coefficient of variation (%) Within populations average genetic distances (%) ITANOS ATHERINOLAKOS FERMA AGIA FOTIA GOUDOURAS KOUFONISI STROGGYLO TRACHYLOS MAKROULO MARMARA Localities Figure 5. Mean coefficient of variation for each population (computed from the coefficient of variation of all shell variables) and within-populations average genetic distances. The islets are presented in order of decreasing area. A slight indication for decrease in morphological variation with decreasing islet area is observed, while the genetic divergence is not effected by the area of the islets. However, due to small sample sizes, statistical support is not provided for either of these observations.