More than one species of Messor harvester ants (Hymenoptera: Formicidae) in Central Europe

It is commonly held that Central Europe harbours but a single harvester ant species, namely Messor structor. Recently discovered bionomic differences between two Central European populations, which may reflect interspecific variation, cast doubt on this assumption. In the present study we test alternative hypotheses – one versus two harvester ant species in Central Europe and adjacent regions – by investigating the genetic diversity of ants determined as M. structor or close to it (“M. cf. structor”). Sequences of the mitochondrial COI gene revealed two major lineages of different but partially overlapping geographic distributions, both occurring in Central Europe. The existence of a cryptic species within M. cf. structor is the most plausible interpretation, since the sequence divergence between the two major lineages equals those between M. capitatus, M. concolor and M. bouvieri. The phylogenetic analyses revealed a distinct substructuring for both of the detected major lineages and the possible existence of additional cryptic species. 469 * These authors contributed equally to the work. species (Bourke & Franks, 1995). Granted that species delimitation is subject to ongoing discussion (see Sites & Marshall, 2003 for review) the species category is nonetheless of unquestioned operational value for systematics and phylogenetics (Avise & Walker, 2000; Mallet, 2001). Hence in this paper we test a pair of alternative hypotheses whether the differing life histories of Central European Messor cf. structor harvester ants reflect (i) a single, bionomically versatile species; or (ii) two fully separated species. Traditional insect classification is morphology-based; in particular morphometric analyses have proved powerful (Seifert, 2002). However, groups with small interspecific and high intraspecific variation are often poorly resolved by such methods alone (for review: Wiens, 1999; for ant examples: Lucas et al., 2002; Steiner et al., 2004, 2005, 2006a; Knaden et al., 2005). A strong, gradual size polymorphism within the worker caste of Central European Messor ants causes pronounced allometric distortions, making morphometric differentiation more complicated than in species with monomorphic workers (Cs sz et al., unpubl.). In such cases molecular genetic analyses may provide complementary information. Morphologically most similar species may differ markedly in mitochondrial DNA (mtDNA), as shown for ants, among others, by Heinze et al. (2005), Knaden et al. (2005) and Steiner et al. (2004, 2005, 2006a). In this paper the one-or-two-species problem in M. cf. structor is addressed by surveying its mtDNA diversity in Central Europe. In order to evaluate geographic, intraand interspecific variation samples from other European regions and other Messor species are included in the analysis. MATERIAL AND METHODS Individuals from 40 Messor colonies from ten European countries were studied (Table 2), including samples from German and Austrian populations, which reportedly differ bionomically (Fig. 1; Heller, 1971; Schlick-Steiner et al., 2005b). Voucher specimens were deposited in the Hungarian Natural History Museum in Budapest, the Babes-Bolyai University in Cluj-Napoca and the private collection of B.C. Schlick470 first available use of Messor structor subsp. striaticeps var. sevani Karawajew, 1926: 103; junior synonym of structor: Arakelian (1994: 39) Caucasus Arnol’di, 1977: 1644 Messor clivorum subsp. sevani first available use of Messor barbarus subsp. capitatus var. jakowlevi Ruzsky, 1905: 750; junior synonym of structor: Dlussky et al. (1990: 224) Ukraine Arnol’di, 1977: 1643 Messor rufitarsis subsp. jakowlevi misspelled as rufitarsis tadzhicorum: Arnol’di (1977: 1644); subspecies of structor: Bolton (1995: 257) Tadzhikistan Arnol’di, 1969: 79 Messor rufitarsis subsp. tadzhikorum first available use of Messor structor st. rufitarsis var. darianus Santschi, 1926: 291; junior synonym of structor: Kuznetsov-Ugamsky (1927: 92); synonymy of turanicus: Pisarski (1967: 383) Turkestan Pisarski, 1967: 383 Messor rufitarsis subsp. darianus junior synonym of structor: Atanassov & Dlussky (1992: 114) Yugoslavia Finzi, 1929: 92 Messor structor subsp. novaki first available use of Messor barbarus st. structor var. aegaea Emery, 1921: 213 Turkey Menozzi, 1928: 126 Messor structor var. aegaea junior synonym of clivorum: Tarbinsky (1976: 56) Turkestan Kuznetsov-Ugamsky, 1927: 92 Messor structor var. subpolitus junior synonym of rufitarsis ssp. darianus: Arnol’di (1977: 1644) in key; of structor: Dlussky et al. (1990: 224) Uzbekistan Kuznetsov-Ugamsky, 1927: 91 Messor structor subsp. turanicus junior synonym of structor: Atanassov & Dlussky (1992: 114) Turkey Emery, 1921: 215 Messor barbarus subsp. varrialei junior synonym of clivorum: Emery (1921: 210) Persia Crawley, 1920: 164 Messor platyceras var. rubella subspecies of structor: Bolton (1995: 256) Persia Crawley, 1920: 163 Messor platyceras subspecies of clivorum Arnol’di (1977: 1644); junior synonym of structor: Dlussky et al. (1990: 224) Turkestan Ruzsky, 1905: 738 Messor tataricus junior synonym of structor: Dlussky et al. (1990: 224) Turkestan Ruzsky, 1905: 735 Messor structor var. clivorum combination in Messor: Emery (1908: 456); junior synonym of structor: Emery (1921: 210) Italy Emery, 1898: 141 Stenamma (Messor) structor var. tyrrhena junior synonym of structor: Mayr (1855: 464) Russia Nylander, 1849: 39 Myrmica mutica junior synonym of structor: Mayr (1855: 464) Poland Schilling, 1839: 56 Formica aedificator junior synonym of structor: Nylander (1856: 85) Austria Fabricius, 1804: 407 Formica lapidum junior synonym of structor: Nylander (1856: 85) Austria Fabricius, 1804: 406 Formica rufitarsis combination in Messor: Emery (1897: 238) France Latreille, 1798: 46 Formica structor Current status / Comments Type locality Author Original combination TABLE 1. List of taxon names currently regarded as junior synonyms or subspecies of Messor structor (Latreille, 1798).

species (Bourke & Franks, 1995). Granted that species delimitation is subject to ongoing discussion (see Sites & Marshall, 2003 for review) the species category is nonetheless of unquestioned operational value for systematics and phylogenetics (Avise & Walker, 2000;Mallet, 2001). Hence in this paper we test a pair of alternative hypotheses whether the differing life histories of Central European Messor cf. structor harvester ants reflect (i) a single, bionomically versatile species; or (ii) two fully separated species.
Traditional insect classification is morphology-based; in particular morphometric analyses have proved powerful (Seifert, 2002). However, groups with small interspecific and high intraspecific variation are often poorly resolved by such methods alone (for review : Wiens, 1999;for ant examples: Lucas et al., 2002;Steiner et al., 2004Steiner et al., , 2005Steiner et al., , 2006aKnaden et al., 2005). A strong, gradual size polymorphism within the worker caste of Central European Messor ants causes pronounced allometric distortions, making morphometric differentiation more complicated than in species with monomorphic workers (Cs sz et al., unpubl.). In such cases molecular genetic analyses may provide complementary information. Morphologically most similar species may differ markedly in mitochondrial DNA (mtDNA), as shown for ants, among others, by Heinze et al. (2005), Knaden et al. (2005) and Steiner et al. (2004Steiner et al. ( , 2005Steiner et al. ( , 2006a.
In this paper the one-or-two-species problem in M. cf. structor is addressed by surveying its mtDNA diversity in Central Europe. In order to evaluate geographic, intraand interspecific variation samples from other European regions and other Messor species are included in the analysis.
PCR products were purified using the QIAquick PCR purification kit (Qiagen, Hilden, Germany) and then either directly sequenced in both directions using the Big Dye termination reaction chemistry (Applied Biosystems, Foster City, USA) or sequenced after cloning (pGEM-T vector, Promega and DH5a E. coli cells, plasmid DNA was extracted with the QIAprep Miniprep-kit (Qiagen), sequenced with the M13 universal primer), and analysed with an ABI 377, or, alternatively, with an ABI Prism 310 automatic sequencer (Applied Biosystems).
1255 bp were used for phylogenetic analyses. Nucleotide sequence alignment was achieved with Clustal X using the default settings (Thompson et al., 1997). To compare the relationships among sequences broken down to haplotypes, distance (Neighbour Joining algorithm, NJ, based on Tamura-Nei distance) and character (maximum parsimony, MP; Bayesian Markov Chain Monte Carlo BMCMC) analyses were performed using PAUP* (test version 4.0b3a; Swofford, 1998) and MrBayes v3.1 (Ronquist & Huelsenbeck, 2003). For MP analysis all characters were assigned equal weights. MP trees were generated with heuristic search using the tree bisectionreconnection branch swapping with 10 random taxon addition sequence replicates and the Multree option in effect. The maximum number of trees in memory was set to 10,000. Bootstrapping was applied for NJ (1000 replicates) and MP trees (100 replicates). Prior to BMCMC analysis the GTR+I+G model (Tavaré, 1986;Yang, 1993) was chosen using Modeltest 3.06 (Posada & Crandall, 1998), which uses hierarchical likelihood ratio tests (Huelsenbeck & Rannala, 1997) to determine how well competing substitution models fit the data. In BMCMC analysis 1,000,000 generations with a sample frequency set to 100 were run twice. As after 750,000 generations stationarity was achieved (average standard deviation of split frequencies stable at 0.004), the last 2,500 trees of each run were used to compute a majority rule consensus tree assigning posterior probabilities of tree topology.

RESULTS
The 1255 bp sequences of the COI gene of all samples were deposited in GenBank under accession numbers DQ074323-DQ074365. No gaps arose in alignment. All phylogenetic analyses (Fig. 1)  Within the 29 samples determined as Messor cf. structor, mutations at 166 sites (17 mutations at the first, two at the second, and 147 at the third codon position) resulted in 16 haplotypes with a maximum sequence divergence of 9.3% (Fig. 1). All phylogenetic trees revealed the monophyly of M. cf. structor and samples determined as M. cf. structor always clustered into the same two major lineages, arbitrarily termed A and B. Number of samples in lineage A was 14 in B 15. Node support for the two lineages were maximum in all trees. Maximum sequence divergence within the lineages was 2.4% in lineage A and 5.1% in lineage B. Minimum sequence divergence between lineages was 7.1%. Moreover, in all trees the two major lineages were substructured and the sublineages were supported by very high node support values (Fig. 1). Lineage A contained two sublineages, comprising haplotypes HT1-HT2 and HT3-HT5, with a minimum divergence of 2.3% between them. Lineage B contained four sublineages, HT6-HT7, HT8-HT11, HT12-HT13 and HT14-HT16, with minimum sequence divergences between the sublineages ranging from 3.6 to 5.1%. To test for a possible saturation effect on the phylogenetic signal at the third codon position, a NJ search based only on the first and second positions of the COI data was conducted (tree not shown). The M. cf. structor lineages A and B, as well as all sublineages, were confirmed by high node support values.
The map of where the samples of M. cf. structor (Fig.  1) were collected suggests a geographic separation of DNA lineages with partial overlaps. Lineage A has a larger distribution area; from France and Germany in the west to Bulgaria and Romania in the east. Lineage B is found from Bulgaria to Romania and Austria with its westernmost record from Slovenia. In the Dinaric and Balkan region the distribution areas overlap.

DISCUSSION
The mtDNA trees indicate that the genus Messor, as currently understood, may be polyphyletic since Aphaenogaster iberica occurs closer to the Palearctic Messor than to the Nearctic outgroup species Messor lobognathus and M. chamberlini. This agrees with morphological findings (male genitalia: P.S. Ward, pers. comm.; presence / absence of strong propodeal spines in the worker caste: Cs sz, unpubl.) and contributes to the dispute on the generic classification of Nearctic "Messor" species (reviewed by Brown, 1974 andBolton, 1982).
Considerable genetic variation was found in the mitochondrial COI gene among samples determined as M. cf. structor. The two major lineages A and B are very far apart. The minimum divergence of 7.1% between the lineages is in the order of magnitude of the minimum interspecific divergences between M. bouvieri, M. Fig. 1) and that between congeners of Cardiocondyla, Cataglyphis, Lasius, Myrmica and Tetramorium ants (Savolainen & Vepsäläinen, 2003;Steiner et al., 2004Steiner et al., , 2005Steiner et al., , 2006aKnaden et al., 2005;Heinze et al., 2005). Thus, the mtDNA data are compatible with the two-specieshypothesis.
For a profound evaluation of genetic variation, however, nuclear markers should be analysed in addition to mtDNA (Beltrán et al., 2002;Lin & Danforth, 2004), and molecular data should be substantiated by other approaches such as morphology, karyology, semiochemistry and ecology (Wetterer et al., 1998;Ross, 2001;Lucas et al., 2002;Ward & Brady, 2003;Janda et al., 2004;Seifert & Goropashnaya 2004;Steiner et al., 2004;Knaden et al., 2005;Maeder et al., 2005;Schlick-Steiner et al., 2005a;Ward & Downie, 2005). This should also rule out hybridization with a hitherto not included Messor species and introgression of its haplotypes resulting in the observed mtDNA pattern (cf. Ross & Shoemaker, 2005). Preliminary microsatellite data of German lineage A and Austrian lineage B populations (Arthofer et al., 2005) suggest a complete separation of allele size at one microsatellite locus (MS2D) and a private allele of the German populations at another (MS2C; Fig. 1). Current morphological analyses (Cs sz et al., unpubl.) additionally corroborate that these populations belong to different species.
Overall, the evidence presented here leads us to reject the hypothesis that there is only one species of Messor harvester ants in Central Europe. The alternative hypothesis of two species, corresponding to lineages A and B, clearly is more plausible. Messor cf. structor thus encompasses a cryptic species in Central Europe. Only by in-depth analyses will it be possible to evaluate whether the distinct substructuring of A and B in the mtDNA trees is due to the existence of even more than two species. The high minimum divergence values between sublineages of A (2.3%) and B (3.6-5.1%) are compatible with this scenario.
Under these circumstances the assignment of species names to mtDNA lineages is futile. It is not even clear which lineage should bear the name Messor structor (Latreille, 1798), since the original description is vague and the type material lost (J. Casevitz-Weulersse, Museum of Natural History, Paris, pers. comm.). Biogeographic considerations tentatively point to lineage A because it is the only lineage presently known from the terra typica in France (Fig. 1). For the remaining lineages any of 19 taxon names currently classified as subspecies or junior synonyms of Messor structor could apply (Table  1).
From a phylogeographic point of view the distributions of lineages A and B are remarkable. The disjunction in Central Europe (Fig. 1) suggests different routes of dispersal, probably from a common Pleistocene refugium: a bypass of the Alps in the west by the ancestors of the German populations, as hypothesized for the butterfly Polyommatus coridon (Schmitt et al., 2002) and for certain dragonflies (Sternberg, 1998), and a bypass in the east by the ancestors of the Austrian and Czech populations.