Phylogenetic Relationships of Palaearctic Formica Species (Hymenoptera, Formicidae) Based on Mitochondrial Cytochrome b Sequences

Ants of genus Formica demonstrate variation in social organization and represent model species for ecological, behavioral, evolutionary studies and testing theoretical implications of the kin selection theory. Subgeneric division of the Formica ants based on morphology has been questioned and remained unclear after an allozyme study on genetic differentiation between 13 species representing all subgenera was conducted. In the present study, the phylogenetic relationships within the genus were examined using mitochondrial DNA sequences of the cytochrome b and a part of the NADH dehydrogenase subunit 6. All 23 Formica species sampled in the Palaearctic clustered according to the subgeneric affiliation except F. uralensis that formed a separate phylogenetic group. Unlike Coptoformica and Formica s. str., the subgenus Serviformica did not form a tight cluster but more likely consisted of a few small clades. The genetic distances between the subgenera were around 10%, implying approximate divergence time of 5 Myr if we used the conventional insect divergence rate of 2% per Myr. Within-subgenus divergence estimates were 6.69% in Serviformica, 3.61% in Coptoformica, 1.18% in Formica s. str., which supported our previous results on relatively rapid speciation in the latter subgenus. The phylogeny inferred from DNA sequences provides a necessary framework against which the evolution of social traits can be compared. We discuss implications of inferred phylogeny for the evolution of social traits.


Introduction
Large scale molecular studies have recently focused on the phylogeny of ant subfamilies [1,2]. Based on these studies, diversification of ants started in early Cretaceous 115-135 Mya, and the age of the subfamily Formicinae is about 80 Mya. One of the internal dating points used by Brady et al. [1] was 44.1 Mya for the fossils of the genus Formica found in Eocene amber [3,4]. Formica ants represent a large group of soil insects that occur mainly in the Holarctic. The genus has currently 176 recognized species, a bigger part of which are distributed in the Nearctic and a smaller part (63 species) in the Palaearctic ( [5], World Catalogue of Ants, www.antweb.org, 27 September 2011). Even though the genus is well studied, these numbers are still changing (e.g. [6,7]). Many species are widespread and abundant, and they play an important role in ecosystems being active predators, tending aphids and improving soil composition. Formica species also demonstrate a great diversity of complex behavior and social organization. The subgenus Raptiformica includes slave-making species, and the subgenera Formica s. str. and Coptoformica use temporary parasitism as a mode of founding new colonies, while the species of the subgenus Serviformica are used as slaves. The organization of colonies ranges from simple monogynous societies to huge supercolonies [8,9,10].
It is often concluded that facultative behavioral responses as well as traits connected to some socially important features have evolved independently several times across lineages (e.g. [11,12]). On the contrary, the army ant syndrome or the complex of specific behavioral and reproductive traits has been shown to have evolved once rather than independently in the Old and New Worlds as thought previously [13]. Variation of social characteristics has made Formica ants useful for studying the evolution of social organization and for testing different theoretical implications of the kin selection theory (e.g. [14]). For example, Boomsma and Sundström [15] made a comparative phylogenetic study on the evolution of polyandry, looking for the correlation between the frequency of multiple mating and the paternity skew in seven Formica species in which single mating is still common. As no phylogenetic data were available, they used ten different phylogenetic hypotheses for testing the statistical significance of the observed relationship. Similarly, Helanterä and Sundström [16] studied whether colony size or colony kin structure had affected the evolution of worker egg-laying, and they corrected the phylogenetic dependencies by using our preliminary results. The phylogeny (as inferred e.g. from DNA sequences) provides the necessary background against which the evolution of the other traits can be compared.
Most taxonomists have distinguished four subgenera in the European Formica species (e.g. [3,17,18]), namely Raptiformica, Coptoformica, Serviformica, and Formica s. str. In addition, the only described species of the subgenus Iberoformica has a restricted distribution in the Iberian Peninsula [19]. Distinctions in morphological characters used for taxonomy of the genus often tend to be clear in local faunas but vague and imprecise when closely related species are studied on a large geographical scale. The previous allozyme study on 13 Formica species of four European subgenera [20] agreed with the subgeneric division based on morphological and behavioral characters with some exceptions. One of the exceptions was the topological position of F. uralensis (subgenus Formica s. str.) that was associated with Serviformica. According to Dlussky [3], F. uralensis belongs to Serviformica on the basis of the morphological similarities of the males. These characters were considered by him evolutionary conservative, while similarities in the worker morphology and nest constructing behavior of F. uralensis and Formica s. str. were believed to be secondary (convergent) traits. Due to a low level of allozyme variation in Formica ants, the resolution of the relationships of the species was poor with these data and the phylogeny of the genus remained unclear [20].
The present study aims to explore the phylogenetic relationships of the Palaearctic subgenera within genus Formica. One aim is to provide phylogenetic information which can be used in comparative studies such as those by Boomsma and Sundström [15] and by Helanterä and Sundström [16]. The other aim is to compare the divergence among species within and between the subgenera. The motivation for this is that the species in the Formica rufa group (Formica s. str.) form a morphologically variable and poorly differentiated group of species which hybridize (e.g. [21,22,23]). We have used mitochondrial DNA (mtDNA) to examine the Formica rufa group [21] and shown that these species form a monophyletic group of closely related species. Our results generally supported the division of this group into distinct species suggested on the morphological basis. Yet the species are geographically widely distributed and show, even within a single species, a wide variation of social structures from monogynous societies to large supercolonies which resemble those in invasive ants (e.g. [9,10,21]). The results from the F. rufa group suggest recent radiation and we want to compare this pattern with that in the other subgenera.

Sampling and molecular techniques
We examined 35 individuals of 24 Formica species representing the four subgenera ( Table 1). As an outgroup, we used five individuals of Polyergus rufescens from two locations in Germany because Polyergus has been suggested to be the sister group of Formica [24]. No specific permits were required for the described field collections. All locations were not privately-owned or protected in any way. The field studies did not involve endangered or protected species. All samples were stored in 70% ethanol until DNA extraction. Total genomic DNA was extracted from only the head and mesosoma of single individuals with the DNeasy Tissue Kit (QIAGEN Inc.). A 1.5 kb mtDNA fragment including the cytochrome b gene (cyt b) was amplified and sequenced with the following primers: Cytb-Fe-F [25], CB1, CB2, CB3, tR S [26], CB-11059, CB-11178, and CB-11449 designed by using of the Oligo Primer Analysis Software v. 6.45 [27]. Polymerase chain reaction (PCR) was carried out in 25 mL volumes containing 16PCR buffer, 2.0 mM MgCl 2 , 0.4 mg/mL BSA, 0.2 mM dNTPs (MBI Fermentas), 0.4 mM of each primer and 2.0 U Taq polymerase (Fermentas). A program for the amplification in a thermal cycler was used as follows: 3 min at 94uC, 35 cycles of 30 sec at 92uC, 30 sec at 45uC, 1-2 min at 68uC, and 10 min at 72uC.
PCR products were cleaned with the QIAquick Gel Extraction Kit (QIAGEN Inc.) and then sequenced on an Applied Biosystems

Data analysis
The simplest model of nucleotide substitution with the best fit to our data was selected on the basis of minimal value of the Akaike information criterion (AIC) that finds balance between goodnessof-fit and complexity of model by using the ModelGenerator software [28]. The transversion model (TVM) with correction for rate heterogeneity among nucleotide sites and empirical base frequencies, proportion of invariable sites (0.39), and alpha parameter of gamma distribution (0.39) was selected for the data set without an outgroup (1441 bp). The simplest model with the best fit to the data set with the outgroup Polyergus rufescens (1416 bp), was TVM with correction for rate heterogeneity among nucleotide sites and empirical base frequencies, proportion of invariable sites (0.40), and alpha parameter of gamma distribution (0.55). Maximum likelihood trees (ML) were constructed by using the nearest neighbor interchange tree search with MultiPhyl software [29]. Neighbor-joining (NJ) trees were constructed with PAUP* version 4.0b10 [30]. Sequence variation and substitution pattern of the 1.5 kb mtDNA fragment were analyzed using the program MEGA v. 3.0 [31]. We compared the log likelihood scores of maximum likelihood trees constructed with and without a molecular clock assumption [32] to evaluate constancy in rate of the sequence evolution among lineages.

Sequence variation and phylogenetic relationships among subgenera
Over the entire 1441-bp region examined in the Formica species (excluding Polyergus), 327 nucleotide positions were variable with 266 parsimony informative polymorphic sites. The Polyergus sequence used as the outgroup was too distant for reliable rooting of the tree as the synonymous nucleotide positions were saturated. The TVM+I+G distances between Polyergus and the Formica species were within a range from 0.50 to 0.65 and exceeded more than fourfold the distances among the Formica species (0.00-0.15) ( Table 2). The use of non-synonymous substitutions alone did not resolve all the groupings within the genus Formica. The ML analyses based on all nucleotide positions with and without the outgroup produced similar trees. The tree including the outgroup showed the species pair F. candida and F. picea as basal to the rest of the species and Serviformica appeared paraphyletic. The species F. candida and F. picea are closely related and their taxonomical position was clarified only recently [33]. The tree without the outgroup separated Serviformica from the other species with only a moderate bootstrap support of 73% (Figures 1, 2). Topology of the NJ tree (not shown) without outgroup rooting was similar to the ML tree. The comparison of the likelihood scores of the ML trees constructed with and without the molecular clock assumption showed that the sequences have evolved at roughly constant rates (P = 0.15). Therefore, variation in mtDNA sequences is suitable for approximate estimation of relative divergence times.
The subgenera Formica s. str. and Coptoformica were represented by several species and formed monophyletic clades respectively. Serviformica formed also a group of its own in an unrooted tree, but we could not reliably infer its monophyly because of problems with the root. The Serviformica species which have been studied have a haploid chromosome number n = 27, except F. picea which has n = 26 like all other Formica subgenera [34,35]. Two species in our unrooted tree fell outside these three subgenera, namely F. sanguinea (subgenus Raptiformica) and F. uralensis. The systematic position of F. uralensis has been controversial in the past (e.g. [17]), and Dlussky [3] concluded that its male genitalia are most similar  Table 2. Genetic divergence estimates within and between the phylogenetic groups in Formica (%): average distances within the groups (the diagonal, in bold), mean uncorrected (below the diagonal) and net (above the diagonal) distances between the groups. to those of Serviformica. We consider this a weak argument as the variation of male genitalia is poorly studied throughout the genus and unknown in many species. Furthermore, the haploid chromosome number (n = 26) departs from that of most Serviformica species [34]. Separate position of F. uralensis in the mtDNA phylogenetic tree agreed with the results from an earlier allozyme study [20] showing that this species represents a separate phylogenetic lineage and could be placed to a subgenus of its own. Rooting of the phylogenetic tree was problematic as the outgroup sequence was too distant and the differences separating the subgenera were small. An earlier study by Hasegawa et al. [24] based on COI sequences included five Formica species from three subgenera and had Polyergus as the nearest outgroup. The branching order of the Formica species could not be solved with great confidence, but the unweighted maximum parsimony tree suggested Serviformica (represented by F. fusca and F. cunicularia in that study) as a basal subgenus with Coptoformica and Formica s. str. clustering together. This pattern agrees with the topology of the tree in Figures 1, 2. However, all the subgenera (Formica s. str., Coptoformica, Raptiformica, Serviformica, and the species F. uralensis) demonstrated a bush-like branching pattern with a low bootstrap support for any cluster containing two or more subgenera.
All genetic distances between the subgenera were around 10% implying approximate divergence time of 5 Myr if we use the average divergence rate of 2.0% per Myr suggested for insect mtDNA [36]. The genetic distance between Formica and Polyergus (0.5-0.65) would then suggest that these two genera can have diverged during the Eocene if the divergence rate has been similar. This agrees with fossil record as several species classified to the genus Formica have been recorded from the Baltic amber [4]. Dlussky [4] further concluded that Formica from the Late Eocene amber constitute an archaic group and most of them are not similar to living congeners. The sequence divergences estimated by us suggested that the subgenera included in this study are young compared to the age of the whole Formica clade since separation of the sister genera. The genus has Holarctic distribution. There are true Nearctic Raptiformica species, but no true Nearctic Coptoformica (the F. exsectoides species group is clearly different from the Palaearcitc species) and no ants comparable to the moundbuilding Formica s. str. If interchange between North America and Eurasia reached a peak in the Middle Pliocene (4 Mya) [37], most of the diversification has happened during or after that period. Although migration between the continents occurred during the Pleistocene, when sea levels dropped significantly during glacial periods, making a dry land connection between Alaska and Siberia, it included the cold-adapted groups only, not the forms restricted to temperate climates like Formica ants.

Divergence within the subgenera
The subgenus Formica s. str. formed a tight cluster of species and the internal topology was similar to that obtained in the previous study on the F. rufa group [21]. The only difference was that the polyctena/rufa clade appeared here as basal to the rest of the species instead of the truncorum/frontalis clade which is generally considered to be basal to the F. rufa group (F. aquilonia, F. lugubris, F. polyctena, F. rufa, F. paralugubris). The reason for this was most probably that the level of variation in this subgenus is low compared to the divergence from other subgenera and the resolution therefore poor. The small interspecific distances in this subgenus, particularly within the F. rufa group, are noteworthy. These interspecific distances were smaller than the sequence divergence observed within the species F. exsecta and similar to the intraspecific divergence in F. pressilabris, F. manchu and F. uralensis. These comparisons support previous findings [21] suggesting that the speciation rate has been high within the Formica rufa group of mound-building ants. This could be in part associated to the changes in the social organization when limited dispersal of females from polygynous societies can lead to strong differentiation between populations and possibly to speciation if male dispersal is also restricted [38,39]. The F. rufa group species show different types of social organization having largely monogynous and monodomous colonies in some populations, whereas the species F. aquilonia, F. paralugubris and F. polyctena are always polygynous and tend to build large supercolonies ( [9,10], but see [40]). However, similar variation in social organization is known in other Formica species, e.g. in F. exsecta [38] and F. cinerea [41] without clear signs of accelerated speciation rate or increase in intraspecific genetic divergence. The relative importance of social organization and phylogeographic history in the speciation process within this group could not be estimated from the present data alone. Genetic studies have recently revealed new species within the F. rufa group, probably differentiated as a result of geographical isolation in the Alps [42,43]. The known species have also been shown to hybridize as detected from transspecific capture of mtDNA [44] or nuclear introgression and phenotypic clustering [22,23], while the sympatric populations generally form separate gene pools [43,45].
According to the present results, at least two highly divergent lineages occurred in the subgenus Coptoformica. Formica exsecta was genetically most distant from the other species. Our previous study on F. exsecta included one sample from Tibet that proved highly divergent from the other samples of the species [46]. In the present phylogenetic tree this Tibetian sample (F. exsecta-II) clustered together with the other conspecific sample showing reciprocal monophyly of F. exsecta relative to the other Coptoformica species. It should be noted that the present data did not include F. mesasiatica [3], a morphologically defined sister species of F. exsecta that occurs only in Central Asia [18]. Our previous study on F. exsecta from the Palaearctic region included F. mesasiatica from Kyrgyzstan and placed its haplotypes within the F. exsecta clade leaving F. exsecta paraphyletic [46].
Unlike the subgenera Coptoformica and Formica s. str., the subgenus Serviformica clustered with only limited bootstrap support (73%) that included three small clades with large differences both between and within them: cinerea/fusca/lemani/selysi; cunicularia/ rufibarbis; and candida/picea. Notably, this subgenus is the largest Palaearctic group within genus Formica ( [5,6,7], www.antweb.org). Divergence among the Serviformica species far exceeded that within the other subgenera and was partly comparable with the subgeneric differences. Further studies with sampling of Nearctic species and subgenera are needed to shed more light to the history of the genus and the possible monophyly of Serviformica.
The shallow phylogeny of the Formica s. str. clade and the low divergence estimate (1.18%) within this subgenus supported our previous results that speciation in this subgenus, particularly among the species of the F. rufa group, was relatively fast [21]. Bernasconi et al. [43] have reported from the Swiss Alps a population which may represent yet another species of the F. rufa group, and Kulmuni et al. [23] showed that some type of incompatibility factors keep populations with apparent hybrid background viable but separate. On the contrary, the Serviformica subgenus was highly diverged (within-subgenus divergence estimate was 6.69%) with possible subdivisions, even though the phylogeny is incomplete as only eight species were studied. Formica uralensis showed low genetic divergence despite its clear distinction from the other phylogenetic groups. Sequencing of the same mtDNA fragment from 11 samples of this species throughout its European distribution range revealed the nucleotide diversity as low as 0.19% [47]. Similar and even lower diversity estimates have been detected in the F. rufa group species [21]. It is possible that F. uralensis represents the only species that survived in this lineage and underwent reduction in historical effective size during the last glacials like the F. rufa group species

Character evolution
All the subgenera (including F. uralensis) except Serviformica share some important social characteristics. They build mound nests with plant materials (even though the mound formation is not always clear in F. sanguinea and F. truncorum) and it is generally considered that they have lost the ability of independent singlequeen colony foundation. When colonizing a new locality, they found colonies either by temporary social parasitism or by colony fission, or they are dulotic (F. sanguinea) and use slaves of another species. The Serviformica species do not build mounds with plant materials, they are capable for single-queen colony foundation, they are used as hosts during the socially parasitic colony founding by F. sanguinea, F. uralensis, and all species of Coptoformica and Formica s. str., and they also serve as slaves in F. sanguinea colonies. The branching order given in Figures 1, 2 suggests that the mound-building behavior and parasitism could have evolved only once before the other subgenera separated from each other. A molecular phylogeny of closely related Formicine species by Hasegawa et al. [24] showed that slave-making behavior has evolved independently in two closely related genera, Polyergus and Rossomyrmex. Both genera are more closely related to the genera they parasitize (Formica and Proformica, respectively) than to each other. Moreover, other phylogenetic studies on ants have indicated that social parasitism has originated independently several times in Myrmica [48]. Even though the phylogenetic clustering (Figures 1,  2) indicates the possibility that the nesting behavior including mound building with plant particles and temporary social parasitism may have evolved once in the ancestral lineage, the short internal branch, uncertainty of rooting, and the high frequency of interspecific parasitism in ants in general [49], make such a conclusion premature.
In addition to inferring the origin of qualitatively defined traits (such as parasitism), a phylogenetic tree and phylogenetic contrasts have been used to evaluate the evolution of quantitative traits. Using this approach, Boomsma and Sundström [15] presented the data from seven Formica species showing a negative association between the paternity skew and the frequency of double mating. In other words, the more frequent double mating is, the more evenly sperm is used by females. The analysis taking into account putative phylogenetic relationships of the species showed that the result was robust. It is therefore likely that the general conclusion holds even when none of the ten hypothetical trees used by Boomsma and Sundström [15] agrees with the result obtained here. Nevertheless, the present results provide a background against which such comparative studies can be made, as demonstrated by the comparative study of worker reproduction by Helanterä and Sundström [16].