Redescritpion of Proformica nasuta (Nylander, 1856) (Hymenoptera, Formicidae) using an integrative approach

The taxonomy of the Palaearctic ant genus Proformica Ruzsky, 1902 is confused and in need of revision. The type specimen for P. nasuta (Nylander, 1856), the type species of the genus, was from Beaucaire, southern France, and is presumably lost. Based on extensive sampling of Proformica nests in southern France, including the type locality, we show that the concept of P. nasuta has been erroneous for more than a century. We integrate information from the morphology of workers and sexual castes, DNA markers, and cuticular hydrocarbons to re-define species in southern France. This allowed us to provide a new, accurate description of P. nasuta and designate a neotype, as well as reference individuals for all castes. In addition, we propose a name, P. longipilosa sp. nov., for a species that since the end of the 19th century has mistakenly been included in P. nasuta.


Introduction
The ant genus Proformica Ruzsky, 1902 is composed of 25 species (Bolton 2014) restricted to dry and open environments such as steppes, mountain meadows and Mediterranean seashores (Agosti 1994). It is endemic to the Palaearctic region, with a disjunct distribution. A first area extends from eastern Europe to eastern Asia and contains most of the species, and a second area, much more limited in species number and distribution, occurs at the southwestern tip of Europe (Portugal, Spain and southern France). This distribution is somewhat reminiscent of that of the meadow and steppe vipers, the Vipera ursinii species complex, which is composed of taxa restricted to steppe-like ecosystems. Asia and Europe show distinct viper taxa that diverged in the early Pliocene, about 4 Mya (Zinenko et al. 2015). The genus Proformica may have experienced the same biogeographic history as these vipers and several other organisms inhabiting steppe-like environments (Ruano et al. 2011;Sanllorente et al. 2015). Only one taxon, P. nasuta (Nylander, 1856), is reported to occur in both Asian and western European areas.
The taxonomy of the genus Proformica is complicated and in need of revision. The situation is particularly complex in the eastern area, with currently 23 species reported. In Western Europe, two distinct zoogeographical areas can be distinguished, the Iberian Peninsula and southern France, which are separated by a barrier formed by the Pyrenees mountain range. Three described species are currently recorded for the Iberian Peninsula (Collingwood 1976), but at least six forms are recognized by ant taxonomists (Xavier Espadaler, Barcelona, pers. comm.) and substantial morphological variation within each form makes species delimitation difficult. In contrast, only one described species, P. nasuta, has been recorded for southern France (P. ferreri Bondroit, 1918 may also be present in the French part of the Pyrenees).
Proformica nasuta is the type species for the genus Proformica and was described from Beaucaire, France. The concept of this species is unclear. For instance, variation in the number of erect hairs on the mesosoma, a character commonly used in the taxonomy of Proformica, has been interpreted either as mere intraspecific variation (Espadaler & Cagniant 1987), or as an indication that the name P. nasuta actually covers two taxa (Santschi 1925;Collingwood & Yarrow 1969). Populations of species of Proformica are small, inconspicuous and patchily distributed, and the species are often considered rare. As a consequence, the genus is poorly represented in institutional collections and most taxonomic work is based on few specimens, rendering the accurate perception of intra-specific variation difficult. Moreover, the type specimen of P. nasuta has not been located. Having not been found in the most likely candidate collections and not explicitly referred to in the literature, it is presumably lost. As P. nasuta is the oldest name in the genus, designation of a neotype and a precise redefinition of this taxon are indispensable before further taxonomic work on this genus can be undertaken. For this purpose, we analysed a sample, unprecedented in its size and geographic extent, of Proformica nests in southern France using an integrative taxonomy approach based on morphological data from workers and sexuals, DNA sequences and cuticular hydrocarbons. Southern France was the best location for this investigation as it encompasses the type locality for P. nasuta and harbours no other known Proformica species. Combining the results of these different characters can increase our ability to provide valid decisions about species delimitations (Schlick-Steiner et al. 2010). Although some of these kinds of data are less relevant than others for the descriptive taxonomy of a particular species group, incongruences between results based on different kinds of data can provide information on the biology of the group studied and insights into ongoing ecological and evolutionary processes. Measurements of workers and queens are given below (Appendices 2-3).
We collected a total of 11 males from three localities (Plaine de la Crau, Pompignan and Tarascon) and examined males from Sainte-Baume that had been collected by F. Bernard (MNHN)  We amplified four DNA markers, two mitochondrial, two nuclear: (i) COI (~600 bp), coding for part of the cytochrome c oxydase subunit 1, (ii) Cytb (~700 bp), corresponding to the end of the sequence coding for NADH dehydrogenase subunit 6 and part of cytochrome b, (iii) 28S (~600 bp), coding for part of the large ribosomal subunit, and (iv) LW Rh (~550 bp), coding for part of the long-wavelength rhodopsin. COI was amplified for 45 Proformica individuals (GenBank accession numbers: KU749600-KU749637 and KU749641-KU749654) using two sets of primers covering the same region: either LepF1 (5'-ATTCAACCAATCATAAAGATAT-3') and LepR1 (5'-TAAACTTCTGGATGTCCAAAAA-3') (Hebert et al. 2004), or CI13 (5'-ATAATTTTTTTTATAGTTATACC-3') and CI14 (5'-ATTTCTTTTTTT-CCTCTTTC-3') (Hasegawa et al. 2002). For some individuals we used the two different primer pairs and compared the sequences obtained for the same individual. For each of four individuals, two highly divergent copies of COI were sequenced. To detect sequences that might come from accidental amplification of numts (copies of mitochondrial DNA transferred into the nuclear genome), we searched for the presence of premature stop codons in the amino-acid sequences. Three sequences (belonging to two individuals) had one premature stop codon. Based on the distribution of these sequences and the divergent copies from the same individual in the COI phylogeny, we identified a clade of putative numts. We amplified the Cytb marker using primers Cytb-FeF (5'-CAGTTTAATTTCTAATGAACAAAC-3') and Cytb-FeR (5'-GGATCTCTAAAAATATATGGG-3') (Liautard & Keller 2001) for a subset of Proformica individuals, and we used these sequences to design internal primers more specific to Proformica in order to increase amplification success: cytbPf (5'-CCTTTTAATAATRTYACTATTGC-3') and cytbPr .
As outgroup we used species for which we obtained new sequences (Appendix 1): Bajcaridris theryi (Santschi, 1936) (Hasegawa et al. 2002;Goropashnaya et al. 2004Goropashnaya et al. , 2007Goropashnaya et al. , 2012Ward & Downie 2005;Moreau et al. 2006): Cataglyphis iberica (Emery, 1906)  A partition scheme was defined with PartitionFinder (Lanfear et al. 2012) for each phylogenetic analysis, using the Bayesian Information Criteria for nucleotide substitution model selection. Prior data blocks were defined by marker and codon position. Three separate phylogenetic reconstructions were performed using both maximum likelihood and Bayesian inference algorithms: one for COI (to highlight the position of the clade of putative numts), one for Cytb (which includes the largest number of individuals), and one for the concatenated nuclear markers (28S + LW Rh) (because nuclear and mitochondrial markers might tell different stories).
Maximum likelihood phylogenies were constructed with RAxML (Stamatakis et al. 2008) on the web server at vital IT, Switzerland (http://embnet.vital-it.ch/raxml-bb/), using the GAMMA model of rate heterogeneity. Node support was estimated by generating 100 trees by bootstrapping. Bayesian inference phylogenies were constructed with MrBayes 3.2 (Ronquist et al. 2012). For the COI phylogeny we used the substitution models SYM + G, F81 and GTR + G for the first, second and third codon position, respectively. For Cytb we used HKY + G, HKY + I and GTR + I + G for the first, second and third codon position respectively. For the concatenated nuclear genes we used K80 for the first codon position of LW Rh, and K80 + I for 28S and the second and third codon positions of LW Rh. Each analysis consisted of two runs of four Markov chains run for 10 million generations. Parameters were unlinked for all partitions. A standard deviation of split frequencies of less than 0.01 between two independent runs was reached after less than 2.4 million generations. A burn-in fraction of the first 25 % of the trees was discarded.

Cuticular hydrocarbons
Colonies from nine localities (Bonnieux, Plaine de la Crau, Montpellier, Grand Luberon, Montagne de Lure, Sainte-Baume, Pompignan, Sisteron, Mont Ventoux) were used for analysis of cuticular hydrocarbons. Using forceps, we gathered three to five workers from each colony and put them into glass vials containing 1 ml of hexane. The containers were stored in a freezer at -20°C until chemical analysis. For chemical analysis, the ants were retrieved from the vials and the solvent evaporated. The extract was re-dissolved in 10 µl of hexane. Two µl of each extract were injected into a Perkin-Meyer GC-MS functioning at 70eV with a source temperature of 230°C and equipped with a ZB-5HT column (30 ml × 0.25 mm ID × 0.252 µm df; 5% phenyl-95% dimethylpolysiloxane). The temperature program was 2 min at 150°C, and then 5°C/min until 320°C, and a 5 min hold at 320°C (total 41 min). Substances were identified using standard alkanes, library data and Kovats retention indices. For the comparisons, we calculated the percentage of each hydrocarbon from the total hydrocarbon content in each ant sample. The data were analysed using Principal Component Analysis. We chose not to transform the data since transformation introduces additional background noise into the data when numerous zero values are present; these have to be replaced to make transformation possible when comparing species. Indeed, reanalysis of the data after transformation (following the procedure of Reyment 1989) gave similar results, but with slightly less efficient separation of groups (Oppelt et al. 2008). Analyses were made with the Statistica software.

7
We also performed chromatograms of cuticular hydrocarbons for two species used as outgroups: Proformica longiseta Collingwood, 1978 from Sierra Nevada (Spain) and Cataglyphis cursor from Aixen-Provence (France). Lists of cuticular hydrocarbons known for these species have been published in Errard et al. (2006) and Nowbahari et al. (1990), respectively, but without quantification.

Nest census and queen reproductive status
Six nests were excavated in July 2011 and the ants counted. Six queens from two nests were dissected to assess their reproductive status. In addition, one apterous queen was obtained by rearing pupae from Sainte-Baume and was dissected to confirm its queen status. Several workers of various sizes were also dissected.

Morphology and altitudinal distribution
Two groups of nests were separated by combining two morphological characters, GHL and PDG for the workers, and GHL and nG for the queens (Fig. 2). The two groups were distinct for both characters, independently of CW, a proxy for size (Fig. 2). One of the groups, coloured in orange in the figures and hereafter denominated as the orange taxon, encompasses the type locality of Proformica nasuta (Fig. 1). The other group is coloured in blue in the figures and is hereafter denominated as the blue taxon.
Further, queens of the blue taxon were all winged or showed wing sclerites, while all queens of the orange taxon were ergatoid. We did not find males of the blue taxon in the field and we could not locate specimens in museum collections. Males of the orange taxon have dense and long hairs on the head, mesosoma and the anterior face of the first gaster segment.
For both the orange taxon and the blue taxon, altitudinal distribution of the nests was bimodal (Fig. 3).
Most nests of the orange taxon were found below 200 m, but those from Sainte-Baume, Sumène, Grand Luberon and Montagne de Lure were found above 800 m, at the tops of medium-sized mountains. In contrast, most nests of the blue taxon were found above 1000 m on plateaus and mountains, but those from Orange, Sisteron and Vinsobre were found lower, below 600 m. Interestingly, within each of these taxa, GHL and PDG are highest for workers from the mountain localities (except for Sumène) (Fig. 2).

Molecular phylogenetic analysis
Maximum likelihood and Bayesian inference produced very similar phylogenies, so we chose to present only Bayesian inference phylogenies. The clade of putative numts in the COI phylogeny is delimited in red in Fig. 4A. The two mitochondrial markers yielded similar topologies ( Fig. 4A-B), showing two main clades which corresponded approximately to the two taxa defined in the morphological analysis.
Mismatch between morphotypes and clades was observed for some specimens. The nuclear markers showed very little variation. As a consequence, the resulting tree is poorly resolved (Fig. 4C).

Cuticular hydrocarbons
Identification Top and middle graphics represent the regression of gaster hair length (GHL), pubescence distance on the gaster (PDG) and unilateral number of hairs on the gaster (nG) against cephalic width (CW), a proxy for size. The graphics at the bottom represent a combination of two morphological characters that highlights two distinct groups of nests. These two groups are coloured in blue and orange respectively. The dots with a black cross correspond to nests from mountain and lowland localities for the orange and blue groups, respectively.
(Appendix 5). The PCA distinguishes the two outgroups, Proformica longiseta and Cataglyphis cursor, from the Proformica samples from southern France (Fig. 5). The blue and orange taxa are segregated along the first axis of the PCA. In addition, the strongest differentiation occurs within the orange taxon, between a group formed by the two mountain localities (Montagne de Lure and Grand Luberon) and the others.

Queen reproductive status, nest census
Excavation of nests of the two taxa revealed the same general structure: the entrance opens directly at the ground surface, sometimes under a small stone; a vertical gallery of 10-20 cm leads to a small chamber where males can be found when present; then, the gallery goes down obliquely and reaches a final chamber, about 50 cm below ground level, where queens are present. Secondary galleries, lateral (perpendicular) to the principal one, may be present and lead to chambers. The content of nests is presented in Table 1. Repletes, i.e. workers with inflated gaster serving as stores of liquid food, were found in colonies of both taxa. Colonies had one to many queens that appeared to be actively reproducing (mated, with numerous mature oocytes and yellow bodies) (Table 2). Workers, even the largest, always had fewer than 3 ovarioles per ovary and never had a spermatheca. In contrast, apterous and winged queens had a spermatheca and many more ovarioles per ovary (~ 10).

Fig. 4.
Bayesian consensus trees of COI (A), Cytb (B) and concatenated sequences of 28S and LW Rh (C) for Proformica workers from southern France and outgroups. Labels are composed of the locality name, the colony code ( figure), the code of the individual (w1 for worker 1, w2 for worker 2) and, for COI, the primer pair used (Ci for CI13 and CI14, Lep for LepF1 and LepR1). Sequences where a stop codon was detected are labelled with a red "STOP". Sequences with an asterisk specify individuals for which another sequence was obtained and fitted outside the clade of putative numts. Colours match the groups defined in the morphological analysis. Posterior probabilities are given for major nodes. Accession numbers are indicated for sequences retrieved from GenBank.  Ruzsky, 1902Nylander (1856 described P. nasuta, the type species of the genus, from Beaucaire. Our analyses assigned workers from the type locality and from two other localities within a radius of 10 km (Jonquières and Tarascon) to the orange taxon. They lack erect hairs on the mesosoma, agreeing with the description of Proformica nasuta by Nylander as "nuda". Although the type is presumably lost (as it could not be found in the following collections: Nylander (Helsinki) (Radchenko 2007), Forel (Geneva), Emery (Genoa), Bondroit (Brussels) and Santschi (Basel)), we are confident that the nest samples we collected in Beaucaire and in the surrounding area correspond to the species described by Nylander. Below we provide a redescription of P. nasuta (the orange taxon), and the description of a new species, P. longipilosa sp. nov. (the blue taxon).

Redescription of P. nasuta (Nylander, 1856) and designation of the neotype
As the type specimen of P. nasuta is presumably lost, we propose fixation of a neotype from a nest sample collected in Beaucaire, France, terra typica of the species, and matching Nylander's concept of P. nasuta. The original description (Nylander 1856: 66) is based on a small worker (" Long. 3 -3.5 mm ") with elongated head (" ... facies producta antice visa subrectangularis... "). This feature is found exclusively in minor workers. Therefore, a minor worker was selected from Beaucaire, France (colony Beaucaire 1) and designated as the neotype. The neotype is deposited as MNHN-1598 with the labels "FRA, N43.83544 E4.61828, Beaucaire, 9 juillet 2011, leg. R. Blatrix & C. Lebas" and "Néotype Proformica nasuta (Nylander, 1856), des. Galkowski, Lebas, Wegnez, Lenoir & Blatrix, 2016". In case of loss or destruction of this specimen, a replacement neotype can be designated from a series of ten other minor workers collected from the same nest and deposited at the MNHN. Other workers from the same nest are deposited at the following collections: AT (no. 15557), LB, SMNH, XE, ZISP and the collections of the authors. A queen from the same nest and a male from colony Tarascon 1 (a few kilometers away from the type locality) are deposited at MNHN.
Body uniformly dark brown to black, appendices and mandibles lighter. Erect hairs rare or absent, short when present (GHL/CW < 0.11). Dense pubescence on dorsal surface of first and second gaster tergites (PDG < 29). Profile of mesosoma sinuous. Petiolar scale erect, thick, slightly notched at summit in large workers. Head of minor workers clearly elongate, rectangular (CL/CW > 1.3). Head of media and major workers less elongate (CL/CW 1.1-1.3), a bit shiny toward occiput, faintly sculptured in anterior part. Clypeus finely striate longitudinally. Mandible with five teeth of increasing size from base to apex.

Remarks
We have not examined the type specimens of the Asian varieties of this species, P. nasuta metalica Kuznetsov-Ugamsky, 1923 and P. nasuta syrdariana Santschi, 1928, described from Kazakhstan, nor the type specimen of the taxon Formica aerea (Roger, 1859), which was described based on a single minor worker collected in Greece and later synonymized with P. nasuta by Emery (1925). Although we decided not to change the status of these eastern taxa until a thorough revision of the eastern Proformica is undertaken, we believe it is very unlikely that they will be conspecific with P. nasuta.

Diagnosis
Workers varying greatly in size, the smallest having a strongly elongated head. Body black. Pubescence sparse and sculpture of tegument weak, giving a shiny aspect. Hairs on mesosoma and gaster very long.

Etymology
The epithet of this species refers to the long erect hairs on the mesosoma and gaster of workers. Body black; only tibiae, scape and mandible brown. All parts of body with long erect hairs (GHL/CW > 0.12). Pubescence on dorsal surface of first gaster tergite sparse in all worker categories (PDG > 24), revealing smooth and shining cuticle. Profile of mesosoma sinuous. Petiolar scale erect, thick, slightly notched at summit in large workers. Head of minor workers clearly elongate, but less than in P. nasuta (CL/CW 1.16-1.28). Head of media and major workers even less elongate (CL/CW 1.046-1.19). Clypeus finely striate longitudinally, with faint trace of median carina. Frontal triangle and space between frontal carina also finely striate. Sculpture disappears toward occiput, cuticle becoming smooth and shining, or faintly punctuated in large workers. Scape long, surpassing occipital border. Color as in worker. Many long and erect hairs on all body parts (nG > 26, GHL > 200 µm). Some erect hairs also on femora and tibiae. Dense pubescence on entire body, masking surface of cuticle. Mesosoma less wide than head. Wing remains indicate winged queens, although wings possibly small given reduced development of scutum and scutellum. Petiolar scale high and wide, deeply notched at summit. Gaster rather small. Head almost as large as long, entirely and finely sculptured, faint riddles of anterior part replaced by puncture on posterior part, giving head an almost dull aspect. Clypeus finely striate longitudinally. Scape surpassing occipital border.

Male
Unknown.

Remarks
We made direct comparisons between specimens of P. longipilosa sp. nov. and P. longiseta (A. Tinaut leg.) from Sierra Nevada, Spain, and P. ferreri (IRSNB-BC, 2 workers, 1 ♂ (type specimen) from Spain). The latter two species, in addition to P. nasuta, were formally described from western Europe. Specimens of P. longipilosa sp. nov. are unambiguously distinguished from those of these two species by the combination of the following characters: erect hairs on the body are longer (GHL/CW > 0.12) and the pubescence on the dorsal surface of the first gaster tergite is sparser. This last character is especially discriminant in media and major workers (PDG 24-68, mean = 43.7). In addition, the cuticle is smoother, giving a shinier appearance, in particular on the head.

Discussion
Based on arguments from morphometric analysis, DNA sequences and cuticular hydrocarbons, we show that the populations of Proformica in southern France belong to two species, one of which is P. nasuta, and the other a new species that we name P. longipilosa sp. nov.

Consequences for the taxonomy of Proformica
There has been much confusion in the concept of the taxon P. nasuta, in part because the type appears to have been lost a long time ago. None of the taxonomic studies published after the description of the species by Nylander (1856) made reference to the type, which was collected in Beaucaire (France). Forel (1886) described the worker and the queen from specimens collected in Orange (France) and sent samples to many of his colleagues throughout Europe and Russia. We examined the specimens from Orange in the Forel, Emery and Bondroit collections, and we collected new samples from the same locality in 2011. All differ markedly from those of Beaucaire and belong to the new species we describe in this study, P. longipilosa sp. nov. All the taxonomic studies after 1886 used the samples from Orange, or descriptions of them, as a reference for P. nasuta. These studies described P. nasuta as having long erect hairs and sparse pubescence (e.g., Ruzsky 1905;Emery 1909;Wheeler 1913;Bondroit 1918;Santschi 1925;Bernard 1968;Dlussky 1969;Collingwood 1976;Agosti & Collingwood 1987), two characters that are typical of P. longipilosa sp. nov. Other specimens from southeastern France (Plateau de Caussols, Tourettes-sur-Loup) have also been used as references for P. nasuta (Collingwood 1956;Stumper 1957;Dlussky 1969), but they come from an area that we now recognise as belonging to the range of P. longipilosa sp. nov., and are thus likely distinct from P. nasuta. This mistake has been perpetuated so that the actual conception of P. nasuta refers to P. longipilosa sp. nov. A consequence of this is that all reports of P. nasuta since 1886, including all those from eastern Europe and Asia, are probably erroneous.
There is a striking similarity between the biogeographic patterns of the genus Proformica and those observed for other steppe elements (Ruano et al. 2011;Sanllorente et al. 2015) such as, for instance, the meadow and steppe vipers (Vipera ursinii species complex). Both Proformica and the V. ursinii complex are known to occupy the same type of habitats and show the same pattern of disjunct distribution, with one area across Asia and one area restricted to western Europe. Moreover, they are both poor dispersers (Sanllorente et al. 2015;Ferchaud et al. 2011). It is likely that the current distribution and speciation patterns of the two groups have been induced by the same climatic and geographic events. The eastern and western clades of the Vipera ursinii species complex diverged about 4 Mya (Zinenko et al. 2015), and the taxa from the two geographic areas are completely distinct. In addition, mutation rates and the number of generations per unit time are expected to be higher in insects than in snakes (Martin & Palumbi 1993). We thus expect that the Proformica species from western Europe are distinct from those from the East.

Intraspecific variation
Localities for P. nasuta can be divided into two subgroups. One is composed of lowland localities that can be close to each other and form an almost continuous distribution in the plain of the Languedoc and the Rhône valley (Montpellier, Sauteyrargues, Pompignan, Grospierres, Collias, Beaucaire, Jonquières, Tarascon, Plaine de la Crau, Aurons and Bonnieux). The other subgroup is composed of localities isolated on the summits of medium-sized mountains (Sumène, Sainte-Baume, Grand Luberon and Montagne de Lure). Beyond ecological differences (mountain vs lowland), the distinction between the two groups is supported by the analysis of morphological characters, DNA and cuticular hydrocarbons. Except for Sumène, individuals from mountain localities have some erect hairs on the mesosoma, whereas hairs are lacking in most lowland individuals. Mountain individuals also have longer hairs and sparser pubescence on the gaster compared to lowland individuals. Interestingly, mitochondrial sequences from Sainte-Baume, Grand Luberon and Montagne de Lure form three monophyletic clades that are highly differentiated from each other, whereas most of the lowland localities show little differentiation. This pattern is consistent with a particularly high degree of isolation of the mountain localities, as already shown for P. longiseta in Spain (Sanllorente et al. 2015). Cuticular hydrocarbons also differ between lowland and mountain localities. At this stage we are reluctant to consider the mountain localities as a separate species because lowland and mountain localities form a consistent group morphologically and do not form two monophyletic clades in the mitochondrial phylogenies. Instead, we consider that the population of each mountain locality is derived independently from the lowland population. Sanllorente et al. (2015) proposed that climate-driven range fluctuation of populations of P. longiseta during Pleistocene glaciations induced a strong isolation among populations that are now restricted to mountain tops in southern Spain, because this species is adapted to cold and arid environments. Extant populations would be derived from an ancient, large population, independently from each other. A similar process may explain the divergence we noted between lowland and mountain populations of P. nasuta, and among mountain populations, and may have induced local adaptation of mountain populations. The morphological features of the mountain individuals of P. nasuta make them more similar to P. longipilosa sp. nov., especially where localities of the two species are close to each other (Grand Luberon and Montagne de Lure). These features might be the result either of introgression of P. longipilosa sp. nov. into populations of P. nasuta, or of morphological convergence in response to a similar environment.
Proformica longipilosa sp. nov. also shows two subgroups. One is composed of populations from localities on plateaus and mountains (Plateau de Calern, Plateau de Caussols, Gréolières and Mont Ventoux) and the other is composed of lower elevation localities (Orange, Sisteron and Vinsobre). Individuals of the lowland localities tend to have shorter hairs and denser pubescence on the gaster than those from highelevation localities, and thus are morphologically closer to P. nasuta than the high-elevation individuals. As proposed for P. nasuta, these features could result from either introgression or convergence. Although lowland P. longipilosa sp. nov. and mountain P. nasuta tend to converge morphologically, they can still be easily distinguished, leaving no doubt regarding their assignment to species. All nests from the three low-elevation localities of P. longipilosa sp. nov. had mitochondrial sequences typical of P. nasuta. We hypothesise that this incongruence between morphological and molecular characters, specifically for low-elevation localities, results from introgression of mitochondrial DNA from P. nasuta to P. longipilosa sp. nov. Such introgression would be most likely to occur in low-elevation localities because P. nasuta is relatively widespread in the lowlands. Complete or gene-specific introgression of maternal DNA is a well-known phenomenon in insects (Ballard 2000;Chan & Levin 2005;Linnen & Farrell 2007). An isolated event of partial mitochondrial introgression is also suggested in Plateau de Calern, where all nests were morphologically classified as P. longipilosa sp. nov., but one had Cytb sequences typical of P. nasuta. All other nests in this locality fitted within P. longipilosa sp. nov. for both mitochondrial markers. It is worth noting that we could not find any locality with nests of both P. nasuta and P. longipilosa sp. nov., although the two can be found in similar habitats.

Interpretation of cuticular hydrocarbons
Cuticular hydrocarbons separate the two species without ambiguity. Populations from all localities of P. nasuta form a homogeneous clade with relatively little differentiation except for two localities isolated at the summits of mountains (Grand Luberon and Montagne de Lure). Proformica longipilosa sp. nov. (Mont Ventoux and Sisteron) appears to be well separated from all other ants studied here, including the outgroups (P. longiseta and Cataglyphis Förster, 1850), confirming its status as a separate species. This classification is globally consistent with spatial distribution of the localities and with the classification based on morphology and DNA sequences. Localities of P. nasuta from the lowlands form continuous populations without important geographical isolation, allowing regular exchange of migrants resulting in little differentiation of cuticular hydrocarbons. However, the mountain localities Grand Luberon and Montagne de Lure are isolated and, probably as a consequence, are divergent for cuticular hydrocarbons. On the mitochondrial tree they also diverge from other localities of P. nasuta. Cuticular hydrocarbons in Proformica appear to be linked to phylogenetic signature but seem to change rapidly with geographical isolation, even faster than mitochondrial DNA. Geographic variation in cuticular hydrocarbons depends on the taxon. For instance, profiles are very stable for Formica ants from Finland to Great Britain (Martin et al. 2008) and for Lasius niger Linnaeus, 1758 from Denmark to the south of France (Lenoir et al. 2009;Lenoir unpubl.). In contrast, rapid spatial changes in hydrocarbons are present in some taxa like Odontomachus Latreille, 1804 (Smith et al. 2013) and Cataglyphis (Dahbi et al. 1996). Interestingly, Rossomyrmex minuchae Tinaut, 1981, a slave-maker parasite of Proformica longiseta, also has different chemotypes in three populations in Sierra Nevada, Spain (Sanllorente et al. 2012). It is noteworthy that the genus Cataglyphis is phylogenetically, biologically and ecologically very close to Proformica. Both have very limited queen dispersal, are specialized on dry habitats and forage on dead invertebrates at the warmest time of the day. It would be worth investigating whether strong divergence in cuticular hydrocarbons within species could be related to one or more of these characteristics.

Conclusions
Although we relied on an integrative taxonomy approach, using several complementary sources of information, we confirm the general view that the taxonomy of the genus Proformica is a complex problem. The nuclear markers chosen for use here evolve too slowly and thus lack resolution. Information from mitochondrial genes is blurred by genetic processes such as transposition and introgression and may be biased by queen philopatry. Our results suggest that morphology is a better tool to resolve taxonomy in this genus than either cuticular hydrocarbons or DNA sequences of the genes commonly used for phylogenetic analyses and barcoding, although genetic markers other than those used in this study should also be investigated. However, for the genus Proformica, the zoogeographical region of southern France is the least complex in taxonomic terms. Thus, the morphological approach developed in this paper may prove unreliable in other regions, such as the Iberian Peninsula and eastern Europe. A population-genetic approach using tools such as microsatellites or single nucleotide polymorphisms from Next Generation Sequencing (e.g., RADseq) may help disentangle this taxonomic knot.