The genus Vespa is one of the four genera of the subfamily Vespinae; it is composed of 22 extant species of hornets (Archer 1991, Nguyen et al. 2006). Most species have a distribution restricted to Asia, with the highest diversity found in northern Indo-Malaya (Matsuura and Yamane 1990, Carpenter and Kojima 1997). Two species are also naturally distributed outside of Asia: Vespa crabro is found in Europe and around the Black Sea and the Caspian Sea and Vespa orientalis in the north of Africa, Mediterranean regions and across the Middle-East. In the mid-19th century Vespa crabro was introduced into North America where it is now established (de Saussure 1898), while more recently, Vespa velutina was accidentally introduced into Europe, where it became invasive (Villemant et al. 2011). Several hornet species have been the subject of various biological studies, either because of their social habits (Matsuura 1991, Foster et al. 2000, Ishay et al. 2008), their threat to human health (Vetter et al. 1999), their impact on apiculture (Abrol 1994, Ranhabat et al. 2009) or even their interest as edible insects (Ying et al. 2010). However, the phylogeny of this genus is not yet well resolved.

The first cladistic study of Vespa was that of Archer (1994a), based on 11 morphological characters (10 binary and one multistate). It distinguished a main clade comprising most Vespa species, and two species unresolved at the base of the entire tree (Fig. 1). One of them, Vespa binghami, is the only Vespa species presenting morphological adaptation to nocturnal habits (van der Vecht 1959). The other unplaced species of Archer’s cladogram, Vespa basalis, is a small hornet readily distinguishable by its reduced punctation, especially on the clypeus. The main clade found by Archer (1994a) presented an unresolved basal node with four lineages, one comprising a single species (Vespa orientalis) and a second he termed the crabro group (Vespa crabro and Vespa dybowskii). The third lineage, or tropica group, included five species in two clades: one with Vespa mandarinia + Vespa soror, and the other with three species (Vespa ducalis, Vespa philippinensis and Vespa tropica). Archer’s last lineage, called the affinis group (Fig. 1), was composed of the 12 remaining species including an unresolvedclade of four species which have very similar morphology (Vespa bicolor, Vespa simillima, Vespa velutina and Vespa vivax). This unresolved clade was termed the bicolor group by Archer in another paper (1994b) and was sister-group of two unresolved species from so-called Sundaland: Vespa bellicosa and Vespa multimaculata.

Archer’s study was based mostly on male characters, which are known to be reliable phylogenetic characters (e. g. Song and Bucheli 2010), but his cladogram has many unresolved relationships. Furthermore, some of the lineages are only supported by single characters from female morphology, such as the tropica group, which is characterized by the female clypeus shape. Corroborating such groups thus requires further analyses.

This study presents results from an ongoing project on the evolution of vespine wasps, focusing on the genus Vespa. Our aims are to confirm the monophyly of the genus Vespa, a question not addressed by Archer (1994a), and to clarify relationships among the species based on a more extensive morphological matrix and combining molecular data.

Phylogenetic relationships were inferred for the 22 species of the genus Vespa and five other species of Vespidae as outgroup: Dolichovespula media (Retzius, 1783), Provespa anomala (de Saussure, 1854), Provespa barthelemyi (du Buysson, 1905) and Vespula germanica (Fabricius, 1793) from the subfamily Vespinae and Polistes dominula (Christ, 1791) from the Polistinae.

The morphological matrix was scored from specimens in the following natural history collections: American Museum of Natural History, Muséum National d’Histoire Naturelle, Nationaal Natuurhistorisch Museum and United States National Museum of Natural History. Specimens for molecular study were collected by J.K. and J.M.C. and preserved in 95 - 99% ethanol.

DNA was extracted from one leg and one antenna per specimen using QIAGEN “DNeasy tissue Kit”. Genes were amplified using PCR with PuReTaq Ready-To-Go beads in a total volume of 25µL including primers (Appendix 1) and DNA. Amplification cycles were specific to genes (Appendix 2). AMPures and CleanSEQ procedures were used for DNA purification and sequencing was performed on ABI PRISM 3730xl machines by Agencourt Biosciences (Beverly, USA). One missing gene fragment of Vespa basalis was obtained from Genbank database (accession number: AB585949).

The analyses are based on 45 morphological characters and multiple nuclear and mitochondrial loci, comprising: 374 sites of 12S, 528 sites of 16S, 2231 sites of 28S (sequenced in 4 fragments), 1442 sites of CO1 (sequenced in 2 fragments), 880 sites of elongation factor 1α (EF1α), and 328 sites of H3. Each of these genes was aligned separately using the MAFFT software with the L-INS-i algorithm (Katoh et al. 2005). Morphological characters were coded with Winclada V. 1.00.08 (Nixon 2002).

Phylogenetic analyses were performed in a parsimony framework using TNT (Goloboff et al. 2008). Analyses were first conducted on the morphological matrix and on the different molecular datasets separately, then on a matrix of all the data combined. Sequence alignments were merged with Winclada, and the final composite molecular matrix contained 5783 aligned nucleotides. Each analysis was performed with the new technology search algorithms including sectorial searches, the parsimony ratchet, tree drifting and tree fusing, default parameters except: 200 ratchet iterations, upweighting percentage 8, downweighting 4; 50 cycles of drift; minimum length hit 25 times. Molecular gaps were treated as missing data. Node supports were computed as GC-values of a symmetric resampling of 1000 replicates (group supported/contradicted values; Goloboff et al. 2003). GC values range from -1 to 1 and are the differences in group frequency between the group found in most parsimonious trees and the most frequent contradictory group. Only supported groups (support value above zero, scaled by TNT 0-100) are shown.

Our analyses of both morphological and molecular characters confirm the monophyly of the genus Vespa. This genus was first diagnosed from other Vespidae on the basis of the shape of the head (Thomson 1869) and especially the vertex length, which is congruent with the other synapomorphies of the genus. In our morphological and total evidence analyses, the genus Vespa is the sister-group to the other vespine genera with similar relationships to those described by Carpenter (1987). The results of Carpenter (1987) based on morphology and ours based on combined analysis contradict Pickett and Carpenter (2010).

While our molecular sample is incomplete, it nonetheless confirms the monophyly of two of Archer’s species groups within Vespa based on the morphology (crabro and tropica groups). Molecular data help to place Vespa orientalis, which both Archer’s and our morphological analyses failed to resolve well. Vespa orientalis is the only Vespa species distributed in arid areas in central Asia and the Middle-East. A close relationship of this species to Vespa affinis + Vespa mocsaryana despite obvious morphological differences begs the question of morphological adaptations to arid climates in Vespa orientalis that may have blurred the morphological phylogenetic signal. However, the close relationship of Vespa orientalis and Vespa affinis is suggested only by the 12S gene in the molecular data. Further gene sequences for this last species and for Vespa mocsaryana are necessary to clarify whether the clade consisting of Vespa orientalis and Vespa affinis + Vespa mocsaryana is definitively supported.

Our results are also consistent with previous authors regarding the close relationships of Vespa luctuosa and Vespa fervida, which are very similar in their morphology (van der Vecht 1957, Archer 1994a, 1999). On the other hand, Vespa bellicosa and Vespa multimaculata, for which close relationships to Vespa luctuosa were suggested based on their distribution and morphological similarities (Bequaert 1934), are not closely related to Vespa luctuosa. They appear to form aclade together with Archer’s bicolor group. Relationships within this last clade are still poorly resolved with low node supports. Molecular data showed a closer relationship between Vespa vivax and Vespa velutina (Fig. 3), while the morphological characters placed Vespa simillima and Vespa velutina as sister species. Such a discrepancy may have resulted from the fact that no molecular data were available for Vespa multimaculata and Vespa bellicosa.

Archer’s finding of a main clade of Vespa excluding Vespa basalis and Vespa binghami has been confirmed both by morphological and molecular data. Our results also suggest that the nocturnal species Vespa binghami is closer to the main clade of Vespa than is Vespa basalis. Morphological adaptations to nocturnal habits in Vespa binghami such as enlarged ocelli are thus autapomorphies. Recognition of the subgenus Nyctovespa, with Vespa binghami as sole included species (van der Vecht 1959), would thus render the subgenus Vespa paraphyletic, and the synonymy of Nyctovespa (Carpenter 1987) is justified on that basis.

Our extended morphological matrix and the molecular sequences partly support Archer’s results, and this analysis confirms that male characters such as the shape of the last metasomal sterna and the genitalia are reliable phylogenetic characters in Vespidae.