Ninety-nine specimens were included in the molecular analysis, representing eleven OTUs. In addition, two specimens each of Vespula germanica (Fabricius, 1793) and Vespula vulgaris (Linnaeus, 1758) were used to root the trees; sequences for Polistes (Polistella) snelleni Saussure, 1862 and Polistes (Aphanilopterus) exclamans Viereck, 1906 were downloaded from Genbank and used with the two species of Vespula to root the trees in analyses of the mitochondrial sequences. Most specimens were collected in 80% ethanol in the field, but we also included some specimens that were killed with ethyl acetate. For specimens collected before 2012, DNA was extracted from the mesosoma, leaving the legs, wings, head and metasoma as vouchers; for specimens collected in 2012 and 2013, as well as specimens selected as type specimens, DNA was extracted from one single leg to preserve a nearly intact specimen. Most specimens were collected in Switzerland, but we also included specimens form Croatia, France, Greece, Italy and Portugal (Table 1). All DNA extractions are deposited in the DNA bank of the Swiss Barcode of Life initiative (Swissbol; www.swissbol.ch).

Full lab protocols can be found in Praz et al. (2008). DNA was isolated using phenol-chloroform extractions; PCR reactions were performed with GoTaq polymerase (Promega) in a Biometra T1 thermocycler. PCR products were purified enzymatically using a mix of the enzymes exonuclease I (Fermentas) and FastAP thermosensitive alkaline phosphatase (Fermentas) and sequenced in both directions with the primers used in the original amplification using BigDye terminator technology (Applied Biosystems). Big Dye products were purified with Sephadex (GE Healthcare Life Sciences) and analyzed on a ABI-3500 DNA sequencer.

We sequenced two fast-evolving genetic markers: the 600 bp fragment of the mitochondrial gene cytochrome oxidase 1 (COX1) used as an universal barcode (Hebert et al. 2003) and the nuclear marker ITS1; we chose ITS1 rather than ITS2 because preliminary analyses revealed that ITS2 was polymorphic in Polistes bischoffi and could not be sequenced directly.

For COX1 we used the universal primers LepF and LepR (Hebert et al. 2004) with the following conditions: an initial denaturation of 1 min at 94 °C, then six cycles of 1 min at 94 °C, 1.5 min at 45 °C, and 1.25 min at 72 °C, followed by 36 cycles of 1 min at 94 °C, 1.5 min at 51 °C, and 1.25 min at 72 °C, with a final step of 5 min at 72 °C. For specimens with degraded DNA, we used another universal forward primer, UAE3 (Zhang and Hewitt 1996) in combination with LepR to amplify a 400 bp fragment of the barcode. The conditions for this 400 bp fragment were as above, except that the extension time at 72 °C was 45 seconds in each cycle.

The presence of nuclear pseudogenes, or NUMTs, was carefully examined by visually detecting “ghost bands” on the agarose gel, and especially by detecting double peaks in the chromatograms. No indication of the presence of NUMTs was found in the specimens analyzed, with the exception of Polistes nimpha. For this OTU, double peaks were found in up to 20 nucleotide positions in every specimen, strongly suggesting the presence of NUMTs; no indels were found, and no stop codons were found in the translated amino acid sequence for these sequences, even when polymorphism was allowed, suggesting that the NUMTs were highly similar to the true mitochondrial sequences and thus of recent origin. The presence of NUMTs in Polistes nimpha was therefore unlikely to affect our results, especially given that Polistes nimpha was not the focus of our study, as it is not closely related to any of the main OTUs.

For ITS1, we used the primers CAS18sF1 and CAS5p8sB1d (Ji et al. 2003) to amplify a 700 bp fragment. For most specimens, the chromatograms were clean, without double peaks, indicating no within-specimen polymorphism in ITS1. In Polistes dominula, a few sites were polymorphic, and one insertion rendered the sequencing difficult in some specimens at position 550; in Polistes nimpha, several sites were polymorphic and insertions or deletions prevented direct sequencing in all specimens, except two (the numbers 57 and 65). Given that Polistes nimpha was not the focus of our study, we did not clone the PCR products to obtain clean sequences of the individual copies of ITS1, and merely included two specimens in our analysis.

Genetic distances between each terminal were computed under the GTR model of nucleotide substitution in Paup 4.0b10 (Swofford 2002). We then performed maximum likelihood analyses of each marker separately using RAXML (Stamatakis et al. 2005), performing 1000 bootstrap replicates. For the mitochondrial marker, the first and second position were combined in one partition, while the third codon position constituted a second partition. For ITS1, we coded each insertion or deletion as an additional, binary character added as a separate partition, hereafter referred to as the “gap” partition; one insertion or deletion was considered as one character, regardless of the size of the indel. In total, the coding of the insertions and deletions resulted in 42 characters, 38 of which were parsimony informative and four of which were autapomorphic. We do not intend to unravel the phylogenetic relationships among the European species of Polistes, and therefore we do not present an analysis of a matrix combining both genes.

We applied a GTR + G model to each DNA partition; the gap partition was analyzed as a binary character with two states, with a gamma shape to accommodate rate heterogeneity. FigTree v1.3.1 (Rambaut 2009) was used to visualize the trees and produce the figures.

We measured a total of 266 specimens, most of them from Switzerland (158), but also some from Italy (30), Greece (24), Croatia (17), France (10), Germany (6), Slovakia (3), Turkey (3), Czech Republic (2), Liechtenstein (2), Austria (1), Azerbaijan (2), Uzbekistan or Tadjikistan (3), Mongolia (2), China (1), and Portugal (1). Sixteen characters were selected (Table 2) for measurements, most of them on the head and antenna, and two on the hind leg. Measurements were made on mounted specimens or parts (head, leg) of them using a pinholding device, permitting rotations around all three axes (X, Y, and Z). An Olympus SZH10 stereo-microscope equipped with eye-pieces Olympus GWH10X-D (with an eye-piece micrometer dividing 10 mm in 100 units) and Leica 10445111 (10x/21B) (with an eye-piece micrometer dividing 5 mm in 100 units) was used at several magnifications (Table 2). For terminology of morphological structures we followed Goulet and Huber (1993), occasionally also Richards (1973).

We applied the multivariate ratio analysis (MRA) of Baur and Leuenberger (2011) to our data. MRA comprises a set of tools for analyzing size and shape of body measurements in a multivariate mathematical framework that is entirely consistent with the customary usage of body lengths and ratios in taxonomic works (e.g., in descriptions, diagnoses). In systematic and taxonomic studies, MRA offers several advantages over conventional explorative multivariate methods, such as principal component analysis (PCA) and linear discriminant analysis (LDA). MRA removes biases from spurious contradictions in the results due to different definitions of size and shape. Furthermore, the numeric output of MRA can be used directly in the descriptive part of a taxonomic study. László et al. (2013) reviewed these issues in an application to parasitic wasps. Following Baur and Leuenberger (2011), we first calculated isometric size (isosize), defined as the geometric mean of all variables. We then performed a shape PCA (i.e., a principal component analysis in the space of all ratios) for evaluating how the morphometric pattern corresponds to the OTUs revealed in the molecular analyses. In order to decide how many components to retain we inspected the scree graph (Rencher 2002: 398–399). We also plotted isosize against shape PCs, because the correlation of size with shape is a measure of the amount of allometry in the data. Two graphical tools, the PCA ratio spectrum and allometry ratio spectrum respectively, were also employed in some cases. Finally, we used the LDA ratio extractor to extract the best ratios, and calculated the standard distance as well as the measure δ.

The R language and environment for statistical computing was used for data analysis (R Development Core Team 2013; version 3.0.1). For the above methods we employed slightly modified versions of the R-scripts provided by Baur and Leuenberger (2011, under “Supplementary material”). Scatterplots were generated with the package “ggplot2” (Wickham 2009).

Of the 99 ingroup specimens included, complete COX1 sequences were obtained for 96 specimens, and ITS sequences for 80 specimens (Table 1). This difference is due to 12 ITS sequences of Polistes nimpha that were polymorphic and excluded, as well as some specimens with degraded DNA, which could be sequenced for the shorter mitochondrial fragment but not for ITS1.

Analyses of the COX1-sequences (Fig. 1) reveal that Polistes helveticus and Polistes bischoffi represent two distinct, well-supported clades (Bootstrap support, hereafter BS, of 100 and 94%, respectively). Sequences of all included specimens of Polistes bischoffi, including the 10 specimens from Switzerland and one specimen from Corsica, were absolutely identical (genetic distance of 0); similarly, sequences of the 20 specimens of Polistes helveticus were identical. The genetic distance between these two clades was 2.6%. The relationship between these two clades, as well as the relationships among the species of the gallicus-group, were not resolved.

Figure 1.

Phylogenetic tree based on maximum likelihood analysis of sequences of the mitochondrial gene cytochrome oxidase 1 (COX1); numbers shown at nodes are bootstrap values based on 1000 bootstrap replicates performed in RAxML v.7.0.4. Branches interrupted by an oblique line have been shortened for better graphic representation.

More generally, most OTUs included in this study were recovered as monophyletic with high bootstrap support >90%, with the exception of Polistes dominula. Sequences for this OTU formed two well-supported clades (see below). The two specimens identified as Polistes gallicus by Arens (2011) did not appear closely related to other specimens of Polistes gallicus. For this reason, this taxon is simply referred to as Polistes sp. aff. gallicus.

Maximal within-OTU genetic distances were 0.2% for Polistes associus, 0.3% for Polistes sulcifer, 1.0% for Polistes hellenicus and for Polistes gallicus, 1.06% for Polistes sp. aff. gallicus, and 1.4% for Polistes biglumis. Within OTU-distances were higher for Polistes nimpha (2.4%) and especially for Polistes dominula (up to 4.9%; see below). For Polistes nimpha, although two weakly supported clades are revealed within this OTU (Fig. 1), the ranges of distance within (0–0.6% and 0–0.7%) and between these clades (0.4–2.4%) overlapped. In contrast, sequences for Polistes dominula formed two distinct clades that did not overlap. All sequences within the first clade were identical, thus the distance within this clade was equal to 0. In the second clade, the distances ranged from 0 to 0.67%; the distances between these two clades were between 3.6 and 4.9%. These two clades were weakly associated with geographic location: specimens originating from western Switzerland (Geneva, Valais and one location in Vaud) and from one site close to Zurich formed one clade, whereas specimens originating from the Grisons, from one location in Vaud and from the southern parts of the canton of Zurich formed the other clade; specimens from one locality in Zurich were distributed in both clades.

The minimal distance between two OTUs was 2.6%, observed between Polistes helveticus and Polistes bischoffi, as indicated above, as well as between the two included social parasites, Polistes semenowi and Polistes sulcifer.

Analyses of ITS1 (Fig. 2) again strongly suggest that Polistes helveticus and Polistes bischoffi represent two distinct, well supported clades (both with BS of 95%). Sequences for all of the eleven specimens of Polistes bischoffi, including one specimen from Corsica, were identical; within Polistes helveticus, the genetic distance was 0.17% due to one single polymorphic site. The genetic distances between both clades were between 2.23% and 2.37%. The relationship between these two species, as well as the relationships among the different species of the gallicus-group, were not resolved.

Figure 2.

Phylogenetic tree based on maximum likelihood analysis of sequences of the nuclear marker ITS1; numbers shown at nodes are bootstrap values based on 1000 bootstrap replicates performed in RAxML v.7.0.4. Branches interrupted by an oblique line have been shortened for better graphic representation.

All other OTUs were recovered as well supported clades, with bootstrap supports > 85% (Fig. 2). No sequence of ITS1 could be obtained for the two specimens of Polistes sp. aff. gallicus from Greece. The two clades observed in analyses of the mitochondrial marker in Polistes dominula were not recovered in analyses of ITS1, although maximal within-OTU distances were comparatively high for this OTU (0.77%). However, no distance correlation between ITS1 and COX1 was observed; for example, some specimens exhibiting high mitochondrial distances (eg, numbers 5 and 43) had identical ITS1 sequences. Other within-OTU genetic distances were as follows: 0% for Polistes sulcifer, Polistes associus and Polistes gallicus; 0.24% for Polistes biglumis; 0.32% for Polistes hellenicus.

The smallest interspecific distance in ITS1 sequences was 0.8%, between Polistes biglumis and Polistes hellenicus; the maximal distance in our ingroup was 11.5%, observed between Polistes nimpha and Polistes biglumis. The minimum distance between Polistes bischoffi and any other OTU was 1.72%, between Polistes bischoffi and Polistes biglumis.