Altogether about 75 specimens from 10 localities of the Balkan Peninsula, from 10 localities of the Apennine Peninsula, and from 5 localities in Central Europe were selected for investigation and comparison of characters related to the head, thorax, wings, legs and genitalia of males and females.

Six Ornatoraphidia flavilabris specimens originating from three countries, namely Austria, Italy and Greece (Table 1) were investigated by molecular methods. Tissue was removed from legs of adult specimens preserved in ethanol (96%).

Coordinates or locations of previously published O. flavilabris records (Navas 1929, H. Aspöck and U. Aspöck 1971, Joost 1973, Kofler 1977, Hölzel et al. 1980, Devetak 1984, Gepp 1986, H. Aspöck et al. 1991, Sziráki et al. 1992, Popov 1993, 1997, 2000, Letardi 1994, 2003, 2018, 2021, Letardi and Pantaleoni 1996, Pantaleoni and Letardi 1998, Letardi and Migliaccio 2002, Letardi and Biscaccianti 2007, Badano 2008, Aistleitner and Gruppe 2009, Klokočovnik et al. 2010, Letardi et al. 2010, Badano and Letardi 2010, Tillier 2013, 2015, Sziráki 2014, Devetak et al. 2015, Devetak and Rausch 2016, Hiermann et al. 2018, Letardi and Scalercio 2018) as well as personal unpublished records (Suppl. material 1) were georeferenced into a distribution map using Quantum GIS 3.4.11 (QGIS Development Team 2019) (Fig. 3).

DNA was isolated with a QIAamp DNA Mini Kit (Qiagen, Hilden, Germany), strictly following the manufacturer’s instructions. DNA was stored in a final volume of 100 μl in elution buffer at -20 °C.

PCR amplification of fragments of three different gene fragments, namely cytochrome c oxidase subunit 1 (cox1), cytochrome c oxidase subunit 3 (cox3) and 28S rRNA gene (28S), was performed with an Eppendorf Mastercycler (Eppendorf AG, Hamburg, Germany) containing 10 × Reaction Buffer B, 2.5 mM MgCl₂, 1.6 mM dNTPs, 1 μM primers, 1.25 units DNA polymerase and 1–5 μl DNA. Sterile H₂O was added to a final volume of 50 μl. For negative controls microbial DNA free water (Qiagen, Hilden, Germany) was added instead of template DNA.

For cox1 a 658 bp fragment was amplified using the primers LCO1490 (5’-GGT CAA ATC ATA AAG ATA TTG G-3’) and HCO2198 (5’-TAA ACT TCA GGG TGA CCA AAA AAT CA-3’) published by Folmer et al. (1994) for cox3 also a 658 bp fragment was amplified using the primers Raph-cox3fw (5’-TAG TCC ATG ACC HTT AAC AGG-3’) and cox3-rev (5’-ACA TCA ACA AAA TGT CAA TAT CA-3’) published by Haring and U. Aspöck (2004). The final 28S sequence was obtained by the amplification of overlapping fragments obtained by the primer pairs Raph28S-1+ (5’-CAG GGG TAA ACC TGA GAA A-3’)/Raph28S-4- (5’-AGC GCC AGT TCT GCT TAC C-3’) and Raph28S3+ (5’-AGC TTT GGG TAC TTT CAG GA-3’)/Raph28S6- (5’-GGA ATA GGA ACC GGA TTC CC-3’) published by Haring et al. (2011). Amplification of all three genes was carried out using the following conditions: 95 °C for 15 min, followed by 35 cycles of 95 °C for 1 min (denaturation), 52 °C for 1:30 min (annealing) and 72 °C for 2 min (elongation), followed by a final extension of 72 °C for 10 min.

The PCR products were subjected to electrophoresis in 2% agarose gels stained with GelRed Nucleic Acid Gel Stain (Biotium, Inc., CA, USA). For further sequencing, bands were analyzed with a Gel DocTM XR+ Imager (Bio-Rad Laboratories, Inc., CA, USA), cut out from the gel and purified with the IllustraTM GFXTM PCR DNA and Gel Purification Kit (GE Healthcare, Buckinghamshire, UK).

Sanger sequencing was performed with a Thermo Fisher Scientific SeqStudio (Thermo Fisher Scientific, MA, USA). Sequences were obtained from both DNA strands and consensus sequences were generated in GenDoc 2.7.0. Obtained sequences were stored in GenBank (MZ313518.1–MZ313535.1) and compared to available sequences using the Basic Local Alignment Search Tool (BLAST) (https://blast.ncbi.nlm.nih.gov/Blast.cgi) in GenBank.

Obtained sequences were aligned with ClustalX 2.1 (Larkin et al. 2007) and manually edited with GeneDoc 2.7.0 (Nicholas 1997) for further analysis. The three included genes (cox1, cox3 and 28S) were analyzed independently. Alignments of cox1 and cox3 both showed a length of 658 bp without any insertions or deletions. Translation to amino acid sequences for both genes (cox1 and cox3) showed intact reading frames for all included sequences. 28S sequences, which are variable in length, were prealigned with ClustalX. Sequence gaps and ambiguously aligned sections were checked and edited manually with GeneDoc. This manual alignment approach was compared to a computer approach by Haring et al. (2011) and was considered appropriate for this study. After removal of these sections the final alignment resulted in a length of 1602 bp. Final sequence alignments of all three genes are given in supplementary files (Suppl. material 2–4).

Based on the best fit evolutionary model selection, pairwise Tamura-3-parameter + G (gamma distributed) distances were calculated individually for all three genes in MEGAX (Kumar et al. 2018). Therefore, available sequences of Raphidioptera species were downloaded from GenBank. Available sequences of species belonging to the “Phaeostigma clade” proposed by Haring et al. (2011), to which also the genus Ornatoraphidia belongs, were chosen as well as sequences of species belonging to the “Puncha clade” which were used as outgroups for tree calculation (Table 2).

To infer the phylogenetic position of the analyzed O. flavilabris samples, maximum likelihood (ML) analysis using the Tamura 3-parameter + G + I model with four discrete gamma values was applied. Nodal support was evaluated by 1000 bootstrap replicates.

Specimens of O. flavilabris from Greece had already been carefully studied and compared with specimens from Italy and Central Europe by us in the late 1960s, and again in the following decades. These examinations consistently led to the conclusion that constant differences do not exist, and thus all populations were assigned to one species (H. Aspöck et al. 1991, 2001). In the course of the present study we again thoroughly examined many specimens from many localities and could only confirm what we had already concluded half a century ago.

Thus, on a morphological basis it is impossible to answer the question whether the Central European populations (at least those of Lower Austria) can be traced back to migrations from the Apennine Peninsula or from the Balkan Peninsula.

Sequences of all six included O. flavilabris specimens were successfully amplified by PCR and sequenced. The sequence length was 658 bp for all cox1 (GenBank accession: MZ313518.1–MZ313523.1) and cox3 (GenBank accession: MZ313524.1– MZ313529.1) sequences. The 28S sequences of specimens originating from Greece had lengths of 1741 bp (Ofla1, GenBank: MZ313530.1) and 1749 bp (Ofla2, GenBank: MZ313531.1), those from Austria had lengths of 1763 bp (Ofla5, GenBank: MZ313534.1) and 1765 bp (Ofla6, GenBank: MZ313535.1) and those from Italy had a length of 1763 bp (Ofla3, GenBank: MZ313532.1; Ofla4, GenBank: MZ313533.1).

After alignment, cox1 sequences showed 96 (96/658; 14.6%) variable positions, of which 81 were parsimony informative. Amino acid sequences of all six O. flavilabris specimens showed no differences. The Cox3 sequences showed 91 (91/658; 13.8%) variable positions, of which 71 were parsimony informative, which resulted in nine differences at amino acid sequence level. The 28S sequences showed 25 (25/1602; 1.6%) variable positions, of which 18 were parsimony informative.

In total, nine cox1 sequences of Raphidioptera with a length of 658 bp were analyzed (Table 2). The overall mean pairwise Tamura-3-parameter distance was 16.0%. Pairwise distances between the six O. flavilabris sequences analyzed ranged from 0.0% to 15.2% with a mean distance of 8.4%. While genetic distances between the O. flavilabris specimens (Ofla5, Ofla6) from Austria and the specimens from Italy were 2.8% (Ofla3) and 3.0% (Ofla4) respectively, genetic distances to the specimens from Greece were considerably higher, being 12.9% (Ofla1) and 13.8% (Ofla2), respectively (Table 3). Pairwise distances of O. flavilabris to other included Raphidioptera species ranged from 15.6% to 19.4% (Suppl. material 5).

Altogether, 15 cox3 sequences of Raphidioptera with a length of 658 bp were included in the analysis (Table 2). The overall Tamura-3-parameter distance was 12.8%. Pairwise distances between nine analyzed O. flavilabris sequences ranged from 0.0% to 13.3% with a mean distance of 6.0%. Pairwise distances between the O. flavilabris specimens from Austria and Italy ranged from 2.5% to 3.2%. Again, much higher pairwise distances were observed between the specimens from Austria with either of those from Greece (11.5% Ofla1 and 12.8% Ofla2) (Table 3). Pairwise distances of O. flavilabris to other included Raphidioptera species ranged from 13.0% to 21.4% (Suppl. material 6).

Overall, 13 28S sequences of Raphidioptera with a length of 1602 bp were included in the analysis (Table 3). The overall mean pairwise intraspecific Tamura-3-parameter distance was 2.9%. Pairwise distances between seven analyzed O. flavilabris sequences ranged from 0.0% to 1.5% with a mean distance of 0.7%. While pairwise distances between the specimens from Austria and Italy ranged from 0.3% to 0.4%, distances between the specimens from Austria and Greece were 1.1% (Ofla1) and 1.5% (Ofla2) (Table 3). Pairwise distances of O. flavilabris to other included Raphidioptera species were comparably higher and ranged from 2.6% to 4.7% (Suppl. material 7).

All sequences included in pairwise sequence calculations were used for the maximum likelihood analysis. Xanthostigma xanthostigma and Puncha ratzeburgi belonging to the Puncha clade were used as outgroups with one exception. For cox1, only X. xanthostigma was used as an outgroup, since no other sequence of the Puncha clade was available from GenBank.

The cox1 tree resulted in two major clades. Clade 1 included all O. flavilabris sequences, clade 2 comprised Phaeostigma notata and Dichrostigma flavipes (Fig. 4A). Clade 1 was further divided into an O. flavilabris lineage from Austria and a lineage from Italy forming a monophyletic group that was well supported by high bootstrap values. Together they represent the sister group of a O. flavilabris lineage from Greece (Fig. 4A).

Figure 4. 

Maximum likelihood trees based on cox1 (A.), cox3 (B.), and 28S (C.) sequences of Raphidioptera species. Puncha ratzeburgi (except cox1 tree) and Xanthostigma xanthostigma were used as outgroups. Vertical colored bars represent Ornatoraphidia flavilabris sequences originating from Austria (green), Italy (red), and Greece (blue). Em.-Rom.=Emilia-Romagna, Italy. Bootstrap values >50% are shown.

The cox3 tree showed three O. flavilabris lineages (Austria, Italy and Greece) with high bootstrap support. Again, the lineages from Austria and Italy formed a sister group, and together they were the sister group of the lineage from Greece. In addition, the lineage from Italy was subdivided into two sublineages originating from the Calabria and one from the Emilia-Romagna regions of Italy (Fig. 4B). Parvoraphidia microstigma arose as the sister species of O. flavilabris and P. microstigma + O. flavilabris formed the sister group of Turcoraphidia amara, all together representing a monophyletic group. However, the phylogenetic positions of P. microstigma and T. amara were only supported by comparably low bootstrap values (Fig. 4B). Also in a recent paper on the phylogeny of Neuropterida based on the transcriptomic dataset (Vasilikopoulos et al. 2020) Ornatoraphidia emerged as sister of Parvoraphidia and both being sister of Turcoraphidia.

The 28S tree comprised two major clades (Fig. 4C). Clade 1 included all O. flavilabris sequences and was subdivided into three O. flavilabris lineages (from Austria, Italy and Greece). Clade 2 comprised all other included members of the Phaeostigma group. As already observed in the cox1 and cox3 tree, the O. flavilabris lineage from Austria formed the sister group of the lineage from Italy, and together they formed the sister group of the lineage from Greece (Fig. 4C). In the 28S tree, T. amara and P. microstigma belonged to clade 2, unlike tree cox3, however, again the bootstrap support was low (Fig. 4C).