We used Arixenia esau Jordan, 1909 as representative of the Arixeniidae and Hemimerus hanseni Sharp, 1895 as representative of the Hemimeridae. Arixenia esau was collected in a cave in the Mulu area, Sarawak, Malaysia (3°57′N, 114°49′E) on 19 January 2009. The specimens were collected individually on rocks under the colony of the naked bulldog bat Cheiromeles torquatus Horsfield, 1824 and were preserved in pure ethanol. Hemimerus hanseni was collected in Big Babanki, Northwest Cameroon (6°7′N, 10°15′E) on 10 February 2007. The specimens were collected individually from the hairs of the Gambian pouched rat Cricetomys gambianus Waterhouse, 1840 that had been killed by automobiles and randomly found on the road. The material was preserved in 70% ethanol. A. esau and H. hanseni are not protected in the countries where they were collected, are not protected by any international law, and were not collected in protected areas. The taxa were unequivocally identified using the criteria in Nakata & Maa [15] and Rehn & Rehn [35]. The studied specimens are kept in the collection of Department of Biology and Ecology, University of Ostrava.

Total genomic DNA was isolated from the femur of the leg with a Qiagen DNeasy Tissue kit (Qiagen) and following the manufacturer’s protocol. Three nuclear markers (18S and 28S ribosomal DNA and histone-3) were analyzed using primers published by Whiting [43] and Jarvis et al. [11] and listed in Table 1. PCR reactions were performed in 20-µl volumes containing 1×Taq buffer, 2.5 mM MgCl₂, 200 µM dNTPs, 0.6 µM primers, 1 U Taq polymerase (Promega), and 50 ng of template DNA. The thermal protocol included predenaturation (95°C, 3 min); 10 cycles of denaturation (94°C, 60 s), annealing (54°C with a decrease of 0.5°C in each cycle, 90 s), and extension (72°C, 75 s); followed by 20 cycles identical to the previous 10 but with an annealing temperature of 48°C; and a final extension (72°C, 4 min). PCR products were examined on a 1% agarose gel, and amplified DNA was isolated from the gel using the QIAquick Gel Extraction Kit (Qiagen) or from a PCR mixture using the QIAquick PCR Purification kit (Qiagen). Purified PCR products were sequenced using the BigDye Terminator v3.1 Cycle Sequencing Kit (Invitrogen) and were analyzed on a 3130 Genetic Analyzer (Applied Biosystems). Results were viewed and edited using the program Chromas Lite v. 2.01 (http://www.technelysium.com.au/chromas_lite.html), and sequences were submitted to GenBank (Table 2).

We used sequences from 35 other earwig species to determine the phylogenetic position of Arixenia esau and Hemimerus hanseni (Table 2). Among these comparative sequences, those from 34 species were obtained from Jarvis et al. [11], and those from one species were obtained from von Reumont et al. [44]. Following Jarvis et al. [11], we chose species from the orders Orthoptera, Phasmida, Embidiina, and Grylloblattodea as outgroups.

Because of large expansion regions in 28S and 18S ribosomal DNA and/or mutations in target sequences of universal primers for histone-3, the PCR amplification of some of these regions in the Dermaptera is problematic [cf. 11]. Sequences were aligned using Multalin [45] under default conditions and were edited in MEGA version 5 [46]. The use of different alignment software, e.g., Mafft [47], generated nearly identical results (data not shown). Regions with low consensus values (resulting from a lack of amplification, a lack of available complementary sequences, and/or a high degree of sequence polymorphisms including indels typical for non-protein coding regions) were excluded to avoid incorrect identification of homologies and/or overweighting of indels [48], [49]. Among the 18S DNA sequences, three regions with low consensus values were excluded: 48 bp at the 5′ end and 33 bp at the 3′ end were excluded because of a large number of missing sequences; a 156-bp region that included parts of the V4 hypervariable region described in De Rijk et al. [50] was excluded because of a combination of missing sequences and alignment failure. Among the 28S DNA sequences and following Jarvis et al. [11], we included a 509-bp region (including the D2 variable region described by De Rijk et al. [51]) at the proximal part (from the 5′ end) and a 378-bp region (including D4) in the middle of the gene, but we used more stringent conditions and therefore did not include the distal part of the gene because only a small number (<35%) of comparative sequences were available. A second alignment was derived from the variant described above by including only those parts with sequence information of Arixenia and Hemimerus. Both alignments are accessible at TreeBASE (http://purl.org/phylo/treebase/phylows/study/TB2:S14160).

A basic description of the genetic variability within the data set was obtained using DnaSP v5 [52]. The data set was divided into three parts corresponding to 18S DNA, 28S DNA, and histone-3 sequences. For each part, the best model of sequence evolution was calculated with the aid of MrModelTest 2.3 [53] using Akaike Information Criterion (AIC).

Bayesian analyses were conducted in MrBayes 3.2.1 [54] using the resultant GTR+I+Γ model. Each analysis consisted of two independent runs for 40 million generations; trees were sampled every 5000 generations, and the first 25% were discarded as burn-in. The AWTY (Are We There Yet?) approach was used to explore the convergence of MCMC in Bayesian inference [55]. The stationarity of the runs was inspected by cumulative plots displaying the posterior probabilities of splits at selected increments over an MCMC run, and the convergence was visualized by comparative plots displaying posterior probabilities of all splits for paired MCMC runs.

The current study is the first to include molecular data for both arixeniids and hemimerids in one phylogenetic analysis. This data set enabled us to apply a rigorous cladistics approach and to test competing hypotheses that were previously scattered in the literature. Based on the results, we conclude that Arixeniidae and Hemimeridae are highly specialized and modified families belonging to the suborder Neodermaptera and infraorder Epidermaptera [11]–[18]. The results placed both groups in the superfamily Forficuloidea and support the sister group relationships of Arixeniidae+Chelisochidae and Hemimeridae+Forficulidae.

The genetic markers used in our study (18S and 28S ribosomal DNA and histone-3) are highly conserved among animals and have been traditionally used in phylogenetic studies [56], [57]. However, complete resolution of the deep branching in the order Dermaptera, which originated in the middle Mesozoic [14], will require analysis of long portions of nuclear DNA. Given the moderate values of posterior probabilities accompanying the nodes for Arixeniidae+Chelisochidae and Hemimeridae+Forficulidae, analysis of long portions of nuclear DNA will be also important for fine-scale resolution of the branching order within the superfamily Forficuloidea. The latter kind of phylogenomic analysis was required, for example, for the fine-scale resolution of mammalian and avian evolutionary histories [58], [59]. Nevertheless, the methods and tests used in our study indicate that our hypotheses regarding the phylogenetic position of groups of our interest, the Arixeniidae and Hemimeridae, are well supported. Our report is the first to use the Bayesian tree-building approach to resolve dermapteran phylogeny. This method is characterized by high power and consistency and is especially useful in crossing deep valleys in a virtual landscape of phylogenetic trees, thereby ensuring an increased probability of reaching global and not only local optima [60]. Exploration of MCMC using AWTY indicates that stationarity and convergence of the chains were reached during particular runs.

The only previous molecular study that included hemimerids [11] indicated that the group nested within the Neodermaptera, but the exact placement was poorly supported. That analysis, which treated parameter values as unity, placed Hemimerus as sister to Forficulidae+Chelisochidae but with relatively low bootstrap and Bremer support. Our study, in contrast, well supports the position of both epizoic Hemimeridae and Arixeniidae in the monophyletic group Forficuloidea representing part of Spongiphoridae +Arixeniidae+Chelisochidae+Hemimeridae+Forficulidae; the Hemimeridae is clearly sister to the monophyletic Forficulidae.

Our results suggest (as do the results of Jarvis et al. [11]) that the Spongiphoridae is not a monophyletic group but rather consists of two distinct lineages. One lineage is paraphyletic and contains the entire Anisolabididae clade, and other lineage is nested in the “Eudermaptera” and is sister to the Chelisochidae. Discussions of relationships within the Dermaptera based on morphological characters have generally focused on the male genitalia and have subdivided the Dermaptera into two infraorders, the Catadermaptera (with two penis lobes) and the Eudermaptera (with one lobe) [61]–[64]. Colgan et al. [41] demonstrated that division of Dermaptera into the “Catadermaptera” and “Eudermaptera” is not supported because the former seems to be paraphyletic. “Eudermaptera” was considered to be monophyletic by these authors. However, polyphyly of Spongiphoridae, which was indicated by Jarvis et al. [11] and was confirmed in this study, is inconsistent with Steinman’s concept of infraorders based on penis characters because it indicates the possibility of repeated evolution of penis lobe reduction during the phylogeny of the Dermaptera.

The morphologies and life histories of the Arixeniidae and Hemimeridae are very different from those of the other Dermaptera. Adaptations to an epizoic life history resulted in the evolution of similar characters and thus the rise of homoplasies in data sets based on morphology.

The supraordinal classification of Grimaldi & Engel [14], which is based on paleontological material, comprises three suborders: the extinct Archidermaptera and Eodermaptera, and the recent Neodermaptera (corresponding to Hemimerina, Arixeniina, and Forficulina together in the traditional view). The internal phylogeny and classification of recent Dermaptera ( = Neodermaptera) has been in constant flux, with dramatically different arrangements of families and superfamilies by contemporaneous authors. Grimaldi & Engel [14] distinguish two infraorders within the Neodermaptera: the Protodermaptera and Epidermaptera. The Protodermaptera is basal including the families Karschielidae, Pygidicranidae, and Diplatyidae (sensu [19]), but may be paraphyletic, e.g., [9], [10]. Although the protodermapterans are well characterized by having ventral sclerites of equal size, carinae on the femora, and a segmented pygidium [10], [14], the validity of the infraordinal status of the Protodermaptera and the phylogeny within the infraorder must be verified by further analyses because the recent phylogenetic studies did not use representative material [11], [39], [42]. In particular, the phylogenetic position of the Karschielidae has not been studied by molecular methods. The Epidermaptera comprises the remainder of the recent families of earwigs. Based on his study of some morphological structures of the abdomen, Klass [12] reported that none of the abdominal characters contradicts a subordinate placement of Hemimerus within the Neodermaptera, and he noted weak support for a close relationship of Hemimerus to Apachyus (Apachyidae).

Haas [10] and Haas and Kukalova-Peck [9] performed extensive quantitative phylogenetic analysis of the Dermaptera based mainly on wing characters (especially the articulation and folding pattern of hind wings). Some of the characters seem to be apomorphic, but these characters cannot be used for the Hemimeridae and Arixeniidae because both of these epizoic groups have secondarily reduced wings. Recently, Tworzydlo et al. [14], [16]–[18] studied the ovary structure and initial stages of oogenesis of several dermapteran taxa, including the epizoic Arixenia esau Jordan, 1909. Their interpretation of the results supports the monophyly of the Dermaptera and the inclusion of the arixeniids in the Neodermaptera. Although the shared characters of the ovaries (elongated oviducts and a few, short ovarioles with a low number of follicles) support the position of the Arixeniina within the clade Forficuloidea comprising Chelisochidae+Forficulidae+Spongiphoridae, a placement of this taxon as a sister group of any of them is not strongly supported.

The families Chelisochidae and Forficulidae exhibit morphological autapomorphic characters on tarsomeres in that the third tarsal segment is inserted dorsally on the second tarsal segment and the second tarsal segment is enlarged [10]. The second tarsomere in the Forficulidae is lobed and very wide, while the second tarsomere in the Chelisochidae is not lobed but is enlarged into a peg-shaped process extending under the ventral surface of the third tarsomere [10], [11]. The third tarsal segment of both the Arixeniidae and Hemimeridae is inserted dorsally, which supports their relationship to the Chelisochidae and Forficulidae. The second tarsomere of the Hemimeridae is of the Forficulidae type (it is lobed and wide) and therefore could support the sister position of Hemimeridae+Forficulidae. The second tarsomere of the Arixeniidae is neither lobed nor enlarged but is a conspicuous process extending under the ventral surface of the third tarsomere. We cannot, however, exclude the possibility that the tarsomere characters in the Arixeniidae and Hemimeridae are adaptations to an epizoic life history rather than examples of synapomorphy.

Application of molecular phylogenetics has often generated hypotheses that are incongruent with results derived from a traditional approach based on morphological analyses [58], [59], [65]. Consequently, the “total evidence approach” may be complicated by conflicting signals between molecular and morphological data sets [66]. The phylogenetic reconstructions within a total evidence framework may be biased by the inclusion of phylogenetically misleading data, which are a priori more frequent in morphological data sets because of the greater potential for rapid adaptive change in particular phenotypic traits than in commonly used genetic traits. For example, the reduction of penis lobes, which was considered a synapomorphy in previous reconstructions of Dermaptera phylogeny [10], [12], [61]–[64], may result from parallel evolution, which is typical for reductive character states. From this point of view, molecular markers and corresponding tree-building methods are especially useful in phylogenetic resolution of groups with morphologically highly derived lineages because these methods effectively reveal relationships based on homologies [67], [68], as demonstrated in present study of epizoic lineages of earwigs.