Fifty specimens of Dicronocephalus belonging to five species and seven subspecies from four countries were obtained (Fig. 1, Table 1), but we were unable to obtain specimens of the remaining two species, D. bieti and D. shimomurai. For examining male genitalia, these were extracted from the abdomens and cleaned by heating with 10% KOH solution in a WiseTherm®HB-48P heating block at 60 °C for 1~2 hours. Male genitalia were preserved in microvials with glycerine after examination. Photographs of external morphology and genitalia were taken with a Canon EOS 10D camera and stacked with a combineZM program (Hadley 2006). Based on previous studies (Pascoe 1863, Pouillaude 1914, Kurosawa 1968, 1986, Young 2012), diagnostic characters were obtained to provide precise criteria for species identification. In this study, the most recent taxonomic scheme by Krajcik (2014) was followed, especially for subspecies treatment of D. wallichii. All examined specimens are stored in the Department of Agricultural Biology, National Academy of Agricultural Biology (NAAS), Jeonju, Korea.

Figure 1. 

The male habitus of species and subspecies of Dicoronocephalus. A D. adamsi adamsi B D. a. drumonti C D. yui yui D D. dabryi E D. uenoi katoi F D. wallichii bowringi G D. w. wallichii H D. w. bourgoini.

Genomic DNA (gDNA) was extracted from middle legs removed from dried specimens of all species and accomplished using a QIAamp DNA Mini Kit (Qiagen, Hilden, Germany) in accordance with the manufacturer’s instructions. Polymerase Chain Reaction (PCR) was performed in order to amplify the cytochrome c oxidase subunit I gene (COI) and 16S ribosomal RNA gene (16S rRNA) using Accupower PCR PreMix (Bioneer, Daejeon, Korea). The universal primer set LCO1490/HCO2198 (Folmer et al. 1994) for amplifying the DNA barcoding region (658bp) of COI sequences was not successful for all samples; this may be caused by the degraded quality of gDNA (Goldstein and Desalle 2003, Hajibabaei et al. 2006; Wandeler et al. 2007). We applied the PCR methodology for retrieving COI sequences from old specimens given in Han et al. (2014) and designed new primer pairs: LCO-Ceto232F (5’–GCHTTYCCYCGAATAAATAAYATA–3’) corresponding to HCO2198 and HCO-Ceto367R (5’–ACDGTYCADCCNGTTCCTGCNCC–3’) corresponding to LCO1490. 16S rRNA was targeted in a 600 bp region with two primers, 16SB/16SA, that successfully amplified in Lucanidae and Elateridae (Hosoya et al. 2001, Hosoya and Araya 2005, Han et al. 2009, 2010). PCR amplification conditions were as follows: for COI, initial denaturation at 94 °C for 5 min, then 45 cycles at 94 °C for 30 s, 46 °C for 25 s, and 72 °C for 45 s followed by a final extension at 72 °C for 3 min, and for 16S rRNA, initial denaturation at 94 °C for 5 min, then 40 cycles at 94 °C for 1 min, 50 °C for 1 min, and 72 °C for 45 s followed by a final extension at 72 °C for 5min. The amplicons were purified using a QIA quick PCR Purification Kit (Qiagen, Hilden, Germany) after the product yield was monitored by 0.7% agarose gel electrophoresis. DNA sequencing was performed using an automated DNA sequencer (ABI 3730xl 96-capillary DNA analyzer; Applied Biosystems, Foster City, CA) with the same primers used for PCR. All sequences (excepting a 198 bp fragment of COI in no. 7282) are available from GenBank under accession numbers KM390855–KM390903 for COI and KM390809–KM390854 for 16S rRNA (Table 1).

For the phylogenetic analyses, three data sets were used, a 658 bp fragment of COI, 520 bp fragment of 16S rRNA sequences, and the concatenated COI and 16S rRNA sequences. The data sets were aligned using ClustalW in MEGA 5.2 (Tamura et al. 2011), and genetic distances were calculated using Kimura’s two-parameter test (Kimura 1980). The phylogenetic analyses were constructed using maximum likelihood (ML), Bayesian inference methods (BI), and maximum parsimony (MP).

ML analysis was performed with GARLI 2.0 (Zwickl 2011), and the analysis was initiated at a random start tree using GTR+I+G model parameters selected by MrModelTest (Nylander 2004), with a 10,000 generation search algorithm and 1,000 bootstrap replications. The frequencies with which to log the best score (“logevery”) and to save the best tree to file (“saveevery”) were set to 10,000 and 10,000 respectively, and the number of generations without topology improvement required for termination (“genthreshfortopoterm”) was set to 5,000. At the end of the analysis, there was no improvement in the tree topology by a log likelihood of 0.01 or better. The bootstrap values were calculated using the SumTrees program of the DendroPy package (Sukumaran and Holder 2010).

BI analysis was performed with MrBayes 3.1.2 (Ronquist and Huelsenbeck 2003). Metropolis-coupled Markov chain Monte Carlo (MCMC) analyses were run with one cold and three heated chains (temperature set to 0.2) for 5,000,000 generations and tree sampling every 100 generations. The posterior probabilities were then obtained and a majority-rule consensus tree was generated from the remaining trees after discarding the first 25% of samples.

MP analysis was performed with TNT 1.1 (Goloboff et al. 2008). The analyses, followed by tree bisection reconnection (TBR) branch swapping, used default options that performed 100 random additional sequences and saved up to ten trees per replication. To obtain the strict consensus tree, symmetric resampling (Goloboff et al. 2003) with a 33% change probability and jack-knifing with a 36% removal probability were implemented using a traditional search with 1,000 replications. Each set of results was summarized in terms of absolute frequency, and the group support values were analyzed. For bootstrap value (BP) in ML and MP, and posterior probability value (PP) in BI, supporting values of <70% as “weak”, 70–79% as “moderate”, 80–89% as “strong”, and ≥ 90% as “very strong” support were used.

The data set of COI, with no evidence of indel (insertion/deletion) events, had 144 (21.9%) variable sites (Vs). Of these, 140 (21.3%) were parsimoniously informative sites (PIs). The data set of 16S rRNA, with indel events at three sites, consisted of 43 (8.3%) Vs, of which 41 (7.9%) were PIs. There was about 2.6 times more variability and the level of PIs was about 2.7 times greater in COI than in that in 16S rRNA.

Phylogenetic inferences based on three analyses (ML, BI, and MP) reconstructed the same topologies for COI (Fig. 2; for BI, ML and MP tree data not shown, see Suppl. material 1 for sequences), and there was separation into two major clades (A and B) with very strong supporting values (100%), except for ML. Eight ingroup taxa representatives including subspecies were clearly clustered into seven monophyletic groups corresponding to nominal species; the two subspecies of D. adamsi formed one cluster. Their terminal nodes were well supported, but the values of ML and BI were very low in D. yui yui (<50% in ML and 53% in BI) and D. wallichii bowringi (<50% in ML and 56% in BI).

Figure 2. 

Phylogenetic relationships among Dicronocephalus species reconstructed with Bayesian inference using COI sequences. Numbers above branches indicate ML bootstrap values and Bayesian posterior probabilities. Numbers below branches are bootstrap, symmetric resampling, and jacknife support from parsimony searches, respectively. Scale bar represents 10% nucleotide mutation rate.

The intra-specific distances of COI were rather low, ranging from 0–2.3%. The inter-specific divergences were highly variable, ranging from 2.7%–16.7%. The distances between the ingroup and outgroup taxa ranged from 16.1%–20.1% (Table 2).

Clade A is composed of D. adamsi adamsi, D. a. drumonti, and D. yui yui with strong bootstrap support (>72%). The two subspecies of D. adamsi did not separate into two distinct subgroups. The genetic divergences between the two subspecies were relatively low (0–1.7%); moreover, D. a. drumonti shared haplotypes with D. a. adamsi from Korea and China. D. yui yui was sister to D. adamsi with distinct inter-specific divergences (5.6%–7.3%).

Clade B is composed of D. dabryi, D. uenoi katoi, and three subspecies of D. wallichii with strong bootstrap supports by ML and BI, but relatively low support (56%–62%) by MP. Among the members of Clade B, D. dabryi and D. uenoi katoi formed a monophyletic group with very strong supporting values in all analyses and with distinct inter-specific divergences (5.6%–8.9%). The intra-specific divergences of these two species (0–1.5% in D. dabryi, 0.2%–2.3% in D. u. katoi) were explicitly lower than their inter-specific values. The three subspecies of D. wallichii were clustered as a monophyletic group and clearly subdivided. D. w. bowringi diverged early from an ancestor, and then D. w. wallichii and D. w. bourgoini underwent subsequent separation with strong bootstrap supports by ML (83%) and BI (99%); however, despite low divergences within each subspecies ranging from 0.3%–0.8%, the genetic divergences between these subspecies were unexpectedly variable ranging from 2.7%–8.1%. Genetic divergences were larger between D. w. bowringi and both D. w. wallichii (4.3%–5.0%) and D. w. bourgoini (4.8%–8.1%), than those between D. w. wallichii and D. w. bourgoini (2.7%–5.7%).

ML, BI, and MP analyses of 16S rRNA resulted in considerably similar topologies to those of COI (Fig. 3 for BI, ML and MP tree data now shown, see Suppl. material 2 for sequences), but a polytomy was found in D. yui yui and paraphyly in D. w. bowringi with respect to D. w. wallichii.

Figure 3. 

Phylogenetic relationships among Dicronocephalus species reconstructed with Bayesian inference using 16S rRNA sequences. Numbers above branches indicate ML bootstrap values and Bayesian posterior probabilities. Numbers below branches are bootstrap, symmetric resampling, and jacknife support from parsimony searches, respectively. Scale bar represents 10% nucleotide mutation rate.

The intra-specific pairwise distances of 16S rRNA were relatively low, ranging from 0–0.4%. The inter-specific divergences ranged from 0.8%–6.3%. The distances between the ingroup and outgroup taxa ranged from 9.7%–11.8% (Table 3). The lowest inter-specific divergence range (0.8%–1.2%) was revealed between D. adamsi and D. yui yui, and this is rather similar to the divergence ranges of the D. wallichii subspecies (0.8%–1.6%).

D. adamsi was clustered as a sister to D. yui yui in Clade A with strong bootstrap support (>90%), while the remaining taxa were clustered into Clade B with relatively low supporting values (>76%) in BI and MP. The monophyly of D. adamsi, D. uenoi katoi, D. w. wallichii, and D. w. bourgoini was well supported by bootstrap analyses (>84%). In contrast, in all analyses a polytomy was found in D. yui yui and ML and BI showed paraphyly of D. w. bowringi. We showed that these phenomena were caused by few parsimony-informative nucleotide variations in conserved regions. A comparison of each of those sequences, showed that D. y. yui has different substitutions at 326 nucleotide position. Two samples (7290 and 7291) have “C”, while one sample (7292) has “T”. On the other hand, D. w. bowringi has a substitution occurred in 196 nucleotide position. The 7693 sample has “G”, while the other samples (7692, 7694, and 7695) and two samples (7274 and 7275) of D. wallichii have “A” at this site (Suppl. material 2).

In the combined data set of COI and 16S rRNA, phylogenetic reconstructions produced topologies congruent with the COI analyses. The nodal supporting values were improved compared with the analyses based on each gene (Fig. 4, see Suppl. material 3 for sequences). Monophyly of the seven taxa including subspecies was strongly supported by bootstrap values >90%, except for low support of 53% and 55% in ML and BI, respectively, for the terminal node of D. w. bowringi. D. w. wallichii was grouped as a sister to D. w. bourgoini based on the results of the COI analyses with a high value in BI (94%) and moderate value in ML (74%), but not in MP (Fig. 4).

Figure 4. 

Phylogenetic relationships among Dicronocephalus species reconstructed with Bayesian inference using COI and 16S rRNA sequences. Numbers above branches indicate ML bootstrap values and Bayesian posterior probabilities. Numbers below branches are bootstrap, symmetric resampling, and jacknife support from parsimony searches, respectively. Scale bar represents 10% nucleotide mutation rate.

The 19 diagnostic characters used to classify species or subspecies were re-examined in order to determine whether they are suitable for identification (Table 4). Of these characters, mentioned in previous studies, 13 are clearly suitable for species or subspecies identification; however, we recognized six characters that are ambiguous and not applicable (Table 5). For example, Pouillaude (1914) mentioned three diagnostic characters as follows: 1) D. dabryi has a different color of the pronotum and the elytra compared with D. wallichii subspecies (Fig. 1); 2) D. w. wallichii can be separated from the others (D. adamsi, D. w. bowringi, D. w. bourgoini, D. dabryi, and D. beiti) by having no angular projection at the base of the anterior edge of the clypeus (Fig. 5); and 3) D. w. bourgoini can be distinguished from the others by the projected apicosutural angle of the elytra (Fig. 6). However, none of these characters has proven to be suitable for species identification. We observed that the color of the pronotum and the elytra of D. dabryi was the same with grayish powder in freshly collected specimens, but it has faded gradually in old specimens (Fig. 1D). Also the anterior edge of the clypeus of D. w. wallichii (Fig. 5G) was sinuate in the middle, similar to that of D. w. bourgoini (Fig. 5H), and did not match the description by Pouillaude. We therefore consider that these characters might have been mistakenly described and illustrated by Pouillaude (1914). In addition, the projection of the apicosutural angle of the elytra of D. w. bourgoini was not distinct and could not separate this taxon from the other species and subspecies (Fig. 6H). We consider that using another character such as “the posterior margin of the elytra is round or truncated” may more diagnostic than the former character as shown in Fig. 6. Pascoe (1863) used the triangular umbone on the shoulder of the elytra (Fig. 7) to distinguish D. a. adamsi from D. w. bowringi. But, we consider that the presence of a triangular umbone is as an unsuitable character. We found this state also in some specimens of D. adamsi, although the size of the triangular umbone was small and variable in each specimen. Kurosawa (1986) used the widest portion of the pronotum as a distinguishing character state, but this was variable in all specimens of D. w. bourgoini and not distinct enough to be used in species and subspecies identification.

Figure 5. 

Anterior edge of clypeus of Dicronocephalus. A D. adamsi adamsi B D. a. drumonti C D. yui yui D D. dabryi E D. uenoi katoi F D. wallichii bowringi G D. w. wallichii H D. w. bourgoini.

Figure 6. 

Apicosutural angle of Dicronocephalus. A D. adamsi adamsi B D. a. drumonti C D. yui yui D D. dabryi E D. uenoi katoi F D. wallichii bowringi G D. w. wallichii H D. w. bourgoini.

Figure 7. 

Umbone (in the circle) of shoulder of Dicronocephalus. A D. adamsi adamsi B D. a. drumonti C D. yui yui D D. dabryi E D. uenoi katoi F D. wallichii bowringi G D. w. wallichii H D. w. bourgoini.

Legrand (2005) used six diagnostic characters to distinguish between the two subspecies, D. a. adamsi and D. a. drumonti. Among them, we found four characters, namely body size, general body shape, longitudinal bands on the pronotum, and the shape of the triangular umbone of the elytra, to be ambiguous. He also illustrated the metasternal process and the parameres and explained in the key to subspecies that the ridge of the metasternal process does not reach the plate, and the process is weakly raised and more rounded anteriorly in D. a. drumonti. Also, the parameres of D. a. drumonti are shorter and with more acute lateral angles than of D. a. adamsi. However, we found that these characters were variable in the specimens from the two geographically isolated populations (Fig. 8). For example, the shape of the lateral angles of the parameres of Tibetan D. a. drumonti (Fig. 8C, D) is similar to that of a D. a. adamsi from South Korea (Fig. 8K, L), and another specimen of D. a. drumonti from Sichuan, China (Fig. 8G, H) resembles a D. a. adamsi from Dandong, China (Fig. 8S, T). We did not find any significant diagnostic characters to separate the two subspecies and therefore the new synonymy is here proposed (Dicronocephalus adamsi drumonti Legrand, 2005 = Dicronocephalus adamsi adamsi Pascoe, 1863, syn. nov).

Figure 8. 

Metasternal process (in the circle) and aedeagi of Dicronocephalus adamsi drumonti and D. a. adamsi. A, B, C, D D. a. drumonti (Tibet) E, F, G, H D. a. drumonti (Sichuan) I, J, K, L D. a. adamsi (South Korea) M, N, O, P D. a. adamsi (North Korea) Q, R, S, T D. a. adamsi (Dandong, China).