On the taxonomic position of the enigmatic genus Tonkinodentus Schileyko, 1992 (Chilopoda, Scolopendromorpha): the first molecular data

Abstract The taxonomic position of the monotypic Vietnamese genus Tonkinodentus Schileyko, 1992 (for T.lestes Schileyko, 1992) has been considered in the light of the first obtained molecular data. Both molecular (28S rRNA) and morphological data support the position of this extraordinary eye-less genus within the family Scolopendridae Leach, 1814, a sighted clade, and thus suggests the polyphyly of blind scolopendromorphs. The species diagnosis has been amended and color images of T.lestes provided for the first time.

Taking into consideration all these facts, Schileyko (2007) supposed Tonkinodentus to be the first blind representative of Scolopendridae. Schileyko (2007: 71) also wrote that a discussion concerning taxonomic position of this genus will be published elsewhere and left Tonkinodentus unassigned to any subfamily. Thus, the aim of this paper is to specify a taxonomic position of this enigmatic genus using the first molecular data as presented below.

Material and methods
All the studied material is deposited in the Zoological Museum of Moscow Lomonosov State University (ZMMU). The work was carried out based on the two specimens of Tonkinodentus lestes (Rc 6358, holotype; Rc 6555, non-type). Abbreviations used are LBS = leg-bearing segment(s), col. = collector. Specimens were examined both wet and dry under various angles of direct illumination; the photos were taken using a Canon EF-S 60 macro lens mm mounted on Canon EOS 300 camera and DeltaPix Invenio-8DII digital camera. Lewis et al. (2005) and Bonato et al. (2010) were followed for standard terminology of centipede morphology.
A tissue sample of T. lestes was taken from the 75% ethanol preserved specimen (voucher number Rc 6555 in ZMMU, collected in 1994). To avoid contamination, extraction and amplification of the DNA were carried out in the ZMMU Laboratory of Historical DNA. This laboratory was specially designed for work with samples from museum specimens, which potentially have their DNA degraded. No previous work on fresh tissues had been performed in this laboratory (Kruskop et al. 2018;Lebedev et al. 2018). DNA was extracted twice; for the first time, it was extracted and purified using the QIAamp DNA MiniKit (Qiagen), which included an overnight lysis step at 56 °C and longer incubation with EB-buffer (5 min) at the purification step. For the second time, DNA was extracted using a non-destructive method (Gilbert et al. 2007) with the following modifications: incubation at 55 °C was performed for 8 h, DNA purification was done with Qiagen PCR purification kit.
We amplified a fragment of the 28S rRNA nuclear gene. The DNA was highly degraded, so short fragments (100-200 bp) were obtained using the combination of internal primers designed for this study (Appendix 1, 2). Primer pairs were developed manually using Bioedit (Hall 1999) and an alignment of candidate centipede sequences from GenBank. First DNA extraction was successfully amplified with 28S-endF/28S-endR primer pairs, but another pair of primers (startF_2 and IntR_2) worked only on the second DNA extraction.
The PCR program for amplification of short fragments included an initial denaturation at 95 °C for 3 min, 45 cycles of 95 °C for 30 s, annealing temperature (see Appendix 1) for 30 s and 72 °C for 30 s, and a final extension of 72 °C for 6 min. All stages of the extraction process included a negative control run in parallel. PCR products were visualized on a 1% agarose gel. PCR product was sequenced via Evrogen on ABI PRISM 3500xl sequencer. All sequences were deposited in GenBank under the following accession number MK517656.
Additional sequences of 28S rRNA and 18S rRNA of various scolopendromorphs (including the members of Scolopendrinae, i.e. potential close relatives of T. lestes) were downloaded from GenBank (see Appendix 3). Craterostigmus tasmanianus Pocock, 1902, a member of Craterostigmomorpha, was used as an outgroup. We did not increase the length and variability of our alignment by adding mitochondrial DNA data available for this set of taxa (excluding T. lestes) in GenBank because the Chilopoda mtDNA sequences are very variable, much more than their nuDNA ones and there is a high possibility of saturation of mtDNA while comparing distant taxa and because the cases of mitochonrial introgression are rather common. We hold to an opinion that in such situations combining both nuDNA and mtDNA data in one alignment can lead to errors and is better avoided, especially when, as in our case, the DNA fragment of the target specimen is short and represents only one type of DNA markers (either nuDNA, or mtDNA).
Sequences of T. lestes were checked and put in contig using Seqman 5.06 (Burland 1999). Than contig and GenBank sequences were aligned with Geneious 11.1.5 (http://www.geneious.com) using Geneious Alignment. Subsequently, the alignment was checked and manually revised if necessary using BioEdit Sequence Alignment Editor v. 7.1.3.0 (Hall 1999). Two alignments were prepared for the following phylogenetic analysis: sequences of 28S only and concatenated alignment of 28S + 18S. We did not cut the sequences of the other taxa to match the length of the T. lestes fragment, the length of 28S alignment was 1743 b.p. and 1913 b.p. for 18S alignment. 18S sequences were added to improve the resolution of the resulting trees. Genetic distances were calculated using MEGA 6.1 (Tamura et al. 2013).
The optimum partitioning schemes were identified with PartitionFinder (Lanfear et al. 2012) using the greedy search algorithm under the AIC criterion: GTR + I + G for 28S and SIM + I + G for 18S. Phylogenetic trees were reconstructed under Bayesian criteria (BI) and the maximum likelihood (ML). Bayesian inference (BI) was performed in Mr-Bayes v. 3.1.2 (Ronquist and Huelsenbeck 2003) with two simultaneous runs, each with four chains, for 8 million generations for 28S and 12 million generations for 28S + 18S. We checked the convergence of the runs and that the effective sample sizes (ESS) were all above 200 by exploring the likelihood plots using TRACER v. 1.5 (Rambaut and Drummond 2007). The initial 10% of trees were discarded as burn-in. Confidence in tree topology was assessed by posterior probability (PP) (Huelsenbeck and Ronquist 2001).

Amended diagnosis and redescription of Tonkinodentus lestes Schileyko, 1992
Schileyko (2007) redescribed T. lestes in detail, but the black-and-white photographs are far from satisfactory. Below we present a new diagnosis and description of this species accompanied by color photographs.  Diagnosis. Cephalic plate lacking any sutures, its posterior margin overlapped by tergite 1; eyes absent (Figs 4,13). Forcipular tooth-plates well developed and relatively short, with 7 teeth arranged in 2 parallel rows in a chess-board pattern (Fig. 5); trochanteroprefemoral process bisected sagittally (Figs 5,6). Sternites 2-20 with paramedian sutures (Figs 7,14). Pleuron with intersclerite membrane clearly visible; spiracles triangular with a 3-part "flap", slit-like entrance and deep atrium. 21 LBS; the ultimate one visibly shorter than penultimate (Fig. 15). Leg with tarsus 1 considerably longer than tarsus 2, with both tarsal spur and pretarsal accessory spines. Ultimate sternite with poorly developed longitudinal median depression in caudal half. Cylindrical coxopleural process well developed, with spines (Figs 10,16,17). Ultimate legs of "common" shape (sensu Schileyko 2009; Figs 8, 12); femur, tibia, and tarsus 1 each with an apically rounded distal ventro-lateral process (Figs 8,12,18 Antennae of 19 articles (in the both specimens left antenna of 19 and right one of 18, as the corresponding apical article seems to be broken off), reaching the anterior margin of tergite 5 [5.5-6] when reflexed. Basal articles 6 or 7, with a very few long setae, subsequent articles densely pilose. Basal antennal articles flattened.
Cephalic plate (Figs 4, 13) without any sutures, rounded and remarkably narrower than tergite 1; its posterior margin covered by the latter. No light spots at the place of ocelli.
Maxillae 2: the second article of telopodite distally with dorsal spur. Dorsal brush very poorly visible, consisting of short, delicate and transparent setae; apical setae no longer than pretarsus. Uniformly brown pretarsus ( Fig. 19) simple (not pectinate) and claw-shaped, as long as 1/3-1/4 of the length of the apical article of telopodite; pretarsus with 2 thin accessory spines.
Forcipular segment: coxosternite with shortly branched medial suture which is as long as 1/3 of coxosternal length; 2 short sutures stretched caudo-laterad from median diastema ( Fig. 5) [all coxosternal sutures very hardly visible] in the form of an angle of ca 60° [ca 70°]; chitin-lines short but well developed (Fig. 6). Tooth-plates definitely wider than long [visibly higher than in the holotype]; height of tooth margin increasing medially. Each tooth-plate with 7 teeth, fused to various degrees and arranged in 2 parallel rows in a chess-board pattern (Fig. 5), the lateral tooth is the shortest and the most isolated. Basal sutures of tooth-plates form a nearly straight line. Trochanteroprefemural process well developed, divided sagittally into 2 (dorsal and ventral) halves (Figs 5, 6), each half with 2 or 3 lateral tubercles [dorsal halves of both processes with 3, ventral ones (which are visibly smaller) with 2]; the apical end of this process is considerably higher than corresponding tooth-plate. Tarsungula (Fig. 6) of normal length (left one broken off apically in the holotype), ventrally with 2 blunt ridges.
Remarks. The known material consists of two specimens only, neither of which are in perfect condition. More material is needed to investigate the anatomy (e.g. peristomatic structures, foregut, gizzard).
All differences between the holotype and the second specimen are explicable by the latter being a subadult. The much paler and considerably softer cuticle the second specimen suggests that it is newly moulted. Because of this, some delicate structures (e.g. forcipular sutures, leg spurs) are less evident than in the holotype. The most delicate parts (maxillae, antennae, legs) are somewhat deformed (wrinkled) in the holotype, but in the second specimen, the ventral surfaces of the apical articles of the ultimate legs are deformed (unnaturally concave).
Eight specimens of Cormocephalus dentipes Pocock, 1891 (Rc 7518, 7013, 7028, 7231, 7233) from India (Assam and Punjab states), Western Nepal and Indonesia (Sumatra, Medan) demonstrate virtually the same structure of the sagittaly bisected process of the forcipular trochanteroprefemur (Fig. 20). As for the chess-board pattern of the arrangement of the teeth of the forcipular tooth-plates in Tonkinodentus (Fig. 5), it is unique among the Scolopendromorpha.
Summing up, the genus Tonkinodentus is morphologically the typical representative of the subfamily Scolopendrinae (and namely of the former tribe Scolopendrini Leach, 1814) and is the most similar to the genus Scolopendra L., 1758, but differs readily from the latter by the absence of eyes and the peculiarities of the forcipular segment.

Sequence characteristics
We obtained 175 b.p. of 28S rRNA of Tonkinodentus lestes. The complete matrix included sequences from 40 species. Information on the length of 28S and 18S fragments and variability is given in Appendix 4 (all data shown for ingroup only). Uncorrected mtDNA genetic distances are given in Appendix 5, 6 (below diagonal).

Phylogenetic analysis
The results of the phylogenetic analysis are presented in Figures 21 and 22

The taxonomic position of Tonkinodentus and the problem of mono-vs paraphyly of the blind scolopendromorphs
The question of the correct taxonomic position of Tonkinodentus (in fact, the first eye-less scolopendrid) is connected directly with the problem of mono-vs paraphyly of the blind scolopendromorphs. An origin of the family Cryptopidae sensu Attems (1930), or the "blind clade" sensu Vahtera et al. (2012a), which includes all three eye-less scolopendromorph families (Cryptopidae, Plutoniumidae, and Scolopocryptopidae Pocock, 1896) is a matter of a long discussion. Schileyko (1992) argued the monophyly of the blind scolopendromorphs, stating that the group "Cryptopidae" is not a natural taxon, and tried to support this by producing the first character matrix for the order (Schileyko 1996). This was, however, quite limited and included only 15 genera and eight characters. This viewpoint was, in part, supported by Shelley (1997: 106), who wrote: "… no longer should the present division [of order Scolopendromorpha], based primarily on the presence or absence of eyes, be uncritically accepted". Shelley (2002: 2) later wrote: "Based on anatomical and biogeographical considerations (discussed by Shelley (1997) In contrast, both molecular and/or combined analyses supported the monophyly of the eye-less clade (Edgecombe and Giribet 2004;Vahtera et. al. 2012a). Edgecombe and Giribet (2004: 125) wrote: "The cryptopid clade is present across most of parameter space for combined morphological and molecular data (…), leading us to favor the hypothesis that loss of ocelli in Cryptopidae occurred once and defines a monophyletic group". Also Vahtera et. al. (2012aVahtera et. al. ( , 2013) considered a single loss of ocelli in Scolopendromorpha as the most parsimonious. Confirming these conclusions Bonato et al. (2017: 2) stated that Plutoniumidae, Cryptopidae and Scolopocryptopidae are a "… well-supported monophyletic subgroup, informally labelled as the 'blind clade'...".
Morphology and Sanger sequence data reviewed above have been inconclusive with regards to the monophyly of a clade uniting the blind scolopendromorphs (except for Tonkinodentus); this grouping is robustly supported by phylogenomic data (Fernández et al. 2016). All of 20 analyses using different gene partitions, optimality criteria (Bayseian Inference or Maximum Likelihood), or tree-inference algorithms recover this group with strong support.
The only mention of Tonkinodentus within this discussion has been made by Vahtera et al. (2012a: 14), who wrote: "Our data are lacking the monotypic blind scolopendrid genus Tonkinodentus Schileyko, 1992. Morphology supports an assignment of this rare genus to Scolopendridae (Schileyko 2007) but this hypothesis remains yet to be tested in terms of the molecular data. As such, although we postulate a single origin for blindness in three families of Scolopendromorpha, an independent loss of ocelli within Scolopendridae (in Tonkinodentus) is probably based on published morphological evidence for the affinities of Tonkinodentus". Summing up, the results of the first molecular approach applied to this peculiar genus should be of the special importance for this discussion.
As it was already noted above, Schileyko (2007) assigned Tonkinodentus to the family Scolopendridae (sensu lato), so the precise taxonomic position of Tonkinodentus within the family remains indefinite. In the most current general review of scolopendromorph genera, Edgecombe and Bonato (2011: 400) included this genus in the former tribe Scolopendrini Leach, 1814, but provided no arguments for doing so. Later, using a combined morphological and molecular approach, Vahtera et al. (2013: 578) showed that "the tribe Asanadini [Verhoeff, 1907] nests within Scolopendrini for molecular and combined datasets", thus reducing both tribes, but without formalizing their new statuses. The molecular data confirms that Tonkinodentus nests in the family Scolopendridae, or in the subfamily Scolopendrinae (Figs 21,  22), and thus, the discovery of the first eye-less scolopendrid is confirmed.

Conclusions
Work with ancient DNA from long-preserved museum collections is now an important and developing, but complicated, phylogenetic approach. In this study of T. lestes, DNA was so degraded and in so small an amount that two different methods of DNA extraction were used, and only two short fragments of 28S rRNA were obtained.
Both morphological and the first molecular data unequivocally support the position of blind Tonkinodentus inside sighted Scolopendridae. The position of Tonkinodentus among the members of Scolopendrinae (i.e. non-blind scolopendromorphs with slit-like spiracles covered by a "flap") is well confirmed by morphological data, but has quite low nodal support in our phylogenetic analysis. More fresh materials are necessary to complete both internal anatomical and molecular studies of this enigmatic scolopendrid.
The position of T. lestes within the sighted family Scolopendridae coincides with hypothesis that blind scolopendromorphs are non-monophyletic, although phylogenomics strongly supports monophyly of a clade of the three obligately blind families.  (2002)

Appendix 1
Primer pairs used in this study.

Appendix 2
Appendix 3 Simplified scheme of the primer positions on 28S gene.
Sequences used in this study.

Appendix 5
Uncorrected p-distances (%) for sequences of 28S nuDNA gene for species (above diagonal). Standard error estimates are shown above the diagonal. Uncorrected p-distances (%) for sequences of 18S nuDNA gene for species (above diagonal). Standard error estimates are shown above the diagonal.
Uncorrected p-distances (%) for sequences of 18S nuDNA gene for species (above diagonal). Standard error estimates are shown above the diagonal.