Five million years in the darkness: A new troglomorphic species of Cryptops Leach, 1814 (Chilopoda, Scolopendromorpha) from Movile Cave, Romania

Abstract A new species of Cryptops Leach, 1814, C. speleorexsp. nov., is described from Movile Cave, Dobrogea, Romania. The cave is remarkable for its unique ecosystem entirely dependent on methane- and sulfur-oxidising bacteria. Until now, the cave was thought to be inhabited by the epigean species C. anomalans, which is widespread in Europe. Despite its resemblance to C. anomalans, the new species is well-defined morphologically and molecularly based on two mitochondrial (cytochrome c oxidase subunit I COI and 16S rDNA) and one nuclear (28S rDNA) markers. Cryptops speleorex sp. nov. shows a number of troglomorphic traits such as a generally large body and elongated appendages and spiracles, higher number of coxal pores and saw teeth on the tibia of the ultimate leg. With this record, the number of endemic species known from the Movile Cave reaches 35, which ranks it as one of the most species-rich caves in the world.

Microphotographs were obtained with a Nikon DS-Ri-2 camera mounted on a Nikon SMZ25 stereomicroscope using NIS-Elements Microscope Imaging Software with an Extended Depth of Focus (EDF) patch. Images were edited in Photoshop CS6 and assembled in InDesign CS6. Material is shared between the ISER -Emil Racoviță Institute of Speleology, Bucharest, Romania; IZB -University of Belgrade -Institute of Zoology, Faculty of Biology, Belgrade, Serbia; NHMW -Naturhistorisches Museum Wien, Austria; NMNHS -National Museum of Natural History, Sofia, Bulgaria and the ZMUT -University of Turku -Zoological Museum, Finland. In addition to the type material of the new species we have morphologically studied material of C. anomalans from Serbia and Romania.

Molecular methods
Altogether 29 specimens from both inside and outside the Movile Cave were included in the phylogenetic analysis. Of these, 14 were sequenced in this study. Total DNA was extracted from the legs using NucleoSpinTissue kit (Macherey-Nagel) according to the standard protocol for human or animal and cultured cells. Samples were incubated overnight. One nuclear (28S rRNA) and two mitochondrial (cytochrome c oxidase subunit I, COI, and 16S rRNA) fragments were chosen for amplification since they have proven informative between closely related taxa (Vahtera et al. 2012(Vahtera et al. , 2013. 28S rRNA fragment was amplified with the primers 28Sa/28Sb (Whiting et al. 1997), COI fragment with the primers LCO1490/HCO2198 (Folmer et al. 1994) and 16S rRNA with the primers 16Sa/16Sb (Xiong and Kocher 1991;Edgecombe et al. 2002). All primers had a universal tail (T7Promoter/T3) attached to them. Polymerase chain reaction (PCR) amplifications were performed with MyTaqTM HS Red Mix. PCR was performed in a total volume of 23 μL containing 7.5 μL of MQ, 12.5 μL of MyTaq HS Red Mix, 2×, 0.5 μL of each primer (10 μM) and 2 μL of DNA template. PCR started with initial denaturation at 95 °C for 1 min and was followed by denaturation at 95 °C for 15 s. Annealing temperature for 28S rRNA and COI was 49 °C and 43 °C for 16S rRNA. Annealing lasted for 15 s and was followed by extension at 72 °C for 10 s. The last three steps were repeated 35 times. A negative control was included. PCR products were run in electrophoresis on 1% Agarose gel using Midori Green Advanced DNA Stain (Nippon Genetics). Samples were purified with an A'SAP PCR clean-up kit (ArcticZymes). Sequencing was performed by Macrogen Europe. The resulting chromatograms were visualized and assembled using the software Sequencher 5 (Gene codes corporation, USA). All new sequences are deposited in GenBank (See Table 1 for accession numbers).

Phylogenetic analyses
Most specimens included in the analysis had all three markers successfully sequenced. To obtain more geographic variation in the dataset, 15 Cryptops specimens (mostly from Wesener et al. 2016) from GenBank (Table 1) were additionally included in the phylogenetic analysis. Of these, 12 had only COI available. Multiple sequence alignments were performed in MAFFT7 online service (Katoh et al. 2019;Kuraku et al. 2013). Sequences were trimmed in Mesquite v 3.10 (Maddison and Maddison 2019) after which the three separate data sets were concatenated with SequenceMatrix (Vaidya et al. 2011) for the phylogenetic analyses. The final molecular matrix including all three data sets (COI, 16S, 28S) consisted of 1561 characters and 29 taxa (excluding outgroup).
Phylogenetic analysis was conducted using both parsimony and maximum likelihood as optimality criteria. Parsimony analysis was done with TNT v. 1.5 (Goloboff and Catalano 2016) treating gaps as missing data. The search strategy consisted of 100 replications, and of 10 rounds of both ratchet and tree drifting followed by tree fusing (Goloboff 1999). Command xmult was executed until 50 independent hits of the shortest tree were found. A strict consensus of the most-parsimonious trees was produced. The command 'blength' was used to report the branch lengths of the resulting trees. Jackknife (Farris et al. 1996) resampling method with 1000 replicates and with a probability of a character removal being 0.36 was applied to estimate nodal support. Maximum likelihood analysis of the combined data was conducted RAxML v. 8 (Stamatakis 2014) in the CIPRES portal (Miller et al. 2010). The three genes were separated into different partitions. Unique general time-reversible (GTR) model of sequence evolution (RAxML implements only GTR-based models of nucleotide substitutions) with corrections for a discrete gamma distribution (GTR+ Γ) was used.
Nodal support values were estimated using the rapid bootstrap algorithm with 1000 replicates together with GTR-CAT model (Stamatakis et al. 2008). The mitochondrial genes (16S+COI) and the nuclear ribosomal 28S were additionally analysed separately using the same search strategy as was used for the combined data. Uncorrected p-distances of aligned COI, 16S and 18S data were calculated with MEGA v. 7.0.21 (Kumar et al. 2016).  : Negrea, 1993: p. 87 and all subsequent records (Negrea 1994(Negrea , 1997(Negrea , 2004Negrea and Minelli 1994;Sarbu et al. 2019).
Antenna relatively long, extending to the middle of tergite 5 when folded backward (Figs 1A, 2A); composed of 17 articles; article length formula: 17<1<2=16<3=4=13=14<5=6=11=12<7-10; basal two articles relatively stout, in general articles increase in length to a maximum at articles 7-10, then gradually shortening; article 17 is more than half length of article 16 (approx. 60%); articles 5-10 much longer than wide, length up to 3 times the width. All surfaces of antennal articles with scattered long setae, densest on articles 1-3; short, fine setae abundant on all articles except for articles 1 and 2, as well as basal part of 3.
Forcipular segment anterior margin of coxosternite convex on each side, with a weak median diastema, fringed by 2 marginal setae on each side. Surface of coxosternite ( Fig. 3A) covered with scarce short setae, 10-15 in total; trochanteroprefemur stout, median margin slightly expanded proximally, with 4 setae; femur and tibia very short; tarsungulum long, curved, almost equal in length to trochanteroprefemur's height.
Maxilla 2 with a well-developed pretarsus; dorsal brush white, dense, situated on the distalmost part of article 3 of telopodite. Proximal side of first maxillary telopodite covered by 10-15 setae (Fig. 3A).
Legs generally long; leg 10: prefemur 1.47 mm long, femur 1.59 mm, tibia 1.76 mm, tarsus 2.35 mm, pretarsus 0.7 mm. All tarsi single (Fig. 6A). Walking legs (Fig. 6A, B) smooth, generally poor in setae; spiniform setae sparsely present on the surface of prefemur, and occasionally also on the femur; all pretarsi long, with an anterior and posterior accessory spines of different size, the larger being 2/3 rd of pretarsus; accessory spines absent on leg 21; 20 leg: prefemur, femur and tibia slightly swollen; femur and tibia being slightly concave at midlength; a specific field of dense, minute setae present on the ventral, lateral and mesal sides of prefemur, femur and part of tibia.
Etymology. The species epithet is a noun in apposition, meaning "king of the cave", referring to the species top position in the food chain of the Movile ecosystem.

Phylogenetic analyses
Parsimony analysis resulted in a single most-parsimonious (MP) tree of length 1586 steps (Fig. 7). Two C. speleorex sp. nov. specimens collected from Movile Cave (samples K3 and K4) group within C. anomalans as a separate clade supported by jackknife resampling value (hereafter JF) of 99. The phylogeny shows the Movile Cave clade being evolutionary most closely related to the clade (JF = 75) including C. anomalans samples from southern Serbia and Belgrade area (JF = 100) and Romania and SW Serbia (JF = 84). This Serbian/Romanian clade forms a sister group with the clade (JF = 95) containing a single C. anomalans specimen (lab code 4) from southeast Serbia (collected from a cave) and identical sequences of C. anomalans from London, UK and different parts of Germany (JF = 100). All specimens above form a clade with strong support (JF = 92). Outside this clade are Cryptops sp. from Austria and an unsupported clade containing Cryptops spp. from Croatia and Slovenia together with C. hortensis (Donovan, 1810). Basal to these are resolved C. parisi Brolemann, 1920 and C. croaticus Verhoeff, 1931 (JF = 82) followed by Cryptops sp. from Germany.
Regarding the placement of C. speleorex sp. nov. and the relationships among the C. anomalans specimens, the likelihood analysis (Fig. 8) resulted in a mostly congruent tree topology with the parsimony tree, the only difference being that in the parsimony analysis C. speleorex sp. nov. is resolved basal to the Serbian/Romanian clade whereas in the likelihood tree it is resolved within it. The C. speleorex sp. nov. specimens form a clade supported by bootstrap value (hereafter BS) of 100. Cryptops speleorex sp. nov. groups together with the C. anomalans specimens from Serbia (excluding a single Serbian C. anomalans specimen, lab code 4) and Romania. All the specimens above form a sister clade to a group including C. anomalans specimens from Serbia (lab code 4), Germany and the UK. As in the parsimony analysis, the additional Cryptops species (other than C. anomalans) were resolved as basal to C. anomalans. Their internal grouping varies from that in the parsimony tree, which is not surprising due to the lack of nodal support in the basal-most nodes.
When analyzed separately (only likelihood, tree not shown), the mitochondrial COI and 16S resolved C. speleorex sp. nov. as a distinct clade (BS = 100) within C. anomalans specimens, the tree topology regarding C. speleorex sp. nov./C. anomalans being identical to that of the parsimony tree. Not surprisingly, the level of variation in the nuclear 28S was low and the likelihood analysis based on it could not resolve the relationships among the C. anomalans/C.speleorex sp. nov. specimens (tree not shown).  C. anomalans SW Serbia (7) C. anomalans southern Serbia (3) Cryptops anomalans southern Serbia (12) C. anomalans SW Serbia (1a) C. anomalans SE Serbia cave (4)

Pairwise distances
Pairwise distances between the samples by each marker are shown in Tables 3-5.
The differences between C. speleorex sp. nov. and the closest clade (Fig. 7) comprising of C. anomalans specimens from Romania and Serbia are 9.2-12.2% (COI) and 6.6-8.7% (16S rDNA). Nuclear 28S rDNA was conservative and showed almost no variation (0-0.3%) between these specimens. The difference between the new species and the rest of the C. anomalans specimens (Serbia (lab code 4), Germany and UK) is 13.8-15.5% (COI). In respect to 16S the differences were 10.7-12.5% and 9.9-11.2% between the new species and the Serbian (lab code 4) and C. anomalans from London, UK, respectively. Intraspecific difference between the two C. speleorex sp. nov. specimens is 8.5% in COI and 6.6% in 16S.

Discussion
Scolopendromorphs are strictly terrestrial and most species are found in forest leaf litter, decomposed wood, under bark of dead trees, in the soil, under stones or in caves in the temperate and tropical areas of the world. Few species are well adapted to eremic environments (Minelli and Golovatch 2013), occasionally in atypical habitats such as forest canopy (Lewis 1982;Phillips et al. 2020) or tropical rivers (Siriwut et al. 2016). Although less common than lithobiomorphs, scolopendromorphs may occur in caves, where they are represented with some highly adapted species, mainly from the family Cryptopidae. Other families are only marginally recorded in caves: Scolopocryptopidae (genera Thalkethops Crabill, 1960 1914)). The genus Cryptops is by far the most frequent in the caves worldwide with some 18-20 species found in caves in South Europe (Spain, France, Italy, Greece), Canary Islands, Cuba, Brazil, Australia and Africa. Troglomorphic species are known from the nominate subgenus, and the subgenera Trigonocryptops and Paracryptops (see Table 6). Several morphological characters traditionally used in centipedes taxonomy could be subject to intraspecific variation related to postembryonic development, animal life stage and ecology (Akkari et al. 2017). This might render species identification problematic in some cases and generates taxonomic errors. This is also true for such a highly variable and widely distributed species as C. anomalans. In fact, nine species and subspecies were hitherto synonymised with this species (see Krapelin 1903;Verhoeff 1931;Crabill 1962;Zapparoli 2002). Three subspecies are still listed as valid for it (Chilobase 2.0). Now the identity of these taxa and the presence of any possible cryptic species within C. anomalans could only be revealed via an integrative study combining morphological and molecular markers. Whereas clear molecular differences are here indicated by the different markers and the high interspecific distance between C. anomalans and the newly described species C. speleorex sp. nov., the morphological comparison was not as straightforward since both species show several similarities, including an overlapping in size. While several of the differences observed between both species (Table 2) could be understood as a clear indication of troglomorphism in C. speleorex sp. nov. such as the elongation of appendages, a few other characters including the number of saw teeth on tibia and tarsus 1 of the ultimate legs, number of coxal pores and the shape of spiracles were diagnostic to separate both species.
Intraspecific distance between the two sequenced Cryptops speleorex sp. nov. specimens is relatively high in comparison to the detected interspecific variation (Tables 3-5) raising a question whether these two specimens could actually be interpreted as two separate species. However, this variation is only shown in the two mitochondrial markers -there are no morphological differences (or any difference in their nuclear 28S marker) between the C. speleorex sp. nov. specimens. As Morgan-Richards et al. (2017) well explains, cryptic speciation should never be used as a null hypothesis in the absence of phenotypic or nuclear data supporting it. Instead, "the origin of the divergent mtDNA haplogroups might result from complex biogeographical scenarios or they might simply represent normal, stochastic processes of mutation and extinction of a non-recombining locus within a large population".
Taxonomic and evolutionary implications of C. speleorex sp. nov.
The type locality of C. anomalans is unknown and therefore it is impossible to conclude which part (if any) of the studied population is the actual C. anomalans described by Newport (1844). Before this study, only a handful of C. anomalans specimens from a limited geographic range had been sequenced (Spelda et al. 2011;Vahtera et al. 2013;Wesener et al. 2016). We acknowledge that describing C. speleorex sp. nov. as a new species leaves C. anomalans paraphyletic and that monophyly is violated by this taxonomic act. However, we view this as an inevitable consequence of speciation with a particular evolutionary implication, i.e., that C. speleorex sp. nov. evolved within what is currently known as C. anomalans. It is worth noting that the closest evolutionary relatives of C. speleorex sp. nov. appear to be the C. anomalans specimens from Serbia (excluding the sample number 4) and Romania (Figs 8,9). This means that they are most closely related to each other than either of them is to the rest of the studied C. anomalans populations. The current situation with C. anomalans should not be seen as a failed taxonomy but as a natural consequence when new data from a widespread species is obtained.