Three new cavernicolous species of dragon millipedes, genus Desmoxytes Chamberlin, 1923, from southern China, with notes on a formal congener from the Philippines (Diplopoda, Polydesmida, Paradoxosomatidae)

Abstract The large Southeast Asian genus Desmoxytes is slightly rediagnosed. A number of troglomorphic, most likely troglobitic, species occur in southern China. A key is provided to all 10 Desmoxytes spp. currently known from China, including three new presumed troglobites: Desmoxytes eupterygota sp. n. from Hunan Province, as well as Desmoxytes spinissima sp. n. and Desmoxytes lui sp. n. from Guangxi Province. “Desmoxytes” philippina Nguyen Duc & Sierwald, 2010, from the Philippines, is formally removed from Desmoxytes, but not assigned to another genus. It probably belongs in a new genus in the subfamily Australiosomatinae, tribe Antichiropodini, close to the Bornean Euphyodesmus Attems, 1931 and Borneochiropus Golovatch, 1996.


Introduction
Originally, the name "dragon millipedes" was proposed as a vernacular name to distinguish Hylomus draco Cook & Loomis, 1924, a Chinese species showing unusually spiny paraterga, Apparently, Cook and Loomis (1924) were so impressed by this feature that they placed the species not only in a new genus, but also created a new family. However, Jeekel (1968Jeekel ( , 1980 demonstrated that actually several Southeast Asian species share this peculiar character, referring to Hylomus as a synonym of the slightly older Desmoxytes Chamberlin, 1923, assigning the latter genus to the tribe Orthomorphini, and sinking Hylomidae under Paradoxosomatidae. Golovatch and Enghoff (1994), followed by Nguyen Duc et al. (2005), adopted the notion "dragon millipedes" as a fairly compact and homogeneous group of species composing the genus Desmoxytes, tribe Orthomorphini, subfamily Paradoxosomatinae, in which the paraterga are wingor antler-shaped, or even spiniform. Since the genus was based only on a couple of equivocal apomorphies, the generic diagnosis of Desmoxytes as presented by Golovatch and Enghoff (1994) remained rather shaky whilst their cladistic analysis showed no meaningful resolution.
After Mesibov (2006) had described Desmoxytoides, a monobasic genus from eastern Australia which also demonstrates antler-like paraterga, the "dragon millipedes" again became a vernacular name meaningless from a taxonomic or evolutionary viewpoint (Enghoff et al. 2007, Nguyen Duc andSierwald 2010). Even though Mesibov (2006) had failed to assign Desmoxytoides to a tribe or subfamily, due to the presence of a thick solenomere lacking a nearby solenophore the type species D. hasenpuschorum Mesibov, 2006 shows, it clearly represented a different subfamily, Australiosomatinae, albeit rather untypical of it because male legs 1 were devoid of adenostyles. This alone suggested multiple origins of conspicuously hypertrophied paraterga within Paradoxosomatidae. Parallelisms have been reconfirmed and further extended by Golovatch and Stoev (2010) who described a similarly spinigerous species of Tectoporus Carl, 1914 (Tectoporini, Paradoxosomatinae) from Papua New Guinea. Based on gonopod features alone, these authors also suggested placing provisionally both Desmoxytoides and Desmoxytes philippina Nguyen Duc and Sierwald, 2010, the latter species from the Philippines, into the Australasian tribe Antichiropodini (Australiosomatinae). This agrees far better with biogeographical evidence as well, because Borneo supports still another few species in Euphyodesmus Attems, 1931 andBorneochiropus Golovatch, 1996, both latter genera again in Antichiropodini, which show spine-shaped paraterga (Golovatch 1996). Furthermore, Golovatch and Stoev (2010) have proposed to restrict the usage of the term "dragon millipedes" solely to the genus Desmoxytes, a group still fairly well diagnosed earlier by Golovatch and Enghoff (1994) despite certain shortcomings (Mesibov 2006;Enghoff et al. 2007). An updated, only slightly modified diagnosis will be provided below.
The genus Desmoxytes s. str. is currently represented by 26 species ranging from southern China in the North to about the middle of Malay Peninsula within both Thailand and Malaysia in the South. Most of the species diversity is found in Vietnam (9 species), followed by Thailand (8), southern China (7) and Myanmar (2), while neither Cambodia nor Laos has hitherto been known to support any Desmoxytes, even the sole pantropical congener, D. planata (Pocock, 1895). Most of the species are quite local in distribution: the few which have been reported from southern China tend to be restricted to a single locality each. This alone suggests that many more new species of Desmoxytes will be revealed in the future.
Several Desmoxytes species show remarkably bright, apparently aposematic live colorations ranging from intense carmine to pink or purple-pink (see review in Enghoff et al. 2007). One such species, D. purpurosea Enghoff, Sutcharit and Panha, 2007, comes from a small karstic area in northern Thailand, originally probably a cave, but later having its roof collapse, resulting in the present cavernous mountain with high humidity throughout (Enghoff et al. 2007). However, all three heretofore known troglomorphic, most likely even troglobitic congeners, D. longispina (Loksa, 1960), D. scutigeroides Golovatch, Geoffroy andMauriès, 2010 andD. scolopendroides Golovatch, Geoffroy andMauriès, 2010, are pallid to pale brownish in coloration, being confined to karst caves in Guangxi Province . Both latter species come from Mulun Karst which probably contains the richest cave fauna not only in China, but even globally (Deharveng et al. 2008). Epigean congeners from China have only been recorded in Jiangxi (D. draco (Cook and Loomis, 1924)), Guangxi (D. cornutus (Zhang and Li, 1982) and D. minutubercula (Zhang, 1986)) and Yunnan provinces (D. planata). At least the latter species shows an intense red, aposematic live coloration which fades out completely during preservation in alcohol, whereas the coloration in the remaining trio is light, yellowish to light brown in alcohol (Zhang and Li 1982, Zhang 1986, Golovatch and Enghoff 1994. Another interesting observation is that the few species of Desmoxytes which have spiniform paraterga are all confined to Guangxi Province, several being presumed troglobites.
Generally speaking, the family Paradoxosomatidae, one of the largest in Diplopoda as a whole, contains surprisingly few cavernicoles. It is only the large genus Desmoxytes that appears to harbour several troglomorphic species, with caves in southern China hosting virtually all of them.
The present note is devoted to descriptions of three new cave-dwelling Desmoxytes from southern China, i.e. two more from Guangxi Province, as well as the first species from Hunan Province. A key is also given to all ten species of Desmoxytes known to occur in China. In addition, a recent error concerning the identity of a formal congener from the Philippines is being corrected here.
The holotypes will be deposited in the collection of the Institute of Zoology, Chinese Academy of Sciences, Beijing, China (IZAS), with paratypes to be housed in the collections of the South China Agricultural University, Guangzhou, China (SCAU), Guangxi Normal University, Guilin, China (GNUG), Zoological Museum, University of Moscow, Russia (ZMUM), and Muséum national d'Histoire naturelle, Paris, France (MNHN).
A dynamic web page for each taxon name mentioned in the paper is generated on the fly by the Pensoft Taxon Profile tool (see Penev et al. 2010). All species descriptions are automatically exported at the time of publication to a wiki platform (www. species-id.net) through the Pensoft Wiki Convertor (see Penev et al. 2011, Stoev andEnghoff 2011).
Gonopods with rather short, subcylindrical, distoventrally setose coxae. Telopodites mostly suberect, only seldom subfalcate. Prefemoral (= densely setose) portion from 1/3 to 1/2 as long to nearly as long as femorite, the latter not twisted, at most only slightly enlarged distad and devoid both of a mesal groove/hollow and any processes. Seminal groove running entirely mesally to pass onto a usually shortened solenomere, the latter mostly sheathed by a usually condensed, rather simple solenophore, much shorter than femorite, composed of a smaller lamina medialis and a larger lamina lateralis. Solenophore demarcated from femorite by a clear-cut sulcus or cingulum at base, poorly or strongly set off from base of solenomere.
Type species. Desmoxytes coniger Chamberlin, 1923 Species composition. Currently 29 described species. Remarks. The above diagnosis largely repeats that given by Golovatch and Enghoff (1994). Here it only emphasizes variation in relative lengths of the femoral and solenophore parts of the gonopod telopodite. Name. To emphasize the paraterga being true wings. Diagnosis. Differs in the paraterga being mostly wing-shaped, rather long and strongly curved, combined with a short gonopod femorite and a condensed solenophore, as well as ♂ legs totally devoid of femoral humps. See also Key below.
Name. To honour Mr Lu Shiyi, one of the collectors. Diagnosis. Differs in the paraterga being spiniform only until segment 5, combined with the gonopods being strongly condensed, ♂ femora 6 humped. See also Key below.

1
At least paraterga on collum and following four segments spiniform, mostly very long and directed evidently more dorsally than laterally (Figs 3, 5). Paraterga long and spiniform only on collum and following four segments, evidently shorter on segment 5, small and coni-to tuberculiform thereafter (Fig. 5)  As we have noted above, spiniform paraterga must have appeared independently in several paradoxosomatid lineages, including the subfamily Australiosomatinae. This latter group is still rather vaguely defined as opposed to the other two subfamilies, Alogolykinae and Paradoxosomatinae, showing no clear-cut apomorphies. Like Alogolykinae, and contrary to Paradoxosomatinae, most of the Australiosomatinae are supplied with adenostyles (= ventral projections) on ♂ femora 1, coupled with the gonopod showing a shorter or longer prefemoral portion, and a usually (but not always) shortened femorite crowned with one to several processes at or near the base of a strong solenomere. Contrary to the conditions observed in the other two subfamilies, this strong, thick, often apical solenomere in Australiosomatinae requires no or almost no support from the neighbouring structures, thus never being set off at or near the distal end of the femorite by a lateral sulcus or cingulum to develop a special sheath in the form of a solenophore. A complex, often membranous, apical solenophore that is clearly set off at least by one sulcus or cingulum at or near the distal end of the femorite must have appeared later in order to protect and sheath a mostly thin, weak, flagelliform solenomere.
In short, we see the main evolutionary trend in the family Paradoxosomatidae as the progressive development of a complex solenophore in conjunction with the soleno-mere's growing flexibility and fragility. The Australiosomatinae would therefore be the basalmost group, with the Alogolykinae (and probably also a few Paradoxosomatinae like the tribe Paradoxosomatini and a few genera of the tribe Chamberlinini which also have solid solenomeres -see Golovatch 2011, Nguyen Duc andKorsós 2011) likely forming the next step by still retaining a strong and thick solenomere, the apical part of which, however, is already protected and sheathed by a solenophore. Finally, in the bulk of the Paradoxosomatinae (and of the Paradoxosomatidae as a whole), a distinctly delimited, complex, mostly membranous solenophore has developed to sheath and protect the weak, flagelliform solenomere. Jeekel (1968) divided the Australian Australiosomatinae into two tribes, Australiosomatini and Antichiropodini, seeing the main differences between them in gonopod structure alone. Thus, the gonopod in Australiosomatini was stated to show a short to longer prefemoral portion and a stout femorite with several strong branches, including a modestly developed solenomere. In contrast, Antichiropodini were said to be characterized through a short prefemoral portion, an elongated femorite and a similarly long, strong, apically or subapically located solenomere. Golovatch (1996), when treating several Australiosomatinae from Borneo, regarded Euphyodesmus Attems, 1931 and a newly described genus, Borneochiropus Golovatch, 1996, as particularly close to each other, having also transferred both to the tribe Antichiropodini. Not only do at least some of their constituent species show spiniform paraterga so vividly reminding of those observed in certain Desmoxytes, but both lack adenostyles while their gonopod traits are especially similar: the prefemoral portion is medium-sized to hypertrophied, accordingly the femorite is elongate to strongly reduced, and the solenomere is apical to subapical. Both Euphyodesmus and Borneochiropus could have as well been assigned to Australiosomatini, but their placement into the Antichiropodini was favoured because their gonopods looked far simpler than those of most of the Australiosomatini in having only uni-or biramous telopodites, and the solenomere invariably apical or subapical.
In other words, the choice was indeed quite arbitrary. There are a few transitional conditions in gonopod conformation to be observed within Euphyodesmus alone which make it difficult to unequivocally assign this genus to either of the tribes.
This background information is necessary to properly assess the identity of Desmoxytes philippina, described recently from the Philippines (Nguyen Duc and Sierwald 2010). That it is nothing else but an Australiosomatinae has already been noted elsewhere, albeit indirectly because the species had not been described yet (Golovatch and Stoev 2010). Moreover, the gonopod structure (a uniramous telopodite with the prefemoral, femoral and solenomere parts being subequal in length, the femorite also twisted so that the seminal groove runs mostly along its lateral face, the absence both of a solenophore and of its basal delimitation sulcus), the lack of adenostyles, as well as the spiniform paraterga strongly suggest affinities to the Bornean Euphyodesmus and Borneochiropus. The proximity of the Philippines to Borneo supports this view. In no way does D. philippina resemble a species of Orthomorphini, a tribe in which the seminal groove always runs along the mesal side of the gonofemorite before passing onto a flagelliform solenomere sheathed by an evident apical solenophore, both latter structures being set off at their bases at least by one distinct sulcus or cingulum.
So we are forced to remove "D". philippina from Desmoxytes and assign it provisionally to a still unclassified genus of Antichiropodini or Australiosomatini, placing it there together with the similar Euphyodesmus and Borneochiropus. For the time being, we refer to this species as "Desmoxytes" in quotation marks to emphasize its highly doubtful generic allocation. The latter is the more so strange as Nguyen Duc and Sierwald (2010) explicitly noted the above basic differences. Placing philippinus provisionally in Euphyodesmus would have been more logical at the subfamily level.