Sechelleptus arborivagus sp. nov., a new arboreal spirostreptid millipede (Diplopoda, Spirostreptidae) endemic to Mayotte Island (Comoros Archipelago), Indian Ocean

A new millipede species of the genus Sechelleptus Mauriès, 1980 is described and illustrated from Mayotte Island, Indian Ocean. This new species, S. arborivagus sp. nov., found on trees, looks particularly similar to the sympatric S. variabilis VandenSpiegel & Golovatch, 2007, but is much larger and has a very different ecological behavior. Phylogenetic analyses based on a concatenated dataset of the COI and 16S rRNA genes and including nine species of Spirostreptidae (including Sechelleptus, Doratogonus Attems, 1914, Bicoxidens Attems, 1928 and Spirostreptus Brandt, 1833), strongly support the monophyly of Sechelleptus. Despite the similarity of their genitalia, the molecular analyses also reveal a clear-cut genetic divergence between S. arborivagus sp. nov. and S. variabilis (22.55% for COI and 6.63% for 16SrRNA) and further suggest the presence of a higher diversity within the genus Sechelleptus on Mayotte.


Introduction
The genus Sechelleptus Mauriès, 1980, a senior synonym of Rubanostreptus Krabbe, 1982(Jeekel 1999, was erected by Mauriès (1980) to reassign Iulus seychellarum originally described by Desjardins (1835) from a specimen collected in the Seychelles. The genus had been considered as monotypic until the extension of the generic concept by Golovatch & Korsós (1992) to include a large group of spirostreptids from Madagascar. First supposed to be restricted to Madagascar and some adjacent islands, the genus European Journal of Taxonomy 755: 1-21 (2021) 2 appears to have a wider distribution ranging from East Africa to Madagascar and species of the same genus have been observed in Tanzania (Enghoff et al. 2016), Mauritius and Zanzibar (Jeekel 1999), and the Comoros (VandenSpiegel & Golovatch 2007). In the latter study, the authors described a new species of Sechelleptus, i.e., S. variabilis VandenSpiegel & Golovatch, 2007, and mentioned the presence of another putative congener. However, they were unable to assign the single female specimen found to a formal species. Based on its appearance and peripheral characters, the unknown female was at least assigned to the genus Sechelleptus (VandenSpiegel & Golovatch 2007).
In 2019, at the initiative of DEAL Mayotte, a new visit was organized to Mayotte. The visit was effectuated in November, during the rainy season, and allowed to collect, for the fi rst time, two mature males of a large spirostreptid, corresponding to the unknown females mentioned in VandenSpiegel & Golovatch (2007). The study of the newly collected material shows that this species and S. variabilis are morphologically quite similar. At the fi rst glance, even the gonopods of the large specie look like a giant form of those of S. variabilis. Nevertheless, a closer observation of the specimens reveals morphological differences, which are corroborated by a molecular study. In this paper, a morphological description of the new species is provided and its phylogenetic affi nities are discussed.

Taxon sampling and morphological examination
This study is mainly based on material from Mayotte Island, collected in 2019 by the fi rst and third authors. Some additional samples were obtained from the Royal Museum for Central Africa (RMCA), Tervuren, Belgium.
All samples are stored in 70% ethanol. Specimens for scanning electron microscopy were air-dried, mounted on aluminum stubs, coated with gold and studied using a JEOL JSM-6480LV scanning electron microscope.
The terminology used to describe the gonopod structures follows that of Hoffman (2008). All measurements are in mm unless otherwise indicated.

DNA extraction, amplifi cation and sequencing
A few legs of seven freshly collected diplopod specimens, plus one older sample from the RMCA collection, all from Mayotte, were detached for molecular analysis (Table 1). The DNA of each sample was extracted with a Macherey-Nagel NucleoSpin tissue kit, slightly adapted from the manufacturer's protocol. Partial fragments of two mitochondrial markers, the cytochrome c oxidase subunit I (COI) and the large subunit ribosomal RNA (16S rRNA), were amplifi ed with polymerase chain reactions (PCRs) using the primers displayed in Appendix 1. All amplifi cations were performed in a 20 μL reaction mixture containing 2 μL of extracted DNA (regardless of initial concentrations), 2 μL of 10X buffer, 0.5 mM MgCl2, 0.2 mM dNTP, 0.8 μM of each primer, and 0.02 units/μL of PlatinumTM Taq DNA Polymerase (Invitrogen TM , Waltham, MA, USA). PCR conditions comprised an initial denaturation at VANDENSPIEGEL D. et al., A new spirostreptid millipede from Mayotte 3 94°C for 3 min followed by 35 cycles with, per cycle, a denaturation at 94°C for 45 s followed by an annealing step of 45 s at 48°C and an extension step at 72°C (1 min) and with a fi nal extension at 72°C for 10 min followed by 10 min at 4°C. PCR products (and negative controls) were checked on a 1.5% agarose gel containing 0.03% of MidoriGreenTM Direct (NIPPON Genetics Europe, Dueren, Germany) using a UV transilluminator. Positive amplifi cations were purifi ed using the ExoSAP-ITTM protocol (following manufacturer's instructions) and then sequenced in both directions, using the same couples of primers (0.05 mM each), by Macrogen Inc. (Seoul, South Korea). The sequences were checked using Geneious R11 (Biomatters Ltd., Auckland, New Zealand): paired bi-directional strands were trimmed, assembled, edited and consensus sequences were extracted for each specimen and each DNA marker. The COI sequences obtained for the three specimens of Sechelleptus arborivagus sp. nov. (DB2, DB3 and DB4) were further manually edited as they presented two additional nucleotides generating a frameshift in the predicted coding region and numerous stop codons. For DB2 and DB4 sequences, one thymine was removed from position 345-346 and another for position 426-427. For DB3 sequence, the same nucleotides were removed from position 315-316 and position 396-397. As control, consensus sequences were compared against the Identifi cation System of BOLD (www.boldsystems.org) and the BLAST web application of GenBank (https://blast.ncbi.nlm.nih.gov/Blast.cgi).

Phylogenetic analyses
For phylogenetic analyses, a few additional Spirostreptidae Brandt, 1833 sequences from GenBank and BOLD (see Appendix 2) were selected based on BOLD and BLAST 's comparative results. Although this set of taxa does not represent an exhaustive list, we estimate it suffi cient to evaluate the relationships of the new species. In addition, some representatives of the family Harpagophoridae Attems, 1909 (Thyropygus spp.) were chosen as outgroup. All sequences of each marker were aligned with MAFFT ver. 7 implemented online (Katoh & Standley 2013) with default settings. Both COI and 16S alignments were cured (by removing ambiguous sites and trimming datasets) using the least stringent settings in Gblocks 0.91b online (Castresana 2000;Talavera & Castresana 2007; Gblocks Server available online at http://molevol.cmima.csic.es/castresana/Gblocks_server.html). Finally, a combined COI-16S alignment was created with Mesquite ver. 3.5 (Maddison & Maddison 2018) by concatenating the Gblocks curated alignments of the two markers. Best partition scheme and best-fi t substitution models (see Appendix 3) were estimated using PartitionFinder 2 (Lanfear et al. 2016) on the basis of one partition for 16S and three for COI (protein coding gene partitioned into single codon positions). Phylogenetic reconstructions were evaluated using statistical approaches including maximum likelihood (ML) using GARLI ver. 2.01 (Zwickl 2006) and Bayesian inferences, the latter utilizing MrBayes ver. 3.2.7a (Ronquist et al. 2012) (MB) and Phylobayes MPI ver. 1.5a (Lartillot et al. 2009) (PB), all performed on the CIPRES Science Gateway ver. 3.3 (Miller et al. 2010). For the MB analysis, two parallel runs (with four chains each) were executed for ten million generations. Parameters were estimated independently for each partition using the following command: unlink statefreq = (all) revmat = (all) shape = (all) pinvar = (all) tratio = (all). Trees were sampled every 1000 th generations and were used to reconstruct a 50% majority rule consensus tree after having discarded the fi rst 25% as 'burn-in'. Analysis in PB was conducted under the CAT+GTR+Gamma substitution model (with maxdiff" value set to 0.1 Minimum Effective Size to 300 and excluding the 1000 fi rst of cycles from convergence checks). Analyses using the ML method were conducted in GARLI with 1000 bootstrap replicates. Values were then summarized on the best ML tree using SumTree ver In addition, estimations of the average evolutionary divergences for each marker (based on a MAFFT alignments of the sequenced specimens only) were calculated in MEGA-X (Kumar et al. 2018;Stecher et al. 2020) as the number of base differences per site (P-distance) from between sequences and species group with the option 'Pairwise deletion'.

Phylogeny
Amplifi cation and sequencing provided a fragment length of about 730-900 bp for COI (partial sequence of 200 bp only for two samples, DS1 and DS2, could be retrieved) and about 500 bp for 16S. The combined COI-16S fi nal dataset comprised 1051 sites (561 bp for COI and 490 bp for 16S).
All phylogenetic analyses provided congruent results with identical tree topologies. The Figure 1 presents the tree resulting from the MB analysis on which the support values of PB and ML are also summarized. The genus Sechelleptus represented here by S. variabilis, S. arborivagus sp. nov. and an undetermined species (i.e., DU1, a sub-adult female collected at Mont Combani on Mayotte) is strongly recovered as monophyletic in all analyses and appears sister to an unidentifi ed dipoplod from Madagascar (for which the COI sequence was retrieved from BOLD, see Appendix 2). Both S. variabilis and S. arborivagus sp. nov. are recovered with high support, but their relationship with DU1 is not clearly resolved.
The relationships among the outgroups were not the primary focus of the present study, but are consistent with previous studies (Mwabvu et al. 2013(Mwabvu et al. , 2015Tinago et al. 2017) that showed paraphyletic groups suggesting the presence of cryptic species (i.e., Bicoxidens spp.) and possible identifi cation errors in the GenBank database.

Genetic distances
The MAFFT alignment of the sequenced specimens used for calculations in MEGA-X provided datasets of 1007 positions for COI and 502 positions for 16S. Little genetic variation is observed between the sequenced individuals of Sechelleptus arborivagus sp. nov. compared to S. variabilis (see Table 2). For S. arborivagus sp. nov., the within mean group distance calculated in MEGA-X is very low: 0.1% for COI and 0.2% for 16S. These values are much higher within S. variabilis, being 4.9% and 1.5% respectively.

Type species
Iulus seychellarum Desjardins, 1835 Diagnosis (adapted from Mauriès 1980 andKorsós 1992) A genus of moderate to large spirostreptid millipedes (up to 120 mm long) characterized by the rather simple gonocoxite and the long, slender, ribbon-shaped gonotelopodite with a spine arising well distad of the knee and a small free solenomerite arising just near the apex.

Key to species of Sechelleptus
Waiting a complete revision work of the genus, only the following additional lines to the key of Jeekel

Remarks
In the original description of the species, the length unit given for the specimens' sizes is wrong and should have been cm instead of mm. Correct measurements are: length ♂♂: 3.4-8.0 cm; ♀♀: 5.0-7.0 cm; midbody width 0.34-0.5 cm and 0.3-0.5 cm, respectively.
Taking this modifi cation into account, the recently collected specimens agree with the description (VandenSpiegel & Golovatch 2007).

Diagnosis
A medium-sized arboreal millipede with relatively long legs, particularly similar to S. variabilis by sharing the structure of the male fi rst leg and rather simple gonopods with the metaplica widened and a little higher than proplica, the latter without lateral cone. The two species differ by the gonotelopodite being apically divided in two branches in S. arborivagus sp. nov. and simple in S. variabilis.

Etymology
Referring to the ecology of the species, which has always been observed climbing trees.

Paratypes
Male similar to holotype. Female coloration as in male, but generally larger in size than male (up to 120 mm long 9 mm wide (58-61 body rings plus telson, no apodous rings). Vulvae located in membranous pouches attached to coxae 2 and 3 and to the inner lateral margin of ring 1, simple, consisting of two simple, subequallysized, moderately sclerotized valves, the aboral valve with an apical cluster of setae; ridge between valves covered with a lateral longitudinal operculum (Fig. 7).

Distribution
The species seems endemic to Mayotte (Fig. 8).

Affi nities
On the basis of the gonopod structure having the telopodite with a spine arising well distad of the knee, a ribbon-shaped distal part, and a small free solenomerite arising just near the apex, the new species is manifestly a new member of the large genus Sechelleptus. Following the key published by Jeekel in 1999, arborivagus keys out close to sulcicollis and macilentus. Indeed the three species have a rather simple gonocoxite with a distally widened metaplica without a strong lateral cone but the new species do not show the small lateral uncus present on the metaplica of sulcicolis and possess a more or less spiniform mesapical projection on the proplica which is not present in sulcicolis neither in macilentus. By the overall shape of the male fi rst leg and gonocoxite, the new species seems to be especially close to S. variabilis, also from the Comoros, but it differs strikingly by the structure of the gonotelopodite (in S. variabilis the gonotelopodite has a simple and pointed tip carrying the terminal opening of the seminal groove whereas in the new species the gonotelopodite has a divided tip, the longer branch carrying the terminal opening of the seminal groove) as well as by the larger body size and the longer and curved claws (Fig. 4K vs Fig. 2C). Other important differences concern the defensives glands, large in S. arborivagus sp. nov. (Fig. 4I) (vs inconspicuous in S. variabilis ( Fig. 2A)), and the size of eyes: in the new specie the eyes are larger and include 60 ± 5 ommatidia (n = 10) arranged in 12 rows; whereas in S. variabilis the eyes, smaller, include 34 ± 3 (n = 10) ommatidia arranged in 9 rows.

Natural history
Most of the specimens belonging to the new species were collected on Mt Tchaourembo (see Fig. 8) in a forest fragment at 500-550 m a.s.l. All specimens were seen in trees and never in pairs, the males being rare (sex ratio > 1/6). The species possesses enlarged ommatidia, relatively long legs with strongly curved tarsal claws, as well as a tendency for specimens to secrete extremely copiously from their defensive glands when irritated. Such modifi cations are considered by several authors as an adaptation to tree climbing and to arboreal life (Enghoff & Enghoff 1976;Hoffman & Howell 1983;VandenSpiegel 2001).

Discussion
Millipede systematics is mainly based on male gonopods because they use to be species-specifi c (Bond et al. 2003). However, studies based on DNA have demonstrated that molecular divergence in different millipede groups may not refl ect divergence in morphology-based identifi cations and may hide considerable variation (Bond & Sierwald 2002;Bond et al. 2003;Adams et al. 2009;Mwabvu et al. 2013Mwabvu et al. , 2015Tinago et al. 2017). Although our relatively small taxon sampling, the phylogenetic analysis strongly recovers Sechelleptus as monophyletic and discriminates at least two or three different groups. Furthermore, the mean inter-specifi c distance values (14.9% for COI and 5.1% for 16S) were remarkably similar to previous studies that reported the presence of high genetic divergence among population of different spirostreptid species (Mwabvu et al. 2013(Mwabvu et al. , 2015, suggesting the existence of more than one species in those taxa. It is argued that high level of divergence between identifi ed spirostreptid species may indicate that changes in genital morphology occur rather slowly relative to the high rate of substitution in mitochondrial sequences (especially for COI), and may underestimate species diversity. This also appears to be the case among the different forms of Mayottan Sechelleptus, which also share strongly similar gonopods. At the fi rst glance, the new species of Sechelleptus seems to be a giant form of S. variabilis. However, although only subtle morphological differences are observed within the gonopods, the comparatively large body size and the behavior of S. arborivagus sp. nov. are remarkable. These observations fi nally corroborate our molecular analyses that clearly show suffi cient genetic difference between the different Sechelleptus species collected on Mayotte (22.6% for COI and 6.6% for 16S between S. arborivagus sp. nov. and S. variabilis).
The genetic analyses also suggest the presence of another different species, i.e., DU1, although its phylogenetic position remains unresolved. This unique specimen found at Mont Combani is a sub-adult female that could not allow a formal identifi cation, but, judging from its general appearance, appears to be an intermediate from between the two Sechelleptus species collected on Mayotte. The genetic divergences, along with adaptations to arboreal life observed in the novel species, may indicate an "adaptive micro-radiation" on Mayotte Island or even the Comoros. However, the inclusion of more specimens, including adult males, in phylogenetic analyses is needed to test this hypothesis and evaluate the status of that putative new species.