To investigate the morphological and molecular diversity of the British Trachysphaera, 28 specimens from Wales and the Isle of Wight were selected. From the mainland population of T. lobata, 10 specimens from a French population (Génis) were used. Additional specimens were included from the northern French-Spanish border region from three sites: (1) one male and female of T. pyrenaica from Grotte de l'Estellas, France, (2) one female of T. cf. rousseti from Leitza, Spain, (3) one male and female of T. cf. drescoi from Sare, France.

As near outgroups, two female specimens of undetermined Italian Trachysphaera, as well as a specimen of the eastern central European species and type of the genus, T. schmidti Heller 1858, were selected because of their divergent morphology not related to the British populations. Two specimens of the British Polyzonium germanicum Brandt 1837 (order Polyzoniida) were selected as far outgroups (NCBI GenBank accession numbers: KJ408478, KJ408479).

All animals were collected by hand. Two slightly different preservation methods were utilized: (1) Immediately after capture, specimens were put into 1.8 ml screw top vials with 98% ethanol; British Trachysphaera were stored in individual tubes, while all French T. lobata and all Italian Trachysphaera specimens were each put together into a single vial. (2) The specimens of T. pyrenaica, T. cf. rousseti and T. cf. drescoi were captured and stored in 80% ethanol, before being transferred to 98% ethanol >12 months later.

Before the disintegration of the specimens for the removal of tissue and SEM preparations, colored multi-layer photographs of the enrolled specimens were taken under a Leica Z6 Imaging-System. A total of 11 specimens from the Isle of Wight population, and 8 from the Welsh population were photographed. For optimal depth of field, the 10-15 single photographs taken from each specimen were put together into one multi-layer photograph using the software Auto-Montage.

The tiny size of the specimens made DNA extraction difficult. In at least one case, the animal was filled with a long and massive nematomorph >4x the length of the host specimen. Due to the possible presence of parasites and the fact that no sequences of the genus (or family) were present on NCBI GenBank for comparison, no whole body extraction was utilized, but every specimen was carefully dissected. Intersegmental muscle tissue was chosen as the extraction target. Due to the small size of the specimens, the specimen was pulled apart along the tergite margins and divided into several parts, where the muscle tissue binding the tergites could be removed with fine forceps. Dissected muscle tissue was washed in a dish of ethanol to remove attached particles of the intestine, which are usually filled with cephaline gregarines.

The muscle tissue was processed with a DNAeasy Blood & Tissue kit from Qiagen following the manufacturer’s extraction protocol, except that two times 50 µl elution buffer were used to heighten the DNA yield of the extraction. DNA was extracted from a total of 44 specimens: 13 specimens of T. cf. lobata from the Isle of Wight and 15 from Wales; 10 specimens of T. lobata from Génis, France; one specimen of T. pyrenaica from the Grotte de L'Estellas; one specimen of T. cf. rousseti from Leitza, Spain; one specimen of T. cf. drescoi from Sare, France; one specimen of T. schmidti from Croatia and two specimens of the Italian Trachysphaera. The dissected specimens were also used for the SEM study (see below), while the remaining parts are conserved as voucher specimens at the ZFMK. Genomic DNA is archived in Qiagen extraction buffer and stored at -20°C at the ZFMK.

To gain insight into the genetic diversity of the British Trachysphaera populations as well as their distance to the French and Spanish taxa, the standard barcoding fragment of the cytochrome c oxidase subunit I (COI), a mitochondrial gene, was chosen as a marker. The COI gene was amplified using polymerase chain reaction (PCR) (Saiki et al. 1988) utilizing the HCO/LCO primer pair (LCO-1490, HCO-2198, Folmer et al. 1994), which corresponds to the COI region 1057–1500. When amplification was unsuccessful, as was the case for the T. lobata specimens from Génis, the Nancy & LCO primer pair was used (Simon et al. 1994). Each reaction volume (20 µl) included 5 µl of DNA extract (since the usual amounts of 1 µl and 2.5 µl did only yield weak bands), 1.3 µl dH water, 1 µl Q-solution (Qiagen), 9.5 µl Qiagen Multiplex Mix, and 1.6 µl (10 pmol/µl) of each primer. PCRs were run with a positive and a negative control. All reactions were procured with a 'touch down' PCR containing an initial denaturation at 95°C for 15 minutes, 15 (25) cycles of 94°C for 35s, 55°C (40°C) for 90s, 72°C for 90s, and a final elongation at 72°C for 10 minutes. PCRs from 34 specimens of T. lobata and the Italian Trachysphaera yielded bands, while those for T. pyrenaica, T. cf. rousseti and T. cf. drescoi were unsuccessful.

Purified PCR products from 34 specimens were outsourced for double-strand sequencing to a contract sequencing facility (Macrogen, Seoul, Korea) on an ABI3730 XL automatic DNA sequencer, using the same primer sets as for PCR. Sequences of a total of 27 specimens (13x Isle of Wight, 12x Wales, 2x Italy) could be obtained, while the sequences for the French T. lobata only contained various contaminations. Overall, PCR and sequencing success was limited (Table 1). We successfully resequenced two tissue samples (Myr924 & 925) of the French T. lobata as well as the T. schmidti sample at BGI (Beijing, China), again on an ABI3730 sequencer (primers: (S0326-0317F)D3A, (S0326-0317R)D3B). The reaction volume (5µl) included 3µl of DNA extract, 1µl of primer (3 pM), 0.5 µl BigDye, and 0.5 µl dH water. The same attempt was unsuccessful for the T. pyrenaica specimen.

Sequencing reads were assembled with Bioedit 7.1.3. (Hall 1999), while the identity (i.e., COI) of all sequences was confirmed with BLAST searches (Altschul et al. 1997) against the NCBI non-redundant database.

The analysis involved 32 nucleotide sequences (COI from 32 specimens), with a total of 660 positions in the final dataset. The number of base substitutions per site from between sequences were determined using MEGA (v. 5.2, Tamura et al. 2011). The phylogeny was inferred by using the Maximum Likelihood method based on the Tamura 3-parameter model (Tamura 1992). Initial tree(s) for the heuristic search were obtained automatically by applying Neighbor-Join and BioNJ algorithms to a matrix of pairwise distances estimated using the Maximum Composite Likelihood (MCL) approach, and then selecting the topology with superior log likelihood value. A discrete Gamma distribution was used to model evolutionary rate differences among sites (5 categories (+G, parameter = 0.7759)). The nucleotide frequencies A = 0.3273, T = 0.3273, C = 0.1727, G = 0.1727. Codon positions included were 1st+2nd+3rd+Non-coding, while all positions containing gaps and missing data were eliminated.

The tree with the highest log likelihood (-2177.4216) is shown below (Analysis section). The percentage of trees in which the associated taxa clustered together is shown next to the branches (bootstrap). The tree is drawn to scale, with branch lengths reflecting the number of substitutions per site.

To evaluate the intra- and interspecific variation of closely related Trachysphaera species, a total of 19 specimens (3 males, 2 females from Isle of Wight; 1 male, 7 females from Wales; 1 male, 3 females of T. pyrenaica; 1 male, 1 female T. cf. drescoi) of Trachysphaera, all of which were also subject to DNA extraction, were studied using scanning electron microscopy. Objects prepared for SEM were: (1) anal shield; (2) anterior body part including head, collum, thoracic shield and tergite 3; (3) midbody tergite (for body part and tergite nomenclature see Fig. 2); (4) male telopods. For scanning electron microscopy, samples were cleaned manually and dehydrated in an ethanol series (80%, 90%, 95% and twice in 100%) and air-dried overnight. The samples were then mounted on aluminum stubs before being coated with gold (layer of ca. 40 nm) in a sputter coater. SEM micrographs were taken using a Hitachi S2460N SEM, based at the ZMFK. Ultrasonic cleaning of the often dirt-covered samples was not attempted, because the risk of destroying the unique surface structure of the tergites was deemed too high. Pictures of the state of all characters in all specimens (as far as they could be obtained) are available on MorphBank (Collection ID 853154, specimen IDs (MBsID) see Checklist Materials, image IDs (MBiID) are given where necessary and not fully listed (326 images in total)).

a

b

Figure 2.

Body part and tergite nomenclature. SEM micrographs of a female of Trachysphaera pyrenaica (GEU157, MBiIDs 852870 and 852871). Abbreviations: Ant = antenna; AS = anal shield; Co = collum (tergite 1); h = head; th-sh = thoracic shield (tergite 2); T# = tergite number.

a: lateral view, showing slightly unrolled specimen  
b: same specimen, view of the opening between anal shield and thoracic shield, exposing head and collum (reduced first tergite) usually held inside the sphere when rolled up  

A total of 15 (11, 4 of which separately given for male and female) morphological characters commonly employed in the taxonomy of Trachysphaera were investigated for their taxonomic value (see Table 2). Some characters were selected because they greatly improved the taxonomy in the giant pill millipede order Sphaerotheriida (VandenSpiegel et al. 2002, Wesener 2009), such as the endotergum, i.e., the underside of the tergites. Each investigated population was scored individually (see character matrix at Table 3); when two different character states were present in the population sample (which was often the case), both character states were coded as 'U' (unresolved). In another step, the telopods of T. lobata populations from Wales and the Isle of Wight, T. pyrenaica and T. cf. drescoi were compared. Note that not all specimens used in this study were analyzed with all used techniques.

Despite differences in the morphology and coloration (see below), no variation in the mitochondrial haplotypes were found within Trachysphaera populations. Genetic distances (uncorr. p-dist.) in the COI sequence between the UK Trachysphaera populations and between T. lobata from France were low (1.9­–2.5%). The Isle of Wight Trachysphaera show a 2.5% divergence to the French population and a 1.9% genetic distance to the population from Wales. The Wales population shows an equal genetic distance of 1.9% to both the French and the Isle of Wight Trachysphaera.

All T. lobata populations show high genetic distances of 17.3–18.8% to the other Trachysphaera species from Italy and Croatia. The latter two, T. schmidti from Croatia and T. sp. from Italy, differ by 16.7%.

The close relationship of the UK and French Trachysphaera populations is also reflected in the phylogenetic tree (Fig. 3). The monophyly of T. lobata is well supported (97%), while the relationships between its three analyzed populations are less clear, with the sister-group of the Isle of Wight population only receiving moderate support of 68%. The Italian Trachysphaera population appears to be closer related to T. lobata (99% support) than to the type species of the genus, T. schmidti, despite a similar genetic distance (17.3% / 18.8%).

Figure 3.  

Maximum Likelihood Tree of 32 COI sequences of Trachysphaera and outgroup specimens. Rate model: discrete Gamma distribution, 5 categories (+G, parameter = 0.7759); nucleotide frequencies: A / T = 0.3273, C / G = 0.1727; tree is drawn to scale, branch lengths reflect number of substitutions per site; 1st, 2nd, 3rd codon positions and non-coding positions were included, gaps and missing data excluded; node values = 1000 replicates ML bootstrap support.

Most of the studied somatic characters show considerable variation (see Table 3), often independent of the observed generic variation (Fig. 4). Often, large variation in a character, such as number of toothed rows on the collum (Fig. 5a, c2), was observed even in the same specimen, with four rows on the left on five on the other side. Nevertheless, many characters were discovered to be constant between populations and species (Table 3). For example the endotergum (Fig. 5d, c5–7) displays no variation within the observed specimens.

a

b

c

d

e

f

Figure 4.

Habitus. Preserved Trachysphaera specimens from two locations, multi-layer photographs, showing high variation in colour and/or encrusted dirt. Not to scale; for further specimen information see Checklists.

a: TW15: T. lobata, male, Isle of Wight  
b: TW18: T. lobata, female, Isle of Wight  
c: TW20: T. lobata, male, Isle of Wight  
d: TW26: T. lobata, female, South Wales  
e: TW31: T. lobata, female, South Wales  
f: TW32: T. lobata, female, South Wales  

a

b

c

d

e

Figure 5.

Trachysphaera SEM characters. A+B: MBsID 852812, C­-E: MBsID 852823. Abbreviations: AS = anal shield; Co = collum; Gr = groove, 'Ohrgrube'; h = head; T# = refers to number of tergite; th-sh = thoracic shield; c2 = collum, number of toothed ridges; c3 = thoracic shield anterior margin, number of rows of sclerotized nodules; c4 = thoracic shield, number of rows of large sclerotized protuberances; c5 = endotergum structure; c6 = endotergum, number of rows of setae; c7 = endotergum, number of rows of sclerotized nodules; c9 = female tergite 10, posterior margin, number of rows of large bacilli (with row counts 1, 2); c11 = female anal shield, shape; c13 = female anal shield, setae at posterior margin; c15 = female anal shield, large circular grooves (for more information on characters see Table 2).

a: TW1: T. lobata, female, Isle of Wight; anterior view; scale bar: 400 µm  
b: TW1: T. lobata, female, Isle of Wight; anterior body part, lateral view; scale bar: 300 µm  
c: TW29: T. lobata, female, South Wales; tergite 10 and anal shield, lateral view; scale bar: 300 µm; c13 points to Fig. 4e  
d: TW29: T. lobata, female, South Wales; endotergum (underside of tergite); scale bar: 20 µm  
e: TW29: T. lobata, female, South Wales; posterior margin of anal shield, detail; scale bar: 50 µm  

The only unambiguous morphological characters that neither show great variation within a species, nor are constant between closely related species, are the characters 8–12 (Table 3). Characters 8 and 9 describe the number of rows of bacilli at the posterior margin of tergite 10 in the female and male sex, respectively. Here, T. lobata has two rows in both sexes (e.g., male: 852997, female: 853015), while T. pyrenaica, T. cf. rousseti and T. cf. drescoi have only a single row in both sexes (e.g., male: 852879, female: 852903). A second character, the presence of a protuberance on the anal shield (character 10 / 11 in Table 2), separates at least the males of T. pyrenaica, T. cf. drescoi and T. cf. rousseti from T. lobata, whose members lack such a conspicuous protuberance (e.g., present: 852829, absent: 852928).

The telopods of the studied Trachysphaera species show extensive variation between species, within species, within populations and even within the same individual (Figs 6, 7, 8). In T. lobata the male telopods show great differences in the shape of the femoral process, which is medially swollen in male TW14 (Fig. 6c), simply tapering in male TW12 (Fig. 6a) and with a special 'bump' in male TW21 (Fig. 7c), the latter being the only male specimen among the samples from Wales. Similar differences can be observed in the tibial process, which is barely visible in some T. lobata specimens from the Isle of Wight, but protruding in the single male from Wales (Fig. 7c, d). Even more pronounced are the differences between the left and right telopod of the T. pyrenaica male (GEU157). Here, the femoral process is triangular on the left telopod (Fig. 8c), while it is well-rounded, identical to those of T. lobata on the right telopod (Figs 6c, 8d). The differences between the left and right telopod are even more pronounced for the tibial process, which is well-developed and triangular on the left (Fig. 8c), but almost absent on the right telopod (Fig. 8d) of the same specimen.

a

b

c

d

Figure 6.

Telopods. SEM micrographs of Trachysphaera lobata males from Isle of Wight. For TW# refer to Checklists. Abbreviations: fem = femur; pre = prefemur; syn = syncoxite; ta = tarsus; ti = tibia.

a: anterior view (MBiID 852943)  
b: anterior view (MBiID 852942)  
c: posterior (anal) view (MBiID 852974)  
d: anterior view (MBiID 852977)  

a

b

c

d

Figure 7.

Telopods. SEM micrographs of Trachysphaera lobata male from South Wales. For TW# refer to Checklists. Abbreviations: fem = femur; pre = prefemur; syn = syncoxite; ta = tarsus; ti = tibia.

a: anterior view (MBiID 853010)  
b: anterior view (MBiID 853007)  
c: posterior (anal) view (MBiID 853009)  
d: posterior (anal) view (MBiID 853008)  

a

b

c

d

Figure 8.

Telopods. SEM micrographs of Trachysphaera cf. drescoi and T. pyrenaica males from France. For TW# refer to Checklists. Abbreviations: fem = femur; pre = prefemur; syn = syncoxite; ta = tarsus; ti = tibia; 18 = leg (pair) 18.

a: T. cf. drescoi, anterior view (MBiID 852849)  
b: T. cf. drescoi, anterior view (MBiID 852848)  
c: T. pyrenaica, anterior view (MBiID 852894)  
d: T. pyrenaica, posterior (anal) view (MBiID 852898)  

The UK Trachysphaera populations belong, based on their mitochondrial DNA (COI sequence), clearly to T. lobata, an observation also corroborated by the morphological data (see below). Both populations analyzed (from the Isle of Wight and South Wales) show unique haplotypes.

Success of DNA extraction and PCR was generally low, indicating the necessity of special treatments to ensure success. Some specimens rolled-up so tight that ethanol was not able to penetrate the specimen, and muscle tissue was already partly decayed. We thus recommend to conserve specimens individually and open them shortly after conservation (i.e., a few minutes after death at best) so that ethanole can enter the organism. Furthermore, it is reasonable to not extract sequences from whole specimens but only from muscular and other tissues free of contamination from gut content or parasites. PCR success was improved (i.e., obtaining positive bands at all) with another deviation from standard protocols, namely using 5µl of DNA instead of the usual 1-2 µl. We were not able to study the reason for this necessity and can thus only speculate. Possibilities are insufficient primer matching, decay, or other, inherent peculiarities.

The differing results of sequencing attempts at the ZFMK and BGI despite of presumably standardized methods highlight the persistent difficulties of obtaining glomerid DNA sequences. Here, we deem a further investigation of best practice methods for Glomerida extremely important.

For studies of the intraspecific variation of Trachysphaera species, faster evolving markers that show a greater amount of variation should provide a better resolution than the COI gene fragment studied here. Unfortunately, such a marker is currently not available in Diplopoda.

Between Trachysphaera species, the genetic distance is quite high, on par with those observed between different genera of the order (see Wesener 2012). Potentially, the observable pattern of substitutions between the COI genes is already saturated. Here, more conservative markers like slowly evolving nuclear or ribosomal genes might allow more robust phylograms than provided. The large interspecific distances of the COI sequences (>15%) observed in different Trachysphaera species highlight the importance of this marker for barcoding purposes. Based on the available data, the barcoding gap is huge (­1.8–2.5% intraspecific vs. 15–­20% interspecific).

Telopods need to be studied and illustrated with great care. Slight variations of the angle of view can provide very different observations (compare Figs 6, 7, 8). Additionally, vast differences of telopod shape do not correspond to observed genetic differences. Thus, we deem telopods no good character of trachysphaeran species discrimination, despite their use in other studies.

Since trachysphaeran surface structures are tiny, complex, and manifold, their evaluation as well as drawing conclusions is obfuscated. Many characters are too fine-scaled or complex to be evaluated in a discrete fashion and it is hard to define categories due to variability in quasi-countable characters. The possible inner variation of extended structures (e.g., endotergum) confounds decisions of species identification based on small details of the structure. However, even small details vary in their peculiarities: large sclerotized protuberances (LSPs; Fig. 5a, c4) form three rows in the middle of the thoracic shield (TS) but dissolve into swirls on the edges of the TS. This means, defining characters as well as states is problematic. The only character sets that seem to allow the discrimination of T. lobata from the rest of the species in our taxon sample are the number of rows of bacilli at the posterior margin of tergite 10 and the shape of the anal shield (characters 8-12 in Table 3). Examples for these are depicted in Fig. 9, using males as they show the features more pronounced.

a

b

Figure 9.

Taxonomically informative character examples. Abbreviations: AS = anal shield; T10 = tergite 10; c8 = male tergite 10, posterior margin, number of rows of large bacilli; c12 = male anal shield, shape. For more information on characters and states see Table 2.

a: TW30: T. lobata, male, South Wales (MBsID: 852824). Character states: two rows of bacilli on posterior edge of T10 (c8/1) and well-rounded anal shield (c12/0). This set of character states is (in our sample) unique for T. lobata.  
b: GL07: T. cf. rousseti, male, Spain (MBsID: 852825). Character states: one row of bacilli on posterior edge of T10 (c8/0) and anal shield with special protuberance (c12/1). This set of character states is (in our sample) shared by T. pyrenaica, T. cf. rousseti, and T. cf. drescoi.  

Problems did not only arise from the structural diversity itself but also from the method of documentation. Too few pictures were taken and not all angles were available due to fixation. Comparing all specimens, not all structures were shot from the same angle. A better chance to find true species-specific characters and evaluate them properly would be given by zoomable 3D images, as can be obtained by micro-CT studies. Additionally, physical influences may have distorted the evaluation. LSPs and similar extended structures (e.g., setae) especially may have been evaluated inconsistently due to prior abrasion (during struggle in life, dissection, mounting and sputtering). Grooves may have been evaluated incorrectly if clogged by dirt. Here, testing whether ultra-sonic cleaning is feasible would be beneficial (we did not try due to the risk of loosing of one specimen). Given these complications and our relatively small sample size, further corroboration of our results is desirable. Meanwhile, our conclusions have to be considered as preliminary, but nevertheless relevant.

Both, barcoding and the morphological study identify the British Trachysphaera populations as T. lobata. We were not able to trace the origin of these British populations because only one French mainland population could be sampled due to external circumstances. However, an independent origin of the South Wales and the Isle of Wight populations could be ascertained. Whether this independent origin is the result of two different recent anthropogenic introduction events or if they represent two relic populations of a once widespread occurrence currently cannot be determined. Once the COI sequence of more French populations of T. lobata becomes known, a clearer picture of the origins of the British Trachysphaera can be drawn.