As outlined by Kenning et al. , centipede ultimate legs are special, and Lithobius forficatus is no exception. Several, not mutually exclusive pathways of morphological modifications and behavioral adaptations can be distinguished in centipedes. Many species possess elongated, even multi-annulated ultimate legs (Scutigeromorpha, Newportia spp.), some species possess thickened or pincer-like ultimate legs (e.g. Scolopendromorpha), and in many species sexual dimorphisms occur (in particular Geophilomorpha). The ultimate legs of L. forficatus contrast with other, more drastic examples of ultimate leg transformations in centipedes  as they in many aspects still resemble a regular walking leg. In general, in L. forficatus there is a continuous increase in leg length, in particular in legs 14 and 15 (Fig. 1a) . Thus, given the presence of telopodal pores and coxal organs, the three most posterior pairs of walking legs (12–14) as well as the ultimate legs (15) may pose as an example of a gradual morphological modification in contrast to regular walking legs. In comparison to the walking leg 10, the ultimate leg features a series of distinctions such as the leg musculature, the glandular organization, sensillar configuration, and the associated nervous system as discussed in the following.
External morphology, musculature, and tendons
In addition to size, one of the major differences in podomere configuration between walking and ultimate legs is the division of the tarsi. According to Zapparoli and Edgecombe , in the genus Lithobius legs 1–13 exhibit a clearly visible tarsal articulation between the tarsomeres, although this articulation is not complete (i.e. the joint is complete around all but a small extent dorsally). This is in concordance with comprehensive anatomical descriptions of Lithobius forficatus by Rilling [23, 34], as well as descriptions of Lithobiomorpha by Eason , Lithobiidae by Barber , and this contribution. Only the tarsi of leg pairs 14 and 15 show a complete separation and thus constitute distinct podomeres, providing new degrees of freedom in leg posture and movement. The functional implications are diverse but conjectural at this time, particularly as most of the leg joints lack fixed rotational axes which allows for a variety of displacements (compare ). However, it seems reasonable to assume that the enhanced maneuverability and flexibility achieved by the fragmentation of the tarsi is significant in light of the defensive performance of the ultimate legs. Which role the presumptive second tendon of the tibio-tarsal articulation plays however, remains uncertain, but likely it supports elevation of the tarsi.
Another prominent difference in podomere configuration is found in the coxa. In the walking leg, the coxa forms an incomplete ring, which allows the trochanter to be pulled in the soft pleural flanks as part of the remotor movement . In the ultimate leg, the entire coxa is turned caudally, leaving the lateral (formerly anterior) side as a rigid cuticular ring. Moreover, while several pro- and remotor muscles are reduced (e.g. transversal muscle m1 as well as sterno-trochanteral muscles m37), levators and depressors (i.e. coxo-trochanteral muscles m40 and m41) are substantially larger . This effectively allows only for up and downward movements, but less so for lateral motion. Although ultimate legs do not participate in locomotion in terms of propulsion, they play a pivotal role in the process by stabilizing the body. Ultimate legs are always stretched backwards. When a centipede turns to one side, the ultimate legs swing to the other side resulting in a faster turn. Experiments with animals that have only a single ultimate leg left give further support to this hypothesis: the remaining ultimate leg is always held along the median plane of the body in order to keep the centre of mass and hence balance .
All centipedes but Scutigeromorpha feature coxal organs – specialized epithelia and pores that are in most cases associated with the ultimate leg coxae (e.g. [4, 11, 36,37,38,39,40]; apomorphic character for Pleurostigmophora ). Remarkably, in most species of Lithobiomorpha the coxal organs are not restricted to the last pair of legs, but are present on the coxae of leg pairs 12–15 (secondarily modified in some species to the last two, three or five leg pairs ). According to Herbst , coxal organs serve a defensive function, facilitating the autotomy of legs, while Verhoeff  attributed them a function in supporting leg regeneration, or even in prey capture. Their relevance still is not conclusively resolved. Littlewood [43, 44] interpreted lithobiomorph coxal organs to be release sites of pheromones. However, the presence of transport epithelia, as well as ecological data and behavioral observations suggest a role in atmospheric water uptake (as proposed by Rosenberg and Bajorat ). In fact, species living in rather dry habitats (Hessebius spp.) feature a low number and diameter of coxal pores, while very large diameters were reported in species living near to streams . In any case, the biological importance of coxal organs is demonstrated by being present on the last four pairs of legs.
Telopodal organs were originally interpreted as pheromone glands, because pores and associated glandular systems are thought to be absent in the anamorphic larval stages [12, 13, 47]. However, Lewis and Yeung  documented the changes in the distribution of telopodal pores during the development of L. microps and have shown that during the larval and postlarval stages, each last pair of legs houses a certain number of telopodal pores. In successive stadia and the anamorphic addition of leg pairs, the number of telepodal pores increases on the respective last pair of legs (e.g. leg 12 in stage 4, with legs 13–15 only present as limb buds [49, 50]), but also decreases on the former ultimate legs.
Also termed defense glands, these closely aggregated glandular organs become effective in predation avoidance by the secretion of a sticky, slowly hardening substance, trapping any opponent [5, 8, 11]. Besides its adhesive quality, little is known about the biochemical composition or physical properties of the secretion. Based on histochemical staining experiments, Blower  described the secretion of the telopodal glands as amber-colored substance that hardens rapidly and forms fibers after being discharged. Staining reactions revealed that the telopodal secretion contains lipoids or proteins, or a mixture of both. According to Verhoeff , a single podomere of Lithobius mutabilis accommodates about 200 glandular pores which total up to approximately 800–1200 pores per ultimate leg. Kästner  stated that L. forficatus has over 500 pores per ultimate leg. Given the 4000 pores in L. forficatus as shown in this contribution, we believe that it is possible, if not probable, that the specimens Verhoeff and Kästner analysed were still in their late anamorphic, or early epimorphic stages.
Sensilla typology and distribution
Epidermal sensilla are widespread on the cuticle of centipedes, and are mostly aggregated on the antennae, mouthparts, legs, and gonopods . However, only few sensilla types are comprehensively understood with regard to their ultrastructure and function (summarized by [52,53,54]). The external morphology of antennal sensilla has been investigated for at least one species in each of the higher-taxonomic groups of centipedes: Scutigera coleoptrata , Lithobius forficatus [5, 21], Craterostigmus tasmanianus , Cryptops hortensis , Scolopendra oraniensis , and Geophilus flavus [54, 59,60,61,62]. The most comprehensive accounts on L. forficatus [5, 21] revealed approx. 2400 sensilla on each antenna. About 2000 trichoid sensilla are present, each with a length of 100–150 μm and featuring a terminal pore. The angle to the antennal cuticle is more or less 90°. The shaft bears numerous spiral ribs and it is basally encompassed by a crescent-shaped socket, and two small glandular pores. Sensory cells possess short dendritic outer segments exhibiting tubular bodies (mechanoreception) and long dendritic outer segments that project towards the terminal pore (chemoreception) . In addition, about 80 sensilla basiconica, 80 s. brachyconica, and 240 s. microtrichodea are present on each antenna.
Primarily, trichoid sensilla were found on walking and ultimate legs. We classified subtypes based on differences in length of their shafts (macro-, meso-, and microtrichoid). The subtype s. macrotrichodeum II with a multidentate socket was exclusively found on walking legs. Apart from the dentate socket, s. macrotrichodea II are slightly smaller than the dominant s. macrotrichodea I (present on all legs). As we detected no terminal pores, an exclusive mechanoreceptive function is presumed. Type I s. mesotrichodea are present on all legs and are assumed to function exclusively as mechanoreceptors. Type II s. mesotrichodea II are only present on legs 13–15 and are the only sensilla on the legs which are associated with two glandular pores, a feature that is usually found on antennal s. trichodea . The only similarity between s. mesotrichodea I and II is their shaft length. Other features like the presence of a terminal pore, the association with two glandular pores, as well as the exclusive distribution clearly differentiates these subtypes. Similar sensilla were not described on the antennae of L. forficatus . The presence of a terminal pore clearly indicates a chemosensory function. The sponge-like, serrated appearance of the shaft between the ribs might suggest an additional olfactory function. Interestingly, this morphology is similar to the beak-like sensilla in Scutigera coleoptrata [55, 63] that are thought to serve an olfactory function in scutigeromorph centipedes. Whether these cuticular specializations are associated with pores or a specialized cuticle remains unclear without ultrastructural investigation. The location of s. mesotrichodea II interspersed between the pores of telopodal glands also indicates that these sensilla might function as a quality assessment of telopodal gland secretion. As no sexual dimorphic characteristics were found, a role in chemical recognition/communication is unlikely, but cannot be ruled out. Sensilla microtrichodea are present on all legs and the antennae , with minor differences in shaft length. Antennal s. microtrichodea are present in all investigated pleurostigmophoran centipedes arranged at the base of antennomeres, and are thought to function as proprioceptors providing information on the position of the antenna [5, 56, 57, 62]. Only in L. forficatus were antennal s. microtrichodea described without a terminal pore and lacking long dendritic outer segments . However, on walking and ultimate legs, we detected terminal pores, thus an additional chemoreceptive function can be assumed.
Apart from trichoid sensilla, small sensory cones were found on all legs, on the claw, as well as the distal femur and tibia. The very short shaft, the slightly sunken appearance, and the presence of a terminal pore are (at least) comparable to sensory cones in Scolopendromorpha [20, 58]. The term sensory cone was introduced to describe a very small, typologically incomprehensive class of antennal sensilla . Therefore, it is unwarranted to refer to them as homologous sensilla. The presence of similar cones on the claw was also detected in Lithobius obscurus and Craterostigmus tasmanianus (termed ‘basiconic sensilla’ ), and a homology between those sensilla in Lithobiidae and either the subsidiary spine or posteroventral spine of Henicopidae was proposed . However, based on size, shaft surface and insertion, we reject this terminology to avoid unjustified homology hypotheses (compare ). While they may function as contact-chemoreceptive sensilla on the claw given their location and presence of a terminal pore, they will hardly come into contact with any substrate on the telopodite. Their function is thus uncertain, but may be thermo- or hygroreceptive in nature, given their external-morphological similarity to thermoreceptors in hexapods [65, 66]. Bauer  showed that in L. forficatus hygroreception is mainly located on the tarsi, while thermoreception is mostly mediated by basal antennomeres [16, 34]. Again, without complementary ultrastructural investigations, any functional conclusions are a matter of speculation.
Neuroanatomy of the ultimate leg-associated ganglion
As demonstrated by Schendel et al. , walking leg ganglia 1–14 of L. forficatus are of rather uniform morphology and architecture which is in accordance with previous investigations and descriptions [34, 67,68,69,70]. The ganglion associated with the ultimate legs (G15), however, shows a variety of anatomical peculiarities.
Only in the ganglion of the ultimate leg, leg nerves are directed posteriad, which corresponds to the transformation of the coxa and posture of the respective telopodites. As in the walking leg ganglia, G15 houses several large projections and termination sites associated with leg nerves N4 + N5. However, the ventral neuropilar domain (vnd) of G15 has a multiglomerular organization as opposed to a uniglomerular organization of G10 (compare Fig. 9e, f, and ). Moreover, the volume of the G15 vnd is at least five times as large as the counterpart in the walking leg-associated G10. Another substantial architectural difference lies within the ventral anteriad projection (vap). While in G10 the vap remains as an inconspicuous bundle of longitudinal fibers giving rise to the vnd, the vap of G15 is condensed to a series of discrete longitudinal, parallel lamellae, hence termed the ventral lamellate domain (vld). In general, tetraconate (Crustacea and Hexapoda) vnc ganglia associated with appendages possess particularly organized neuropilar domains. For example, the thoracic ganglia of Hexapoda feature sensory neurites and associated neuropils in the ventral part of the ganglion, the so-called ventral association center (vac) [25, 71,72,73]. This organization and corresponding tract patterns are also found in crustaceans (e.g. [74,75,76]). Within the vac, different sensory modalities terminate in different regions, e.g. afferents from mechanoreceptive sensilla project to distinct regions of the neuropil and given the specific sensory modality, an ordered structure of neuropilar areas is usually observed along a gradient [25, 26, 77]. While in locusts and crayfish the vac is either fused or strongly interconnected by commissural fibers, each hemiganglion of L. forficatus features distinct neuropilar regions without obvious contralateral connections.
Intriguingly, the neuropilar domain in the G15 of L. forficatus is organized into two domains (ventral neuropilar and lamellate domains). The principle of a bipartite architecture in processing afferents is mostly known for and well documented in the mandibulate deutocerebrum. Innervated by a variety of antennal sensilla of different modalities, it is typically composed of a glomerular neuropil processing chemosensory information and (at least one) mechanosensory neuropil [22, 25, 27, 28, 31, 32], which is also the case in the pectines-associated neuromere of scorpions [78, 79]. In terms of sensory neurobiology, glomeruli are (with few exceptions) usually associated with the processing of chemoreceptive stimuli in a chemotopic manner (e.g. [25, 28, 29, 32]), whereas a highly ordered rectilinear, striated and multilayered arrangement of fibers with perpendicular arborizations can be attributed to the processing of mechanosensory information in a somatotopic manner [25, 80,81,82]. The bipartite organization of the deutocerebrum of Lithobius forficatus features the deutocerebral chemosensory lobes (typically composed of 43 glomeruli) and the corpus lamellosum (mechanosensory neuropil) that is composed of at least four neuropilar lamellae [19, 22, 67]. Although numbers are much lower, the organization and the specific architecture of the ventral neuropilar and lamellate domains show a strong resemblance to the deutocerebral neuropils. Taking the characteristics and elaboration of primary processing centers associated with antenna, walking leg and the ultimate leg into account, it is evident that as a general consequence, a pronounced input from cuticular sensilla shapes primary processing neuropils. The increase in total volume and in volume of the associated neuronal substrate in G15, the increase in glomerular number, and the differentiation into lamellae thus indicates an enhanced sensory performance that is mirrored by the higher abundance of cuticular sensilla. Given the incremental enlargement of legs 12–14 as well as the augmentation with sensilla and epidermal glandular pores, a likewise gradual increase in sensory performance as well as volume of associated processing centers in ventral nerve cord ganglia is to be expected, but requires further examination.