The velvet worm brain unveils homologies and evolutionary novelties across panarthropods

Background The evolution of the brain and its major neuropils in Panarthropoda (comprising Arthropoda, Tardigrada and Onychophora) remains enigmatic. As one of the closest relatives of arthropods, onychophorans are regarded as indispensable for a broad understanding of the evolution of panarthropod organ systems, including the brain, whose anatomical and functional organisation is often used to gain insights into evolutionary relations. However, while numerous recent studies have clarified the organisation of many arthropod nervous systems, a detailed investigation of the onychophoran brain with current state-of-the-art approaches is lacking, and further inconsistencies in nomenclature and interpretation hamper its understanding. To clarify the origins and homology of cerebral structures across panarthropods, we analysed the brain architecture in the onychophoran Euperipatoides rowelli by combining X-ray micro-computed tomography, histology, immunohistochemistry, confocal microscopy, and three-dimensional reconstruction. Results Here, we use this detailed information to generate a consistent glossary for neuroanatomical studies of Onychophora. In addition, we report novel cerebral structures, provide novel details on previously known brain areas, and characterise further structures and neuropils in order to improve the reproducibility of neuroanatomical observations. Our findings support homology of mushroom bodies and central bodies in onychophorans and arthropods. Their antennal nerve cords and olfactory lobes most likely evolved independently. In contrast to previous reports, we found no evidence for second-order visual neuropils, or a frontal ganglion in the velvet worm brain. Conclusion We imaged the velvet worm nervous system at an unprecedented level of detail and compiled a comprehensive glossary of known and previously uncharacterised neuroanatomical structures to provide an in-depth characterisation of the onychophoran brain architecture. We expect that our data will improve the reproducibility and comparability of future neuroanatomical studies. Supplementary Information The online version contains supplementary material available at 10.1186/s12915-021-01196-w.

Studies of the nervous system are often conducted to gain insights into evolutionary events [20][21][22][23][24]. For most major arthropod groups, in-depth investigations in recent times have led to a clear picture of the anatomical and functional organisation of their nervous systems. The nervous systems of onychophorans, in contrast, were studied mainly at the end of the nineteenth (e.g. [25,26]) and beginning of the twentieth centuries (e.g. [27][28][29][30][31]), with only few recent exceptions (e.g. [13,15,32,33]). Accordingly, the range of techniques used to investigate the onychophoran nervous system is largely limited to classical methods. Furthermore, most studies have focussed on specific aspects of the nervous system, such as neurogenesis and neural development [13,15,[34][35][36][37] or on the localisation of specific structures and neurons using selective markers [14,18,33,[38][39][40][41][42]. Thus, a comprehensive overall picture is still lacking, especially in the light of the array of new methods currently available.
Unfortunately, further research on the onychophoran nervous system is currently hampered by inconsistencies in nomenclature and interpretations between different authors. A review of the literature reveals cases where a single structure is given different names by various authors. For example, the same part of the brain referred to as "arcuate body" by Strausfeld et al. [42] is named "central body" by Schürmann [32], "Zentralkörper" by Hanström [31], "corpus striatum" by Holmgren [27], and "bourrelet dorsal" by Saint-Remy [26]. In addition, the same term is sometimes applied to different structures, e.g. the term "pedunculus" to either a connection between mushroom body and central body [32], or a structure anterior to the mushroom body lobes [33], or even the mushroom body lobes themselves [43].
Similarly, there are numerous inconsistencies in descriptions of specific traits of the onychophoran brain. For example, the pathway of optic tracts and the number of optic neuropils as reported by Strausfeld et al. [33] and by Mayer [44]. In other cases, contradictions are evident even within a single study (e.g. "bridge" in [32]), or questionable characters are used to claim homology of structures with those in other animal groups (e.g. "frontal ganglion" in Cong et al. [45,46], but see Mayer et al. [47] for an opposing view).
Finally, previous investigations of the onychophoran brain suggest that various structures remain to be described and characterised (e.g. an "anterior neuropil" referred to in [38,48]), and emphasises that current understanding of the onychophoran brain is far from complete. Furthermore, the hitherto known architecture of the onychophoran brain has not been critically reappraised, although some features have already been questioned (e.g. the number of mushroom body lobes in [32]), while other features have been reported, but as yet not analysed in detail. For example, the onychophoran central body has been described to be stratified [26,27,42,48], but the number and architecture of strata have not been addressed. Therefore, a revision of the onychophoran brain followed by a uniform designation of its components is needed in order to bring consensus to its anatomy and to clarify unresolved controversies.
In the current study, we report new insights into the structural organisation of the brain of the onychophoran Euperipatoides rowelli (Peripatopsidae), gained from using a range of specific markers against acetylated αtubulin and the presynaptic protein synapsin in conjunction with high-resolution confocal microscopy. We then used the data from immunolabelled whole-mount preparations of the brain to three-dimensionally reconstruct its architecture and visualise previously unknown volumetric and spatial relationships among neurite regions. We supplement our data with three-dimensional (3D) reconstructions based on Azan-stained histological serial sections and X-ray micro-computed tomography (μCT). These analyses allowed us to gain an in-depth understanding of the onychophoran brain anatomy, revealing all major neuropil regions and allowing for their detailed characterisation. Together with neuroanatomical information available in the literature for the onychophoran nervous system, we compiled a glossary with proposed standardised terms and synonyms in order to bring consistency for future studies.
The multi-methodological approach enabled us to characterise in greater detail known neuropil regions and tracts and describe neuropil regions that were previously unknown. This in turn allows for a more comprehensive comparison of onychophoran brain structures to corresponding features in arthropods. Finally, we interpret our results to address several controversies surrounding the onychophoran brain. This reappraisal of the onychophoran nervous system should facilitate future comparative studies, and thereby shed light on the evolution of the nervous system in Panarthropoda and the brain structure in their last common ancestor.

Results
We used a multiscale approach to gain a broader and detailed picture of the brain structure in the onychophoran E. rowelli (Figs. 1,2,3,4,5,6,7,8,9,10,11,12 and Additional file 1: Fig. S1, Additional file 2: Fig. S2, Additional file 3: Fig. S3, Additional file 4: Fig. S4, Additional file 5: Fig. S5, Additional file 6: data S1, Additional file 7: Data S2, Additional file 8: Data S3, Additional file 9: Data S4, Additional file 10: Data S5, Additional file 11: Data S6). X-ray μCT scans of specimens contrasted with osmium tetroxide and ruthenium red proved useful for visualising the overall shape of the onychophoran central nervous system within the whole animal. In addition, histological sections of entire heads in conjunction with Azan staining provided detailed structural information on the central nervous system with its individual neuropil areas without dissection artefacts. Finally, an in-depth picture of brain architecture was provided by synapsin immunolabelling of dissected brains to reveal synaptic terminals and, hence, details of neuropil areas, while acetylated α-tubulin immunolabelling showed pathways traversed by fibres of the nervous system.
Neuroanatomical terms suggested herein as a standard nomenclature for studies of the onychophoran nervous system are preceded by an arrow (→). The corresponding explanations, definitions and synonyms are compiled in a glossary (Table 1). Whenever possible, the Fig. 1 Organisation of the central nervous system of E. rowelli. 3D renderings based on SRμCT (a, b) and Azan staining (c, d). Dorsal is up in all images. a Overview. b Detail of suprarectal commissure. Asterisk indicates passage of genital duct under ventral nerve cord. c, d Frontal and lateral views of anterior part of nervous system. Cell body rind (blue) partly covers neuropils, fibre tracts, commissures and peripheral nerves (light-brown). c', c" Selected frontal slices from (c) at the level of the brain (c') and ventral nerve cord (c"). ANC, antennal nerve cord; BR, brain; CNC, circumpharyngeal nerve cord; DC, dorsal commissures; EY, eye; HO, hypocerebral organ; JAN, jaw nerve; LEN, leg nerve; LE, leg; LIN 1, LIN 2, lip nerve1 and 2; PHN 1, pharyngeal nerve 1; SP, slime papilla; SPN, slime papilla nerves; SRC, suprarectal commissure; VC, ventral commissure; VNC, ventral nerve cord nomenclature follows the terms and definitions suggested by Ito et al. [49] for insects and Richter et al. [20] for invertebrates in general. Note that corresponding terms do not necessarily imply homology with homonymous (i.e. like-named) structures in arthropods and other animals.

General features of the nervous system
The central nervous system of E. rowelli consists of an anterodorsal →brain composed of a →protocerebrum and a →deutocerebrum, a pair of →circumpharyngeal nerve cords and the →ventral nerve cords (Fig. 1a, c, d and Additional file 1: Fig. S1a-c). The brain is associated ventrally with a pair of →hypocerebral organs (Fig. 1c). The ventral nerve cords run along each side of the body and are connected with each other via a posterior →suprarectal commissure looping above the hindgut (Fig. 1a, b). In addition, the nerve cords are linked along the body's length by two types of serially arranged commissures: the →ventral commissures and the →dorsal commissures (Fig. 1c, d). The ventral commissures connect the two ventral nerve cords across the ventral midline, whereas the dorsal commissures project laterally, then cross one of the two →dorsolateral trunk nerves and finally connect the ventral nerve cords along the dorsal midline above the →heart nerve (cf. [14]). The dorsal commissures are missing in pedal regions, where instead a pair of segmental →leg nerves emanates from the ventral nerve cords on each body side (Fig. 1c, d). Each leg is supplied by an anterior and a posterior leg  Organisation of brain neuropils in E. rowelli revealed by anti-synapsin immunolabelling. Confocal laser scanning micrographs of frontal vibratome sections. Dorsal is up in all images. Anti-synapsin immunoreactivity (glow) and DNA labelling (blue) in anterior (a), median (b), and posterior sections (c) through the left brain hemisphere. Note strong immunoreactivity in the ponticulus in a and near lack thereof in b. Also note strongest immunoreactivity in μ-lobe as opposed to weakest immunoreactivity in δ-lobe (c) of mushroom body. Arrowheads (b) point to discontinuous row of somata between mushroom body lobes. ANP, antennal neuropil; CB, central body; CB-IL, inner lamina of central body; CB-OL, outer lamina of central body; CBR, cell body rind; δ, dorsal or δ-lobe of mushroom body; E-ONP, extracerebral part of optic neuropil; FB, frontal body; HO, hypocerebral organ; λ, lateral or λ-lobe of mushroom body; μ, median lobe or μ-lobe of mushroom body; MG, macroglomerulus; OLF 4, olfactory fascicle 4; OLG, olfactory glomeruli; PDN, posterior dorsal neuropil; PNT, ponticulus; SAN, subantennal neuropil; v, ventral or v-lobe of mushroom body; VH, ventral horn; PVN, posterior ventral neuropil. Scale bars: 50 μm nerve. The same holds true for the slime papillae-specialised limbs of the third body segment-that are innervated by an anterior and a posterior →slime papilla nerve (Fig. 1d). The two slime papilla nerves are preceded by the →lip nerve 3 and succeeded by the →salivary gland nerve (cf. [13]). The circumpharyngeal nerve cords differ from the ventral nerve cords in that they are oriented dorsoventrally rather than anteroposteriorly and are not associated with commissures, except for a single, prominent →subpharyngeal commissure ( Fig. 1a and Additional file 1: Fig. S1a, c).
Volume measurements reveal that the central nervous system of E. rowelli occupies~3% of the total body volume, of which the brain constitutes only~0.7% and the ventral nerve cords together with the circumpharyngeal nerve cords~2.3% ( Table 2). The central nervous system exhibits a clear subdivision into central neuropil and a peripheral cell body rind. Most somata are located in a ventral cell body rind and are largely lacking in most dorsal areas of the central nervous system, particularly in the ventral nerve cords (Fig. 1c, d and Additional file 1: Fig S1b, c and Additional file 6: Data S1, Additional file 7: Data S2). The brain consists of two ovoid, bilateral brain hemispheres fused along the midline (Fig. 1a, c and Additional file 1: Fig. S1b). Besides the posterior circumpharyngeal nerve cords, the most prominent bundles of neurites leaving the brain are the bilateral, anterior neuropils extending into the antennae, i.e. the →antennal nerve cords (Fig. 1a, c and Table 1). Among additional structures associated with the brain are a pair of anterolateral →optic neuropils supplying the eyes, a pair of ventral →tongue nerves projecting into the tongue (cf. [13]), and two pairs of lip nerves supplying the mouth (the mediodorsal →lip nerves 1 (Fig. 1c, d) and the posterolateral →lip nerves 2 ( Fig. 12a-c)), whereas the lip nerves 3 are associated with the circumpharyngeal nerve cords ( Fig. 12c; cf. [40]). In addition, the brain is associated with a pair of lateral →jaw nerves extending ventrally into the jaw musculature, and a pair of posterior pharyngeal nerves (→pharyngeal nerves 1) emerging at the indentation between the two brain hemispheres and entering the dorsolateral wall of the pharynx (Figs. 1c and 12 a, b). A second pair Fig. 6 Organisation of mushroom body in E. rowelli. Three-dimensional reconstruction based on anti-acetylated α-tubulin immunolabelling (a) and horizontal optical sections from stacks of confocal micrographs of brains stained against acetylated α-tubulin (b) and synapsin (c). Anterior is up and median is left in all images. a Volume rendering of mushroom body in ventral view. Note connection with central body via ponticulus. b Detail of fascicles of globuli cells with two main branches (dotted line) giving rise to four digitiform branches (arrows). c Detail of the four lobes, all of which are sectioned in inset. Note strongest synapsin immunoreactivity along μ-lobe. ANP, antennal neuropil; CB, central body; δ, dorsal or δ-lobe; FGC, fascicles of globuli cells; λ, lateral or λ-lobe; μ, median or μ-lobe; OLL, olfactory lobe; PED, pedunculus; PNT, ponticulus; v, ventral or vlobe. Scale bars: 50 μm (→pharyngeal nerves 2) is associated with the circumpharyngeal nerve cords and enters the pharynx laterally (Table 1; cf. [39]).

Organisation of the brain and associated structures
The brain of E. rowelli is comprised of an outer cell body rind and an inner neuropil region (Fig. 1c, d). The neuropil region constitutes 40-48% of the total brain volume, depending on the method used for fixation and further specimen treatment (Table 3). It is organised into distinct tracts, layers and neuropils (Figs. 2,3,4,5,6,7,8,10,11,12 and Additional file 8: Data S3, Additional file 9: Data S4, Additional file 10: Data S5, Additional file 11: Data S6), which are characterised in the following.

Frontal body
The →frontal body is a prominent anteriormost protocerebral midline neuropil occupying 3 to 4% of the total volume of the brain (Figs. 2, 3 and 5 c-e; Table 3). It borders anteriorly and dorsally the cell body rind and posteriorly the →central body but fuses laterally with the →antennal neuropil and ventrally with the →median lobes and →ventral horns (Figs. 4a, c, 11 and 12 and Additional file 8: Data S3, Additional file 9: Data S4, Additional file 10: Data S5, Additional file 11: Data S6). Parallel running fibres (or neurites) traversing the central body, connect the →ventral neuropil with the frontal body (Additional file 8: Data S3 and Additional files 9: Data S4). Anti-tubulin immunolabelling revealed an additional anteroventral connection to a branch Fig. 7 Organisation of brain commissures in E. rowelli. Three-dimensional reconstructions based on anti-acetylated α-tubulin immunolabelling (ac) and horizontal optical sections of confocal micrographs of brains stained for acetylated α-tubulin (a'-c'). Anterior is up in (a, a', b, b' and c'), top right corner in c and dorsal is up in c. Reconstructed brain commissures are depicted in green and shown in spatial relation to mushroom body (grey in a, b and c). Brain commissures 1 (a; ventral view) and 2 (b; dorsal view) originate from antennal neuropil and proceed adjacent to mushroom body lobes, whereas brain commissure 3 (c; posterolateral view) proceeds dorsal to mushroom body lobes. Arrowheads point to corresponding commissures. ANP, antennal neuropil; BRC 1, BRC 2, BRC 3, brain commissure 1 to 3; CBR, cell body rind; δ, dorsal or δ-lobe of mushroom body; FGC, fascicles of globuli cells; λ, lateral or λ-lobe of mushroom body; μ, median or μ-lobe of mushroom body; PED, pedunculus; PNT, ponticulus; PVN, posterior ventral neuropil; v, ventral or v-lobe of mushroom body. Scale bars: 50 μm associated with the antennal neuropil (Additional file 8: Data S3 and Additional files 9: Data S4). Although a histological subdivision of the frontal body into distinct layers is not evident, anti-synapsin labelling reveals highly immunoreactive strata in its posterodorsal and anteroventral regions (Fig. 5c, e; Table 1).

Central body
The →central body is the dorsalmost protocerebral midline neuropil occupying 2.4 to 3.0% of the total volume of the brain (Figs. 2, 3, and 5a-e; Table 3). It is an arcuate (lateral sides bent anteriorly), stratified structure situated posterior to the →frontal body. The central body can be subdivided into an →anterior division and a →posterior division. Anti-tubulin and anti-synapsin immunolabelling shows that both the anterior and posterior divisions are further stratified into an →inner lamina and an →outer lamina. Inner and outer laminae of Fig. 8 Organisation of olfactory and glomerular structures in E. rowelli. Three-dimensional reconstruction based on anti-synapsin immunolabelling (a) and horizontal optical sections of confocal micrographs of brains stained for synapsin (b) and acetylated α-tubulin (c-f). Anterior is up in all images. a, b Olfactory fascicles 1-3 adjacent to mushroom body's lobes. Olfactory fascicle 4 connects to v-and δ-lobe (see Additional file 10: Data S5 and Additional file 11: Data S6). a-d Size and arrangement of olfactory glomeruli (a and c). Note their similar size and presence of a macroglomerulus. e, f Cap-like arrangement of microglomeruli anterior to olfactory glomeruli. ANP, antennal neuropil; CBR, cell body rind; δ, dorsal or δ-lobe of mushroom body; FB, frontal body; λ, lateral or λ-lobe of mushroom body; LO, lobes of mushroom body; MG, macroglomerulus; MIG, microglomeruli; μ, median or μ-lobe of mushroom body; OLF 1, OLF 2, OLF 3, OLF 4, olfactory fascicle 1 to 4; OLG, olfactory glomerulus; OLL, olfactory lobe; ONP, optic neuropil; v, ventral or v-lobe of mushroom body. Scale bars: 50 μm (b and c) and 20 μm (d-f) Fig. 9 Diameter range of glomeruli in E. rowelli. The diameters of 65 out of a total of~80 glomeruli of one brain hemisphere were measured. Dashed red line indicates average diameter of glomeruli; dashed blue line demarcates diameter of macroglomerulus both divisions are fused with each other at the lateral ends of the central body and therefore show an inverted anteroposterior arrangement in the anterior and posterior divisions (Fig. 5c-e). An additional posterior stratum is associated with the posterior division (Fig. 5c), from which it is delimited by a discontinuous layer of somata. Two relatively small, bilaterally symmetric, cylindrical →auxiliary neuropils are embedded ventrally into the inner lamina of the central body (Fig. 5a, d and e). Each auxiliary neuropil is linked via a bundle of neurites with the →antennal neuropil of the ipsilateral brain hemisphere (Additional file 8: Data S3 and Additional files 9: Data S4).

Mushroom bodies
The →mushroom body is a bilaterally paired neuropil located lateroventrally in the →protocerebrum (Figs. 2, 3 and 6 a-c). With 4.2-5.3% of the total brain volume, the mushroom bodies together constitute the largest individual neuropils of the brain (Table 3). They are embedded in the cell body rind of each brain hemisphere and adjoin dorsally the →subantennal neuropils (Fig. 4b). Each mushroom body is composed of three major subunits including (i) the →fascicles of globuli cells, (ii) the →pedunculus, and (iii) the four →lobes (Fig. 6a-c). The fascicle region comprises the anteriormost part of the mushroom body and lacks the presence of anti-synapsin labelled synaptic terminals ( Fig. 2; Additional file 10: Data S5, Additional file 11: Data S6). It receives axon fibres from the →globuli cells located in the anterior cell body rind that converge to numerous digitiform branches. These smaller branches, in turn, anastomose to four larger branches that join the pedunculus (Fig.  6a-c and Additional file 2: Fig. S2a-c). The pedunculus gives rise posteriorly to four lobes that are sickle-shaped, laterally convex structures. Based on their position, we distinguish a dorsal (δ-lobe), lateral (λ-lobe), medial (μlobe) and a ventral lobe (v-lobe; Fig. 6a, c). While the δlobe and the v-lobe are stacked on top of each other, the μ-lobe and the λ-lobe encompass them medially and laterally, respectively. All four lobes are separated from each other by thin, discontinuous layers of somata (Fig.  4b, c). We were unable to determine whether these belong to glial cells. Immunoreactivity against synapsin is consistently stronger in the μ-lobe and weaker in the δlobe than in the v-lobe and the λ-lobe (Fig. 4b, c).

Ponticulus
The →ponticulus extends from the →pedunculus of the →mushroom body to the →outer lamina of the →central body (Fig. 6a). The ponticulus shows a mixed tractlike/neuropil-like organisation, as the anti-synapsin signal is high near the pedunculus but progressively decreases and finally disappears towards the central body ( Fig. 4a, b), whereas acetylated α-tubulin is distributed along its entire length (Figs. 2, 3; Additional file 8: Data S3 and Additional file 9: Data S4). Moreover, synapsin is not distributed evenly but shows a granular arrangement (Fig. 4a). The ponticuli of both brain hemispheres con-stitute~0.1% of the total brain volume ( Table 3).

Brain commissures 1-3
The brain of E. rowelli exhibits three, individually identifiable branched commissures, identified by Azan-labelled sections and anti-tubulin immunostaining: the →brain commissures 1-3 (numbered from posterior to anterior), formed by contralaterally projecting bundles of axon fibres crossing the →protocerebrum (Figs. 2, 3 and 7a-c; Additional file 2: Fig. S2a-c). These prominent commissures occupy together~0.6% of the total volume of the brain ( Table 3). Two of them, the brain commissures 1 and 2, crosslink the →antennal neuropils of each brain hemisphere, whereas the brain commissure 3 connects the →subantennal neuropils of both brain hemispheres ( Fig. 7a-c). None of these commissures are directly associated with the →mushroom bodies, from which they are spatially separated by fibres belonging to the →posterior ventral neuropil and subantennal neuropil (Fig.  7a-c and Additional file 3: Fig. S3a-f). The brain commissure 1 originates from the antennal neuropil, then passes through the space between the →pedunculus and the →ponticulus and follows the μ-lobe of the mushroom body to finally cross the midline of the brain (Fig.  7a). The brain commissure 2 fasciculates immediately after leaving the antennal neuropil into three branches (Fig. 7b). All three branches flank the posterior margin of the pedunculus and proceed further posteriorly above the →lobes of the mushroom body. At this point, the lateral branch terminates, whereas the other two branches merge and continue as a single commissure towards the contralateral brain hemisphere. The brain commissure 3 is situated anterior to the brain commissure 2 and connects the contralateral subantennal neuropils ( Fig. 7c and Additional file 8: Data S3 and Additional files 9: Data S4). After leaving the subantennal neuropil, this commissure passes dorsal to the μlobe of the mushroom body and then splits into two branches that finally cross the posterior ventral neuropil to the contralateral side.

Antennal pathways
The →antennal nerve cords, which are associated with the protocerebral antennae [13,15], enter the brain frontally and continue without any recognisable neuroanatomical border into the two →antennal neuropils (Fig. 11a, b) and the →olfactory lobes (Figs. 2, 3 and 8ad; cf. [15]). Together with the →mushroom bodies, the olfactory lobes represent the ventralmost neuropils of the brain. Each olfactory lobe is an elongated, irregular assemblage of~80 subunits, the →olfactory glomeruli, which are largely, albeit not exclusively, located in the periphery of the lobe (Fig. 8a-d). The olfactory glomeruli from both olfactory lobes constitute together 0.5% of the total brain volume ( Table 3). Each olfactory glomerulus represents a synaptic complex with numerous synaptic sites (Additional file 4: Fig. S4a, b). The olfactory glomeruli appear spherical, ellipsoid or ovoid in shape (Fig. 8a-d). They also vary in size, ranging in diameter from~9 to 40 μm (average diameter 19 μm),  with most glomeruli falling into the range of 10-25 μm (Fig. 9). The median posterior half of each olfactory lobe harbours a single, particularly large glomerulus, which measures 40 μm in diameter and is therefore referred to as the →macroglomerulus (Figs. 8a, c and 9). Volume measurements reveal that the macroglomerulus is~7 times larger than a medium sized olfactory glomerulus (Table 4). Synapsin immunolabelling further revealed four →olfactory fascicles (1-4) associated with each olfactory lobe, three of which (1-3) are thinner (diameter 5.5 μm) than the fourth one (diameter 20.3 μm; Fig. 8a and b). While the olfactory fascicles 1-3 are associated with the anterior part of the olfactory lobe and join the ipsilateral →median lobe of the →brain (Additional file 10: Data S5 and Additional file 11: Data S6), the olfactory fascicle 4 originates further posteriorly and fuses with the v-lobe and the δ-lobe of the mushroom body ( Fig. 8a, b and Additional file 10: Data S5 and Additional file 11: Data S6).

Microglomeruli
We further identified microglomerular complexes in the anterolateral part of each brain hemisphere, where they are embedded in the →subantennal neuropil. Each complex consists of~40 units, the →microglomeruli, which show a cup-shaped arrangement anterior to the →olfactory lobes and are revealed by both, anti-synapsin and anti-acetylated α-tubulin immunolabelling (Figs. 2, 3 and 8 a, b, e, f; Table 1). Each microglomerulus is spherical in shape and represents a complex of numerous, prominent synaptic sites, as revealed by anti-synapsin immunolabelling (Additional file 4: Fig. S4c and d). A single microglomerulus measures 4.5-6.5 μm in diameter and comprises only one-twentieth the volume of a mid-sized →olfactory glomerulus; all microglomeruli together constitute only 0.01-0.02% of the total brain volume (Tables  3 and 4 and Additional file 4: Fig. S4c and d).

Visual pathways
The →optic neuropil is comprised of extra-and intracerebral parts that label consistently with anti-synapsin (Fig. 10a). The extracerebral part appears as a cupshaped structure within each of the two eyes and continues via a short, bottleneck-shaped connection into the brain (Figs. 2, 3 and 10 a and b). The intracerebral part of the optic neuropil then splits into an anterior and a posterior branch (Fig. 10a, b). The anterior branch projects anteromedially between the →antennal neuropil and the →subantennal neuropil, but could not be traced further in our datasets (Additional file 8: Data S3, Additional file 9: Data S4, Additional file 10: Data S5, Additional file 11: Data S6). The posterior branch crosses the →antennal neuropil and passes dorsally towards the →central body (Fig. 10a, d). It then splits into an anterior and a posterior bundle of fibres (Fig. 10a, e). The anterior bundle runs towards the →anterior division of the central body, passing it ventrally without entering before disappearing in the fibre meshwork of the →central neuropil (Fig. 10a, c). The posterior bundle proceeds towards the →posterior division of the central body and enters the associated posterior stratum, after which the anti-synapsin signal vanishes (Fig. 10a, e). The optic neuropil constitutes only~0.2% of the total brain volume (Table 3).

Discussion
Our multiscale approach unveiled the organisation of the onychophoran brain at different hierarchical levels of detail (Fig. 13). While the analysis of SRμCT data and Azan-stained sections of entire heads did not require brain dissection and, thus, revealed the least distorted shape of the individual structures, anti-acetylated αtubulin and anti-synapsin immunolabelling of dissected brains were most informative for demonstrating the intracerebral elements and their components. We used this information for reconstructing the architecture of the brain in E. rowelli and refining the neuroanatomical terminology for Onychophora ( Table 1). The results help to clarify persisting controversies surrounding the identity and homology of specific structures and regions in the onychophoran brain.
This uncertainty might be due to considerable variation across arthropods. For example, the calyx is lacking in myriapods [147], paleopteran and some neopteran insects [118], and partially also in stomatopod crustaceans [152]. Moreover, a structure called the hemiellipsoid body-a potential homologue of the mushroom body calyces (Bellonci 1982 apud [153,154]) or the mushroom body itself [155,156]-has so far been identified in only three crustacean clades: Malacostraca [152,[157][158][159][160], and Remipedia [161], which both lack the Table 3 Volumes of cerebral structures in E. rowelli relative to total brain volume (TBV) based on SRμCT (entire head scan), and light (LM; serial sections of entire head) and confocal microscopy (CLSM; whole-mount preparations of brains) Total volumes of structures from both brain hemispheres were estimated using corresponding numbers of voxels 2 Excluding mushroom body's fascicles of globuli cells mushroom body pedunculi and lobes, as well as Cephalocarida [162]. This suggests either an independent origin in these clades or an evolutionary loss in other crustaceans. In a study using an antibody (anti-DC0) against the catalytic subunit of protein kinase A [163,164], which selectively labels the columnar neuropils in the mushroom bodies of insects (e.g. [165,166]) and other arthropods [147,152,158,159], Strausfeld et al. [148] provided evidence that the mushroom bodies, or at least the mushroom body calyces, generally occur in crustaceans. This suggests that these structures are homologous at least across arthropods [147]. Hence, the question arises whether mushroom bodies were present in the last common ancestor of panarthropods. While comparable structures have not been identified yet in tardigrades (e.g. [85,167]), mushroom bodies (corpora pedunculata) were recognised early in onychophorans [27,28,30,31]. As in arthropods, they are located in the protocerebrum, a homologous brain region in the two clades [15,33,42,48,54,69,168]. Moreover, they show a subdivision into a fascicle region, a pedunculus and four lobes ( Fig. 13; see also [33]). The fascicle region is associated with characteristic globuli cells, which most likely correspond to the globuli cells that give rise to processes within the calyces of arthropods [23,143].
However, as opposed to most insects and crustaceans, the globuli cells in onychophorans do not contribute to a cup-like structure called calyx [119]. As in calyx-less insects [118,119], the globuli cells of onychophorans instead contribute to fibre fascicle that converge to the pedunculus. This resembles also the situation in the calyx-less mushroom bodies of myriapods, paleopterans and some neopterans, where globuli cells form neurite bundles that extend into the pedunculus [118,147]. It is largely accepted, however, that lack of calyces has occurred independently in paleopterans and some neopterans within the Hexapoda lineage [118]. In chelicerates, mushroom bodies including calyces (or mushroom body heads) have been described for several species (e.g. [150,[169][170][171][172]). However, the structure of their calyces is substantially different from the typical insect/crustacean calyces [e.g. 169]. Considering that the mushroom body calyces, as defined for insects [119] and crustaceans, are absent in chelicerates, myriapods and onychophorans, one may conclude that this structure first emerged in the pancrustacean lineage [118] and that onychophoran mushroom bodies most likely resemble the ancestral condition within Panarthropoda.
Accordingly, positional and structural similarities currently speak in favour of the proposed homology of mushroom bodies in onychophorans and arthropods [32,42,54]. Even so, analysing the expression patterns of genes that are involved in learning [147] might assist in substantiating this hypothesis. Localising anti-DC0immunoreactive centres in the onychophoran brain might further help to identify homologous regions, as shown in numerous comparative studies across pancrustaceans (e.g. [148]) and other arthropods [147].
Besides these similarities, our study also revealed differences and previously unnoticed details in the organisation of mushroom bodies in velvet worms. One of these details concerns the number of mushroom body lobes. While only three of them were reported previously ( [26,27,31,33,42,48,80]; but see 32 for suggested existence of additional compartments), we identified four lobes (δ, λ, μ and ν) in each mushroom body of E. rowelli with three independent methods: Azan staining, anti-synapsin and anti-tubulin immunolabelling. Notably, while the μ-lobe exhibits the strongest antisynapsin immunoreactivity, suggesting dense synaptic connectivity, the signal is nearly undetectable in the δlobe. Interestingly, the μ-lobe of E. rowelli also expresses high levels of arthropsin, an opsin (G protein-coupled receptor) with unknown function [86]. Immunolocalisation of pigment-dispersing factor neuropeptides, Er-PDF-I and Er-PDF-II, revealed a similar conspicuous pattern, with most fibres occurring in the μ-lobe [38]. This hints at a functional specialisation of the individual lobes of the mushroom body, which remains to be clarified, e.g. by gene knockouts combined with behavioural assays or electrophysiological recordings (cf. [55]).
Such experiments might also help to clarify the role of the onychophoran mushroom body as a whole, which is still unknown. Does this neuropil serve as an integrative centre for olfactory, visual and place learning and memory, with the additional role as a general behavioural control centre, as in insects [173][174][175][176][177][178][179]? Answers may emerge from expression studies of specific proteins, such as Leo and pCaMKII, that are present in paired lobed centres in the brain of representatives of all four major arthropod groups [147] and are required for learning and memory in the fruit fly Drosophila melanogaster [164,180,181].
Another contentious issue concerns the identity of the "accessory lobes", also referred to as "accessory stalks" or "Nebentrabekel", that have been repeatedly alleged to be associated with onychophoran mushroom bodies [30,32,33,42,48,54,69]. As in chelicerates, but in contrast to other arthropods, these structures are believed to link the mushroom body of both brain hemispheres across the midline. This feature has also been used to unite onychophorans with chelicerates, which exhibit a similar connection [42]. However, our 3D reconstructions clearly show that the corresponding heterolateral connections are associated with the antennal and subantennal neuropils (Fig. 13), but not with the mushroom bodies directly. We refer to these connections as brain commissures 1-3 (Table 1), as they do not seem to contain many synaptic sites and were only discernible with anti-acetylated α-tubulin immunolabelling. Since these commissures are not part of the onychophoran mushroom bodies, they cannot be used for justifying the sister group relationship of onychophorans and chelicerates [42,182]. The connection between the mushroom bodies of both brain hemispheres in chelicerates might instead have evolved independently within their lineage.

Controversies surrounding the frontal body
The frontal body comprises the second largest individually identifiable cerebral neuropil in E. rowelli, but has rarely been noted in onychophorans. Saint-Remy [26] initially described this midline neuropil in the brain of Peripatopsis capensis as "bourrelet médullaire antérieur" (i.e. anterior medullary ridge). Almost five decades later, Hanström [30] rediscovered it in P. capensis, Kumbadjena occidentalis and Ooperipatellus insignis and homologised it with the protocerebral bridge of crustaceans and insects. However, his descriptions are inconsistent, as differing regions are labelled as "Protocerebralbrücke" in his illustrations (cf. figs. 12 and 13 in [30]). Likewise, Schürmann [32,54] identified a V-shaped midline neuropil anterior to the central body as "bridge" in E. leuckartii, a species closely related to E. rowelli, but at the same time he assigned the two divisions of the central body to it (cf. fig. 8.2C in Schürmann [32]). It is worth mentioning that Strausfeld et al. [33] recognised the corresponding region as "dorsal heterolateral neuropil" and "dorsal superior protocerebrum", whereas Mayer [48] and Martin et al. [38] referred to it as "anterior neuropil", although none of these studies provided a detailed description. Beyond this, we know of no further reference to the frontal body in onychophorans. This is astonishing, given its considerable size and conspicuous shape and position in the onychophoran brain (Fig. 13).
Our 3D reconstructions of the frontal body in E. rowelli match superficially Schürmann's [32,54] illustrations of the "bridge". Although his schematised "bridge" is considerably smaller and widely separated from the central body, Schürmann [32,54] correctly depicted its shape and superficial, anterior position relative to other brain neuropils. The issue of whether the frontal body of onychophorans is homologous with the protocerebral bridge of crustaceans and insects, as suggested by Hanström [30], is currently difficult to address due to limited information on its internal organisation. We have shown here that synaptic proteins are mostly accumulated in the anterior and posterior regions of the frontal body, indicating at least some degree of regionalisation and layering. Contrary to onychophorans, the protocerebral bridge of insects and certain crustaceans shows a clear subdivision into distinct columns, slices or glomeruli [183][184][185][186][187] that do not seem to correspond to the transverse layers found in E. rowelli. Furthermore, the onychophoran frontal body lacks a connection to the central body, which differs from the situation in insects, where the protocerebral bridge is clearly associated with the central body [183]. Although differences per se do not invalidate homology, given the lack of compelling evidence and the absence of the corresponding neuropil in chelicerates and myriapods, we regard the proposed homology [31,32] of the frontal body of onychophorans with the protocerebral bridge of pancrustaceans as unlikely.

Organisation of the central body revisited
The central body is the third largest individually identifiable neuropil in the brain of E. rowelli (Fig. 13). Our 3D reconstructions correspond well to previous descriptions of this midline neuropil [26,27,32,33,42,48,56,69], except that we identified a lateral fusion of the bilobed structure at the ventral part of the central body. Antiacetylated α-tubulin and anti-synapsin immunolabelling further revealed two prominent strata-the inner and the outer laminae-which show an inverted anteroposterior arrangement in the two divisions. The two laminae most likely correspond to the successive strata of looped axons and tangential fibres described by Strausfeld et al. [42], suggesting a complex wiring. We further detected a posterior stratum associated with the posterior division, which receives fibres from the optic neuropil, as well as a pair of auxiliary neuropils linked to the antennal neuropils. This indicates processing of visual input (from the eyes) and olfactory and/or mechanosensory information (from the antennae) in the central body.
Besides bilateral input from the optic neuropils, our findings confirm the existence of a thick fibrous connection, the ponticulus, which links each mushroom body to the central body in E. rowelli. This connection was first described as "pédoncule" by Saint-Remy [26] and since then is assumed to be associated with either the anterior division or the posterior division of the central body [26-28, 30-32, 54]. Our 3D reconstructions revealed a specific connection of each ponticulus to the posterior region of outer lamina of the central body.
Interestingly, while the ventral part of the ponticulus contains numerous granular synaptic complexes, these are completely lacking in its dorsal part. The progressive dorsoventral increase of synaptic sites suggests a combined tract-like/neuropil-like organisation. Similar distributions of synapsin have been observed in the olfactory globular tract of crustaceans toward the olfactory globular tract neuropil [188][189][190] and along the mushroom body pedunculus of insects and spiders [191][192][193]. Nonetheless, the ponticuli, which directly link the mushroom bodies and central body, appear to be a unique feature of Onychophora. While direct connections between the mushroom bodies and central complex are conspicuously absent in insects [194,195], recent connectome data from Drosophila revealed multiple indirect connections from mushroom body output neurons primarily to the fan-shaped body of the central complex [178,179], a centre proposed to serve a role in spatial orientation and navigation [196][197][198]. The central body of insects is generally believed to be an integrating centre that processes information from both brain hemispheres and, together with the protocerebral bridge and noduli, controls goal-directed locomotion, including directional control of walking and flight, spatial visual memory, place learning, sky compass orientation and path integration (e.g. [183,199,200]). Behavioural experiments have demonstrated that velvet worms avoid light and are instead attracted by darkness [55,201]. Given the complexity of their locomotor behaviour [202,203], myoanatomy [19] and neuronal supply to individual legs [3,18], it is conceivable that the central body of onychophorans might integrate visual and antennal information related to locomotion. A possible function in directional control of locomotion might explain the relatively large size of this midline brain neuropil (Fig. 13).
The central body of onychophorans has been homologised with that of arthropods [30,149]. This was, however, not always the case, due to dissenting opinions on the homology of the arcuate body of chelicerates with the central body of other arthropods ( [42]; discussed in [204]). Similarities in the organisation into multiple layers as well as its superficial position within the brain and direct connection to the visual system might indeed support homology of the onychophoran central body with the arcuate body of chelicerates [42]. Myriapods might play a key role in elucidating the evolution of midline neuropils across arthropods, as they show similarities to both chelicerates and pancrustaceans [84,[204][205][206]. While the central body of myriapods on the one hand shares to some extent a layered organisation with the arcuate body of chelicerates [205,207], it is on the other hand embedded deeper within the protocerebral neuropil, as in pancrustaceans [161,184,186,208]. Furthermore, to our knowledge, nothing is known about a potentially corresponding lobular midline neuropil in tardigrades (but see, e.g. [85,167,209] for an uncharacterised "central brain neuropil"). Taking all available information into account, we propose that the last common ancestor of Onychophora + Arthropoda most likely possessed a transverse, arcuate midline neuropil, which was composed of successive strata and had a superficial position in the protocerebrum (i.e. abutting the other protocerebral neuropils), since these features are found in both onychophorans and chelicerates [42]. Whether the individual strata in chelicerates correspond to the inner and outer laminae in onychophorans might be clarified by specific labelling using antisera against neurotransmitters and neuropeptides (e.g. [152,207,210,211]).
The association of microglomeruli with olfactory lobes has also been described for the accessory lobes of decapod crustaceans [190,223] and the antennal lobes of distinct insect groups, including orthopterans and coleopterans [214,227,228]. The accessory lobes of decapods are second-order association centres located in the deutocerebrum and always occur in close proximity to the olfactory lobes. Within decapods, accessory lobes most likely evolved in the reptantian lineage [229]. Amongst others, these higher-order neuropils receive olfactory information from the olfactory lobes via interneurons [230,231]. Similarly, microglomeruli are present in the antennal lobes of two distantly related representatives of coleopterans, the diving beetles and lady bugs [228]. Several coleopteran sub-groups also show a microglomerular compartmentation of glomeruli, as revealed by either immunolabelling for synapsin and/ or tachykinin-related peptides [228].
We have shown here that microglomeruli are closely associated with olfactory lobes and antennal neuropils in the brain of E. rowelli, suggesting that they might be connected with olfactory pathways, as for certain microglomeruli of insects [214]. Despite the close association with the olfactory glomeruli, a potential role of microglomeruli in olfaction requires further exploration. An involvement in visual pathways, as described from several arthropod groups [193,213,215,216,232], seems unlikely, as we did not find signs of any connection to the visual system.
Olfactory glomeruli: a remarkable case of homoplasy across panarthropods Olfactory glomeruli are defined as dense neuropils, in which the olfactory receptor neurons terminate and form the first synapses of the olfactory pathway [20]. They occur in various metazoans and have been extensively studied in arthropods (e.g. [206,227,233,234]). To our knowledge, comparable structures have not been recorded in tardigrades. As in arthropods, the olfactory glomeruli of onychophorans are arranged in clusters referred to as olfactory lobes [33,42,48,57,69]. They most likely represent the primary processing centres of olfactory stimuli, as they are supplied by neurites associated with the antennae [15,124,168]. While the olfactory lobes of onychophorans are located in the protocerebrum [31,32], those of arthropods are situated either in the deutocerebrum, as in most mandibulates [206,233], or exhibit great variability with respect to segmental identity, as in chelicerates [42]. For example, in solifuges, scorpions, amblypygids and pycnogonids, the olfactory lobes are associated with post-cerebral locomotory appendages that serve as tactile organs [235][236][237]. Solifuges possess in addition a glomerular neuropil in the tritocerebral brain segment connected to the pedipalps [42] and a further set of olfactory glomeruli occur in close proximity to the mushroom body calyces, at least in solifuges and amblypygids [150,235]. Taken together, the variable position and different segmental identity do not support homology of olfactory lobes across panarthropods. The similar organisation of olfactory neuropils in distantly related groups might rather be due to functional constraints associated with olfactory processing [238,239].
Although a recent study revealed that onychophorans might have lost some chemosensory gene families found in protostomes including arthropods [240], the prominent size of olfactory lobes in E. rowelli indicates that olfaction nevertheless plays an important role in these animals. While each olfactory lobe of different onychophoran species, including our study species, was previously thought to contain only about 30 glomeruli arranged in two clusters [31][32][33], our 3D reconstructions revealed that each lobe of E. rowelli constitutes a single, elongated cluster of 80 glomeruli that differ in size and exhibit a stereotypic arrangement (Fig. 13). The high number of~80 olfactory glomeruli per body side of different sizes and shapes in E. rowelli imply a prominent role of olfactory processing in onychophorans, which might be correlated with predatory behaviour and nocturnal lifestyle combined with lowresolution vision and chemical communication in these animals [70,201,241,242].
We have shown here that each olfactory lobe of E. rowelli contains a median macroglomerulus (∅ 40 μm), whereas most other glomeruli do not exceed half this diameter. This variation in size, already pointed out by Schürmann [32], might indicate functional specialisation of the individual glomeruli. Similar macroglomerular complexes occur in various insects, in which they process pheromone signals, including those from trail pheromones [243][244][245]. The single macroglomerulus in each brain hemisphere of E. rowelli might have a similar function, as trail pheromones have been proposed to play a role in colonising new habitats in this species [241], although the responsible molecules have not been identified yet. Since we only analysed males, a study of both sexes in this and other species would help to clarify whether the composition of olfactory lobes exhibits sexual dimorphism and interspecific variation among velvet worms.
In arthropods, olfactory information is processed in the olfactory lobes and forwarded via projection neurons to mushroom body calyces and a neuropil called lateral horn [246][247][248][249]. A connection between olfactory lobes and mushroom bodies has also been reported in onychophorans [27,[31][32][33], albeit previous studies described an association with the calyces, which we did not detect. Instead, each olfactory lobe of E. rowelli is associated with four fascicles, only the thickest of which-the olfactory fascicle 4-is connected to the lobes of the mushroom body, whereas the olfactory fascicles 1-3 extend towards the median part of the brain (Fig. 13). This suggests that, as in insects, olfactory information passes to at least two higher-order processing centres, but the organisation of these connections bears only sparse resemblance to the situation in insects. As proposed for insects [250], the presence of multiple olfactory centres in the onychophoran brain might be related to the need for processing memory-related and memory-unrelated olfaction. This hypothesis remains to be substantiated by physiological and behavioural experiments.

Potential homologues of onychophoran antennae across panarthropods
The segmental identity of the anteriormost appendages of onychophorans ("primary antennae" sensu Scholtz and Edgecombe [251]) and their potential homologues in other panarthropods have been subject to controversy (e.g. [13,31,34,45,47,58,75,79,[252][253][254][255][256][257][258][259][260][261]). In the last two decades, considerable progress has been made on two issues related to this controversy: (1) It is likely that the onychophoran antennae are serially homologous with trunk limbs. This is suggested in onychophorans by the occurrence of a transitory nephridial anlage at the antennal basis, similar to the segmental nephridia in the legbearing body regions [124,262], as well as corresponding expression patterns of appendage patterning genes in the antennae and other limbs [263]. (2) Head development [34], neuronal tracing [15] and gene expression data [16,116,168,263] revealed that the onychophoran antennae belong to the anteriormost (ocular or protocerebral) segment. This corroborates the hypothesis that the onychophoran antennae are not homologous with the first antennae of mandibulates ("secondary antennae" sensu Scholtz and Edgecombe [251,255]) or the chelicerae of chelicerates, which are both innervated by the deutocerebrum and instead correspond to the onychophoran jaws [15,168].
The question of whether or not there are protocerebral homologues of the onychophoran antennae in living arthropods is difficult to address. The prevalent hypothesis considers the labrum of arthropods as the most likely candidate ( [168,259,261]; Fig. 14b), as evidenced by its paired, appendicular origin in the protocerebral segment [267][268][269][270][271][272][273]. An alternative hypothesis homologises the onychophoran antennae with the frontal filaments of crustaceans [58,257]. Although it seems plausible that frontal filaments are homologous across crustaceans [274], the appendicular nature of these structures is still questionable and casts doubt on their status as putative homologues to the onychophoran antennae [259,261].
The frontalmost limbs of various stem-group arthropods called great appendages or frontal appendages have also been homologised with the onychophoran antennae (e.g. [45,256,[259][260][261]). However, their homology is uncertain even across arthropods, with divergent views regarding their correspondence to either primary [280,281] or secondary antennae [261,[282][283][284]. An alternative recent hypothesis [255,259,260,285] suggests that the frontal appendages of early-branching, stem-lineage arthropods, including representatives of †Kerygmachela, †Opabinia and †Radiodonta, might have been protocerebral structures homologous with the onychophoran antennae (Fig. 14a) and the labrum of arthropods, whereas the short great appendages of †Deuteropoda might have been associated with the deutocerebrum and are thus homologous with the chelicera/first antenna of arthropods.
A particular case is represented by the frontal appendage of the radiodont †Lyrarapax unguispinus. It has been interpreted as a pre-protocerebral structure associated with a pre-protocerebral ganglion rather than with the brain itself [45,75]. The authors also reported a similar arrangement in the onychophoran E. rowelli, concluding that the antennal nerve cords ("frontal appendage nerve" sensu Cong et al. [45]) are associated with preprotocerebral ganglia. The complete reconstructions of brain neuropils presented herein revealed no evidence for such ganglia in E. rowelli, as also demonstrated by previous immunolabelling and dye fill data [15]. Irrespective of whether the frontal appendages of †L. unguispinus are pre-protocerebral [45] or protocerebral, as in other radiodonts [260,285], we conclude that frontal ganglia cannot be used to homologise the antennae of †L. unguispinus and E. rowelli.

Open questions regarding the organisation of visual neuropils
The optic tracts (sensu [15,32,33]) or optic neuropils (this study) have been subject of ongoing debates  [256] and Legg and Vannier [264]. Diagrams of †O. ferox †A. pedunculata and †Anomalocaris spp. after [256], [265], and [264,266], respectively. Anterior end of stage IV embryo of E. rowelli redrawn from [59] to illustrate the position of developing limbs prior to their reorganisation and incorporation into the head. b Alignment of anterior segments with corresponding appendages in onychophorans, tardigrades and arthropods. Combined after [15,59,85,251,255,259,260]. Note that the condition illustrated for myriapods is also found in hexapodes within Pancrustacea. DE, deutocerebrum; IC, intercalary segment (without appendages); PR, protocerebrum; TG1, TG2, TG3, trunk ganglia 1 to 3; TR, tritocerebrum regarding homology and evolution of visual organs and their associated brain centres across panarthropods [33,42,44,69]. The presence of second-order visual neuropils flanking the central body as well as a connection of the optic neuropil to the mushroom body has led to the assumption that the nervous system of onychophorans might be more similar to that of chelicerates than other arthropods [42]. Our datasets did not reveal a connection with the mushroom bodies, suggesting this is an autapomorphy of chelicerates rather than a synapomorphy of chelicerates and onychophorans. While our results confirm connectivity between the optic neuropil and the central body [30][31][32]54], we found no evidence of a "second optic neuropil" in E. rowelli [42], in which optic fibres would terminate. The optic neuropil with all its branches instead represents a unitary structure, which connects to the posterior division of the central body ( Fig. 13), suggesting that visual signals are directly relayed by first-order interneurons into this higher-order processing centre. This again contrasts with the situation in arthropods, in which both median ocelli and compound eyes are associated with a second-order visual neuropil [169,172,286,287]. The absence of secondorder neuropils may indicate that these were either lost in Onychophora, assuming that they were already present in the last common ancestor of Panarthropoda, or that they first arose in the arthropod lineage. Reconstructing the visual system [288] and, in particular, the optic connectivity in tardigrades could help to clarify this issue.
The proposed connection of the optic neuropil to the mushroom body in E. rowelli [33] still requires corroboration, as we were unable to reconstruct all branches of the visual pathway in this species. A similar connection has been reported from the secondary eyes (i.e. derivatives of compound eyes) of spiders [33 83, 143, 149]. This would suggest homology of the onychophoran eyes with the compound eyes [33] rather than the median ocelli of arthropods [44]. The use of more specific methods, such as neuronal tracing [39,40], in conjunction with 3D reconstruction might help to demonstrate the connectivity of the visual system in onychophorans and clarify the homology of visual organs and their neuropils across panarthropods.

Conclusions
The present multiscale approach comprising 3D reconstruction revealed unprecedented details of the onychophoran brain, including the shape, position, spatial relationship and relative volume and size of individual tracts, lobes and neuropils. We have shown here that each mushroom body of E. rowelli is comprised of four rather than three lobes and that the accessory stalks [32] are not part of the mushroom bodies, but rather independent commissures associated with the antennal and subantennal neuropils. Despite detailed analyses using multiple methods, we were unable to endorse the existence of a frontal ganglion in the brain of E. rowelli [45], further questioning its proposed presence in onychophorans [47].
Our data further revealed a higher complexity of the onychophoran central body than previously reported, as this midline neuropil not only comprises anterior and posterior divisions (occurring only dorsally) but is additionally stratified into an inner and an outer lamina exhibiting an inverted arrangement in the two divisions and surrounding a pair of cylindrical auxiliary neuropils. Using 3D reconstruction, we characterised in detail the olfactory lobes and the individual glomeruli and, in addition, identified novel microglomerular structures, the function of which remains unknown. Most importantly, we identified a largely neglected anterior neuropil called the frontal body, which is one of the most prominent structures of the onychophoran brain. The previously suggested homology of the frontal body ("bridge" sensu [32]) with the protocerebral bridge of pancrustaceans is not well substantiated and therefore doubtful.
In summary, the present study improves our understanding of the anatomy and evolution of the onychophoran brain. It also clarifies several controversies surrounding the existence and composition of specific cerebral structures and their connectivity, thus shedding light on their evolutionary origins and homology. Our improved, refined and expanded nomenclature of neuroanatomical terms (Table 1) may serve as a basis for future studies of onychophoran neuroanatomy.

Collection and maintenance of specimens
Specimens of Euperipatoides rowelli Reid, 1996 were obtained from decaying logs and leaf litter in the Tallaganda State Forest (New South Wales, Australia; 35°30′ 31″ S, 149°36′ 14″ E, 934 m) in October 2016. They were collected under the permit number SPPR0008 issued by the Forestry Commission of New South Wales and exported under the permit number WT2012-8163 provided by the Department of Sustainability, Environment, Water, Population and Communities. The animals were maintained in culture as described elsewhere [289]. Only male specimens were used for the present study.

Synchrotron radiation-based X-ray micro-computed tomography
For synchrotron radiation-based X-ray micro-computed tomography (SRμCT), two specimens were prepared following the protocol of Jahn et al. [87]. The first specimen, used for reconstructing the brain and the nerve cords, was anaesthetised in chloroform vapour for 10 s, fixed overnight in glutaraldehyde (2.5 % in 0.1 M sodium cacodylate buffer), washed several times in cacodylate buffer and stained overnight in 2% osmium tetroxide (OsO 4 , 4 %, Carl Roth, Karlsruhe, Germany). For the second specimen, used for reconstructing the anterior neuropils, ruthenium red (RR) was used as a staining agent, as it has proven useful for visualising the nervous system of onychophorans using SRμCT [87]. The specimen was first fixed and stained for 2 h in a solution containing 4% paraformaldehyde (PFA) and 0.3% RR (Carl Roth) in phosphate-buffered saline (PBS) and then decapitated, after which the head was left in the same solution for 5 days. Both specimens were washed several times in the corresponding buffers. Thereafter, the RR-stained specimen was kept in buffer, whereas the OsO 4stained specimen was dehydrated through an ascending acetone series (30,50,70,90,95, 3 × 100% in H 2 O; 15 min each) and embedded in Araldite® (Huntsman Advanced Materials, Salt Lake City, UT, USA). For SRμCT analyses, the Araldite®-block containing the OsO 4 -stained specimen was glued onto a standardised sample holder (HZG, Geesthacht, Germany), whereas the RR-stained head was scanned in a small container, which was filled with PBS and then glued onto a sample holder. Specimens for SRμCT were scanned at the beamline P05 of the storage ring PETRA III (Deutsches Elektronen-Synchrotron -DESY, Hamburg, Germany) operated by the Helmholtz-Zentrum hereon [290,291] as described by Jahn et al. [87] with the following modification: the RR-stained specimen was scanned using inline grating differential phase-contrast (=PBCC) with a photon energy of 20 keV. We have applied a transport of intensities (TIE) phase retrieval approach to the recorded projections prior to the tomographic reconstruction in a custom-implemented reconstruction pipeline [292] using Matlab (Math-Works) and the ASTRA toolbox [293][294][295]. For the processing, raw projections were binned two times, which resulted in an effective voxel size in the final tomographic reconstructed volume of 2.4 μm for the RR-stained specimen, and 4.92 μm for the OsO 4 stained specimen.

Immunohistochemistry and confocal laser scanning microscopy
To obtain details of individual tracts and neuropils, neural markers against acetylated α-tubulin and synapsin were used in conjunction with confocal laser scanning microscopy (CLSM). Synapsin occurring in synaptic terminals was labelled using the monoclonal anti-synapsin serum (Developmental Studies Hybridoma Bank, Iowa City, IA, USA; DSHB Product 3C11 [antiSYNORF1]) raised in mouse against a fusion protein consisting of glutathione-S-transferase and the SYN1 protein of Drosophila melanogaster [299]. Acetylated α-tubulin was targeted using a monoclonal antibody (Sigma-Aldrich, St. Louis, MO, USA, cat. nr. T7451) raised in mouse against acetylated α-tubulin from the outer arm of the sea urchin Strongylocentrotus purpuratus. The antibodies against synapsin and acetylated α-tubulin have both been shown to recognise neuropil and neurite regions in the nervous systems of various invertebrates, including onychophorans, tardigrades and non-insect arthropods (e.g. [3,15,18,69,167,169,235,300]). The staining protocol was based on the technique developed by Ott [301].
Additional anti-synapsin labelling was conducted on vibratome sections. For this, specimens were anaesthetised in chloroform vapour and fixed in ZnFA overnight. After several rinses in HBS, specimens were cut in halves and embedded in a solution containing 3.75 g albumin from chicken egg (Sigma-Aldrich) and 0.5 g gelatine from porcine skin (Type A; Sigma-Aldrich) in 12 mL distilled water. Albumin-gelatine blocks were fixed in 10% PFA overnight and cut in 100-μm-thick sections using a vibratome (MICROM 650 V, Microm International, part of Thermo Fisher Scientific, Walldorf, Germany). Obtained sections were subsequently labelled as for wholemount samples with the following modifications: (i) dehydration and rehydration steps were omitted, and (ii) the sections were mounted between two coverslips in ProLong Gold TM Antifade Mountant (Invitrogen, Carlsbad, CA, USA).
Both whole-mount preparations and vibratome sections were analysed using a confocal laser scanning microscope (Zeiss LSM 880; Carl Zeiss Microscopy) equipped with an Airyscan module. Images were acquired using the ZEN 2 (black edition) imaging software (Carl Zeiss Microscopy). The images were taken using 488 nm (for the secondary antibody) and 405 nm lasers (for DAPI) with the emission wavelengths of 516 nm and 450 nm, respectively.

Data and image processing
Three-dimensional reconstructions were conducted using the following five datasets. Stacks of SRμCT images were used for reconstructing (i) the overall structure of the central nervous system as well as (ii) its anterior part subdivided in perikaryal and neuropil regions. After manual segmentation of structures of interest using Amira 6.0.1 (FEI Visualization Sciences Group, Burlington, MA, USA), label fields were separated using Fiji v.1.52 [302,303]. (iii) Series of histological micrographs were used for reconstructing the anterior nervous system. For this, digitised images were aligned and target structures segmented and reconstructed using Amira 6.0.1. Finally, CLSM stacks of entire brains labelled with DAPI and antibodies against (iv) synapsin and (v) acetylated α-tubulin were used for reconstructing individual tracts, cell layers and neuropils. Raw Airyscan datasets were processed using the ZEN 2 software (black edition) and handled further with Amira 6.0.1 as for the other datasets. Three-dimensional volume and surface renderings of SRμCT, histological and CLSM data were carried out with VGStudio MAX 3.0 (Volume Graphics, Heidelberg, Germany).
Relative volume estimates of brain neuropils were conducted taking into account the number of voxels in each segmented structure relative to the number of voxels in the entire brain. Potential shrinkage artefacts (e.g. [304,305]) were not considered, as we expect that the structures (i.e. nervous tissue) shrink relative to each other, resulting in similar or even the equal volume ratios. Prior to the relative volume estimates, the brain was segregated from the remaining parts of the central nervous system, such as the antennal nerve cords, circumpharyngeal nerve cords and ventral nerve cord, using the Regions of Interest function (ROI) in VGStudio MAX 3.0. Raw datasets were processed (Airyscan mode: Superresolution) using the ZEN 2 software (black edition). Selected substacks of entire brains and vibratome sections were adjusted for optimal brightness and contrast in Fiji v.1.52. Final panels and diagrams were compiled with Adobe Illustrator CS5.1 (Adobe Systems, San Jose, CA, USA).