A new look at the ventral nerve centre of Sagitta: implications for the phylogenetic position of Chaetognatha (arrow worms) and the evolution of the bilaterian nervous system

Structure of the chaetognath central nervous system: current knowledge

A brief description of the layout of the adult arrow worm central nervous system will serve as the basis for illustrating our own results. It consists of six ganglia in the head, one unpaired ventral ganglion in the body, nerve tracts connecting these ganglia and peripheral nerves passing out of these ganglia [2,19,20] (Fig. 2). The ganglia in the head are the cerebral ganglion (the brain), a pair of vestibular ganglia, a pair of oesophageal ganglia, and a suboesophageal ganglion (Fig. 2A, B). The brain is located immediately below the surface epithelium of the head and consists of a neuropil core with numerous synapses surrounded by a layer of neuronal cell somata [14,19,20]. Connective nerve tracts are found between the cerebral and the vestibular, the cerebral and the ventral, and the vestibular and the oesophageal ganglia. A commissural nerve bundle behind the oesophagus connects the vestibular ganglia. The anterio-posteriorly running nerves which connect the cerebral and the ventral ganglia are called the main connectives, and those which connect the cerebral and the vestibular ganglia, the frontal connectives [19]. Sensory organs associated with the brain are a pair of eyes [24,33-35], a ciliated loop, the corona ciliata, localised in the dorsal part of the head [2,20,29-32] and the retrocerebral organ, a structure with an unknown putative sensory function [2].

The ventral ganglion is an elongate structure lying between the basement membrane and the epidermis. Two main connectives link it with the brain ganglia and two other nerve tracts continue caudally (Fig. 2A, B) [19,23]. Similar to the brain, the ventral ganglion consists of a central fibrillar neuropil core, flanked by lateral clusters of cell bodies [2,19,20]. Along the length of the ventral ganglion a series of smaller nerves pass out radially (Fig. 2A, B) which branch in the periphery and form a dense ramifying plexus just external to the basement membrane to provide motor innervation to the body musculature and to innervate the ciliary fence receptors [23]. The ventral ganglion controls swimming by initiating contractions of the body wall musculature and co-ordinating mechanosensory input from the numerous ciliary fence receptors in the epidermis [2,20,23]. The available descriptions of the lateral nerves that exit the ventral ganglion in closely related species of the genus Sagitta display a remarkable degree of variability over the last 125 years. Hertwig [17] described an irregular array of nerve fibres ("Nervenfibrillen") to emerge from the ventral ganglion (Fig. 2C). Kuhl [18], presenting a modified drawing from Burfield [71], depicts 12 distinct, stout nerve trunks to exit the ganglion on both sides (Fig. 2D). Yet, Goto and Yoshida [19] draw only six radial nerves on each side (Fig. 2B). In Shin [2], again, 12 bilaterally arranged radial nerves exit the ganglion on both sides in a nicely ordered way, much like the arrangement of the segmental nerves in the ventral nerve cord of an annelid or arthropod (Fig. 2E). In the same year, Duvert et al. [39] published a report in which they described 20 – 30 irregularly arranged aspartate-immunoreactive fibre bundles that pass out radially from the ventral ganglion on both sides and spread out diffusely and branch in the periphery, similar to the description already provided by Hertwig [17].

The structure of the ventral ganglion as revealed by the histochemical localization of actin and a nuclear marker

In whole mounts of juvenile Sagitta sp. that were processed with the nuclear marker bisbenzimide the general arrangement of the ventral ganglion is visible (Fig. 3A). Scattered concentrations of nuclei on the body surface mark the location of ciliary fence receptors (arrows in Fig. 3A). Double labelling with bisbenzimide and phallotoxins to visualize actin shows that the ventral ganglion is composed of a central neuropil core that has a rectangular, elongated shape and contains longitudinal fibre tracts (Fig. 3B, C). The neuropil core is on both sides flanked by coherent zones of small (ca. 10 μm) neuronal cell somata (ganglion cells; Fig. 3B, C). Most of these somata are restricted to the sides of the ganglion rather than being located dorsally or ventrally to the neuropil. Fibre bundles that emerge from the cells in the lateral clusters approach the neuropil core from a lateral direction (Fig. 3D'). The ganglion cell somata appear to be arranged in rows between those fibre bundles (Fig. 3D"; and Fig. 6C). At the lateral margins of the soma clusters, large cell bodies are interspersed between the smaller ganglion cells at regular intervals (asterisks in Fig. 6D). These large ganglion cells were already recognized by Hertwig ("große Ganglienzellen" [17]) and also in later studies using methylene blue staining techniques [23,27].

Tubulin labelling: the ventral ganglion as a condensation of the intraepidermal nerve plexus

Immunolocalization of acetylated-alpha tubulin in adult Sagitta setosa visualized the extensive intraepidermal nerve plexus that extends throughout the entire epidermal surface of the animals (Fig. 4A, B). This dense network of fine nerve fibres is embedded between the basement membrane and the outer epithelial layer [20,27,39] and is known to be involved in the control of muscle contractions [23,28,72]. The network consists of fibres of various sizes many of which extend in a roughly anterior-posterior direction but many transverse fibres are also present (Fig. 4B). Some of the fibres have a beaded appearance suggesting the presence of synaptic varicosities. Our preparations show that the ciliary fence receptors which are also strongly immunolabelled are innervated by tiny branches of the fibre network (arrows in Fig. 4A) suggesting a sensory function of the intraepidermal plexus in addition to the known role in motor control. Hertwig [17] and Bone and Goto [20] observed numerous multipolar neurons associated with the fibres of the intraepidermal plexus. Tubulin immunolocalization revealed conspicuous small concentrations of labelled material (arrows in Fig. 4B) that may correspond to these peripheral neurons but we failed to establish the identity of these concentrations with sufficient certainty.

An irregular array of ca. 20 – 30 tubulin-labelled fibres and fibre bundles emerges laterally from both sides of the ventral ganglion (Fig. 4C, D). These bundles contribute to and are confluent with the peripheral nerve plexus. We did not notice a regularly ordered spacing or distinct bilaterally symmetrical arrangement of these radial nerve bundles as earlier reports had indicated [2,19]. Rather, our findings resemble the pattern of aspartate-immunoreactive fibre bundles that pass out from the ventral ganglion as described by Duvert et al. [39]. Caudally, two thick bundles of tubulin-labelled fibres emerge from the ganglion to connect it to the intraepidermal nerve plexus in the posterior part of the animal (Fig. 4C). These caudal bundles are superficially similar to the radial fibre bundles, yet they contain more closely packed neurites. In summary, in tubulin-labelled preparations the ventral ganglion very much appears to be a condensed continuation of the peripheral intraepidermal nerve plexus.

Synapsin immunoreactivity: the ventral ganglion as a highly ordered structure

Immunohistochemistry with the SYNORF 1 antibody (Fig. 5) that is directed against presynaptic proteins reveals an overall shape of the synaptic neuropil core in adult Sagitta setosa that is similar to that in specimens labelled for actin (Fig. 3B). Clearly, synaptic contacts are confined to the central neuropil whereas the lateral soma regions are devoid of synapses. Synapsin labelling is weak at the anterior and posterior ends of the ventral ganglion suggesting the presence of more fibre tracts than synaptic neuropil in these areas. Interestingly, synapsin immunolocalization reveals a strikingly different picture of the ventral ganglion than tubulin immunohistochemistry ("unorganized nerve plexus") in that it shows a highly ordered subdivision of the neuropil core into serially arranged compartments. The ganglion seems to be composed of an anterior-posterior sequence of ca. 80 transverse microcompartments (Fig. 5, 6A, B). At higher magnification it becomes apparent that, overlying the pattern of microcompartments, synapsin immunolabelling is particularly strong in a lateral longitudinal stripe on both sides of the neuropil whereas a narrow medial longitudinal stripe is devoid of labelling (Fig. 6A, B). Yet, at a slightly more ventral focus level, synapsin immunoreactivity is more evenly distributed across the neuropil core and microcompartments are not visible (Fig. 6B). A comparable serial arrangement of nervous elements in the chaeotognath nervous system so far has been reported in only one other study. Bone and Pulsford [23], using methylene blue and reduced silver staining techniques, provided evidence for the presence of many serially arranged transverse fibres that cross the neuropil core.

RFamide-like immunoreactivity: individually identifiable neurons

In addition to synapsin labelling, immunohistochemistry against RFamide provides further evidence for serially arranged nervous elements in the ventral ganglion of Sagitta setosa (Fig. 6E–G). RFamide immunoreactive neurons previously have been described in Sagitta setosa [27] and Paraspadella gotoi [38], yet these authors did not map the labelling pattern in greater detail. Our study provides conclusive evidence for the presence of individually identifiable neurons in Chaetognatha that can be homologized between different specimens (Fig. 7). Double labelling that combined RFamide immunolocalization with synapsin immunohistochemistry (Fig. 6E) or a nuclear marker (Fig. 6F, G) showed that within the lateral soma zones the cell bodies of the RFamidergic neurons are located mostly close to the interface between the neuropil core and the soma clusters whereas longitudinal RFamidergic fibre tracts are restricted to the ganglion core (Fig. 6E). In Fig. 7, photomontages of the complete ventral ganglia (whole mounts) of two specimens are shown to illustrate the full extent of the RFamidergic system. The labelling pattern was consistent among the more than 20 specimens that we examined. Anteriorly, typically three fibres are present within both main connectives (CO) that link the ventral ganglion to the brain (Fig. 7, 8H, I). Three main longitudinal tracts of RFamide immunolabelled fibres can be distinguished: the medial bundle (MB), the bilaterally paired intermediate bundles (IB), and the paired lateral bundles (LB; Fig. 6G, 7B, 8A) that run along the lateral borders of the neuropil cores as double labelling with the synapsin marker shows (Fig. 6E). All fibres within these longitudinal tracts have a typical beaded appearance. One of the fibres that enter the ganglion anteriorly crosses the midline close to the entry point of the main connectives to join the medial bundle (solid arrow in Fig. 8H). The second fibre in the main connectives is also associated with the medial bundle whereas the third one joins the intermediate bundle (Fig. 8H, I). In the anterior part of the ganglion, fibres in the intermediate bundle run towards the midline and seem to cross over the median nerve bundle towards the contralateral side (Fig. 7, and open arrows in Fig. 8H, I). In contrast to the observations by Goto et al. [38] on Paraspadella gotoi, we did not record any RFamidergic fibres that leave the ventral ganglion posteriorly.

The RFamidergic neurons, all of which are unipolar, can be subdivided into two different populations: a first series of neurons with lateral somata (L1–4; small circles in Fig. 7) in the anterior third of the ganglion and a second series of slightly larger dorsal neurons (D1–5; large circles in Fig. 7) in the posterior two thirds of the ganglion the somata of which are located clearly more dorsally than those of the lateral neurons as is apparent in lateral views of the nervous system (Fig. 8F, G). Within the cell somata, the immunolabelled material is typically concentrated in a cap of cytoplasm opposite to the point of exit of the single neurite (Fig. 6G, 8D). The arrangement of the most anterior lateral neurons displayed some variation between individual specimens and in addition to the neurons L1 and L2 that we identified consistently in all six specimens that we analysed at the single-cell level, additional neurons are present in some specimens (small circles labelled with a question mark in Fig. 7). Contrary to L1 and L2, the neurons L3, L4, and D1–D6 were reliably present in all analysed specimens and hence represent typical examples for individually identifiable, bilateral symmetrically arranged neurons. The neurons D1–D5 have an identical morphology and appear to be serially repeated clones. Their neurites exit the soma in a medial direction to enter the neuropil core at a right angle to the anterior-posterior axis (Fig. 6G, 8B, C). The neurites of cells D1–D5 all cross over the lateral longitudinal bundle to join the intermediate bundle. The intermediate longitudinal bundle is composed of an inner and an outer portion (Fig. 8C). The neurites of cells D1–D5 (and also of D6; Fig. 8E) consistently contact the inner fibres of the intermediate bundle which is another indication of their serial identity (arrows in Fig. 8C). In some specimens, the presence of faintly immunolabelled material suggested the neurites of some D neurons to proceed even more medially to join the medial bundle. Screening these regions with high-magnification laser-scan microscopy did not provide conclusive evidence that this is the case (inset in Fig. 8D, arrow).

The ventral ganglion is a condensation of the intraepidermal nerve plexus

A conspicuous feature of the chaetognaths is that in addition to the central part of the nervous system, a peripheral part is present which is exclusively intraepidermal [2,19,20,39] and which mediates motor innervation to the body musculature and innervates the ciliary fence receptors [23]. By using anti-aspartate immunohistochemistry, Duvert et al. [39] visualized this extensive intraepidermal nerve plexuses both in the head region and in the trunk. In the trunk, muscle contraction seems to be controlled by this profuse intraepidermal network [23,28,67]. Compared to other Bilateria, an unusual feature of the arrow worm neuromuscular system is that in the trunk axonal varicosities lack specialized junctions and are separated from underlying muscles by a thick connective stratum. Acetylcholine is the major neuromuscular transmitter, which reaches muscle cells in the trunk by diffusion through the intervening body wall extracellular matrix [2,20]. Acetylcholinergic fibres deliver their transmitter by boutons that in the trunk region terminate on the epidermis side of the connective stratum and therefore diffusely bath the muscle cells with acetylcholine [27,39]. It should be noted, however, that in the head, nervous fibres which target oesophageal and somatic head muscles, have conventional nerve endings and neuromuscular junctions that display ultrastructural features similar to classic motor end plates [2,28].

Our experiments on tubulin immunolocalization confirm and extend the observations on the structure of the peripheral plexus and ventral ganglion already published by Hertwig [17]. We showed an irregular array of ca. 20 – 30 fibres and fibre bundles to emerge laterally from both sides of the ventral ganglion. These fibres pass out radially to spread out diffusely and branch in the periphery thus being a major source for the peripheral nerve plexus. Duvert et al. [39] described a similar irregular array of aspartate-immunoreactive fibre bundles to spread out from the ventral ganglion. These findings together contradict earlier reports that in Sagitta, six [19] or twelve [2,18], distinct, bilaterally arranged radial nerves exit the ganglion on both sides in a serially ordered way. Our findings therefore defy any similarities of the chaetognath ventral ganglion with the internalized ventral nerve cords of e.g. annelids or arthropods that have segmentally arranged nerves projecting into the periphery. Rather, the arrow worm ventral ganglion appears to be a condensation of the intraepidermal nerve plexus. Hertwig [17] and Bone and Goto [20] observed numerous multipolar neurons to be embedded within the intraepidermal plexus. These neurons in the plexus may be the evolutionary precursors of the lateral clusters of cell bodies that flank the central fibrillar neuropil core of the ventral ganglion. During the emergence of Protostomia, the neurons in the plexus may have aggregated to form the centralized nerve centres that we observe in the Protostomia (Fig. 1).

Considering its role to coordinate sensory input from the fence receptors and efferent control of the trunk muscles for swimming behavior, Bone and Goto [20] proposed the ventral ganglion to be part of the central nervous system. From a histological point of view, the ventral ganglion can be seen as a condensation of the epidermal nerve plexus that displays a high degree of centralization. It represents a centralized, yet peripherally located center for complex sensory-motor integration. It is important to note that the chaetognath ventral ganglion is not internalized such as the subepidermal ganglia of Annelida, Arthropoda and Mollusca but remains in an intraepidermal position. It appears that, compared to other Protostomia, the Chaetognatha, by transforming a diffuse nerve net to a more centralized neuronal structure, followed their own distinct evolutionary pathway to generate a ventral nervous center in the trunk for sensory integration and motor control. Therefore, it may be appropriate to term this structure "ventral nerve centre" in order to stress the difference to ventral ganglia of other Protostomia. We suggest that this centralized nerve centre with its specific architecture and intraepidermal location is an autapomorphy of Chaetognatha.

The serial organization of the ventral nerve centre

Despite the suggested origin of the ventral nerve centre from the peripheral nerve net its central neuropil core displays a high degree of internal organization. Synapsin immunolocalization revealed a highly ordered system of serially arranged synaptic microcompartments in the ganglion core. This compartmentalization may anatomically be linked to a system of serially arranged transverse fibres that cross the neuropil core as reported by Bone and Pulsford [23]. It does not have any equivalent in the ventral ganglia of other Protostomia [73-78] and can be considered another autapomorphy of the chaetognath ventral nerve centre. Our data on the immunolocalization of the neuropeptide RFamide provide further evidence for serially arranged nervous elements in the ventral nerve centre of Sagitta setosa and extend previous reports on the localization of this substance in Sagitta setosa [27] and Paraspadella gotoi [38]. Most importantly, we provide evidence for serially arranged, individually identifiable neurons that can be homologized between different specimens. Although Goto et al. [38] (their Fig. 1d) did not map the pattern of RFamide-immunoreactive neurons in the ventral nerve centre of Paraspadella gotoi in detail, a comparison with our data nevertheless suggests that some of the L and D neurons may be evolutionary conserved across different chaetognath species. Papillon et al. [79] described the median Hox gene SceMed4 in embryos and early hatchlings of Spadella cephaloptera. This gene is expressed in two lateral stripes in the middle of the developing ventral nerve centre. The SceMed4 mRNA is localized in the bilateral soma clusters but not in the neuropil. These authors suggested that this gene may contribute to the diversity of neuronal subpopulations and to the establishment of distinct axon projection patterns [79]. It would be interesting to explore the expression of this gene in the organism that we studied, Sagitta setosa, to see if the region of expression coincided e.g. with the anterior-posterior transition from the "L" to the "D" type of RFamidergic neurons that we describe in the current study.

Individually identifiable neurons (see [80] for a discussion of this concept) seem to be present in the nervous systems of all major taxa within the Protostomia: e.g. Arthropoda [60,61], Annelida [74,77,78,81-83], Nemathelminthes/Cycloneuralia [84,85], basal Mollusca [73,74,86], Plathelminthes [87-89], and Gnathostomulida [90]. Therefore, we suggest that developmental programmes generating neurons with an individual identity must be present in the ground pattern of the Protostomia. A serial arrangement of individual nerve cells cannot only be found in protostomians with typical segmentation such as annelids and arthropods [60,77,78], but also in unsegmented organisms such as Nematoda [85], Plathelminthes [87-89], Chaetognatha (present report), and in organisms the segmental organization of which is unclear such as basal Mollusca [73,86,91]. Our concepts of segmentation in Protostomia mainly rely on work done in annelids and arthropods (reviews e.g. [92,93]). Budd [92] points out that "because segmentation is an evolutionary feature, it must have been acquired in a series of functional intermediates, and the sudden imposition of eusegmentation on a non-segmental precursor seems highly unlikely." In this view, the organization of individual organ systems such as the central nervous system (but not the entire organism) into serially repeated structures may be the starting point in an evolutionary trajectory from which segmentation as we see it in annelids and arthropods emerged [92].

Evolution of the bilaterian nervous system and the position of Chaetognatha

An intraepidermal nerve plexus is a prominent feature of many basal deuterostomes [55,68]. It is present e.g. in enteropneusts [94], in urochordates such as tunicates [95], and in the basal chordate Amphioxus [96]. An extensive intraepidermal nervous system also characterizes many Protostomia [3,55]. Recent investigations on Annelida [77,78,81] and Onychophora (basal arthropods; Mayer and Harzsch, unpublished data) provide new evidence that, in addition to the internalized parts of the nervous system, the peripheral plexus is a prominent feature in these organisms that seem to have retained more motifs of a flatworm-like orthogonal nervous system than has been perceived before. It has recently been proposed that an epithelial (epidermal) nerve plexus (as is also present in Cnidaria; [97-99]) without concentrations into longitudinal cords characterizes the ground pattern of Bilateria [56,64,65] (Fig. 1). Yet, the presence of at least some individually identifiable neurons in basal deuterostomes such as tunicates (Urochordata; [100-102]), and the lancelet Amphioxus (Chordata; [96]) indicates that the potential to establish individual identities of the neurons in the plexus may not only be present in the ground pattern of Protostomia (see above) but may date back to the ground pattern of Bilateria (Fig. 1).

The evolution of the deuterostome nervous system is a field of intense research [56,63-66,95,96,103-107] but it is beyond the scope of this contribution to embark into this issue. For the Protostomia, Nielsen [3,58] suggested that a perioral/circumesophageal brain in addition to the intraepidermal nerve plexus characterizes their ground pattern (Fig. 1). Such a circumoral brain can be recognized in most protostomian groups (e.g. [3] Fig. 12.2 therein; and [61,71]) and is by Nielsen [3,58] considered to be one of the key apomorphies of the Protostomia that may have evolved from a circumoral concentration of nerve cells around the mouth as present in Cnidaria. The brain components in Chaetognatha are also arranged in such a typical circumoral pattern (compare Fig. 2).

Experimental animals

Juvenile specimens of an unidentified species of the genus Sagitta were obtained from the coastal waters around the Mediterranean island Ibiza (Spain) in March 2006. A plankton net was towed across the surface waters of the Cala Llenya, Cala Vadella, Penyal de s'Aguila, and the Punta Grassio. Adult specimens of Sagitta setosa were obtained during a collection trip to the Biologische Anstalt Helgoland, German Bight http://www.awi-bremerhaven.de/BAH/ in August 2006. Specimens were obtained by horizontal (surface water samples) as well as by vertical (down to 20 meters depth) plankton hauls with the research vessel "MS Aade".

Histochemistry and immunohistochemistry

Specimens were fixed overnight at 4°C (or for 4 h, room temperature) in 4% paraformaldehyde (PFA) in phosphate buffer (PB; 0,1 M, pH 7.4). Histochemistry and immunohistochemistry were carried on free-floating whole mounts of adult specimens with fluorochrome-conjugated secondary antibodies using standard protocols. After fixation the tissues were washed in several changes of phosphate buffered saline (PBS) for at least 4 h, preincubated in PBS-TX (1% normal goat serum, 0,3% Triton X-100, 0,05% Na-acide) for 1 h and then incubated overnight in the following histochemical reagents and primary antibodies diluted in PBS-TX (room temperature):

• Phalloidin Alexa 488 (1:50; probe for actin; see e.g. [108]; Molecular Probes, obtained from MoBiTec, Göttingen, Germany)

• anti-acetylated alpha-tubulin from mouse (Sigma 1:100; see e.g. [109]).

• anti-FMRFamide from rabbit (1:1000; Diasorin; see e.g. [110]).

• anti-synapsin SYNORF 1 from mouse (1:10; [111]; antibody kindly provided by Prof. Dr. E. Buchner, Universität Würzburg)

For double labelling, combinations of these antisera were used, or the specimens were stained with the nuclear dye bisbenzimide (0.1%, 15 min. at room temperature; Hoechst H 33258), prior to the antibody incubations. Specimens were then washed for at least 2 h in several changes of PBS and subsequently incubated in secondary antibodies against mouse and rabbit proteins conjugated to the fluorochrome Alexa Fluor 488 (Molecular Probes, obtained by MoBiTec, Göttingen, Germany) for 4 hours. Finally the tissues were washed for at least 2 h in several changes of PBS and mounted in GelMount (Sigma). In control experiments, the omission of primary antibodies resulted in the absence of any neuronal labelling.

Microscopic analysis

Digital images were obtained with a Zeiss Axioskop fitted with a CCD-1300B digital camera (Vosskühler GmbH) and processed with the Lucia Measurement 5.0 software package (Laboratory Imaging Ltd.). Alternatively, samples were scanned with a Leica TCS SP2 AOBS confocal laser-scanning microscope (Lichtmikroskopiezentrum, Institut für Zellbiologie und Biosystemtechnik, Universität Rostock). Those images are based on stacks of between 15 and 20 optical sections (single images are averages of four laser sweeps) of a z-series taken at intervals of 1 μm.