Comparison with adult Oweniidae and outgroup taxa
The cns of Magelonids is located inside the epidermis (intraepidermal). The neuropil layer appears attached directly to the epidermal basal lamina. The ventral nervous system initially consists of lateral medullary cords which are located rectangular to the brain, but fuse in their further course to a single medullary cord. Ganglia or somata-free connectives are not present. The same is true for the putative sister taxon Oweniidae [8, 9], indicating that an intraepidermal cns with lateral medullary cords which fuse caudally might be the plesiomorphic condition for Annelida.
Contrary to the similarities found for the ventral nerve cord, the brain of magelonids is more complex in terms of neuron types. The anterior brain neuropil in magelonids is compact and the dorsal part of the brain is enlarged when compared to the brains of Oweniidae. In oweniids the brain is a simple ring surrounding the mouth (Beckers et al. 2019), while in magelonids the anterior brain is compact and posteriorly encircles the prostomial coelomic cavities. However, compared to the brain of errant polychaetes [1], the cns of Magelonidae is rather simple. There are no ganglia present, neither in the anterior cns nor in the ventral nervous system. The whole cns is medullary. No tracts such as ventral or dorsal commissural tracts described by Orrhage for M. papillicornis [15] were found; neurites in the neuropil are more or less homogenous. Neither glomeruli in the brain as described for errant polychaetes [14, 22] or hoplonemerteans [23] nor mushroom bodies [1, 24] were found during our investigation. In Oweniidae only one type of neurons is known within the cns [9]. This neuron type is defined by the very little cytoplasm surrounding the nucleus and is also the most abundant type in Magelonids. However, in Magelona mirabilis and M. alleni a second and third type of neuron is present, which are located in the dorsal or dorso- lateral parts of the brain. The cytoplasm of these neurons is enlarged in comparison to the first type, whereas the second one is smaller than the third one. Furthermore, these neurons appear in clusters. Clustered polymorphic neurons are also found in Amphinomidae and Sipuncula as well as in pleistoannelid taxa [1]. However, polymorphic neurons are not described for Chaetopteriformia [1], which form another basally-branching clade within the annelid tree and therefore hamper the explanation of the evolution of these neuron types [8].
In basally branching taxa of the putatively outgroup taxa Nemertea and Mollusca the cns is also medullary and the brains of these taxa are more or less circular shaped. Polymorphic neurons and ganglia are not present [25, 26]. The same is true for Bryozoa, Brachiopoda and Phoronida [27,28,29]. Accordingly, the anatomy of the cns of spiralian relatives shows more similarities to the adult anatomy of Oweniidae and questions the sister-group relationship of oweniids and magelonids [8].
One might wonder why the nervous system of Magelonidae is more complex than the brain of the sister taxon Oweniidae, although adults of both taxa are sedentary and feed on detritus [30]. We can only speculate and suggest that the life style of the larva in magelonids might play a role in this respect. Thus, the larva of Magelonidae is predatory and actively feeds on bivalve larvae at least for 7 month in the plankton [31] and might need a more sophisticated nervous system for hunting. The larva of Owenia fusiformis, however, passively feeds on algae [32].
Peripheral nervous system
A small dorsal nerve, as described for several other annelids [1], originates in the posterior dorsal part of the brain and extends caudally inside the epidermis. Since a dorsal nerve is not present in Oweniidae [9] a convergent evolution or a consecutive branching of oweniids and magelonids has to be assumed. Since there is no nuchal organ, the assumption that this nerve innervates the latter, seems not likely (see 1 for review).The epidermis is innervated by a dense network of mainly dorso- ventrally orientated neurites. These neurites connect the lateral medullary cords and the dorsal nerve like in other investigated annelids [1, 2]. An epidermal plexus as described for Oweniidae [9] is also present, indicating the plesiomorphic condition for Annelida.
Sensory structures
In previous investigations, a lateral organ was mentioned to be present in adult specimens of Magelona mirabilis [19, 20], but only few details were known. It is composed of a densely ciliated spot between neuro- and notopodium. The spatial location and general anatomy resembles the lateral organs described for other polychaetes [19]. The evolutionary considerations behind that finding are difficult to evaluate with present knowledge, since Oweniidae as well as Chaetopteridae lack such an organ, but Apistobranchidae and Amphinomidae (both part of the radiation outside the Pleistoannelida) and at least some Pleistoannelida possess such an organ [19]. Further ultrastructural investigations are needed to clarify the evolutionary origin and value of the lateral organ in Annelida, because the dataset especially for basally-branching taxa is still quite scarce.
Nuchal organs are absent in adult and juveniles of Magelonidae thus far investigated (16; this study). Since these organs are also absent in outgroup taxa as well as in Oweniidae and Chaetopteriformia [9] this complex sensory organ evolved most likely in the stem lineage of Amphinomidae/Sipuncula and Pleistoannelida.
Adult stages of Magelonidae do not possess pigmented eyes, although they are present in larval stages. This situation is comparable to the situation in e.g. Owenia fusiformis, where the larva bears pigmented eyes of the cup- shaped type, while adults only possess rows of pigmented cells [9]. Why eye complexity or even the complete eyes are reduced during ontogenesis of these species remains unclear.
Palps
Palps of adult Magelonidae are located ventro- laterally and palp nerves are connected to the dorso- and ventro- lateral parts of the brain. Notably, the palps of the basally branching oweniid taxon Myriowenia sp. are also innervated by two nerves each – in this case- originating in the dorsal and dorso- lateral part of the brain [9]. Nevertheless, we assume that the palps and palp nerves of both taxa are homologous and that spatial differences concerning the innervation points of the palp nerves in the brain might be caused by different head morphologies of both annelid taxa [9]. Two nerves also innervate the palps in Magelona papillicornis, Spionidae and Chaetopteridae [15], indicating that this pattern resembles the plesiomorphic condition in Annelida. Nevertheless, further investigations including more taxa spread all over the annelid tree are needed to investigate the evolution of palp innervation or general homology of palps throughout annelids.
Glial cells
Glial cells contain intermediate filaments which classifies these cells as radial glia [33]. Furthermore, elongated cells traversing of the epithelium and basal endfeet in contact with the basal lamina, are characteristics of these cells. Radial glia seems to be especially important in species with an intraepidermal nervous system. These cells maintain the structure of the epidermis with an underlying nervous system. Although radial glial cells can be ascertained, a prominent glial layer surrounding the neuropil and somata as described for errant polychaetes [1] is not present. Since a prominent glial layer is suggested to protect the nervous system against mechanical distortions [33, 34] we suggest that the nervous system of magelonids is not that subjected to mechanical stress as the nervous system of e.g. errant species. Glial cell processes which wrap the neuronal somata are also not present due to our investigations.
Giant fibre
Giant fibres are supposed to be involved in rapid signal conduction [35, 36] and are described for a variety of Pleistoannelida [1, 14] and for the putative sister taxon Myriowenia sp. (Oweniidae) [9]. The structure of this fibre in Magelona mirabilis resembles the anatomy of this fibre in Myxicola infundibulum (Sabellidae, Pleistoannelida) [37, 38], Siboglinidae [39, 40] and Lumbricus terrstris (Clitellata, Pleistoannelida) [36] indicating that a giant fibre was already present in the last common ancestor of Annelida. However, the giant fibre in Magelona is not composed of two intersecting large axons with anterior located somata as is the one in Myxicola [37]. Somata are most likely scattered along the course of the fibre. Whether giant fibers present in other invertebrate taxa are homologous remains unknown [41] but the absence of these fibres in nemerteans and platyhelminthes as well as in Brachiopoda and Bryozoans hint on a convergent evolution. Giant fibres in Octopus (Cephalopoda, Mollusca) [42], must have evolved convergently, since Polyplacophora and Solenogastres (Mollusca) lack such a fibre type.
Comparative analysis of Orrhages reconstructions of the nervous system in adult Magelona papillicornis and this study
In his work on the nervous system of sedentary polychaetes, Orrhage described the anatomy of the brain in Magelona papillicornis based on histological sections [15] (Additional file 2). However, since Magelona papillicornis is originally described from Brazil, Fiege et al. [43] doubt the presence of this species in Europe and suggest that this species is not Magelona papillicornis, but most likely M. mirabilis [43]. In order to compare the results of Orrhage with our data, we re- investigated and digitalized the original sections he used for his schematic drawings of his 1966 paper [15]. Information content of his sections and staining are of the same quality like ours (Additional file 2). However, not all of his descriptions are in congruence with our observations. Although we carefully investigated the anatomy of the neuropil of the brain with different methods, we did not find any of the four commissures described by Orrhage (dKvS, vKvS; dKhS, vKhS in [15]) (Fig. 12). Instead, the neurites of the brain neuropil are homogenously arranged. Additionally, he did not recognize the medullary nature of the nervous system and found several ganglia. So, he wrongly described a cluster of neurons in the dorso-lateral part of the brain (DG in Fig. 12b) but did not mention the cluster of polymorphic neurons in the median dorsal part of the brain (S2 in Fig. 12a). The finding of the cluster of neurons with very prominent somata is validated (S3 in Fig. 12a; hg in Fig. 12b). Additionally, his observations on the innervation of the palp correspond to our results. Orrhage mentioned that the lateral medullary cords are connected to the brain by two nerves (hS and vS, frontal esophageal connective and posterior esophageal connective; Fig. 12). He interpreted this as two roots of the circumesophageal connectives (cc). However, this arrangement differs from the anatomy of roots of the cc in e.g. Eunicida. In these taxa the cc split close to the brain into two or more nerves (roots) which are connected to different commissures in the brain [2]. Since commissures in the brain of adult Magelonidae are absent and the connection of the second nerve cord leading to the cc is located more posterior and associated with the palp nerve, it is difficult to give any homology assumptions of this part of the brain to the different roots of the cc described for Errantia (Pleistoannelida). As shown in our investigations, different brain commissures forming roots inside the neuropil or a circumesophageal connective are only observable in larval stages of magelonids. Due to the medullary nature of the whole central nervous system in adult magelonids, terms like circumesophageal connectives have to be avoided.
Comparative larval neuroanatomy
Based on our analyses the ontogenesis of the larval neural structures is comparable to conditions observed in other annelid larvae. Thus, a well-developed apical organ is present in magelonid larvae, possessing flask-shaped FMRFamidergic perikarya as well as serotonergic cells both early in development and throughout the entire larval ontogenesis. Comparable apical cell clusters showing both types of immunoreactivity can also be observed in most annelid larvae (Bleidorn et al. 2015), but serotonergic somata are absent in the putatively closely related early branching annelid Owenia fusiformis [18]. Notably, this lack of apical serotonergic somata is supposedly caused by the unusual larval shape and related re-arrangement of neural structures in larval Oweniidae.
Besides the developmental similarities to various annelid taxa, the larval magelonid nervous system is also well comparable to that of other spiralians. Even though the annelid sister group is not resolved yet, potentially closely related groups are represented by the Mollusca, Nemertea, Platyhelminthes, Phoronida and Brachiopoda [44, 45]. Prominent serotonergic and FMRFamidergic perikarya are also present in the larval apical organ of Polyplacophora [46, 47], Scaphopoda [48], Solenogastres [49], Brachiopoda [50] (shown only for serotonin), Phoronida [51] and Nemertea [25] and are also reported for Entoprocta and Platyhelminthes [51,52,53,54]. Absence of serotonergic somata has been only reported for Ectoprocta, so far [55] (and for Oweniidae). Thus, apical presence of 5-HT and FMRFamide-lir seems to be a plesiomorphic spiralian characteristic.
Furthermore, all the latter taxa, as well as the observed Magelonidae and most other annelid larvae, also develop at least two main nerve cords with numerous commissures and connection to the apical neuropil in early stages, and possess a distinct nerve ring surrounding the mouth opening [1]. Instead, a nerve ring underlying the prototroch is lacking in investigated magelonids.
The occurrence of different roots of larval brain commissures and the presence of two ventral nerve bundles early in the development hint towards an early onset of adult structures even in early larval stages of these annelids, in agreement with comparable structures that are known for adult nervous systems of other Annelida (Orrhage and Müller, 2005). There are two explanations for this situation: on the one hand we can argue that developmental investigations herein show a reduction of neuronal complexity from larval towards adult conditions. Although larval brain roots and well-developed circumesophageal connectives are present in early magelonid stages and well-known for the anterior adult neuroanatomy of different pleistoannelid taxa [1, 2], comparable structures are absent in adults of M. mirabilis. Just a dorsal and ventral part of the brain can be examined in this case. Similar conditions are obvious in oweniids – larval stages possess several brain commissures whereas adult specimens bear a simple ring-shaped brain without distinct roots [9, 18].
On the other hand, these findings can be interpreted in a different way: brain commissures can, per definition, only be present, if they are part of a compact brain neuropil. Any tracts that are found in different parts of the brain and that are not part of a uniform neuropil do thus not form a commissure. Since there is no compact mass present in tubulin- lir immunostainings in larval stages of Magelona mirabilis, we can also suggest that the dorsal and ventral part of the brain (including the roots) present in larval stages are retained in the adult stage and that the anterior compact neuropil develops later during ontogenesis. Additionally, the innervation of the palp from both brain parts (dorsal and ventral), like in the larvae, also hints towards the scenario that this larval brain may be part of the adult brain which surround the coelomic cavities. Such an ontogenetic transformation of the larval morphology with retention of the general assemblage concerning the respective brain areas in adult magelonids might be an explanation for the described conditions.
Nevertheless, one has to keep in mind that different brain commissures as described for the larval brain neuropil in several Pleistoannelida [2] are not present in all taxa, e.g. not in Capitella telata [56].
Evolution of the nervous system in Annelida
The anatomy of the anterior nervous system of Magelonidae shows several characters comparable with the neuroanatomy of the closely related Oweniidae and possible annelid outgroup taxa as well as with the complex brains of pleistoannelid taxa. Thus, the intraepidermal position of the entire cns and the lack of ganglia presents the plesiomorphic annelid condition, which can also be investigated in annelid sister groups and in other basally-branching annelid taxa. On the other hand the enlargement of the dorsal part of the magelonid brain, the compact anterior neuropil and the presence of clusters of polymorphic neurons are characters only present in more derived annelid groups. Additionally a lateral organ is present, which is lacking in Oweniidae but present in several other derived Annelida. In contrast nuchal organs, complex cup-shaped eyes must have developed in the stem lineage of Sipunculida+ Amphinomidae and Pleistoannelida (Errantia+ Sedentaria).