Annelid muscles have been studied in depth in the adults of several species [8, 28, 29] but investigations of the ontogeny of these muscle systems are largely lacking. Phalloidin labelling of F-actin has facilitated studies of myogenesis in many invertebrate phyla, including acoels , molluscs [30, 31], entoprocts , phoronids , cycliophorans  and sipunculans . One study of phalloidin labelling in Capitella sp. 1 larvae  shows that myogenesis in this polychaete annelid first begins with a ring of muscle around the stomodaeum and is followed by the development of numerous longitudinal and then circular body-wall muscles. These muscles persist in the adult. Capitella, however, possesses non-feeding larvae that are brooded and emerge as highly derived metatrochophores, skipping the typical trochophore stage [16, 32]. In P. lamarckii, myogenesis is first seen in the developing gut. Oesophageal muscles and the prototroch and metatroch muscle bands form in the trochophore larva, and body wall musculature is not added until the trochophore elongates to form the metatrochophore. It therefore appears that a change in timing has occurred in the development of P. lamarckii and Capitella, with Capitella developing body wall musculature first, and P. lamarckii forming the musculature of the digestive system before any body wall muscles. This is an example of heterochrony in development and appears to correlate with the different feeding strategies of the two species, with the planktotrophic P. lamarckii developing feeding musculature earlier in development whereas in the non-feeding Capitella larvae the process is delayed. Which, if either, represents the ancestral pattern of myogenesis is a moot point, as it is a matter of debate as to whether lecithotrophic or planktotrophic larvae are basal [33–35]. In both cases, however, the body wall musculature that develops in the larva persists through to adulthood. This is likely to be a common feature of polychaete annelid larval development.
A prototroch muscle ring has been observed in many molluscan larvae, including polyplacophora, bivalves and gastropods, but is distinctly lacking in the scaphopod Antalis entalis [30, 31]. In addition, the sipunculan Phascolion strombus lacks a muscle ring associated with its prototroch . Although the homology of the mollusc and annelid prototroch (and its muscle) is likely, based on embryological considerations , the possible (although still uncertain) placement of the Sipuncula as a sister group to the annelids [36, 37] raises the possibility that the prototroch muscles are independent innovations in molluscs and annelids. A more likely explanation is that the sipunculan prototroch and its associated structures are secondarily reduced. In sipunculans, the prototroch does not develop to its full extent, and the animal uses a post-oral ring of cilia for locomotion . This could explain the lack of muscles (and nerves) in the sipunculan prototroch, and does not invalidate the hypothesis that the prototroch is a homologous structure in trochozoans.
In addition to the prototroch muscle ring, P. lamarckii also possesses a metatroch muscle ring. This muscle ring has been described in other polychaete species [19, 24]; however its distribution outside the Annelida is undocumented.
P. lamarckii exhibits no circular muscle in the body wall (except for the prototroch and metatroch muscle bands) in larvae or juveniles. Muscles repeated in the segments of later stages are components of the chaetal musculature (see Fig. 3F), and should not be confused with circular muscles . The widely accepted view on adult annelid musculature is that the body wall consists of an outer layer of circular muscle and an inner layer of longitudinal muscle . This general assumption has been challenged by the finding that members of several groups of polychaete annelids lack circular muscle altogether [28, 29]. The transverse muscles seen in P. lamarckii are not likely to be rudiments of these circular muscles as they are associated with the chaetae and not with the body wall. Thus reinvestigation into the ancestral state for body wall muscles in the Annelida is warranted, which must be combined with a more robust annelid phylogeny [10, 28, 29].
P. lamarckii larvae possess a conspicuous prototroch, which is the first ciliary feature to appear on the developing embryo. The prototroch ring is initially open on the dorsal side of the larva, as has been documented in other polychaetes [18, 26], and appears to form from two rows of cells. Additional cells may contribute to the prototroch as seen in other annelids , but these were not visible here. Later ciliation in P. lamarckii is quite typical for trochophore larvae and consists of an apical tuft, adoral ciliary zone, metatroch, and neurotroch, i.e., all the components of the 'opposed-band feeding' system . A ventral break in the metatroch, allowing the neurotroch to reach anteriorly to the mouth, has been described previously for some annelids [24, 26]. This does not appear to be the case in P. lamarckii, instead the metatroch appears to be continuous with cilia that run inside the oesophagus. P. lamarckii lacks a telotroch at all stages of development. Early in development, cilia become visible in the developing gut and tubulin staining clearly highlights a pair of protonephridia, which either lead to the anus or a paired opening in the vicinity of the anus. These protonephridia can not be seen in later stages, and possibly degenerate, as is the case in some phyllodocid larvae . Anti-tubulin immunohistochemistry has been used in numerous studies to visualise axon fibres in developing larvae (for example see [18, 41, 42]), however in P. lamarckii intense staining of ciliated bands made visualisation of neural elements impossible until the late stages of development.
Neurogenesis in P. lamarckii
Immunohistochemical analyses of neurogenesis have been carried out on the larvae of three polychaetes, Polygordius lacteus (Polygordiidae) , Phyllodoce maculata (Phyllodocidae)  and Pomatoceros lamarckii (Serpulidae) (this study). These larvae show basic similarities; all develop paired ventral nerve cords, possess a suboesophageal ganglion, an apical organ and later a cerebral ganglion, develop a prototrochal nerve ring and a complex of nerves in the oesophagus (described in detail in ). The larvae differ in terms of the timing of formation of various features, and there are some differences in neuronal structures. Each of these larvae are morphologically and developmentally different, with P. lamarckii possessing a free-swimming planktotrophic larva, P. maculata possessing an initially encapsulated larva  and P. lacteus possessing a free-swimming larva that develops adult segments inside the hyposphere in later larval stages [19, 43]. P. maculata and P. lamarckii are most likely phylogenetically distant from one another, while the position of the Polygordiidae is still unresolved . Neurogenesis in a species of Myzostomidae has also been described but will not be discussed here as the phylogenetic placement of this species is uncertain and their body plans are highly derived, due to their parasitic lifestyle [17, 42].
The finding that neurogenesis in gastropods proceeds from the posterior periphery of the organism to the anterior and centre challenged the widely held view that neurogenesis follows a strictly anterior to posterior pattern in these animals [44, 45]. In response to this Voronezhskaya and colleagues  examined neural development in the polychaete Phyllodoce maculata to investigate whether this posterior-first pattern was more widespread. In P. maculata the first immunoreactive cell is a posterior serotonergic cell that gives rise to the ventral nerve cords. Immunostaining in the apical region is not seen until twelve hours later. Therefore, neurogenesis in P. maculata is similar to that seen in molluscs, i.e. progressing from the periphery to the centre.
The posterior serotonergic cell seen first in P. maculata neurogenesis is strikingly similar to the posterior serotonergic cell seen early in P. lamarckii development. Both cells are in a similar position in the embryo and contribute fibres to the ventral nerve cords. In P. lamarckii, however, this cell arises at a very similar time to (if not after) the first anti-FMRFamide signals in the apical organ. It is unknown whether neurogenesis in P. lacteus begins at the anterior or posterior of the embryo because data on the early stages is not available . In light of these studies, peripheral to central neurogenesis does not appear to be the case in all polychaetes. We suggest the mode of neurogenesis is variable and subject to heterochronic shifts depending on the life history of the organism, i.e., that planktotrophic species require the development of anterior and posterior neural elements early in development for coordination of swimming and/or feeding, whereas lecithotrophic species do not require the anterior elements until later in development.
The neuroanatomy of the polychaetes studied showed several features that are shared between many trochozoans. A prototroch nerve ring is found in all three species, and a metatroch nerve ring is found in P. lamarckii and P. lacteus (P. maculata lacks a metatroch). In both P. lacteus and P. maculata these nerve rings are reactive for both serotonin and FMRFamide, whereas in P. lamarckii only serotonin staining is observed. These nerve rings possess multiple fibres, an observation that has also been made in the prototroch of bivalve molluscs , and the marginal ciliary band of the pilidium larvae of nemerteans , while polyplacophorans and ectoprocts have a more extensive nerve net [48, 49]. Therefore, complex innervation of ciliary bands may be a common feature of Lophotrochozoa. A medial nerve, situated between the ventral nerve cords, is only seen in P. lamarckii, although it appears to develop in the very late larvae and may have been present in stages outside the scope of the study by Voronezhskaya et al. (2002). This median nerve is likely to be homologous to the 'neurotroch nerve' of Spirobranchus polycerus described by Lacalli  using transmission electron microscopy. Both the prototroch nerve ring and the median nerve have been identified in several other trochozoan taxa, although the neurotransmitter content of these structures varies [11, 38, 45, 50]. Neurotransmitter content of a neuron can change during development [51, 52], therefore the difference in neurotransmitters does not necessarily mean these structures are not homologous. This also applies to the ventral nerve cords, which are considered to be homologous in protostomes but also differ in neurotransmitter content .
Comparisons with earlier reconstructions of the serpulid nervous system using light microscopy  shows that visualisation using antibodies against FMRF-amide and serotonin gives a comprehensive picture of neurogenesis in these animals. From his reconstructions, Lacalli  stated that the larval nervous system in S. polycerus consisted of two separate parts: the pre-trochal part comprising the apical organ, prototroch, and connections between, and the oral part comprising pharyngeal nerves, the metatroch nerves and the neurotroch nerve. He also proposed that the adult nervous system develops entirely separately from the larval nervous system. This latter finding was disputed by Voronezhskaya and colleagues , and is not supported by this study. Both Voronezhskaya et al.  and ourselves were able to detect the posterior serotonergic cell and the rudiments of the ventral nerve cords, which were either missed in the S. polycerus reconstructions or are not present in this particular polychaete annelid . In the immunohistochemical studies on P. maculata connections were seen between the oesophageal nerves and the prototroch . In our P. lamarckii study connections between what Lacalli  calls the pre-trochal and oral systems were also detected, notably between the metatroch and ventral nerve cords. The larval nervous system of polychaetes is thus interconnected, and reorganises into the adult nervous system at metamorphosis.
In the metatrochophore of P. lamarckii, the serotonergic ventral nerve cords have a double appearance (Fig. 6F). The pattern seen here strongly resembles that seen in Lacalli's reconstruction of S. polycerus , where the inner cords give rise to the anterior-most commissures and the outer cords give rise to those toward the posterior pole. However, it is not known how integrated the two parts of the cords are to each other, and therefore whether they should be treated as separate nerve cords.
Molecular phylogenetic analyses have shown that Annelida and Sipuncula are closely allied taxa [54, 55]. Indeed, the nervous systems of the two groups are similar . The ventral nerve cords and cerebral ganglia are reactive for serotonin and FMRFamide in both taxa, and the dominant neurotransmitter in the apical organ in early stages is FMRFamide. Both P. lamarckii and the sipunculan Phascolion strombus possess an FMRFamidergic medial nerve and peripheral nerves associated with the viscera. In late larval stages convergence of serotonergic fibres in the ventral nerve cords can be seen posteriorly in both species. Obviously, these features are not shared by all groups of annelids and sipunculans, for example the annelids Phyllodoce maculata and Polygordius lacteus do not show any evidence of a medial nerve [18, 19]. However the fact that members of both phyla possess these characters is consistent with a close relationship between the two taxa. Similar comparisons with other lophotrochozoan taxa can be made when more data are available.