Gene expression in the developing nemertean brain indicates convergent evolution of complex brains in Spiralia

Background Nemertea is a clade of worm-like animals, which belongs to a larger animal group called Spiralia (together with e.g. annelids, flatworms and mollusks). Many of the nemertean species possess a complex central nervous system (CNS) with a prominent brain, and elaborated chemosensory and neuroglandular cerebral organs, which have been suggested as homologues to the annelid mushroom bodies. In order to understand the developmental and evolutionary origins of complex nemertean brain, we investigated details of neuroanatomy and gene expression in the brain and cerebral organs of the juveniles of nemertean Lineus ruber. Results In the hatched juveniles the CNS is already composed of all major elements present in the adults, including the brain (with dorsal and ventral lobes), paired longitudinal lateral nerve cords and an unpaired dorsal nerve cord. The TEM investigation of the juvenile cerebral organ revealed that the structure is already composed of several distinct cell types present also in the adults. We further investigated the expression of twelve transcription factors commonly used as brain and cell type markers in bilaterian brains, including genes specific for annelid mushroom bodies. The expression of the investigated genes in the brain is region-specific and divides the entire organ into several molecularly distinct areas, partially overlapping with the morphological compartments. Additionally, we detected expression of mushroom body specific genes in the developing cerebral organs. Conclusions At the moment of hatching, the juveniles of L. ruber already have a similar neuroarchitecture as adult worms, which suggests that further neural development is mostly related with increase in the size but not in complexity. Comparison in the gene expression between L. ruber and the annelid Platynereis dumerilii and other spiralians, indicates that the complex brains present in those two species evolved convergently by independent expansion of non-homologues regions of the simpler brain present in their common ancestor. The similarities in gene expression in mushroom bodies and cerebral organs might be a result of the convergent recruitment of the same genes into patterning of non-homologues organs or the results of more complicated evolutionary processes, in which conserved and novel cell types contribute to the non-homologues structures.

Most nemerteans are active predators, which hunt for their invertebrate prey using a specialized eversible proboscis, a morphological apomorphy of the clade [1,[14][15][16][17][18]. This active lifestyle is accompanied by a relatively complex nervous system, composed of a large, multilobed brain (with two ventral and two dorsal lobes), a pair of lateral medullary nerve cords, extensive peripheral network and multiple specialized sensory organs [17][18][19][20][21][22][23][24][25][26][27][28][29]. Among the latter, the most conspicuous are the so-called cerebral (or cephalic) organs -paired structures of neurosecretory and either chemo-or mechanosensory function, located on the lateral sides of the head [17-23, 28, 30-33]. The exact arrangement of the cerebral organs varies between nemertean clades from relatively simple ciliated pits present in some Tubulaniformes, to the complex neuroglandular structures connected both directly to the brain and, through the convoluted ciliated canal, to the external environment in lineid heteronemerteans [17-23, 27, 28, 32, 33]. The phylogenetic analysis of morphological traits in nemerteans indicated that cerebral organs were already present in the last common nemertean ancestor [20]. However, it remains unclear, whether the cerebral organs represent an autapomorphy of nemerteans or homologs to some organs present in other spiralians such as ciliated pits of flatworms [30,34] or mushroom bodies of annelids [19,35,36].
In the present study, we describe the detailed morphology of the nervous system and gene expression in the brain and cerebral organs of the juveniles of Lineus ruber (Müller, 1774), a directly developing lineid heteronemertean. L. ruber has been studied in past for both adult morphology [20,[22][23][24][25][26][29][30][31] and some aspects of its development [29,37,38], including the molecular patterning of anterior-posterior axis, germ layers and lateral nerve cords [39,40]. Comparison of our data with the existing morphological descriptions of the adult nervous system in L. ruber [20,[22][23][24][25][26][29][30][31] and other closely related species, allows a better understanding of the ontogeny of the complex nemertean nervous system. Additionally, juxtaposition of gene expression profiles in the developing brain of L. ruber with that of other Spiralia [39,[41][42][43][44][45][46][47][48] can pinpoint similarities and differences in the molecular patterning of the spiralian brains in general, which in turn can inform evolution of the complex nemertean brain. Moreover, by comparing gene expression in cerebral organs of L. ruber and mushroom bodies of a comprehensively studied annelid P. dumerilii [49], we can provide new data to test the homology hypothesis of the cerebral organs of nemerteans and mushroom bodies of annelids.

Morphology of the nervous system in the juvenile L. ruber
The investigated juveniles of L. ruber were freshly hatched from the egg mass, 42 days after oviposition [40]. We visualized the nervous system of the juveniles by applying antibody staining against tyrosinated tubulin, FMRF-amide and serotonin (5-HT), as well as Sytox green nuclear staining and fluorescent in situ mRNA hybridization of the choline acetyltransferase (ChAT), a genetic marker of the cholinergic neurons [50].  (dbc Figs. 1 and 2A, C, E, G, I) and ventral (vbc Figs. 1 and 2B, D, F, H-J) brain commissures, respectively. Thus, the brain neuropile forms a ring around rhynchocoel and proboscis (Fig. 1). Posteriorly, each dorsal brain lobe is further divided into an inferior and a superior branch. The former connects directly to the cerebral organ (see below), while the latter ends blindly on the dorsal side of the animal (Figs. 1 and 2E). The neuropiles of the ventral lobes posteriorly give rise to the LNCs (Fig. 2D, H, J). FMRF-amidelike immunoreactive (FLIR) perikarya and ChAT + cells have been observed in both dorsal and ventral brain lobes ( Fig. 2A -F Compared to the LNCs, it is much finer and does not seem to be associated with any perikarya (Figs. 2A, G, I). It is composed of only a few TLIR and SLIR neurites, while anteriorly, a pair of fine FLIR dorsal accessory nerves branch out from it (adn, Figs. 1 and 2A). At the level of the pharynx, a fine, SLIR and TLIR dorsal commissure connects dorsal and lateral nerve cords (dc,

Figs. 1 and 2I).
The SNS is composed of thick TLIR, FLIR and SLIR pharyngeal nerves, which originate from the ventral brain lobes and meander around the pharynx (phn, Figs. 1, 2C, D, J). Numerous sensory FLIR and SLIR cells are located along the pharyngeal nerves (psc, Fig. 1; double arrowheads A pair of conspicuous cerebral organs is located on the lateral sides of the head, just behind the brain (co, Figs. 1, 2C and E). More details of their morphology can be found in the following section. Other sensory structures, detected in addition to the cerebral organs, includes FLIR and ChAT + anterior sensory cells (asc, Figs. 1 and 2 C, E), which likely contribute to the socalled frontal organs [19,22,23,25], and numerous SLIR cephalic nerves extending anteriorly from the brain (an, Figs. 1 and 2H). Although 42 days old juveniles already possess rudiments of eyespots [40], we were not able to conclusively detect them in our investigation. EdU staining in 60 days old juveniles showed that most of the brain cells at this later developmental stage are not mitotically active in contrast to the cells in other organs, such as proboscis, rhynchocoel or cerebral organs ( Fig. 3A and B).

Detailed morphology of the cerebral organs
Each cerebral organ is composed of two parts: a distal ciliated canal (cc, Figs. 1, 2C, 4B, C), which opens to the exterior on the side of the head (in the posterior part of the so called lateral cephalic slit), and a proximal neuroglandular portion (co, Figs. 1, 2C). The lumen of the ciliated canal is slightly curved in 42 days old juveniles, but the characteristic triple right-angle bends, present in the adult lineids [30][31][32] are not yet evident (cc, Fig. 2C). The ciliated canal connects the external environment with the neuroglandular part, which itself is firmly attached to the superior branch of the dorsal brain lobe (Fig. 1, 2E, and 4B, C). A thick TLIR and FLIR nerve of cerebral organ extends from the most posterior part of the dorsal lobe neuropile and penetrates the neuroglandular portion of the cerebral organ (con, Fig. 4C). We detected a few FLIR and much more numerous ChAT + cells in the neuroglandular portion of the organ (arrowhead, Fig. 4C and arrow, Fig. 4B, respectively), while serotonin-like immunoreactivity was not detected (data not shown).
To gain further insight into the morphology of the cerebral organs, we supplemented the afore-mentioned confocal laser scanning microscopy (CLSM) based methods with ultrathin sectioning of resin-embedded specimens (60 days old juveniles) and TEM examination of the organ. That allowed us to describe the fine structure of the cerebral organ and ultrastructure of the particular cell types contributing to it. Since all detected cell types correspond directly to the ones described previously by Ling in his investigation of adult L. ruber [30], we adopted the terminology used therein. We investigated cross-sections through the neuroglandular portion of the cephalic organ. The mass of the organ is located between the proboscis and the lateral nerve cords (Fig. 4A) and it is penetrated by both the cerebral organ nerve (con) and the ciliated canal (cc). The ciliated canal is divided into two parallel parts: a larger major ciliated canal (mjc) and a smaller minor ciliated canal (mnc) (Fig. 4G). Based on the ultrastructure, six distinct cell types can be

Gene expression in the head
We investigated expression of 12 transcription factors (TFs), which have a role in CNS development of many bilaterians. Those genes include the conserved general brain markers (otx, bf1), genes involved in brain regional specification (pax6, nk2.1, nk2.2, rx, otp) and other neural genes, which are co-expressed in the annelid mushroom bodies (dach, emx, arx, svp, tll).
Expression of otx has been previously described for earlier developmental stages of L. ruber, in which the gene has a general anterior expression in the head [40]. In the 42 days old juveniles, which we investigated, the gene otx is predominantly expressed in the brain ( Fig.   5A and B) and cerebral organs (Figs. 5A, 6B). In the brain, otx is broadly and uniformly expressed both in dorsal and ventral lobes ( Fig. 5A and B). In the cerebral organs it is also widely expressed, both in the ciliated canal and neuroglandular part (Fig. 6B). A similar expression pattern of otx in the brain and cerebral organs has been also reported from developing juveniles of closely related Lineus viridis [52].
bf1 is expressed in the brain, cerebral organs, scattered cells in the anterior epidermis and in the rhynchocoel (Figs. 5C and D, 6C). In the brain bf1 is broadly expressed in the dorsal lobe ( Fig. 5C), but in the ventral one it is only detectable in the lateral clusters of cells (Fig. 5D). The detected expression of bf1 in the cerebral organ is very strong in the neuroglandular part, whereas we did not detect a signal in the ciliated canal (Fig. 6C). Abbreviations: DA dorso-anterior brain domain, DL dorso-lateral brain domain, DM dorso-median brain domain, VA ventro-anterior brain domain, VL ventro-lateral brain domain, VM ventro-median brain domain.
Expression of pax6, nk2.1 and nk2.2 has been previously investigated in the juveniles of L. ruber in relation to the nerve cord patterning [39], however, the expression of those three genes in the brain was not described in the details that we provide here. In the head region, pax6 is expressed in the brain, the epidermal cells of the lateral cephalic slits and in the cerebral organs (Figs. 5E and F, 6D). The gene is broadly expressed in the dorsal lobes (Fig.   5E), while in the ventral ones its expression is restricted to the lateral portions of the brain (Fig. 5F). In the cerebral organs the gene is expressed in the stripe of cells on the lateral side of the neuroglandular portion (Fig. 6D).
In the head region, nk2.1 is expressed in the brain and proboscis ( Fig. 5G and H). In the dorsal lobes the gene is expressed only in the small lateral clusters of cells (Fig. 5G), while on the ventral side the gene is broadly expressed both in the median and lateral domains (Fig. 5H).
nk2.1 is not expressed in the cerebral organs.
nk2.2 is expressed in the brain, proboscis and cerebral organs ( Fig. 5I and J). In the dorsal brain lobes, the gene is expressed in large clusters of posterior cells and in scattered anterior domains (Fig. 5I), whereas ventrally, it is expressed in median and lateral cell clusters (Fig 5J).
Expression in the cerebral organs is detected in isolated domains of both ciliated canal and neuroglandular portion (Fig. 5I).
The gene rx is expressed in the brain, anterior sensory organs, epidermal cells of lateral cephalic slits and in the cerebral organs (Figs. 5K and L, 6E). Dorsally, the gene is expressed in isolated cells distributed relatively uniformly throughout the brain lobes (Fig. 5K). In the ventral lobes, rx is expressed only in a pair of postero-lateral cell clusters (Fig. 5L). In the cerebral organs, the gene is specifically expressed in the cluster of epidermal cells at the anterior side of the ciliated canal opening (Fig. 6E). Expression of otp is detectable in the brain, LNCs, and numerous anterior sensory cells ( Fig.   5M and N). In the dorsal lobes, the gene is expressed only in a relatively few lateral cells (Fig.   5M), while ventrally it is also predominantly expressed in the lateral cells of the brain lobes, but its expression was also detected in the more median cells contributing to the mouth innervation and anterior part of the LNC (Fig. 5N).
In the head region, the gene dach is expressed in the brain, cerebral organs, proboscis and few isolated anterior cells (Figs. 5O and P, 6F). The expression in the brain is rather uniform and transcripts of the gene were detected in all regions of both dorsal and ventral lobes ( Fig.   5O and P). In the cerebral organs, the gene was detected in some of the cells of both the ciliated canals and the neuroglandular portion (Fig. 6F).
Expression of the gene emx was detected in the brain, cerebral organs, proboscis, and cells along anterior cephalic nerves (Figs. 5Q and R, 6G). In the brain the gene is expressed only in a few cells in the ventro-median domain (Fig. 5R). In the cerebral organs the gene transcripts were detected in the cells at the posterior side of the ciliated canal opening and in a single median cell in the neuroglandular part of the organ (Fig. 6G).
The TF arx has a broad expression in the anterior body of the juvenile L. ruber. It is expressed in the brain, rhynchocoel, epidermal cells, anterior sensory cells and in the cerebral organs (Figs. 5S and T, 6H). In both dorsal and ventral brain lobes, its expression was detected in numerous anterior, lateral and median cells (Figs. 5S and T). In contrast, the expression in the cerebral organs was restricted to a single cell at the posterior side of the ciliated canal opening (Fig. 6H).
The gene svp is also broadly expressed in anterior structures; its expression was detected in the brain, cerebral organs, LNCs, anterior sensory cells and proboscis (Figs. 5U and V, 6I). In the dorsal brain lobes, it is expressed in cells distributed through the lateral and median regions (Fig. 5U), while ventrally it is expressed uniformly in the entire ventral lobes (Fig. 5V).
In the cerebral organs, expression of svp was detected in some anterior and lateral cells of the neuroglandular part (Fig. 6I).
Transcripts of the gene tll were detected in the brain, cerebral organs and proboscis (Figs. 5W and X, 6J). Expression in the brain was restricted just to a few cells posteriorly to the ventral commissure (Fig. 5X). Signal from the probes against tll was extremely strong in the cerebral organs ( Fig. 5W and X) and was observed throughout the entire structure in cells of both the ciliated canal and the neuroglandular portion (Fig. 6J).
The brain of the juvenile L. ruber is divided by commissures and lobe neuropiles into eight regions: unpaired dorso-anterior, dorso-median, ventro-anterior and ventro-median regions as well as paired dorso-lateral and ventro-lateral areas (Fig. 5Y). Mapping of the abovedescribed gene expression patterns onto those brain domains reveled that most of the regions express unique combination of the TFs (Fig. 5Y). The only brain regions which seem to express the same sets of TFs are dorsal and ventral lateral domains (Fig. 5Y).

Gene co-expression during brain development
To further explore co-expression of some of the TFs in the brain, we performed double in situ hybridization of the selected brain patterning genes (nk2.1, nk2.2, pax6 and rx). In addition to the investigation of 42 days old juveniles, we also examined co-expression of those genes in the earlier developmental stage, 25 days old early juveniles, in order to test whether the observed co-expression patterns are conserved throughout ontogenesis.
The CNS of 25 days old juveniles shows much simpler morphology when compared to the hatched juveniles (Fig. 7A). It is composed of LNCs, which merge anteriorly in the brain with two commissures -a thicker ventral and thinner dorsal -that form a ring shaped neuropile around the developing proboscis rudiment. At this developmental stage, the brain is not yet divided into the dorsal and ventral lobes and the cerebral organs are not fully formed, being mainly composed by the ciliated canal, that is not directly connected with the brain [40].
In the brain of 25 days old juvenile, nk2.1 is expressed along the ventral commissure and in the lateral parts of the brain (Fig. 7B, C, E, G). In its lateral domains the gene is co-expressed with pax6 (blue arrowheads, Fig. 7B and C) and rx (blue arrowheads, Fig. 7G). Additionally, some of the lateral nk2.1 + cells also express nk2.2 (blue arrowheads, Fig. 7E). The more median nk2.1 + cells that are associated with the ventral commissure are devoid of pax6, nk2.2 and rx expression (white arrowheads, Fig. 7C, E, G). In addition to the expression in lateral domains, pax6, nk2.2 and rx are also expressed in cells associated with the dorsal commissure, which do not co-express nk2.1 (white arrowheads, Fig. 7B, D, F). The analysis of gene co-expression in the 42 days old juveniles generally corroborates the expression map based on single gene hybridization, however it allows more detailed description of the brain molecular regionalization. In the dorsal brain pax6 is broadly expressed in the lateral and median domains (white arrowheads, Fig. 7H) and only small clusters of lateral cells co-express pax6 and nk2.1 (blue arrowhead, Fig. 7H). In the ventral lobes, the lateral cells co-express pax6 and nk2.1 (blue arrowheads, Fig. 7I), while cells in the median domain express only nk2.1 (white arrowhead, Fig. 7I). nk2.1 and nk2.2 are not coexpressed in the dorsal brain (Fig. 7J). nk2.1 is expressed in the most lateral cells of the dorsal brain (white arrowheads, Fig. 7J), while nk2.2 is expressed in the large, more posterior domains and in scattered cells in the anterior brain region (red arrowheads, Fig. 7J). In the ventral brain, both genes are co-expressed in the postero-lateral and median domains (blue arrowheads, Fig. 7K), however nk2.1 has much broader ventral expression with many nk2.1 + cells devoid of nk2.2 expression (white arrowheads, Fig. 7K). rx is expressed in scattered anterior, median and lateral cells in the dorsal brain, which do not co-express nk2.1 (white arrowheads, Fig. 7L). In the lateral parts of the brain some cells co-express rx and nk2.1 (blue arrowheads, Fig. 7L), while some nk2.1 + cells do not express rx (red arrowheads, Fig. 7L). In the ventral brain the antero-lateral and median nk2.1 + cells do not express rx (white arrowheads, Fig. 7M), while small clusters of postero-lateral cells co-express both genes (blue arrowheads, Fig. 7M).
On the whole, comparison of gene co-expression between 25-and 42-days old juveniles shows that the general molecular patterning of the developing brain is retained throughout development. The ventro-median region expresses nk2.1 but not pax6 nor rx. The lateral brain includes cells co-expressing nk2.1 with pax6, nk2.2 and rx, while dorsal brain is mainly composed of pax6, nk2.2 and rx positive cells which do not co-express nk2.1. The differences between both life stages are primarily associated with the more complex architecture of the brain in 42 days old juveniles, which requires a more intricate developmental control, nevertheless the most general gene expression patterns are conserved.

Comparison of juvenile and adult morphology
Nervous system has been investigated in great detail in adult Lineus ruber [20,[22][23][24][25][26][29][30][31] and Lineus viridis [19,20,24,25], a morphologically similar species that belongs to the same species complex [53,54]. Comparison between the juvenile and adult worms reveals that all major nervous structures described in the adults are already present in the 42 days old juveniles, indicating that at this stage the general neuroarchitecture is already fully formed and that further development is mostly related with increase in the size but not morphological complexity. The same pattern is observed in number and diversity of cell types contributing to the cerebral organs. There are, however, some minor differences in immunoreactivity patterns between both life stages. For instance, SLIR perikarya have been reported in the dorsal brain ganglia of adult L. ruber [23], while we observed immunoreactivity against serotonin only in the ventral brain ganglia of the juveniles (Fig. 2I). This indicates, that even though the general morphology of the brain is already established at the moment of hatching, the following growth of the brain is not purely quantitative, but also new cell types are added in some brain regions during further development. Moreover, staining of mitotically active cells showed that in 60 days old juveniles cell proliferation in the brain is lower than in the other organs, while the cells of the cerebral organs are still intensively dividing (Fig. 3), indicating allometric growth of the CNS.
The major postpharyngeal commissure, which ventrally connects the lateral nerve cords, is the only juvenile neural structure which does not correspond directly to any of the elements of the adult nervous system of L. ruber [20,22,29] or, to our best knowledge, of any other nemertean, which nervous system has been studied thus far [e.g. 17-19, 20, 21, 28, 55-58]. In adult nemerteans, the lateral nerve cords are connected by numerous delicate ventral commissures, that are composed just of bundles of neurites and are considered as part of the peripheral nervous system. Conversely, the postpharyngeal commissure described in this study is associated with few SLIR and numerous ChAT + perikarya and has typical medullary arrangement, markedly different from the remaining ventral commissures (Figs. 1 and 2).
There are two possibilities to explain this discrepancy in morphology of both stages: either the commissure degenerates during ontogeny or, due to the allometric growth, becomes much less prominent in later developmental stages and was overlooked in previous investigations.
Nevertheless, the observation of the postpharyngeal ventral commissure in a nemertean is interesting since similar structures are present in numerous annelids (e.g. the first commissure connecting ventral nerve cords [59][60][61][62][63][64]), as well as in all major clades of gastrotrichs [65][66][67] and gnathiferans [68][69][70][71]. Therefore, the distribution of this character on the phylogenetic tree raises the possibility that the ventral postpharyngeal commissure connecting the major nerve cords might represent a plesiomorphic spiralian trait retained in some form in numerous clades.

Expression of brain patterning genes in Spiralia
Molecular patterning of the brain has been investigated in relatively many spiralians, representing diverse clades with broad spectrum of morphological complexity of their brains (Tab. 1). Among those species, the best studied is the annelid Platynereis dumerilii, which possesses a relatively complex brain with multiple morphologically, functionally and developmentally distinct regions [42,49,[72][73][74]. One of the important characteristics of gene expression patterns during the development of the P. dumerilii brain is regional restriction of nk2.1 expression to the ventro-median region and pax6 expression in the lateral domains (including eyes and mushroom bodies), with only the minimal overlap of expression of both genes ( Fig. 8A; [42,49]). This expression pattern resembles the one observed in vertebrates [75,76] and has been proposed as an ancestral bilaterian trait [42]. Although a comparable expression of those two genes is also witnessed in some Spiralia (Tab. 1), including other annelids [43,47,48], rotifers [39] and brachiopods [39,46,[77][78][79], we did not retrieve a similar pattern in neither 25-nor 42-days old juveniles of L. ruber (Figs. 7B, C, H, I and 8B).  [44,45], while one of the pax6 paralogs, pax6A, is broadly expressed in the brain tissue [44,80]. A further parallel between planarians and Lineus is associated with a seemingly diminished role of pax6 in eye formation: pax6 is not expressed during eye development neither in L. ruber (this study) nor in L. viridis [81] (although it seems to have a role in eye regeneration in L. sanguisues [81]), while in flatworms eye regeneration has been demonstrated to be pax6 independent [80]. The role of pax6 in eye patterning is otherwise highly conserved among bilaterians [e.g. 82,83,84]. Due to the unstable position of Nemertea on the spiralian phylogeny [e.g. [6][7][8]10], it is currently impossible to determine whether those similarities between platyhelminths and nemerteans are due to the convergent evolution, a common evolutionary innovation or retention of ancestral plesiomorphic conditions in both lineages.  [42,49] ventro-median domain [42,49] no expression in the brain [41] anterior brain [42] Dimorphilus gyrociliatus Annelida compact paired lateral domains [43] ventro-median and mediolateral domains [43] no expression in the brain [ [77,79] ventral domain [46,78] no expression in the apical organ [39] N/A

Phoronopsis harmeri
Phoronida larval apical organ no expression in the apical organ [85] no expression in the apical organ [86] N/A N/A
Another important differences in expression of brain patterning genes between L. ruber and other Spiralia includes the expression of nk2.2 within numerous brain domains of L. ruber (while the gene lacks brain expression not only in annelids [41,43], but also in brachiopods [39] and flatworms [87]) as well as broad expression of rx in the dorsal lobes of the nemertean brain (versus their more rostral expression in P. dumerilii [42,49]). Altogether this comparison shows that complex brains of nemerteans, and especially their dorsal lobes, show little resemblance in the molecular patterning to the complex brains of P. dumerilii (Fig. 8), which in turns seem to share more molecular similarities with simpler brains of other annelids and apical organs of brachiopod larvae (Tab. 1). This observation, in concert with morphological data [21,64,88], indicates that complex brains of nemerteans and errant annelids evolved convergently, due to e.g. similar selective pressure associated with predatory/active life style [89]. We propose that the increase in the brain size and complexity in those two lineages was achieved by independent expansions of non-homologous regions of simpler brains present in their respective ancestors.

Are mushroom bodies and cerebral organs derived from the same ancestral organs?
In numerous annelid brains, morphologically distinct structures, referred to as mushroom bodies, are present, which have allegedly chemosensory and cognitive functions [35,36,49,73,[102][103][104][105][106][107]. There is an ongoing discussion on whether those structures are part of the ancestral annelid body plan or whether they evolved more recently in one of the annelid subclades [35,73,108]. However, their phylogenetic distribution (especially the lack of comparable structures in Palaeoannelida and Sedentaria [64,88,109]) favors the latter option [64,88,108,110].
Nevertheless, morphologically similar structures are also present in Panarthropoda [36,105,[111][112][113][114], which lead some authors to the idea that mushroom bodies-like structures were already present in the common protostome ancestor [36,49,73,105]. Although similarities in molecular patterning of annelid mushroom bodies and vertebrate pallium led to the assumption that both structures originated from the same sensory and associative brain center of hypothetical ancestral bilaterians [49], such homology statements, based on observation of only two phylogenetically distant clades, are always at the best case highly tentative [89,115].
Cerebral organs of nemerteans, in contrast to the annelid mushroom bodies, can be unequivocally reconstructed as present in the last common nemertean ancestor [20,22,28].
However, it remains unresolved whether they are nemertean evolutionary novelty or rather homologs of the mushroom bodies of annelids [19,35,36] or the lateral ciliated pits present in catenulids and macrostomids [30,34,116,117], the two earliest sequentially branching platyhelminth clades [118]. Similarities between the mushroom bodies of annelids and the cerebral organs of nemerteans are rather superficial: the former are integral parts of the brain and are not connected to the external realm, while the latter are always contacting ambient environment and, especially in Hoplonemertea, might be spatially separated from the CNS [17,28,32,33]. On the other hand, the function, general morphology, connectivity and fine structure of cerebral organs of nemerteans and ciliated pits of flatworms bear a strong resemblance [30,34,116,117], making their homology much more likely. Taking into account the arrangement of the cerebral organs in various nemertean clades, the "ciliated pit" organization seems to represent an ancestral character state also in nemerteans [19][20][21]. If one accepts that the cerebral organs of nemerteans and ciliated pits of catenulids and macrostomids are homologues [34], then, depending on the phylogenetic position of nemerteans, there are two possible scenarios of their evolution: 1) If nemerteans are sister group to platyhelminths (Parenchymia hypothesis [7,119]), then the ciliated pits-like structures represent a synapomorphy of Parenchymia. 2) On the other hand, if nemerteans are closer to annelids than flatworms [5,6,8,10], then the presence of ciliated pits might represent a plesiomorphic condition, present also in the annelid ancestor.
In the face of the above-discussed concerns about the homology of mushroom bodies and cerebral organs, we were surprised to find that cells constituting the cerebral organs express

Conclusions
In this study, we investigated the morphology and gene expression in the developing CNS of the nemertean Lineus ruber. At the moment of hatching, juveniles of L. ruber have already all major components of the adult nervous system, which indicates that further development is mostly related with increase in the size but not morphological complexity. This likeness corelates well with a similar predatory lifestyle of both juveniles and adults [40]. Comparison of gene expression in the brain of L. ruber and the annelid P. dumerilii [41,42,49] indicates that complex brains, observed in those two animal species, evolved convergently by independent expansion of non-homologues regions of simpler ancestral brains. Such scenario corresponds with the similar conclusions drawn by comparative morphology [21,64,88]. In contrast to the discrepancies in gene expression in the brains, we observed that the same set of transcription factors, which is expressed in the mushroom bodies of P. dumerilii [49] is also expressed in the cerebral organs of L. ruber. These similarities might be a result of convergent recruitment of the same GRN into patterning of non-homologue organs or indicators of the homology of some cell types contributing to mushroom bodies and cerebral organs that could evolve from the cell type present in the lateral chemosensory ciliated pits of the hypothetical spiralian ancestor. Further studies on the cell-type level and functional interactions of the studied TFs are needed to fully resolve the level of homology, or convergence, between mushroom bodies and cerebral organs.

Animal collection and morphological investigation
Adult specimens of Lineus ruber were collected near Bergen, Norway (Fanafjord; GPS coordinates: 60.251845N, 5.320947E). The animals had dark red coloration with wide pigment-free areas in the terminal part of the head. Animals were kept in the laboratory in filtered seawater at 14°C with a daytime cycle: 13 hours of sunshine and 11 hours of darkness.
Collection of egg masses and desired developmental stages, animal fixation as well as antibody, nuclear and EdU stainings followed the already established protocols [40].
Specimens for TEM investigation were fixed in 4% PFA in PBS, rinsed in the same buffer, postfixed in 1% OsO4 diluted in PBS for 120 min at 4˚C, rinsed again and dehydrated in graded ethanol/acetone series. The samples were embedded in Epon 812 resin (Sigma Aldrich) and cut to semi-and ultrathin sections with a diamond knife (Diatome Histo Jumbo) using ultramicrotome Leica EM UC6. The ultrathin cross sections of cerebral organ were placed on formvar-covered (Fluka) single slot copper grids and stained with 1% uranyl acetate and lead citrate.

Gene expression analysis
Coding sequences for analyzed genes were identified in the transcriptome of L. ruber with the reciprocal TBLASTN search using orthologous protein sequences from P. dumerilii. Sequence of all of the newly identified genes were translated into protein sequences and aligned with reference sequences from other animals ( Table S1). The alignments were trimmed either manually or with TrimAl software [120] and analyzed with FastTree v2.1 [121] in order to assess orthology of the analyzed genes (Figs. S1-5). All newly obtained sequences were submitted to GenBank (Accession numbers MW720144-MW720151).
Fragments of genes were amplified from cDNA library using specific primer pairs, cloned into pGEM-T Easy vectors (Promega, USA) and then transformed into competent Escherichia coli cells for amplification. Plasmid DNA was isolated and sequenced in both forward and reverse directions using T7 and SP6 primers to assure that the desirable genes were cloned. The antisense probes were transcribed from linearized DNA and labeled either with digoxigenin (for hybridization of single mRNA) or with dinitrophenol (for detection of second gene in double in situ hybridization). Whole mount in situ hybridization followed the same procedure as described for L. ruber juveniles in other studies [39,40].

Imaging and image processing
Samples for confocal laser scanning microscopy (antibody staining and in situ hybridization) were mounted in Murray's clear and scanned in either Leica SP5 or Olympus FV3000 CLSM.
Z-stacks of confocal scans were projected into 2D images in IMARIS 9.