Ancient origin of somatic and visceral neurons

Background A key to understanding the evolution of the nervous system on a large phylogenetic scale is the identification of homologous neuronal types. Here, we focus this search on the sensory and motor neurons of bilaterians, exploiting their well-defined molecular signatures in vertebrates. Sensorimotor circuits in vertebrates are of two types: somatic (that sense the environment and respond by shaping bodily motions) and visceral (that sense the interior milieu and respond by regulating vital functions). These circuits differ by a small set of largely dedicated transcriptional determinants: Brn3 is expressed in many somatic sensory neurons, first and second order (among which mechanoreceptors are uniquely marked by the Brn3+/Islet1+/Drgx+ signature), somatic motoneurons uniquely co-express Lhx3/4 and Mnx1, while the vast majority of neurons, sensory and motor, involved in respiration, blood circulation or digestion are molecularly defined by their expression and dependence on the pan-visceral determinant Phox2b. Results We explore the status of the sensorimotor transcriptional code of vertebrates in mollusks, a lophotrochozoa clade that provides a rich repertoire of physiologically identified neurons. In the gastropods Lymnaea stagnalis and Aplysia californica, we show that homologues of Brn3, Drgx, Islet1, Mnx1, Lhx3/4 and Phox2b differentially mark neurons with mechanoreceptive, locomotory and cardiorespiratory functions. Moreover, in the cephalopod Sepia officinalis, we show that Phox2 marks the stellate ganglion (in line with the respiratory — that is, visceral— ancestral role of the mantle, its target organ), while the anterior pedal ganglion, which controls the prehensile and locomotory arms, expresses Mnx. Conclusions Despite considerable divergence in overall neural architecture, a molecular underpinning for the functional allocation of neurons to interactions with the environment or to homeostasis was inherited from the urbilaterian ancestor by contemporary protostomes and deuterostomes.


Background
For several decades, molecular data have been used to define homologous regions in the nervous system of distant phyla. More recently, homology search has moved to the cell level, using conserved neuronal-type specific molecular signatures [1][2][3][4][5] (and reference [6] for review). This approach provides a novel window on the complexity of ancestral nervous systems, and sets the stage, with unprecedented detail, for an understanding of what has changed or been conserved during their large-scale evolution. In this paper, we undertake a comparison of sensorimotor circuits across the protostome/deuterostome boundary, which is, thus, informative about the nervous system of Urbilateria.
Ever since Bichat distinguished the 'organic' and 'animal' lives [7], the vertebrate body has been construed as a dual entity, one part somatic (engaged with the outside world), the other visceral (concerned with bodily homeostasis). Paralleling this distinction, the sensorimotor circuits passing through the spinal cord and brainstem are divided into somatic and visceral. Somatic circuits are responsible for the somesthetic, visual or auditory perception of the environment and locomotory responses. The visceral circuits are responsible for sensing parameters of the interior milieu, such as arterial pressure, blood gases and various chemosensory modalities including taste, and feedback regulation of the cardiovascular, respiratory and digestive organs. Specific anatomic features and embryonic origins have been progressively discovered for these circuits. For example, somatic and visceral neurons in the vertebrate hindbrain are born and settle at distinct dorso-ventral levels [8]. In the peripheral nervous system, first-order visceral sensory neurons emerge from epibranchial placodes, whereas somatic ones derive from dorso-lateral placodes or the neural crest [9]. In his landmark monograph on the subject, A.S. Romer synthesized a century of observations on the 'duality' of the vertebrates, made up of two bodies (and nervous systems), one visceral and one somatic, 'imperfectly welded' on each other [10]. In the past years, after decades of neglect, the proposed dichotomy of the vertebrate nervous system has found, piecemeal, an unexpected molecular basis, in the form of transcription factors that globally distinguish somatic from visceral neurons ( Figure 1) and are largely, if not completely, restricted to them. Most somatic sensory neurons, first-and second-order, express the POU domain genes Brn3: the touch and pain receptors of the dorsal root and cranial ganglia, auditory and vestibular neurons, ganglionic cells of the retina, and many relay sensory neurons of the dorsal horn of the spinal cordeven if, in the latter, expression is transient and not entirely charted [11,12] (and references therein). Brn3 is required in all the peripheral cell types, its role in the central nervous system (CNS) being yet unexplored. Somatic motoneurons express and require the combination of the homeobox genes Mnx1 [13] and Lhx3/4 [14]. Finally, the vast majority of visceral neurons, sensory and motor, express and require the homeobox gene Phox2b: visceral sensory ganglia and their target, the nucleus of the solitary tract, all autonomic ganglia; visceral motoneurons of the medulla; branchial motoneurons (respiratory -that is, visceral-in the ancestral vertebrates) and chemosensory structures, such as the carotid body and the retrotrapezoid nucleus [12,[15][16][17]. This simple, non-combinatorial code provides new tools to probe the ancestry of these broad neuronal categories.
For this study, we turned to mollusks, for two reasons: first, they belong to the phylum Lophotrochozoa which, genetically less derived than Ecdysozoa, is particularly valuable for comparisons across Bilateria [18]; and second, mollusks have been studied by neurophysiologists for decades and consequently provide a unique catalogue of identified neurons with somatic or visceral functions [19] that allow rigorous tests of association with specific molecular signatures. Using as model systems two gastropods -the opisthobranch Aplysia californica and the pulmonate Lymnaea stagnalis-and the decapodiform cephalopod Sepia officinalis, we show that physiologically defined somatic motor and sensory neurons and visceral motoneurons share, respectively, the Mnx/Lhx3/4, Brn3 and Phox2 transcriptional signature of their vertebrate counterparts, and we discuss the evolutionary implications of this conservation across Bilateria.

Shared molecular signature of gastropod and vertebrate mechanosensory neurons
In vertebrates, the vast majority of sensory neurons that perceive the environment (with the exception of olfactory neurons) express paralogues of the POU-IV/Brn3 homeogene family, henceforth collectively called Brn3. These include mechanoreceptors, proprioceptors, thermoreceptors and nociceptors of the dorsal root and trigeminal ganglia, retinal ganglion cells, vestibular and cochlear sensory neurons and inner hair cells [11] (and references therein), as well as many second-order neurons in these somatic sensory pathways [12] (and references therein). The evolutionary stability of this genetic signature across Bilateria has been uncertain so far, based on the study of model ecdysozoans: mechanoreceptors have not been reported to express the orthologue of Brn3 in Drosophila, where it labels olfactory neurons instead [20], and they form only a small fraction of the 57 neurons that express the Brn3 orthologue in Caenorhabditis elegans [21]. In the gastropod Haliotis asinina, expression of Brn3 in patches of larval ectoderm has been interpreted as marking peripheral sensory structures [22]. We examined the case of A. californica, where mechanoreception and nociception are mediated, at least in part, by clusters of small-size sensory neurons in several ganglia of the CNS, which synthesize the neuropeptide sensorin A [23]. We cloned the Aplysia orthologue of Brn3 [see Additional file 1: Figure S1a] and found it expressed in the Sensorin + neuronal clusters and restricted to them in the CNS (Figure 2a-d,q for the pleural and cerebral clusters). In another gastropod, the pulmonate L. stagnalis, similar clusters of small Brn3 + neurons occurred in several ganglia ( Figure 2e for the pleural ganglion), many of which expressed the Lymnaea orthologue of Sensorin (Figure 2f and Additional file 1: Figure S2) and are thus most likely homologous to the mechanoreceptors of Aplysia. To further elucidate the molecular signature of these cells, we cloned the Lymnaea orthologues of Drgx [see Additional file 1: Figure  S1b], largely restricted to sensory neurons in vertebrates [24], and of Islet1 [see Additional file 1: Figure S1c], expressed in all vertebrate sensory neurons [25] (and motoneurons). Both were coexpressed with Brn3 in Lymnaea (Figure 2g-j,r) as is the case for their orthologue in the touch and pain receptors of the mammalian dorsal root ganglia (Figure 2m-o,s). In addition, Lymnaea sensory clusters, like their vertebrate counterparts, expressed the  In the cerebral ganglia, Brn3 is expressed in the Sensorin + J and K sensory clusters (white arrowheads) [23] but not in the scattered Sensorin + cells. Projections from the sensory clusters contain the Sensorin mRNA and are detected by in situ hybridization [23]. vesicular glutamate transporter VGluT (Figure 2k,l,p), in line with the glutamatergic phenotype of molluscan mechanoreceptors [26]. Thus, the unique Brn3 + /Drgx + /Islet + /VGluT + transcriptional signature of first-order somatic sensory neurons in vertebrates, is also a hallmark of their molluscan counterparts.

Shared molecular distinction between somatic and visceral motoneurons in gastropods, cephalopods and vertebrates
In vertebrates, locomotion depends on somitic muscles, innervated by spinal cord motoneurons that depend on the homeogenes Mnx1 [13] and Lhx3/4 [14] and use acetylcholine as neurotransmitter. In gastropods, the muscles of locomotion (in the foot and body wall or attached to the columella) are innervated by the pedal ganglia [27]. We cloned Ls-Mnx, the Lymnaea orthologue of Mnx1 [see Additional file 1: Figure S1d] and found that it was largely restricted to two clusters of neurons in the pedal ganglia (Figure 3a), which coexpressed the orthologues of Lhx3/4 ( Figure 3c and Additional file 1: Figure S1e) and the vesicular acetylcholine transporter VAChT (Figure 3b,d). At least some of these cells projected in the pedal nerves, as assessed by retrograde filling (Figure 3e-g) and most likely are locomotory neurons. These data extend the observations that some motoneurons in Drosophila also express Lim3 (the Lhx3/4 orthologue) and Mnx [28,29], that some C. elegans motoneurons express a Mnx orthologue [30], and that the Nkx6 + domain of the annelid nerve cord gives rise to Mnx + /VAChT + putative motoneurons [31]. Thus, the Mnx + /Lhx3/4 + molecular signature has been associated with somatic motoneurons since the origin of bilaterians.
We next examined the visceral nervous system, whose sensorimotor circuits in vertebrates largely coincide with the expression pattern of the paralogous homeogenes Phox2a and Phox2b (hereafter collectively designated as Phox2) and depend on the latter for their formation [12,[15][16][17]. Among deuterostomes, we previously found that the Ciona intestinalis orthologue of Phox2 is specifically expressed in neurons of the cerebral ganglion of postmetamorphic animals that motorize the respiratory and digestive 'branchial basket' [2]. Concerning protostomes, the C. elegans orthologue of Phox2 marks specifically five neurons [32] whose function is either unknown (the SIAs) or hard to classify as somatic or visceral (ALA, involved in behavioral quiescence [33]). The inconclusiveness of the latter finding is compounded by the fact that most visceral organs in C. elegans either lack innervation (such as the intestine) or are missing altogether (such as a cardiovascular or a respiratory apparatus). In A. californica, several neurons with cardiovascular, respiratory or excretory functions have been identified by size and location in the abdominal ganglion. One of the best documented, the multimodal motoneuron L7, directly innervates the muscles of the gill, siphon, epineural sheath, heart and abdominal aorta [34] (and references therein), and serves as premotor neuron for the branchial ganglion [35]. We cloned Ac-Phox2, the Aplysia orthologue of Phox2 [see Additional file 1: Figure S1b] and found it expressed in L7, recognizable by its large size and position at the left border of the abdominal ganglion, rostral to the other large (and also Phox2 + ) neuron in the region, L11, which projects in the genital nerve but whose function is unknown (Figures 4a,b and 2q). Among the other neurons expressing Phox2 was a large cluster occupying the 'right upper quadrant' of the ganglion (Figure 4a,b), which includes the giant R3-R13 cells [36] that innervate the heart, major arteries and veins, digestive gland sheath and kidney [37]. Many neurons that are smaller than the R3-R13 cells were also positive for Phox2 in this region. These might correspond to the extra cells found to share with R3-R13 electrical activity [38], immunoreactivity for the cardiomodulator peptide NdWFamide [39] and, thus, possibly a function. A giant Phox2 + cell located close to the branchial nerve corresponded to R14 (Figure 4a), which shares its targets with the R3-R13 group [37]. The R3-R14 neurons are thought to modulate heart beat frequency and local aspects of circulatory physiology using, among others, the peptide HRBP [40] as neurotransmitter. The kidney also receives input from giant neurons of the 'left upper quadrant'. Among those, L5 is uniquely identified by expression of the peptide encoding the LUQ-1 gene [41], allowing us to show on consecutive sections that it was Phox2 + (Figure 4c,d). In Lymnaea, large 'Light Yellow Cells' (LYCs) [42] are considered homologous to the R3-R14 neurons on the basis of their size, color, location, synthesis of a peptide similar to HRBP [43] and projections to the heart, aorta, kidney, and connective tissue of the CNS [44]. We cloned Lymnaea Phox2 [see Additional file 1: Figure S1b] and found that it was expressed in all LYCs of the ventrolateral lobe of the right parietal ganglion, identified by their position and expression of the LYC pro-peptide (Figures 4e,f and 2r). Retrograde filling showed that they projected in the right parietal nerves (Figure 4g), which innervate the mantle cavity and pneumostome [45], suggesting an additional role in ventilation.
As a further test of the vertebrate-like molecular signature of visceral and somatic motoneurons in mollusks, we explored a second molluscan clade, Cephalopoda. In decapodiform cephalopods (squid and cuttlefish), the brachial ganglion and anterior part of the pedal ganglion contain motoneurons for the prehensile and locomotory arms, while the palliovisceral ganglion contains motoneurons for the visceral mass [46,47]. We cloned the orthologues of Mnx and Phox2 in S. officinalis [See Additional file 1: Figure S1b,d] and found that Mnx, but not Phox2, was expressed throughout the anterior pedal lobe (Figure 5a-c) while Phox2, but not Mnx, was expressed in the palliovisceral ganglion (Figure 5a-c). Some of the palliovisceral ganglionic neurons are presynaptic to the motoneurons of the mantle, which make up most of the stellate ganglion, a synapomorphy of this clade. Phox2, but not Mnx, was expressed in most stellate ganglionic neurons (Figure 5d-f). Thus, mantle motoneurons, which control water flow over the gills, express Phox2, like the branchial motoneurons that perform the same function in fish [16,48]. Apart from their respiratory function, motoneurons to the mantle (whose axons fuse to form the so-called 'giant nerve fibers'-reference 49 and Figure 5g) allow rapid water ejection and jet-propulsion of the animal. The molecular identity of these escape motoneurons thus correlates with the ancestral function of the target organ (respiratory, that is, visceral) rather than with its additional, derived function (locomotory, that is, somatic). A conceptually similar but inverse situation can be described in terrestrial (that is, air breathing) vertebrates, whose respiratory motoneurons (projecting to the intercostal, abdominal and diaphragmatic muscles) are somatic, that is, have kept the identity matching the embryonic origin (somitic) and ancestral function (locomotory, except for the diaphragm of mammals) of their target muscles.
Hence, molluscan motoneurons that innervate the viscera are distinguished from those that innervate the locomotory muscles by the same transcriptional code (respectively Phox2 and Mnx/Lhx3/4)a st h e i rv e r t ebrate counterparts.

Discussion
We have shown that molecular signatures for neurons with somatic (that is, relational) versus visceral (that is, homeostatic) functions are conserved between vertebrates and mollusks. Visceral motoneurons, (such as  Figure S1f). Apart from the expressions described in the text, the brachial ganglion contains So-Mnx + neurons, in line with its role in movement of the arms, but also So-Phox2 + neurons in its posterior wall, whose function has not been described so far. (g) Drawing, reproduced with permission from Figure thirteen of reference [49], showing the way in which the axons of stellate ganglion neurons progressively fuse to form the third order giant fibers (gf3) that allow jet-propulsion. (h) Schematic of the central nervous system and stellate ganglion of Sepia. br, brachial ganglion; D, dorsal; P, posterior; pa, anterior pedal ganglion; pp, posterior pedal ganglion; pv, palliovisceral ganglion; so, supraesophageal mass; st, stellate ganglion. Scale bars, 500 μm(a-c); 200 μm(d-f). cardiorespiratory neurons) express the orthologue of the vertebate pan-visceral determinant Phox2b in opisthobranch and pulmonate gastropods and a decapodiform cephalopod. No other transcription factor or neurotransmitter phenotype marks these neurons, specifically or exhaustively, in vertebrates, precluding a more complex signature. However, Phox2b expression is highly selective for visceral neurons (both motor and sensory) in vertebrates [16] and, thus, in combination with hodological criteria, constitutes a strong argument for homology. Neurons with modalities clearly equivalent to those of visceral sensory neurons in vertebrates (which monitor taste or blood pressure for example) are not described to our knowledge in mollusks, precluding exploration of this broad neuronal identity. Of note, in vertebrates, the viscerosensory phenotype is imposed by Phox2b on a somatosensory default identity [12], suggesting that the former is evolutionarily more recent than the latter. Locomotory (somatic) motor neurons express the orthologues of the homeobox genes Mnx1 and Lhx3/4 and VAChT in both gastropods and cephalopods, like their counterparts in the spinal cord of vertebrates. Finally, somatic sensory neurons (such as mechanoreceptors), characterized in gastropods by previous electrophysiological studies or expression of the peptide sensorin A, selectively express the orthologues of the homeodomain genes Brn3, Drgx,a n d Islet1 and VGluT, like their physiological counterparts in the dorsal root and trigeminal ganglia of vertebrates.
What evolutionary relationship can explain the conservation of molecular signatures in neurons with visceral versus somatic functions between deuterostomes and protostomes? The simplest hypothesis is that the cells are phylogenetically homologous, that is, one could trace their ancestry to an original neuron or neuronal cluster in the common ancestor, as was proposed for ciliary photoreceptors in annelids and vertebrates [1] or for branchial motoneurons in vertebrates and urochordates [2]. This might also be the case for somatic motoneurons in vertebrates and lophotrochozoans: their distribution is spatially discrete in each phylum and reconcilable between phyla. In vertebrates, locomotory Mnx + neurons are born in a ventral Nkx6.2 + domain of the spinal cord, topologically and molecularly similar to the ventral domain of the nerve cord of annelids, which produces Mnx + /VAChT + neurons, presumably motor [31]. In mollusks, they are restricted to the pedal ganglia, conceivably homologous to the ventral nerve cord of annelids.
On the other hand, phylogenetic homology between vertebrates and mollusks is unlikely, at least in most cases, for somatic sensory neurons and visceral motor neurons, due to their anatomical distribution, widespread in each species (Figures 1, 3, 4) and hard to reconcile between them. Moreover, Phox2 is expressed in synapomorphic structures of vertebrates (the neural crest-derived autonomic ganglia) and cephalopods (the stellate ganglion). In this case, the most likely evolutionary scenario is that, in the last common ancestor, a 'seminal regulatory interaction' [50] arose between Brn3 and the somatic sensory phenotype and between Phox2 and the visceral phenotype. Although no target gene has been uncovered yet that would explain the physiological dichotomy these transcription factors specify, one can hypothesize, for example, that they direct axonal projections towards somatic versus visceral targets. Subsequently, this regulatory interaction would have been conserved, while Brn3 and Phox2 acquired additional expression sites, giving rise to novel groups of cells of the same broad type along each evolutionary lineage. According to this view, the relationship between the different kinds of somatosensory neurons or visceromotor neurons would be neither of phylogenetic homology (between species) nor of 'sister cell types' (within a species [6]), but instead, both between and within species, of 'deep' [51,52] or 'generative' [53] homology, akin to that proposed for bilaterian appendages.

Conclusions
Regardless of the exact nature of what has been conserved across the protostome-deuterostome boundary, either neuronal groups or regulatory links between transcription factors and neuronal traits, our data show that the viscerosomatic duality of the nervous system, as described in vertebrates, was already part of the urbilaterian body plan. Some of its components might be even more ancient, as suggested by expression of a Brn3 orthologue in the sensory 'rhopalia' of scyphozoan and cubozoan jellyfish [54]. To our knowledge, Mnx has not been analyzed in the nervous system of cnidarians. Finally, no straightforward orthologue of Phox2 emerges from sequence analysis of paired-like homeobox genes in Cnidaria [see Additional file 1: Figure S1b], which complicates the search for a pre-bilaterian origin of the visceral nervous system.

Immunofluorescence for detection of mouse Brn3a
Sections were dried, washed for 5 minutes in PBS, permeabilized 2 × 10 minutes in PBS/ 0.3% Triton X-100, blocked for 20 minutes in blocking solution (PBS, 10% FCS, 0.1% Triton X-100), incubated overnight at 4°C with the mouse anti-Brn3a monoclonal antibody (MAB1585) diluted 1/200 in blocking solution, washed 3 × 10 minutes in PBST, incubated for 2 hours with the fluorescent goat anti-mouse Cy3-coupled secondary antibody (115-165-003 Jackson ImmunoResearch, Newmarket, U K ) ,w a s h e d3×1 0m i n u t e si nP B S Ti nt h ed a r ka n d mounted in Mowiol (Calbiochem). The image was converted to gray scale, inverted and superimposed on a Nomarski photomicrograph in Photoshop CS3.