Gastric pouches and the mucociliary sole: setting the stage for nervous system evolution

A rich body of fossil and comparative evidence indicates that mucoid-ciliary particle feeding evolved in early metazoans. Feeding traces indicate that one of the most characteristic fossils of the Ediacaran fauna, Dickinsonia, was grazing on the sea floor, which was covered by ubiquitous microbial mats [3] (figure 4a). It is assumed that their entire ventral body surface was a digestive ‘sole’ covered with mucus and cilia [3]. We refer to this new tissue as ‘mucociliary sole’. A digestive ventral sole, used for external digestion and movements, also exists in today's placozoans [104] (figure 4b). Placozoans move by ciliary gliding with the thick mucociliary sole facing the sea floor, and by concomitant contraction of the syncytial fibre cells (figure 4b). Capturing of microparticles with the help of cilia and mucus persists in some cnidarians; and mucoid-ciliary food transport is also found in macrophageous cnidarians and ctenophores. For example, in the sea anemone Calliactis parasitica non-neuronal control of ciliary beating involves a direct chemosensory response of pharyngeal epithelial cells [105]. Ctenophores possess an elongated, ciliated pharynx with numerous gland cells secreting digestive enzymes [106]. As in cnidarians, the control of ciliary beating for food transport is non-neuronal. Nutrients are transported into a complex system of gastrovascular canals for nutrient distribution. Residual water is expelled to the outside via anal pores. Excitingly, the pharynx can be everted and used as a mucociliary sole in benthic ctenophores [62]. Finally, mucociliary epithelia are frequently found in foregut and neural midline tissue in bilaterians (that can be derived from tissue surrounding the gastric opening, see below).

This includes the dorsal ciliary folds in the annelid foregut [107], the buccal field in the rotifer Dicranophorus [108], the endostyle/thyroid in the chordate foregut, the neurotrochoid along the neural midline in various invertebrates [109,110] and the enigmantic Reissner's fibre in chordates (see below).

The invention of the mucociliary sole was likely paralleled by cell-type diversification (figure 4d), specializing in the production of mucus, the secretion of digestive enzymes, the transport of mucus and the uptake of dissolved nutrients [45]. Mucus is largely made up of large glycoproteins called mucins [111]. Gel-forming mucins, Otogelin and von Willebrand factor are evolutionary related [111], and proteins with properties of the gel-forming mucins were identified also in the starlet sea anemone Nematostella vectensis [111]. Furthermore, mucus-secreting cells (‘mucocytes’, or ‘goblet cells’) are common to most animal taxa. The secretory vesicles of mucocytes are of low electron density, in contrast to those of the zymogenic gland cells (‘zymocytes’, figure 4d), which secrete digestive enzymes. The latter are found in the bilaterian foregut, in the ctenophore and cnidarian pharynx (with associated filaments) and around the digestive ventral sole of the placozoans (figure 4b) [45]. The zymocytes have apical microvilli and/or cilia in some groups [45], reminiscent of their choanocyte origin. Within the mucociliary sole, the motile ciliary cells specialize in moving the secreted mucus along the epithelial surface.

At the molecular level, the evolutionary divergence of mucociliary cell types in metazoans appears to be closely linked to the diversification of the Fox gene family. Most prominently, foxj has been identified as an ancient key regulator of motile cilia biogenesis [72,112,113], conserved in cnidarians [72] and even sponges [114]. Foxj is active wherever motile cilia are present; for example, in Xenopus, foxj genes are expressed in otic vesicle, pronephros, gill structures, notochord, floorplate and ventral neural tube [115]. Another member of the Fox family, foxi, specifies the osmoregulatory ionocytes, another characteristic cell type of vertebrate mucociliary epithelia [116]; moreover, foxA has been found to specify small 5HT-secreting cells in the mucociliary epithelium that regulate ciliary beating [117,118]. Additional members of the Fox family have been implicated in the specification of mucociliary cell types: the thyroid-related transcription factors TTF-2/foxE4, foxq1 and foxA are jointly expressed in the amphioxus and ascidian endostyle, a presumed homolog of the thyroidea that harbours mucus-secreting and motile ciliated cells; and this function of the Fox genes is considered ancestral for the chordates [119]. Interestingly, the metazoan foxA, B, C, D and E subfamilies are evolutionarily related [120–122] and resulted from duplications of an ancient precursor gene; in Nematostella, foxB [122] and foxA [123] are expressed in the ectodermal ciliated pharynx and foxC [122] in pharyngeal endoderm; in Hydra, foxA/budhead [124] is likewise expressed in the endoderm around the oral opening, suggesting that the mucociliary epithelium around the gastric opening represents the ancient site of activity of the foxA, B, C, D, E superfamily.

With the invention of the mucociliary sole around the gastric opening, the endochoanocytes were no longer needed for mucoid particle trapping. In consequence, they diversified into more specialized cell types—enterocytes, excretory cells and contractile cells (figure 4e). Enterocytes are optimized towards the uptake of nutrients; either directly from the mucus or from the nutrient-enriched water of the gastric cavity. Morphologically, they still resemble the endochoanocytes in that they often bear motile cilia in combination with a collar of microvilli, at least in most invertebrates [45]. Microvilli increase the membrane surface facing the gut lumen and thus facilitate nutrient uptake. Finally, excretory cells are specialized in removing the nutrient-deprived, nitrogen-enriched water from the body cavity. While in sponges the choanocytes engender fluid motion and water removal, specialized cells exert this function in many other animals, such as the protonephridial flame cells. A choanocyte-origin has been previously postulated for enterocytes [45] and for flame cells [46], based on morphological, functional and molecular resemblance.

Having reasoned that a benthic, gastraea-like ancestor with a mucociliary sole populated the Proterozoic sea floor, the stage is set for the evolution of the first nervous system. We will elaborate here that the first neurons emerged in this ancestor in the context of an ectodermal nerve net (figures 4f and 5). The initial function of this nerve net might have been the control of amoeboid locomotion of the gastraea, grazing on algal mats (resembling the movement of an amoeba, but at multicellular scale). Figure 5.Evolution of the nerve net. (a) Evolution of a sensory-contractile network of neurons and muscle cells by division of labour [5,6,125], following an initial scenario proposed by Mackie [126]. Starting point is a network of multifunctional, sensory-contractile cells in the gastraea ectoderm. Contractile fibres are oriented along the apical-blastoporal body axis. Individual sensory-motor neurons innervate many muscle cells. (b) Cell types of the cnidarian nerve net, modified from [127]. (c) The neuromuscular orthogon. Muscle cells have congregated into muscle strands. The evolution of true sensory, inter- and motorneurons enables the differential and antagonistic contraction of entire muscle strands as a response to remote stimuli. (d) Ganglion cells in the ectodermal nerve net in Cerianthus, after [128]. (e) Mechanosensory ciliary-cone receptors in the ectodermal nerve net in Cerianthus, after [56]. (f) Bipolar cells in the ectodermal nerve net in Cerianthus, after [128]. (g) Connections and signaling directions (arrows) in the nerve net of Cerianthus, after [128].

Nerve net-like arrangements of neurons cover the almost entire ectoderm in cnidarians and ctenophores [61,129,130–136]. In ctenophores, the ectodermal nerve net is a polygonal mesh of neurite bundles; another nerve net spans the mesoglea between the ectoderm and the endomesoderm [129]. In the cerianthids, a basal cnidarian lineage, the nerve net appears to be of purely ectodermal nature [128], which suggests that the evolution of inner layer neurons occurred independently in ctenophore, cnidarian and bilaterian lineages [136]. As sponges and placozoans do not form a nerve net, the most parsimonious assumption appears to be that a nerve net existed in the last common ancestor of cnidarians, ctenophores and bilaterians (a grouping referred to as Neuralia by Nielsen [8]); yet, in light of the current uncertainties about the placement of the ctenophores [137], the existence of the ‘Neuralia’ at the exclusion of sponges and placozoans remains unsettled: cf. figure 1a).

The presence of a well-developed nerve net correlates with the presence of basiepithelial muscle fibres directly innervated by nerve net neurons, in cnidarians and in ctenophores [61,67,128,132,133,135,136,138–142]. Therefore, the evolutionary origin of the nerve net appears to be closely linked to that of the musculature [44]. As a rule, the ectodermal muscle fibres associated with the nerve are oriented longitudinally, that is in parallel to the primary body axis (while the inner layer forms circular fibres [61,62]). This matches the situation in the basal cnidarian cerianthids [61], as well as the medusozoan ground pattern [62], and longitudinal muscle fibres are the first to develop underneath the epidermis in the ctenophore Mnemiopsis [143], followed later by circular fibres [129]. It is thus plausible that the benthic gastraea evolved longitudinal muscle fibres innervated by an ectodermal nerve net, by division of labour as illustrated in figures 4f and 5a,b. The perpendicular arrangement of myofibres innervated by an ectodermal nerve net is referred to as ‘neuromuscular orthogon’ ([44]; figure 5c). Importantly, neuromuscular orthogon and mucociliary sole appear to have overlapped, because longitudinal muscles reach the gastric opening (at least in cnidarians).

Such an ectodermal nerve net essentially controlled body wall contractions, possibly mediating local to more widespread ‘avoidance’ responses to mechanical stimuli [141]. Most probably, the nerve net first thus coordinated some very primitive form of amoeboid locomotion, which required alternating and antagonistic contraction of perpendicular myofibres, in order to elongate the body towards the desired direction and/or to change body shape in reaction to mechanical stimuli that were met on the locomotor trajectory. In essence, we hypothesize that such locomotion required more sensory integration and coordination with increasing body size (and that the increase in body size in turn resulted from the invention of the mucociliary sole with external digestion). It will be interesting to work out how an animal of this organizational grade may relate to beginnings of the ctenophore line of evolution [144], and whether placozoans pre- or postdated the invention of the nerve net (that is, whether they are ancestrally simple or secondarily miniaturized [3], having potentially lost or modified their gastric cavity, neurons and myocytes [137]).

What were the cell types that constituted the first nerve net? Core units of the basal metazoan nerve nets appear to be local two- to three-cell sensory-effector reflex arcs [138] that directly innervate the effector cells (myoepithelial cells or cnidocytes in cnidarians [138], muscle cells and ciliated cells in ctenophores [129]). These arcs are interconnected at the level of the sensory neurons or ganglion cells, so that the whole epithelium functions as a single tissue-type reflex arc [145] that triggers contraction of myoepithelial cells (and/or local discharge of cnidocytes in cnidarians). Chemo- and photosensation as well as neuropeptide and indolamine release may have modulated nerve net activity.

The cellular composition of the nerve net has been studied by light- and electron microscopy as well as immunocytochemistry in numerous anthozoans [61,128,132,146,147], hydrozoans [142,148,149] and ctenophores [129,134,150] and proved rather homogenous and similar, which allows ancestral state reconstruction. Mechanosensory ‘ciliary-cone cells’ (figure 5e) occupy the ectodermal surface of cerianthids and sea anemone [56] (and similar cells exist in the nerve net of hydrozoan polyps [149,151,152] and ctenophores [150]). An apical collar of microvilli surrounds a single cilium [49,153], identifying these cells as ectochoanocyte descendants (see above). Many of the sensory cells form synapses [56,138], thus representing true sensory neurons. In addition, two types of ganglion cells are commonly observed, in cerianthids [128], sea anemone [132], hydrozoans [148] and ctenophores [129,134]. The small multipolar neurons (figure 5d) make up a large part of the nerve net in sea anemone, projecting few, thin neurites in different directions [130]. In cerianthids, these are the only neurons to directly receive sensory innervation, to connect to each other, and to directly innervate, via multiple collaterals, the underlying myofibres [128]; they can thus be regarded as the core of the ancestral nerve net [130]. By contrast, the large bipolar cells (figure 5f) send out neurites that are much thicker, less frequent and follow the muscle fibres [130]; they represent second order interneurons transmitting nervous excitation over longer distances, mediating for example whole body retractions [128,130]. The information flow in a prototype nerve net is exemplified for cerianthids in figure 5g.

The prevalence of mechanosensory neurons with actin collar and cilium, multipolar ganglion cells and myoepithelial cells in extant nerve nets supports the idea that the first neurons arose by division of labour from mechanosensory-contractile ectochoanocytes, as hypothesized in the diversification scheme in figure 4f. Similar ideas have been put forward repeatedly [5,66,67] and imply that the first synapse arose from the aggregation of vesicle-related intercellular signalling modules, including synaptotagmin, SNARE-complex components and postsynaptic receptor complexes, as extensively reviewed elsewhere [15,32,34,46,60]. This aggregation process might have occurred around adherens junctions, as suggested by the shared presence of both N- and E-cadherin in adherens and synaptic junctions [154]. In line with a division of labour scenario, cnidarian sensory neurons and ganglion cells are related to the extent that they cannot be clearly separated into two distinct cell types [139]. To test these hypotheses further, the molecular profiling of cell types constituting the nerve net in cnidarians and ctenophores will be especially rewarding—and also solve the yet unsettled question of nervous system homology between ctenophores and other neuralians [155].

In line with the postulated origin of nerve net neurons from ectochoanocytes, nerve net synaptic transmission appears to be largely glutamatergic, as recently postulated for ctenophores [156] and as indicated by many lines of circumstantial evidence in cnidarians. In sea anemone, NMDAR 1 immunoreactivity localizes to both nerve and epitheliomuscular cells [157] and glutamate immunoreactivity has been detected in distinct cells of the ectodermal layer and in the nerve plexus, but was absent from mesoglea and endoderm. Immunoreactivity of the nerve plexus forms a continuous ring on top of the epitheliomuscular cell layer in tentacle cross-sections [158]. Electron optics reveal the presence of glutamate in individual neurites within the nerve plexus, and in large dense-core vesicles within putative neural processes (but evidence for glutamate in synapses is yet lacking [158]). Glutamate exerts inhibitory activity on circular muscle contraction of the sea anemone Actinia equina [159], indicating that glutamate exhibits opposite effects on ectodermal and gastrodermal musculature. Adding to this, numerous studies report paracrine release of structurally related, amidated neuropeptides of the RFamide and other families from cnidarian nerve net neurons [136,139,157,160].

A smaller subset of cnidarian nerve net neurons has been reported GABA immunopositive [146]. In sea anemone [158], GABA and GAD immunoreactivity is prominent in the outer and inner sublayers of ectoderm and endoderm, but not in the nerve plexus. In the basal ectoderm, the epithelio-muscular layer displayed an intense label, but the nerve plexus remained scarcely stained. In the outer ectoderm, GABA labelled some types of sensory neurons [158]. Just like cnidarians, ctenophores use glutamate in nerve net signalling, while GABA has been detected in muscle cells [15,156]. These data suggest that glutamate was the first excitatory transmitter in nerve net neurons, and that GABA acted as an inhibitory signal, possibly released in a refractory manner by the innervated muscle cells themselves.

It is reasoned here that one decisive step in the evolution of animals was the formation of a paired fold in the inner lining of the gastraea, the so-called primary mesentery (figure 6a). This mesentery then duplicated into two metameric, bilateral series, which enclosed gastric pouches in each half of the body (figure 6b). This matches the ground pattern in cnidarians assumed to be ancestral [106]. The obvious advantage of this innovation was the extension of the internal surface, i.e. the surface of the resorptive epithelium, and would have intensified body physiology and probably allowed the maintenance and movement of larger body forms [163] (as were present in the Ediacaran, see below). It has long been suggested that the invention of gastric pouches predated the CBA [164]; this notion can now be revived based on molecular evidence [44]. Beyond that, a plethora of molecular data from sea anemone suggest that bilateral symmetry indeed evolved prior to the cnidarian–bilaterian divergence [165]. Figure 6.The evolution of gastric pouches. (a) Hypothetical modified gastraea with an anterior and a posterior pouch, separated by the primary mesenteries. Establishment of bilateral symmetry. Schematic lateral views (upper panels) and ventral views (lower panels). (b) Duplicated pouches leading to two bilateral sets of gastric pouches separated by mesenteries. Schematic lateral views (upper panels) and ventral views (lower panels). (c) Gbx expression in the gastric pouches of the Nematostella late planula (left) and in the amphioxus embryo (right). Midlle panels show gbx expression in a general cnidarian versus bilaterian (with slit-like blastopore). (d) The metameric gastric diverticula of Dickinsonia, adapted with permission from [161]. Stippled circle indicates inferred zone where additional diverticula have been added.

(e) The metameric gastric pouches of Cerianthus, adapted from [162]. Stippled circle indicates growth zone where additional mesenteries develop.

In anthozoan cnidarians, the first manifestation of bilateral symmetry is the formation of a slit-like oral opening, with ciliated grooves on either side that establish water flow [166]. Within the gastric cavity, the mesenteries channel circulation into the pouches. In scyphozoan medusae, such as the moon jellyfish Aurelia, the gastric circulation is likewise channeled by the mesenteries (called gastric septa in medusa) into equal streams that pass through the gastric pouches [165].

In contrast with anthozoan cnidarians, the gastric opening of the CBA was most probably facing the substrate, grazing the algal mats with a mucociliary sole (see above). If so, some additional advantages of the newly acquired bilateral symmetry were obvious. Plausibly, body movement changed from multidirectional to unidirectional towards the new anterior end. This would have allowed more targeted food search and avoided passing through areas that had already been grazed (cf. figure 4a). Also, a slit-like blastopore would have allowed defecation of the digested and nutrient-deprived mucus with larger non-digestible particles at the new rear end. This is a very straightforward way to evolve the first forward movement: to leave the mucus behind. Apart from that, we can assume that the overall establishment of internal currents passing through the surface-extended pouches might have closely resembled that of some extant cnidarians.

Key to our hypothesis is the presumed homology of cnidarian gastric pouches with the bilaterian coelomic cavities [164]. As outlined elsewhere in more detail [44], new molecular evidence lends strong support to this scenario. Most striking is the spatially restricted expression of gbx genes in all but the first pair of pouches, as similarly observed in sea anemone and amphioxus (figure 6c). Remarkably, these first pouches give rise to a reduced set of musculature in both sea anemone [106] and amphioxus [167]. Moreover, cnidarian gastric pouches and bilaterian somites share the staggered and collinear expression of hox genes with expression boundaries coinciding with boundaries between pouches/somites [44]; and finally, cnidarian gastric pouches express genes that specify somitic musculature in bilaterians, such as mox and twist [168].

It is tempting to speculate that these bilateral pouches probably corresponded to the ‘digestive diverticulae’ present in Ediacaran fossils such as Dickinsonia [169] (figure 6d). These diverticulae have been compared with cnidarian pouches repeatedly [169,170] (figure 6e), which would imply that modern cnidarians resemble ‘a grade of Late Precambrian organization from which bilaterians evolved’ [171]. Among the many interpretations that Ediacaran fossils have been subjected to (e.g. [4,172] for an overview), the comparative approach conducted here may thus help to finally select the most plausible variant and place this fossils right in the centre of metazoan evolution and the bilaterian stem [172]. Strikingly, the Dickinsoniids continuously add new mesenteries and diverticules at their ‘posterior’ end [172]; and in a strikingly similar manner additional mesenteries are added in extant cnidarians such as cerianthids or other anthozoans (stippled circles in figure 6d,e).

The last important step in the evolutionary scenario developed here is the median closure of the gastric opening, leaving mouth and anus at opposite ends (figure 7). This step is often recapitulated in bilaterian development, in that both mouth and anal tissue develop from blastoporal tissue (amphistomy) [173] (for alternative views on the interrelationships between gastric opening, mouth and anus, see [44,174]). This closure of the gastric opening along the new ventral midline would have firmly joined together the two bilateral halves of the mucociliary sole. During this transition, specialized subregions of the former nerve net would have given rise to the bilaterian nerve cord. In figure 7, red and yellow demarcate longitudinal subdivisions of nerve cord tissue that derive from different parts of the nerve net: The medial section corresponds to the tissue of the former mucociliary sole; this tissue most probably evolved into the medial motor column of the bilaterian nerve cord [44], and the more lateral section into the lateral sensory-integrative columns. Conserved expression of mediolateral patterning genes supports this scenario, as detailed elsewhere [44]. For example, nk homeobox genes genes (nk2.2 and nk6) specify the medial motor centre, wherease pax paired-type homeobox genes (pax37, pax258) and msx specify the more lateral tissue (figure 7). Figure 7.The evolution of the bilaterian nervous system. (a) Neuralian ancestor. A nerve net covers the gastrula-shaped animal. Nerve net neurons situated around the digestive opening (red) control feeding. White arrows indicate direction of water flow. (b) Cnidarian–bilaterian ancestor.

Specialized parts of the nerve net are centred on the slit-like digestive opening: the nerve of the digestive sole (red) and the sensory-integrative centre (yellow), controlling contraction of a bilaterial series of contractile gastric pouches. White arrows indicate direction of water flow. (c) Urbilaterian. Motor centre of the ventral nerve cord (red) and sensory-integrative centre (yellow). White arrows indicate direction of water flow.

The proposed closure of the slit-like opening of the gastric cavity necessarily impacted on mucociliary feeding, as trapped particles could no longer be internalized alongside the entire gastric slit, but only through the new mouth (and, presumably) anus. In consequence, mucus is no longer transported towards the new ventral midline but towards mouth and anus (white arrows in figure 7b,c, lower panels). Given that the mucus sticks to the substrate, the anterior-to-posterior movement of the mucus on the ventral body side would have pushed the animals forward. In line with this, creeping forward by means of a mucociliary sole is a prevalent mode of locomotion in many bilaterian phyla. Strikingly, an anterior-to-posterior movement of a mucoid thread propagated by motile cilia, the so-called Reissner's fibre, is observed until today in virtually all chordates (vertebrates and amphioxus) [175]. Our scenario explains the evolutionary origin of this fibre, yet cannot explain why it is so persistent in extant vertebrates.

The evolutionary scenario developed here sheds new light on the evolution of animal nervous systems, highlighting and postulating two major morphological transitions that increased the efficiency of feeding and locomotion and allowed larger body sizes. First, external digestion with the mucociliary sole would have enabled external digestion and thus enhanced the concentration of nutrients in the gastric cavity. In these animals, a first net of interconnected sensory neurons and contractile effector cells may have evolved that sped up and steered grazing on algal mats. Second, the substantial increase of the internal nutrient-resorbing surface via folding into gastric pouches might have boosted physiology and, consequently, energy supply for locomotion, thus triggering the regionalization and specialization of parts of the nerve net needed for the control of more sophisticated and targeted movements. This ultimately led to convergent centralization events in ctenophores, cnidarians and bilaterians. Necessarily, in the absence of comparative molecular data for most of the cell types discussed here, the postulated sequence of cell-type diversification events underlying these transitions remains tentative. To change this, it is the aim of this review to open up and encourage a new phase of evo-devo, which looks at animal evolution from a cell-type perspective, and to combine these new data with classical comparative anatomy, developmental biology and palaeontology. It is expected that the rapid progress in the profiling and functional characterization of cell types in different animals will boost these efforts.