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

Prerequisite for tracing nervous system evolution is understanding of the body plan, feeding behaviour and locomotion of the first animals in which neurons evolved. Here, a comprehensive scenario is presented for the diversification of cell types in early metazoans, which enhanced feeding efficiency and led to the emergence of larger animals that were able to move. Starting from cup-shaped, gastraea-like animals with outer and inner choanoflagellate-like cells, two major innovations are discussed that set the stage for nervous system evolution. First, the invention of a mucociliary sole entailed a switch from intra- to extracellular digestion and increased the concentration of nutrients flowing into the gastric cavity. In these animals, an initial nerve net may have evolved via division of labour from mechanosensory-contractile cells in the lateral body wall, enabling coordinated movement of the growing body that involved both mucociliary creeping and changes of body shape. Second, the inner surface of the animals folded into metameric series of gastric pouches, which optimized nutrient resorption and allowed larger body sizes. The concomitant acquisition of bilateral symmetry may have allowed more directed locomotion and, with more demanding coordinative tasks, triggered the evolution of specialized nervous subsystems. Animals of this organizational state would have resembled Ediacarian fossils such as Dickinsonia and may have been close to the cnidarian–bilaterian ancestor. In the bilaterian lineage, the mucociliary sole was used mostly for creeping, or frequently lost. One possible remnant is the enigmatic Reissner's fibre in the ventral neural tube of cephalochordates and vertebrates.

Reconstructing the evolution of the nervous system is one of the most exciting challenges of current evolutionary research. Almost certainly, a nervous system was present in the last common ancestor of cnidarians and bilaterians (cnidarian-bilaterian ancestor, CBA); yet, since the early metazoan tree remains unsolved (figure 1a), we cannot tell when a nervous system first emerged. Also, it is possible that the body plan of sponges or placozoans was secondarily simplified [1–3], which might have implied secondary loss of nervous system components. One way to approach the early evolution of the nervous system is to [4] track the emergence and diversification of its constituent cell types—sensory receptors, neurons and effector cells. The sister cell-type concept [5] posits that cell-type evolution can be resolved into a scheme of stepwise diversifications, giving rise to sister cell types [5]; and a recent study comparing cellular transcriptomes indicates that a hierarchical tree structure indeed dominates cell-type interrelationships [6]. This review accordingly addresses cell-type diversification in early metazoans and how it paved the way towards the birth of the first neurons. In favour of this approach, the support for a given branching order of cell-type trees is quantifiable and thus testable, meaning that the hypotheses on cell-type interrelationships put forward here can be challenged by comparative whole-organism single-cell profiling approaches [7] in the future. Figure 1.Cell-type diversification in metazoan evolution. (a) A simplified tree of metazoan evolution, with major inventions mapped. CBA, cnidarian–bilaterian ancestor; LMCA, last metazoan common ancestor. (b) The interrelationship between species, cell type and molecular evolution.

The scheme shows how the cell-type tree is embedded in the species tree, and the inset shows how cell-type diversification is paralleled by subfunctionalization (yellow and dark blue lines), divergence (bright blue line) and neofunctionalization (green line) of molecules and cellular modules. It contains examples for each possible mode of cellular diversification—via functional segregation, divergence or co-option of differentiation modules. Functional segregation implies that molecular machines are differentially distributed between emergent sister cell types. At gene level, functional segregation leads to the loss of entire differentiation modules from the specializing cells. This is distinct from functional divergence, where functions (and modules) are retained and modified in different directions.

Cell-type diversification involves the segregation and divergence of existing cellular modules or functions [5]. One prominent example is the division of labour that accompanied the birth of the first neurons and muscle cells (see §2). The general scheme in figure 1b illustrates how the evolution of cellular modules may have underlain the evolution of cell types, which in turn triggered species evolution—visualized here by encapsulated module, cell types and species trees. Following this scheme, this review provides a first survey on how, in practice, the interplay between molecular, cell type and species evolution may have shaped early metazoans, that is, it tries to track the cellular and molecular diversification events that underlay the metazoan species tree depicted in figure 1a. As much as it is clear that we are very far from certainty about almost all of these events, it is also true that first insights can be gained (and tested by comparative cell-type analysis, see above). A plethora of available comparative structural and ultrastructural data, recently complemented by molecular studies on basal metazoans (and on the metazoan outgroup, the choanoflagellates), allows a preliminary sketch of early metazoan cell-type evolution (that is attempted here).

The starting point will be the first metazoan, which most probably resembled a sphere of epithelialized choanoflagellate-like cells, the ‘choanoblastaea’ [8]. The recent characterization of differentiation modules in the unicellular choanoflagellates has provided rich information on the molecular composition of these cells and their multifunctional nature. The inward fold of one-half of the sphere then led to the next important ancestor, the gastraea. We will hypothesize that the simple, bilayered, cup-shaped anatomy of this ancestor was reflected by a first diversification at the cellular level, including: (i) the external ‘ectochoanocytes’, mediating interactions with the environment; (ii) the internal ‘endochoanocytes’ that specialized towards particle capture and nutrient resorption; and (iii) the ‘kopeocytes’, regulating water currents around the gastric opening. Subsequently, the kopeocytes may have further diversified into cell types specializing in ciliary movement and control, mucus production and secretion of digestive enzymes, thus enabling external digestion and forming a ‘mucociliary sole’ around the gastric opening; the endochoanocytes into cell types specializing in internal nutrient distribution and resorption; and the ectochoanocytes into cell types specializing in mechanosensation, information integration and body contractions in response to environmental stimuli. These latter cells would have constituted the first nerve net, which possibly controlled some form of amoeboid locomotion. Such equipped animals may have been close to some simple Ediacaran fossils that have been interpreted as grazing on algal mats with a mucociliary sole for external digestion [3].

Building on this, we will hypothesize that two important changes in the body plan enhanced efficiency of mucociliary sole feeding: (i) the transition to bilateral symmetry, which enabled directed movement, facilitated food search and avoided grazing the same area twice and (ii) the extensive folding of the inner surface into metameric pouches, which optimized nutrient distribution and resorption. We propose that, with these changes, basal metazoans had reached the organizational grade of the Ediacaran fossil Dickinsonia, which may have been close to the last common ancestor of cnidarians and bilaterians (that we collectively refer to as ‘Metameria’, figure 1a). With these animals, the stage was set for nervous system evolution: further diversification of the cell types making up the nerve net and the mucociliary sole gave rise to the plethora of receptor cells, neuron types and muscular and ciliary effector cells present in today's cnidarians and bilaterians.

Another comprehensive view of early metazoan body plan and nervous system evolution has recently been put forward [8]. While initial steps of metazoan evolution (such as the blastaea, see below) are in line with the scenario presented here, the CBA is subsequently derived from modified sponge larva—while the scenario presented here favours an uninterrupted sequence of benthic adults, starting with the gastraea (which each developed via swimming larval stages).

Stem metazoans were likely composed of one main characteristic cell type, resembling, structurally and functionally, the choanoflagellates [8–10; see however 11]. Figure 2 sketches a possible trajectory from choanoflagellate-like pre-metazoans to the last metazoan common ancestor, involving a transition from multiple solitary and colonial forms [12,13] to a fixed life-cycle. These ancestors represented an epithelium [14] that took the shape of a sphere, called choanoblastaea [8], which later gave rise to the cup-shaped gastraea (figure 2; see also §3). They were swimming or lived partially benthic in shallow waters, filtering and concentrating dissolved organic carbon and bacteria as microphageous suspension feeders [3,15]. Building on the resemblances between choanoflagellates and sponge choanocytes, classical authors agreed that the first metazoan cells had an apical collar with an undulating flagellum [16,17] (figure 2b). However, a modern blastaea concept can be much more detailed on the ultrastructure of these cells, as inferred from electron microscopy and from the impressive list of metazoan proteins shared with the choanoflagellates [18]—which were necessarily present in the earliest metazoans. Figure 2.The origin of metazoans, involving a change from facultative transitions between forms (orange arrows) to fixed life-cycle transitions in development (blue arrows), and a move from temporal differentiation of cell types to spatial differentiation. Block arrows indicate direction of evolutionary change. (a) Pre-metazoan state with transitions between chain colonies (cc), amoeboid cells (ac), slow swimmers (ss), fast swimmers (fs) and rosette colonies (rc). 2n and n indicate life cycle transition between diploid and haploid state. (b) Stem metazoan. Variant (a): The zygote develops into a swimming choanoblastaea (cb).

Variant (b): The zygote develops into a benthic, epithelial, biofilm-like organism (be). Transitions between the two forms might have occurred by epithelial curvature (stippled arrows) and closure. (c) Last metazoan common ancestor (LMCA). The zygote (2n) develops via cleavage into a blastula (bl; variant a), or, into a sheet-like embryo (se; variant b), which inverts into an amphiblastula larva (as seen in calcareous sponges; stippled arrows). The amphiblastula gastrulates into the adult gastraea (ga), which produces gametes and closes the life cycle.

One important detail in the context of this review is the fine glycocalyx mesh of the microvillar collar that served as a mucoid filter [12], as well as another mesh of mucus that joined the collars of neighbouring cells [19,20]. This means that, from the very beginning, metazoan feeding was mucociliary—using both motile cilia and mucus for food capture. However, digestion was intracellular at first and took place in the same multifunctional cells that also trapped the food—just as in unicellular eukaryotes. (These two steps of feeding were spatially segregated only in more advanced metazoans, by division of labour, as will be outlined below.) Food particles were taken up by phagocytosis or pinocytosis, followed by intracellular lysosomal degradation [21]. Most probably, nutrients were shared with cells inside the choanoblastaea via the primary body cavity [8].

Other functions related to the reception and integration of, and reaction to, environmental stimuli [15]. (These were later inherited by the first neurons, as will be outlined below.) The flagellar cilia were both motile and sensory (as reported for unicellular eukaryotes and human airway cilia), and thus contained a very short and direct sensory-effector arc [22]. Chemosensation involved guanylyl cyclases signalling via cGMP [23]. Five subfamilies of transient receptor potential (TRP) channels were present [24], among which TRPV (vanilloid TRP) channels likely played an ancient role in mechanosensation, as they mediate mechanosensation in Paramecium and Chlamydomonas [25], human airway epithelia [26] and Caenorhabditis elegans sensory cells [27]. In Chlamydomonas, these channels localize to the proximal region of the cilia, where active bending is restricted (and self-activation avoided) [25]. Proteins required for action potential generation and propagation were likewise present in early metazoans, such as voltage-gated sodium [28,29], potassium and calcium channels [30]. Mechanical and chemical stimuli triggered depolarization of the ciliary membrane and calcium influx into the cilium that had a direct effect on ciliary beating such as reversal or inhibition, as reported for unicellular eukaryotes, cnidarians, ctenophores and bilaterians [31]. Adding to this, choanoflagellates and metazoans share synaptic adhesion/signalling molecules, a primordial neurosecretory apparatus active on the apical cell surface (but no presynaptic active zone proteins) [32], postsynaptic density proteins such as Homer or Shank [33,34] and a variety of receptors including the metabotropic glutamate receptor [35] (but not ionotropic glutamate receptors, which only appeared in the metazoan stem).

Many of these molecules and associated functions were key to the evolution of the first neural cell types, as will be detailed below.

On the effector side, early metazoan cells already possessed a complex machinery for cellular contractions and shape changes based on a dynamic actomyosin cytoskeleton. Actin-based movement involved a slow and a fast version of myosin II that responded in a different manner and with different velocities to Ca2+ signalling [36]. The slow non-muscle myosin probably played a role in various cellular processes involving contraction, such as integrin-mediated basal adhesion, cell–cell adhesion dynamics via adherens junctions or apical constriction [37]—processes that were key for the evolution of metazoan epithelia. The slow myosin is also involved in amoeboid movement [38], which probably belonged to the behavioural repertoire of early metazoan cells. In addition, we can infer that early metazoan cells formed lamellipodia and filopodia, using proteins that reorganize cortical actin filaments such as the Arp2/3 complex, the actin cross-linking protein fascin, and myosin X, a myosin with pleckstrin homology domains that associates with regions of dynamic actin [39,40]. Filopodia are highly dynamic structures that probably played a role in anchoring and stabilizing cells in the blastaea (figure 2a,b).

It is a long-standing notion that the second step in the evolution of the metazoan body was the transformation of the spherical blastaea into the gastraea [8,9], which may have been close to the last metazoan common ancestor (figure 1b). The term ‘gastraea’ means ‘animal with a primitive gut’ and was coined by Ernst Haeckel [9]. Its most characteristic feature is the infolding of the lower body surface, resulting in an outer ectoderm, an inner gastroderm and a gastric opening (figure 3a). From Haeckel's times until today, strong support for the gastraea hypothesis comes from the prevalence of gastrula-like stages during animal development that are interpreted as recapitulation of the gastraea-like ancestral body plan. Even in sponges, a transitory cup-like stage forms during metamorphosis, as recently confirmed [41]; and in cnidarians and ctenophores, the overall body form resembles that of a gastrula during the entire life cycle—at larval, polyp and medusa stages. The ancient gastraea is originally assumed as creeping on the sea floor, with the gastric opening facing downwards (while other authors consider the gastraea a predominantly planktonic organism [8]): plausibly, this would have closed off the gastric cavity from the environment and thus allowed higher nutrient concentrations. In line with this, some sponge larvae settle with the blastopore facing downwards [41]. Sponges might be regarded as representatives of a gastraea-type organization (which then would have undergone surface extension independent of other metazoan lineages; however, it cannot be excluded that, alternatively, sponges lost ancient features [1,2]). Figure 3.The gastraea, the last common ancestor of living phyla. (a) The gastraea. This cup-shaped animal resulted from simple infolding of the initial blastaea and first cell-type diversification.

Colours relate to similar cell types in (b) and (c). (b) Infolding of the amphiblastula larva of calcareous sponges into a cup-shaped transitory stage that resembles a gastrula. The inner cells express endomesodermal markers, after [41,42] and various sources. (c) The initial diversification of metazoan cell types into sensory-contractile cell types and cell types related to feeding. The colour scheme matches the cells of the gastraea in (a). A multifunctional choanocyte-like cell type gives rise to ectochoanocytes, endochoanocytes and kopeocytes (black arrows). These ancestral cells had an apical, undulating cilium or flagellum, propelling water away from the cell, surrounded by a circle of long, contractile, actin-containing microvilli. A well-developed Golgi complex produces lysosomes for intracellular digestion, and a prominent endocytotic machinery mediates phagocytosis (Pc). Mm, mucoid mesh; Fp, filopodia.

Modern support for the classical gastraea hypothesis comes from molecular data. A larger set of genes specifying endomesoderm has recently been found expressed in the ciliated micromeres of the amphiblastula larva of the calcareous sponge Sycon ciliatum, which later invaginate and transdifferentiate into the choanocytes of the adult sponge [42] (figure 3b). This supports homology of sponge inner layer with the endomesoderm (i.e. the gastroderm) of other animals. Additional support comes from the conserved expression of conserved transcription factors such as brachyury, and signalling molecules such as hedgehog, in the blastoporal tissue of various metazoans [43,44].

Given the structural and functional resemblance between sponge choanocytes and choanoflagellates, it seems obvious that the gastraea, just like the blastaea, used choanocyte-like cells to feed on dissolved organic carbon or microbes suspended in the inflowing seawater (microphagy [3]). However, it is also plausible that these cells had already started to diversify, with inner cells specializing more on feeding functions and outer cells on functions important for the interaction with the environment—such as protection, different sensory modalities and primitive body movements. For these outer versus inner cell types, we propose the names ‘ectochoanocytes’ and ‘endochoanocytes’, reflecting their choanocyte-like appearance and their developmental origin from ecto- and endoderm, respectively (figure 3c). Another founder cell type that will be key to the scenario developed here is hypothesized to have existed at the interface between ecto- and endochoanocytes, around the gastric opening (the ‘kopeocytes’; see §3c below). These cells specialized in the generation and control of water currents (and later gave rise to the cell types constituting the mucociliary sole; see §4).

Besides the functions related to feeding, such as flagellar pumping, microvillar-mucoid filtering and pino- and phagocytosis [12,45], the endochoanocytes likely also retained some gustatory chemosensory functions, as recently proposed [46], together with the capacity for cellular contraction—mostly via the ‘slow’ myosin II, which was likely present in the ancient gastroderm as it is broadly expressed in sponge choanocytes and in the cnidarian gastroderm [36].

Ectodermal mechanosensory cells with obvious resemblance to sponge choanocytes have frequently fuelled cell-type comparisons across metazoans [47–49]. For example, the structural resemblance of the hair cells of the ear has long been noted [8,46,50,51]. Just like choanocytes, these cells bear a central cilium, called ‘kinocilium’, surrounded by or adjacent to interconnected microvilli, called ‘stereocilia’ (figure 3c). Besides the vertebrate acustico-lateralis system, similar mechanoreceptors are found in annelids [52] or acoels [53], and in great diversity in cnidarians: in cerianthids, sea anemones, corals and medusozoans [49,54–56]. Remarkably, the mechano-‘sensory’ cilium of these cells appears to derive from a motile cilium and sometimes retains residual motility (e.g. in the insect chordotonal neurons [57] or in hair cells [58]); the axoneme of the kinocilium accordingly possesses a central doublet of microtubules as otherwise characteristic for motile cilia. Interestingly, mechanosensory cells such as vertebrate hair cells and insect chordotonal cells express TRPV receptors [57], which might be inherited from early metazoans (see §2a). In line with an ancient role of these receptors in mechanotransduction, a mammalian TRPV receptor rescues the mechanosensory (and osmosensory) properties of a TRPV channel in nematode worms [59]. Notably, these cells have not (yet) been described for sponges—as they occur in the ectoderm of virtually all other phyla, however, they were very likely present in the ancient gastraea, too.

Contractile ectodermal fibres are found in sponges [60], cnidarians [61,62], ctenophores [63] and in many bilaterians, making it likely that they were also present in the ectoderm of the gastraea (figure 3c). Sponge contractile movements are largely mediated by the slow ‘non-muscle’ myosin, which is expressed in most cell types in Tethya wilhelma including the contractile pinacocytes [36,60]. Ectodermal contraction appears to have always involved intercellular communication by glutamate and GABA: in sponges, contraction waves involve paracrine intercellular signalling, with glutamate triggering and GABA inhibiting contraction [64,65]. Intercellular propagation of the wave is slow and involves the metabotropic glutamate receptor [15].

As will be explained in more detail below, the segregation of sensory-integrative mechanosensory cells versus contractile epidermal cells (figure 3c) may have represented the first step towards the evolution of the later nerve net. This is in line with the classical idea that mechanosensory-contractile cells gave rise to the first neural circuit by division of labour [5], as proposed by Kleinenberg [66] or Mackie [67]; and with more recent suggestions that the sponge contraction waves can be regarded ‘forerunners’ of nervous activities [15,60].

A key characteristic of the gastraea would have been a flow of water into the gastric opening, to ensure fresh supply of suspended particles and to maximize feeding. In contrast with the movement away from the epithelial surface generated by a standard flagellum, this new flow was needed in parallel to the epithelial plane. Therefore, and inevitably, the motor cilia generating this current and the cells bearing these cilia must have shown planar polarity [68] and an oar-like motion (effective stroke) [69], similar to the motile cilia of swimming larvae. (Note that flagella and effective stroke cilia differ structurally [70].) It appears plausible that such cells, shoveling water and suspended particles into the gastric cavity with effective stroke cilia, were first situated around the gastropore, red and brown in figure 3c. We propose to refer to these cells collectively as kopeocytes (from greek ‘kopeon’, the oar). Besides the effective stroke cilium, the kopeocytes presumably retained the apical secretory apparatus and basal contractility. By division of labour, such cells may have generated a number of important sister cell-type descendants, such as effective stroke ciliomotor cells, primary cilia sensory cells, specialized contractile cells and the various cell types of the mucociliary sole, as will be further outlined below.

In sponges, pigmented ring cells and apopylar cells may represent direct kopeocyte descendants. The apopylar cells, regulating water outflow from the choanocyte chambers [20], are contractile and express the fast type of ‘myosin II’ [36]; and ‘fast’ myosin II is also expressed around the pigmented ring in Amphimedon larvae [36], which also bears effective stroke cilia [71]. In other animals, the occurrence of effective stroke cilia often relates to the production of water currents through openings that may ultimately relate back to the gastropore, suggesting that the respective cells are kopeocyte descendants. However, it is not yet clear whether effective stroke ciliated cells represent a uniform cell type or, alternatively, evolved independently one or even many times. Two conserved transcription factor families have been identified that control the motile ciliogenic program in all metazoans: the forkhead-related FoxJ [72] and Rfx, a winged helix factor [73–75]. However, the FoxJ/Rfx network appears to specify effective stroke and flagellar motile cilia alike [76] and thus cannot solve this question.

Sensory cells with primary cilia, such as the mammalian kidney epithelia, are likewise widespread in animals and have recently been found in sponges—inside of the outflow opening, the osculum, which senses chemical signals and water turbulences and initiates contractile responses [15,60,65,77] (see below). This indicates that they may be part of the cellular repertoire situated around the gastric opening and may have arisen from kopeocytes by sensory-motor division of labour (figure 3c). Lacking the central pair of microtubules and the spoke apparatus (as well as the actin collar), primary cilia are non-motile. They sense flow (vibration) via TRP channels [78], responding to a deflection of the cilium owing to water current with a calcium signal. In vertebrates, the two TRPP duplicates PC1 and PC2 comprise a rheosensory complex that translates deflection of the primary cilium into signals [78]. Sponge primary cilia also appear to employ a TRPP (PC2) channel, as indicated by sensitivity to three pharmacological inhibitors blocking this channel in vertebrates [77]. Interestingly, primary cilia also respond to chemicals, and given that both rheosensation and effective chemosensation are necessarily linked to planar flow of water (and/or mucus, see below), it is plausible that primary cilia may relate back to the proposed diversification of kopeocytes (figure 3c). However, because the major characteristic of primary cilia is loss—of the actin collar and of the central microtubule doublet and the spokes—it is also possible that they evolved several times independently (e.g. in sponges versus other animals). To find out, it will be interesting to determine whether a common cellular module exists that is not found elsewhere and positively defines all primary cilia, or subsets of them, across animals.

While the distinct presence of mechanosensors (with actin collar and kinocilium) and of presumed rheo- or chemosensors has been postulated based on the frequent parallel occurrence of both sensory types in basically all metazoans, this notion finds support from the molecular evolution of the class A superfamily of bHLH of transcription factors. While this class is exclusive to metazoans [79], it already comprised three members—an achaete-scute (ASC), an atonal (ATH) and an E12/E47 superfamily member [80]—before the sponges branched off the metazoan lineage. Two of these—the achaete-scute and atonal families—are strongly associated with the specification of distinct sets of sensory cells across metazoans [57].

The atonal family genes specify mechano- and photoreceptor cells in a broad range of animals [57,81], such as the hair cells [82], Merkel cells [83,84] and ganglion cells in vertebrates, the chordotonal stretch and auditory receptor cells [57] and ommatidial photoreceptor cells [85] in insects and, as extrapolated from gene expression, the mechanosensory and photoreceptive cells in jellyfish [86]. (Note that the atonal-dependent rhabdomeric photoreceptor cells couple opsin-based phototransduction to mechanosensitive TRP channels [87].) An ancestral role of atonal in sensory cell specification is supported by mutual rescue of atonal function between Drosophila and mouse [88,89]. Atonal family members have been reported to directly control a gene regulatory network implementing mechanoreceptor morphology. In the Drosophila chordotonal mechanoreceptors, atonal activates Rfx and Fd3F (a Foxj orthologue [90]) that synergize to implement cellular modules characteristic for the mechanosensory cilium. This includes modules required for motility and the TRPV channels [91]. We can thus infer that atonal genes probably played an early role in ectochoanocyte specification. In line with this, the atonal superfamily member bHLH1 appears to be involved in the specification of flask-shaped sensory cells in the demosponge Amphimedon.

By contrast, the achaete-scute-like bHLH factors have been frequently implicated in the specification of chemosensory neurons. In the Drosophila chemosensory system, ASC is required for the specification of gustatory receptor cells in larvae [92] and in adults [93]. In vertebrates, ascl proteins are required for the specification of Merkel-like, 5HT+ taste bud cells [94] and for the 5HT+ sour taste type III taste bud receptors in mice [95,96]. Also, mouse Mash1 has been implicated in the control of catecholaminergic, chromaffin cell differentiation [97]. During early postnatal life in mouse and rodents, these cells possess ‘direct’ hypoxia- and CO2/H+-chemosensing mechanisms, and acute hypoxia causes depletion of adrenal catecholamines via a ‘non-neurogenic’ mechanism [98]. Remarkably, sensory ectodermal cells specifically express an achaete-scute homolog (ASH) gene in Hydra [99] and, most probably, Nematostella [100], consistent with the notion that achaete-scute family members play an ancient role in the specification of at least some kind of sensory neurons [101]. These findings indicate that the distinct genetic control of major sensory modalities was already in place in the last common metazoan ancestor (and thus predated the evolution of the nervous system).

Interestingly, the E2A proteins (E12/E47) have been shown to specifically bind the E-box-containing region of the smooth muscle actin promoter, regulating a smooth muscle cell differentiation marker in vivo [102]. Just like sponges, Trichoplax has an E12/E47 orthologue, indicating that the conserved function of this ancient class A bHLH protein was in the specification of cells with smooth muscle fibres (see above).

Various lines of evidence indicate that a new feeding mode evolved in the Precambrian ocean, referred to as mucoid-ciliary particle feeding [3] (figure 4a,b). This mode replaced the microphagy of individual cells. Food was predigested externally in a layer of mucus and the predigested nutrients transported into the body cavity with water currents. While the ancient choanocytes probably already used an intercellular mucoid mesh for particle trapping (see above), this mesh now covered the entire epithelium around the gastric opening, facing the substrate. Digestive enzymes were secreted into this mucus layer for external digestion. The coordinated and directional beating of effective stroke cilia distributed and transported the mucus over the digestive epithelium and washed the nutrients into the gastric cavity. Such hypothetical animal, which may have been close to the Eumetazoan ancestor, is depicted in figure 4c. Figure 4.Evolution of the mucociliary sole. (a) Body fossil of Dickinsonia costata associated with a series of feeding traces, reproduced with permission from [3]. Numbers delineate the order of their formation in relation to the body fossil at the end of the series of traces (trace #3 made last). Scale bar, 2 cm. (b) Cell types of the placozoan Trochoplax, reproduced with permission from [103]. The thick ventral body surface is composed of ventral epithelial cells (vci), each bearing a cilium and multiple microvilli; lipophil cells (lc); and gland cells (gc), distinguished by their contents of secretory granules and prevalence near the margin. Dorsal epithelial cells (dec) constitute the dorsal surface. In between the dorsal epithelium and ventral plate are fibre cells (fc) with branching processes that contact each of the other cell types. cc, crystal cell; dci, dorsal ciliated cell; gc, gland cell.

(c) A hypothetical gastraea-like ancestor with mucociliary sole. In these animals, mucocytes, zymocytes and serotonergic cells (brown) and motile ciliary cells (red) are located around the gastric opening. Contractile cells with ring-shaped fibres, excretory cells and enterocytes (all green) populate the gastric cavity. Mechanosensory cells (yellow) and contractile epithelial cells (orange) are hypothesised for a primordial nerve net. Blue arrows indicate direction of water flow across the mucociliary sole and into the gastric cavity. (d) Hypothetical diversification scheme for mucociliary cell types. (e) Hypothetical diversification scheme for endochoanocytes. (f) Hypothetical diversification scheme for ectochoanocytes into nerve net cells.