Independent origins of neurons and synapses: insights from ctenophores

Of course, we do not exclude the possibility that some genes (and/or neuron-specific enhancers and other non-coding genome elements acting as master regulators) could be uniquely and differentially expressed in all metazoan neurons when more taxa are included in such analyses, but careful co-localization studies are needed.

The presence of chemical transmission in ctenophores is well supported by functional studies [69,70,75]. However, the ultrastructure of their synapses is remarkably different from those in the majority of metazoans. Structurally, the asymmetrical synapses in ctenophores are quite organized, forming a so-called presynaptic triad, with little or no development of presynaptic density elements [95–97]. Each presynaptic element contains a tripartite complex of organelles: a single layer of synaptic vesicles lining the presynaptic membrane, a cistern of agranular endoplasmic reticulum just above the row of vesicles, and one or more mitochondria. The postsynaptic density, however, is less prominent in ctenophore synapses [95].

To our surprise, we found no biosynthetic enzymes for known ‘neuro'transmitters (i.e. acetylcholine, serotonin, histamine, dopamine and octopamine), with the exception of glutamate and GABA. Importantly, the genomic predictions were experimentally validated using ultrasensitive capillary electrophoresis with attomolar limits of detection [28,94]. This suggests that the ctenophore nervous systems are remarkably distinct among metazoans in terms of their most fundamental characteristics—chemical neurotransmission. The potential signalling roles of small secretory peptides, ATP, glycine and nitric oxide (NO) can be expected (e.g. the presence of genes encoding the relevant synthetic enzymes and receptors; table 1), but their functions as interneuronal messengers in ctenophores must be confirmed experimentally.

Currently, the only molecules identified as candidates for neuromuscular transmission are l-glutamate and perhaps l-aspartate [28]. Mistakenly, the ‘sensitivity to some classical neurotransmitters (e.g. l-glutamate)’ was interpreted as a characteristic uniting the nervous systems of ctenophores, cnidarians and bilaterians [71]. Conversely, the deeper analysis of the relevant pathways illustrates the extensive parallel evolution of the glutamate synthetic, transport and receptor pathways in ctenophores.

l-Glutamate has the highest affinity to induce action potentials, elevate [Ca2+] in ctenophore smooth muscle cells and cause muscle contractions, whereas other transmitter candidates are ineffective, even at concentrations up to 5 mM [28]. To our surprise, we found that ctenophores possess an enormous diversity of all components involved in glutamate signalling, including the synthetic, transport, receptor and inactivation pathways.

Specifically, we cloned and localized fourteen ionotropic glutamate receptors (iGluRs) and eight sialin-like glutamate transporters in P. bachei [28,93] (see figure 3a for relevant genes). We showed an unprecedented diversity of iGluRs in ctenophores, forming a distinct branch on the genealogical tree for metazoan iGluRs topology [28]. Thus, iGluRs might have undergone a substantial adaptive radiation in Ctenophora as also evidenced by unique exon/intron organization for many subtypes [28]. Importantly, ctenophore iGluRs could not be classified in the terms of vertebrate iGluR families (i.e. α-amino-3-hydroxy-5-methyl-4-isoxazolepropionic acid, kainate, delta or N-methyl-d-aspartate (NMDA), but our analysis suggests that ctenophore iGluRs could be related to an ancestral family including earlier NMDA-type receptors (see Extended Data fig. 7 in [28]), and potentially might be activated or blocked by other endogenous ligands (e.g. glycine in addition to l-glutamate). Figure 3.Glutamate metabolism and processing. (a) Table of pertinent genes used in glutamate metabolism and processing. Red numbers indicate expansions of those families in ctenophores. (b) The predicted metabolites, cofactors and enzymes of glutamate metabolism in ctenophores. The asterisk indicates the expansion of glutaminases in ctenophores. Crystal structure of GAD 67, PDB: 2okj is indicated as an insert [96]. (c) Genealogical relationships between the predicted glutamate decarboxylase (GAD) enzymes in several ctenophore species compared with prokaryotic and eukaryotic GAD enzymes. The ctenophore GAD enzymes share highest identity to each other forming a distinct branch (highlighted in grey). Maximum-likelihood trees were constructed based on the pyridoxal-dependent decarboxylase conserved domain (PF00282) amino acid sequences of the identified GADs.

See the electronic supplementary material for methods.

There are eight glutaminases—the enzymes that convert glutamine to glutamate. The glutamine aminohydrolase, or l-glutaminase (GLS), converts glutamine to glutamate (figure 3b). Glutamate is also involved in bioenergetics via the tricarboxylic acid cycle and the production of ATP, and ADP is a strong activator of the GLS enzymes [99,100]. Pleurobrachia bachei and M. leidyi have eight and nine GLS enzymes, respectively. In contrast, most metazoans have one GLS enzyme, and humans have two GLS enzymes (figure 3a,b). The apparent expansion of the glutamate synthetic enzymes in ctenophores might be related to the observed diversification of glutamate signalling and the high bioenergetic demands of the extensive development of ciliated structures and associated behaviours [70], with multiple examples of lineage-specific diversification events [28,40]. Thus, although ctenophores (including Pleurobrachia, Bolinopsis and Mnemiopsis) have muscle sensitivity to l-glutamate, their synthetic, transport, receptor and inactivation pathways are notably distinct from those identified in cnidarians and bilaterians.

GABA is the downstream product of glutamate metabolism and inactivation. Glutamate decarboxylase (GAD), a GABA synthetic enzyme (figure 3b,c), is widely distributed across both prokaryotes and eukaryotes [101]. GAD is related to the group II pyridoxal-dependent decarboxylase enzymes (PF00282), which include the glutamate, histidine, tyrosine and aromatic-l-amino acid decarboxylases. Previous studies have shown that these pyridoxal-dependent decarboxylases have evolved in parallel along multiple lineages [101,102]. The predicted ctenophore GAD enzymes have all of the critical amino acids essential for GAD enzyme function, as defined by the crystal structure [98]. The functional GAD activity in ctenophores is supported both by immunohistochemical visualization of GABA in muscles [66] and by direct microchemical detection of GABA in the extracts from different ctenophore organs [28].

Surprisingly, in ctenophores, GABA is localized only in muscles and not in neurons [28,66]. In addition, there were no observable pharmacological effects of GABA on the Pleurobrachia behaviours and major effector systems, including cilia, muscles and colloblasts [28]. These observations suggest that GABA functions as metabolic/bioenergetic intermediate, and as a possible mechanism to inactivate glutamate. A product of GABA metabolism can itself be a usable source of energy in ctenophore muscles. Indeed, GABA transaminase, which is also found in Pleurobrachia, is the enzyme that catalyses the conversion of GABA back into succinic semi-aldehyde and glutamate, following the formation of succinic acid, which enters the citric acid cycle—the universal aerobic bioenergetics pathway (figure 3a,b). By contrast, GABA is specifically localized in selected neuronal populations of both bilaterians and cnidarians, suggesting that it was subsequently co-opted in evolution as a neurotransmitter.

In summary, our data suggest that the observed diversification of glutamate processing is coupled with an enormous functional complexity of glutamate signalling in ctenophores. Interestingly, the expression levels of the majority of genes associated with glutamate processing were substantially increased on day three of Pleurobrachia development, when the first neurons appeared [28,91]. Thus, the neurogenesis and the fate specification of glutamatergic neural signalling might be mechanistically coregulated by the orchestrated expression of all of the molecular components that support l-glutamate synthesis (glutaminases), its inactivation (transporters/sialins), and its reception (iGluRs). Owing to a possible ‘deficiency’ of other classical transmitters, ctenophores might have ‘taken advantage’ of the versatile l-glutamate molecule and further developed an unprecedented complexity of glutamatergic systems to accommodate their complex behaviours. As a result, extant ctenophores possess one of the most unusual and molecularly diverse complements of glutamate signalling pathways of all animals studied to date. This is also an illustrative example of the extensive parallel evolution of the nervous system.

Historically, pore-forming proteins are broadly divided into two groups: connexins (only identified in chordates) and pannexins, which are present in both chordates (including tunicates and vertebrates) and invertebrates (figure 4). Before their discovery in humans [103], the invertebrate pannexins were initially named innexins—i.e. the invertebrate counterpart to gap junctions. Despite the lack of apparent sequence similarity with connexins, the innexin/pannexin (PANX/INX)-type gap junctions share the same overall membrane topology as connexins. Each pore is composed of four transmembrane regions in which six individual subunits form a channel or ‘innexon’ in the plasma membrane [104]. Two opposing innexons on interacting cells form a functional electric synapse. All gap junctions (connexins and pannexins) mediate fast electrical coupling between cells, which can be symmetrical or asymmetrical, allowing directional information processing. Importantly, pannexins form gap junctions with electrophysiological and pharmacological properties that are distinct from connexins [105]. Thus, these two dissimilar families of proteins represent a perfect example of the convergent evolution of electrical synapses. Figure 4.Convergent evolution of gap junction proteins. Connexins are a chordate-specific family of genes encoding gap junction proteins; they share no identity to the pannexin/innexin (PANX/INX) family. However, both families have the same membrane topology and perform the same function as electrical synapses and membrane pores. The table shows the complement of connexins and pannexin genes in different metazoans with sequenced genomes. The right part illustrates examples of the organization of representative genes. Pleurobrachia PANX/INX genes have the largest number of exons compared with all other metazoans.

We did not find any gap-junction orthologues in choanoflagellates or other eukaryotic groups, suggesting that pannexins/innexins are metazoan innovations [103,106,107]. In addition to metazoans, PANX/INX genes have been found in a few viruses, possibly as a result of lateral gene transfer between a host and its intracellular parasites [108]. For example, the endoparasitic wasp Campoletis sonorensis transmits a polydnavirus to its caterpillars during egg laying [108]. The sequenced ichnovirus (IV) genome has significant similarity to some parts of the wasp genome, including the presence of PANX/INX genes [108]. Predictably, these IV PANX/INX are closely related to the pannexins from arthropods (electronic supplementary material, figure S1).

The pannexin/innexin superfamily has a highly mosaic representation across metazoans, with several examples of gene gains and gene losses. There is no conservation in the exon–intron organization across pannexins. For example, only three PANX genes have been identified in mammals, all of which have four to five exons. The Drosophila genome encodes eight PANX/INX genes with three to eight exons [109]. Twenty-five PANX/INX genes are present in the C. elegans genome, with 3–11 exons [110]. In Aplysia californica, 18 PANX/INX genes have been identified, with 1–10 exons (figure 4). Non-bilaterian metazoans have the most unusual complements of PANX/INX genes. The A. queenslandica [30] and T. adhaerens [29] genomes and the publically available sponge transcriptome data [111] have no identified pannexins.

Cnidarians appear to have both losses and expansions of the PANX/INX genes. The hydrozoan Hydra magnipapillata has at least 19 pannexins [112,113]. By contrast, the anthozoan Nematostella vectensis has only one recognized PANX/INX gene (figure 4 and electronic supplementary material, figure S1), which may be involved in the electrical coupling between blastomeres in embryos [114]. However, no PANX/INX genes have been detected in the genomes of three other anthozoans, Aiptasia [115], Acropora digitifera [116] and Stylophora pistillata. There are also no identified PANX/INX genes in the scyphozoan Cyanea capillata (based on our transcriptome profiling). It is still unknown whether the cubozoans have PANX/INX genes. Gap junction proteins (both innexins and connexions) tend to be present in multiple copies in almost all metazoans, with the exception of Nematostella. Interestingly, the predicted Nematostella PANX/INX protein is clustered at the base of the chordate clade (electronic supplementary material, figure S1; this has also been observed by Abascal & Zardoya [106]). Given these observations, it has been suggested that there has been a horizontal transfer of a PANX/INX gene from an ancestral chordate to Nematostella [106].

In contrast to the cnidarians, all sequenced ctenophores have an enormous diversity of electrical synapses or gap junctions [28], forming a distinct branch in tree topology (electronic supplementary material, figure S1). However, no connexins are encoded in the genomes of Pleurobrachia and Mnemiopsis, or in any of the other ctenophore transcriptomes analysed. In general, the ctenophore PANX/INX genes contain more exons than their orthologues in Hydrozoa and bilaterians (the number of exons varies from 1 to 14; figure 4). Although both P. bachei and M. leidyi have 12 PANX/INX genes in their genomes, only four of them form genealogical sister pairs between species (electronic supplementary material, figure S1), further suggesting the widespread lineage-specific radiation and parallel evolution of this family.

Functional analysis of the electrical synapses in ctenophores is in its infancy [70,117]. Interestingly, the PANX/INX genes are one of the most highly expressed transcripts in the adult aboral organ of P. bachei (figure 5d; [28]), but they are also expressed in the combs and conductive tracts and in the neuron-like subepithelial cells. Gap junctions have previously been identified in the ciliated grooves by electron microscopy [95], which run from the aboral organ to the first comb plate of each comb row and through the endoderm of the meridional canals [117]. These data suggest that a significant fraction of the synaptic transmission in the conductive tracts and the aboral organ is electrical. In addition to conducting electrical synapses, gap junctions also mediate mechanoreception [118] and the direct exchange of small molecules among neighbouring cells by forming channel aggregates in the plasma membrane without any synapses to release ATP or other molecules [119–123]. This possibility might explain the functional significance of the expression of selected types of PANX/INX during early Pleurobrachia development (figure 4a; [28]). Figure 5.Key molecular innovations underlying evolution of neural organization in ctenophores and the cnidarian/bilaterian clade (modified from Moroz et al. [28]). Bars indicate the presence or relative expansions of selected gene families in all basal metazoan lineages from the inferred urmetazoan ancestor. The data suggest that sponges and placozoans never developed neural systems, or, highly unlikely assuming the presence of neuronal organization in the urmetazoan ancestor, sponges and placozoans lost their nervous systems. Either hypothesis points towards extensive parallel evolution of neural systems in ctenophores versus the Bilateria + Cnidaria clade. See text for details.

In summary, gap junction proteins are wonderful illustrations of the convergent evolution and independent lineage-specific diversification of synaptic organization. The chordate-specific connexins share no sequence identity with the pannexin-type gap junctions but have a similar membrane topology and perform the similar functions. Interestingly, some chordate pannexins have lost their gap junction functions [124]; therefore, the expansion of the connexin family in this lineage, with more than two dozen members, provides a complementary solution to support the diversity of electrical synapses in human brains, muscles and secretory tissues. The pannexin-type gap junctions may have originally evolved in the common metazoan ancestor to mediate the release of ATP or other metabolites to support intercellular communications during early development and in multicellular coordination. This might provide a foundation for the independent recruitment of pannexins and connexins into neuronal functions over the course of ctenophore, cnidarian and chordate evolution. The cladistic analyses suggest that Amphimedon, Trichoplax and some cnidarians lost their recognized gap junction proteins, but the significance of this type of gene loss is unclear. It is also possible that other classes of proteins can form functional gap junctions in these metazoan lineages indicating that the early evolution of pore-forming proteins is more complex than was previously anticipated.

Complementary studies [14,28,40,66] have revealed a number of unique features in the molecular organization of ctenophore signalling (figure 5), which is consistent with the hypothesis of the convergent evolution of neural and integrative systems. The most remarkable trait of Ctenophora is the apparent absence of the majority of the conventional low-molecular-weight transmitter systems. This is paralleled by the considerable development and diversification of the synthetic, uptake and receptor components of glutamate signalling, with extensive lineage-specific adaptations within the clade. Given the current placement of ctenophores as the sister group to other animals [60], the most plausible scenario suggests that acetylcholine, serotonin, histamine, dopamine, octopamine and GABA were recruited as transmitters in the common cnidarian/bilaterian ancestor. We believe that GABA originally evolved as a passive product of the inactivation of glutamate and as a bioenergetically important intermediate, as in extant ctenophores. Later in the cnidarian and bilaterian lineages, GABA was recruited into neuronal signalling as a synaptic messenger molecule. A similar scenario for the independent recruitment of NO in neuronal signalling can also be observed in various lineages of cnidarians and bilaterians [125–127]. It may well be that all other classical low-molecular-weight transmitters originally evolved as metabolic intermediates and non-neuronal, in part injury related [128] signalling molecules in early animals. This process was followed by multiple co-option events in different secretory cells, including the various (proto)neuronal cell lineages. The roles of the different transmitters in early embryonic development [129–137] might partially reflect these ancestral functions of the vast diversity of signalling molecules, from glutamate to small secretory peptides.

The origin and subsequent evolution of the neural systems in the ctenophore lineage occurred independently from those in all other animals (figures 1 and 5). Such processes frequently used a distinct array of ‘available’ intercellular messengers, including secretory peptides, which not only are unique for these groups, but also are undergoing very rapid evolution in ctenophores. ATP and related purines, glycine, nitric oxide, carbon monoxide and hydrogen sulfide might also be used as intercellular messengers [28,40], but with multiple ctenophore-specific innovations (figure 5). The predicted pre- and postsynaptic gene complements in Pleurobrachia and Mnemiopsis have reduced numbers of components compared with those of cnidarians and bilaterians (table 1). As a corollary of the hypothesis of the independent origins of neurons, our analyses suggest that both electrical and chemical synapses evolved more than once (figure 4).