Transcriptome profiling of Trichoplax adhaerens highlights its digestive epithelium and a rich set of genes for fast electrogenic and slow neuromodulatory cellular signaling


 Background Trichoplax adhaerens is a fascinating early-diverging animal that lacks a nervous system and synapses, and yet is capable of directed motile feeding behavior culminating in the external digestion of microorganisms by secreted hydrolytic enzymes. The mechanisms by which Trichoplax cells communicate with each other to coordinate their activity and behavior is unclear, though recent studies have suggested that secreted regulatory peptides might be involved.Results Here, we generated a high quality mRNA transcriptome of Trichoplax adhaerens , and predicted secreted proteins to identify gene homologues for digestion, development, immunity, cell adhesion, and peptide signaling. Detailed annotation of the expressed Trichoplax gene set also identified a nearly complete set of electrogenic genes involved in fast neural signalling, plus a set of 665 G-protein coupled receptors that in the nervous system integrate with fast signalling machinery to modulate cellular excitability. Furthermore, Trichoplax expresses an array of genes involved in intracellular signaling, including the key effector enzymes protein kinases A and C that functionally link fast and slow cellular signaling. Also identified were nearly complete sets of pre- and post-synaptic scaffolding genes, most encoding appropriate protein domain architectures. Notably, the Trichoplax proteome was found to bear slightly reduced counts of synaptic protein interaction domains such as PDZ, SH3 and C2 compared to other animals, but abundance of these domains did not appear to predict the presence of synapses in early-diverging groups.Conclusions Despite its apparent cellular and morphological simplicity, Trichoplax expresses a rich set of genes involved in complex animal traits. The transcriptome presented here adds a valuable additional resource for molecular studies on Trichoplax genes, exemplified by our ability to clone cDNAs for nine full-length acid sensing ion channel proteins with almost perfect matches with their corresponding transcriptome sequences.


Abstract
Background Trichoplax adhaerens is a fascinating early-diverging animal that lacks a nervous system and synapses, and yet is capable of directed motile feeding behavior culminating in the external digestion of microorganisms by secreted hydrolytic enzymes. The mechanisms by which Trichoplax cells communicate with each other to coordinate their activity and behavior is unclear, though recent studies have suggested that secreted regulatory peptides might be involved.
Results Here, we generated a high quality mRNA transcriptome of Trichoplax adhaerens , and predicted secreted proteins to identify gene homologues for digestion, development, immunity, cell adhesion, and peptide signaling. Detailed annotation of the expressed Trichoplax gene set also identified a nearly complete set of electrogenic genes involved in fast neural signalling, plus a set of 665 G-protein coupled receptors that in the nervous system integrate with fast signalling machinery to modulate cellular excitability. Furthermore, Trichoplax expresses an array of genes involved in intracellular signaling, including the key effector enzymes protein kinases A and C that functionally link fast and slow cellular signaling. Also identified were nearly complete sets of pre-and postsynaptic scaffolding genes, most encoding appropriate protein domain architectures. Notably, the Trichoplax proteome was found to bear slightly reduced counts of synaptic protein interaction domains such as PDZ, SH3 and C2 compared to other animals, but abundance of these domains did not appear to predict the presence of synapses in early-diverging groups.
Conclusions Despite its apparent cellular and morphological simplicity, Trichoplax expresses a rich set of genes involved in complex animal traits. The transcriptome presented here adds a valuable additional resource for molecular studies on Trichoplax genes, exemplified by our ability to clone cDNAs for nine full-length acid sensing ion channel proteins with almost perfect matches with their corresponding transcriptome sequences.

Background
Electrical signaling in the nervous system is achieved by an array of well characterized electrogenic genes, including those that establish ion and charge gradients across the cell membrane (i.e. ion pumps and exchangers), and those that exploit these gradients to produce transient, fast-propagating electrical signals (i.e. ion channels and ionotropic receptors) [1]. It is also evident that fast information flow in the nervous system is dynamically regulated by slower modalities of cellular communication, referred to as neuromodulation, which can alter neural circuit properties to produce different outputs for changes in behavior [2]. A major source for neuromodulation are ligand-activated G-protein coupled receptors (GPCRs), which activate intracellular G proteins and signaling pathways to alter the functional properties of ion channels and other electrogenic genes through effector kinases such as protein kinases A and C (PKA and PKC respectively) [3]. Studies of the crustacean stomatogastric ganglion have revealed that even simple neural circuits are heavily neuromodulated [4]. For instance, a single neuromodulatory ligand can cause distinct changes in the electrical properties of neurons within a neural circuit, by differentially altering the functional properties of specific ion channel types in each cell. Neuromodulators can also differentially alter synaptic strength, and hence, the effective wiring of neural circuits. Furthermore, combinations of different neuromodulatory ligands can work together to impose complex effects on neural circuits. Given the extensive library of known neuromodulators and GPCRs for both vertebrates and invertebrates, the potential for complexity is staggering. It is therefore striking that the molecular machinery for fast ionotropic and slow neuromodulatory cell signaling produces a finite set of coherent neural output patterns associated with specific behaviors and transitions in behavior [4]. Indeed, understanding how the nervous system evolved must include a consideration of how fast and slow forms of cellular communication became functionally intertwined [5].
Trichoplax adhaerens (Trichoplax sp. H1; phylum Placozoa) is a simple, early-diverging marine invertebrate that remarkably, lacks a nervous system and yet is able to coordinate its different cell types to conduct motile behavior including feeding [6], chemotaxis and geotaxis [7]. Locomotion by this small, flat, disc-shaped animal is achieved through beating cilia located on its ventral epithelium, which interestingly, undergo coordinated cessation of beating upon detection of food [6], and in response to applied regulatory peptides encoded within its genome [8][9][10][11]. Furthermore, observation of spontaneous co-ordinated pauses among adjacent Trichoplax animals not in contact with each other points to regulated secretion of signaling molecules that elicit pausing, and that these signals can travel from animal to animal [6]. Based on ultrastructure [12] and single cell transcriptome [13] studies, Trichoplax only has 6-11 cell types, which lack direct connections with each other in the form of electrical or chemical synapses. Hence, the extent and manner by which Trichoplax cells communicate with each other is unknown. Interestingly, protein-coding sequences predicted from the genome of Trichoplax adhaerens identified a large set of "neural" gene homologues, including ion channels and genes involved in synaptic transmission [10]. Indeed, given Trichoplax's cellular simplicity and strategic phylogenetic position near the base of Metazoa, it might prove useful for understanding how networks of genes underlying fast and slow neural communication co-evolved in the nervous system. A first step is therefore determining the presence of relevant genes in the Trichoplax expressed gene set.
Previously, an mRNA transcriptome for Trichoplax adhaerens was generated in an effort to better annotate its genome, and to compare its genome to the related placozoan Hoilungia hongkongensis [14]. Here, we build on this previous work by generating a deep, high quality mRNA transcriptome for Trichoplax adhaerens, sequenced in quadruplicate from whole animal total RNA. Through a combined de novo and ab initio assembly strategy, we produced a high quality converged dataset with the vast majority of transcripts bearing complete protein-coding sequences. Based on gene ontology annotation and inferred average mRNA expression levels, we report an enrichment of Trichoplax secreted proteins at the whole animal level. Furthermore, prediction and annotation of the Trichoplax secretome identified several novel neuropeptide precursors, and homologues for proteins associated with digestion, development, immunity/defense, and cell adhesion. Also identified in the transcriptome was a near-complete set of genes involved in fast electrical neural signaling, and over 665 GPCRs whose intracellular signaling pathways can alter fast signaling machinery to impose lasting changes in cellular excitability. Trichoplax also expresses an array of genes that link fast and slow signaling machinery, including adenylyl cyclase, phospholipase C, PKA and PKC. Furthermore, the transcriptome was found to contain a complete set of genes involved in intracellular Ca 2+ signaling, including voltage-gated calcium channels, Ca 2+ pumps and exchangers, intracellular calcium channels IP 3 and ryanodine receptors, calmodulin kinase II, calcineurin, and very high level expression of the Ca 2+ sensor protein calmodulin.
Also notable, the vast majority of genes involved in synapse function were identified as expressed, with most bearing appropriate domain architectures required for their synaptic interactions and functions. Comparing the content of synaptic protein interaction domains present within the proteomes of various animals and eukaryotes revealed that Trichoplax has a slightly reduced number of domains compared to other animals, including PDZ, SH3, and C2. However, counts of these domains do not predict the presence of synapses, since sponges and choanoflagellates, which lack synapses, contain more than ctenophores, which have synapses. The transcriptome dataset presented here represents a valuable addition to the toolsets available for conducting molecular and phylogenetic studies on Trichoplax, exemplified by our cloning of nine cDNAs for acid-sensing ion channels using primers designed against their respective transcriptome sequences.  Figure 1A).

Results
A histogram of the assembled transcripts, arranged according to sequence length in base pairs, revealed a somewhat normal distribution, with only minimal enrichment of shorter fragmented sequences ( Figure 1B). The high quality of the assembly is also evident in a histogram of unique protein sequences binned by length in amino acids ( Figure 1C), where the vast majority of transcripts contain compete open reading frames with both start and stop codons, with very few bearing truncated protein-coding sequences at the 5' end (N-terminus), 3' end (C-terminus), or both ( Figure   1C). To further assess the quality of the transcriptome, the 11,981 genes were aligned against protein sequences in the Swiss-Prot, RefSeq and TremBL databases using BLASTp ( Figure 2A). As expected, the highly curated protein database Swiss-Prot produced few alignment hits with high bit-scores, reflecting a general absence of experimentally validated Trichoplax gene in the scientific literature.
Accordingly, the BLAST results were largely unchanged when the transcriptome was re-aligned against a filtered Swiss-Prot database lacking Trichoplax genes. In contrast, alignment with RefSeq and TrEMBL produced generally higher bit-scores, and removal of Trichoplax genes from these databases considerably reduced the overall bit-score quality. A plot of bit-scores of each aligned transcriptome gene, sorted by descending bit-score ( Figure 2B) shows that bit-scores arising from alignment with RefSeq and TrEMBL undergo marked shifts towards lower values when Trichoplax genes were filtered out, while the Swiss-Prot results remained unchanged. We note that a considerable number of Trichoplax transcriptome genes either failed to find BLAST homology in these databases or produced poor alignment scores (i.e. bit-scores below 50), likely representing uncharacterized Placozoa-specific genes. Altogether, these analyses indicate that our transcriptome is of very high quality, contributing a valuable set of complete open reading frames for both molecular and phylogenetic/bioinformatics studies.
The transcriptome expands the known complement of expressed Trichoplax genes To directly compare our transcriptome gene set to genes from the previously published genome [10], coding sequences of our 11,981 genes were matched against their single best counterparts in the genome-predicted gene-set via BLASTn (filtering criteria: best hit per query where alignment length ≥100 bp and sequence identity ≥95%). A total of 9,498 transcriptome genes aligned with a converged set of 9,260 genome-predicted genes, indicative of either fragmented assembly of some transcriptome sequences (i.e. two broken gene fragments mapping to a single genome-predicted gene), and/or of the presence of chimerically fused gene sequences in the genome-predicted set (i.e. two separate transcriptome genes mapping to a chimeric genome-predicted gene; Figure 3A). Furthermore, 2,483 transcriptome genes did not find a match in the genome-predicted set, while 2,260 genome-predicted genes were left unmatched. However, reciprocal BLASTn using genomepredicted genes as query against transcriptome sequences, using the same filtering criteria as above, resulted in only 1,418 genome-predicted genes not finding a match, while 10,102 found matches converging onto 9,250 transcriptome genes, leaving 2,731 transcriptome genes unmatched. Again, this convergence can reflect either fragmented assembly of some of the genome predicted genes, or chimeric assembly of transcriptome genes. Nevertheless, we note that for all de novo assembly programs used in our assembly pipeline, outside of Trinity, longer k-mer lengths were selected according to thresholds identified in a previous study that characterized the effect of k-mer length on chimeric transcript assembly [31].
We next sought to evaluate how many of the 11,981 transcriptome gene sequences mapped to the published Trichoplax genome scaffolds [10], using the program GMAP [21] (v2016-05-01) with a stringent threshold sequence identity of ≥ 90 base pairs, and a trimmed coverage of ≥ 90%. Of the 2,483 transcriptome genes that did not find BLASTn homology in the genome-predicted gene set, 1,768 found a match in the genome scaffolds, representing novel genes likely present in the genome scaffolds but that previously failed to be predicted ( Figure 3B). Instead, 715 transcriptome genes did not map using our cut-offs, representing completely novel genes either located within un-sequenced gaps in the genome scaffolds, or alternatively, genes present in the genome but not meeting our cutoff. Indeed, of the 9,498 transcriptome genes that previously aligned to the genome predicted-gene set via BLASTn, 329 failed to meet our GMAP cut-offs. As a crude confirmation that the 715 novel genes are indeed expressed transcripts, we note that their average TPM expression level 166 ±88 standard error (SE), compared to 48 ±26 SE for the 1,768 novel gens that mapped to the genome. Importantly, separate mRNA transcriptomes for Trichoplax sp. H1 (i.e. Trichoplax adhaerens) and sp.
H2 were recently released [14], in an effort to better annotate the sequenced genome. The work presented here complements and improves upon this previous study by having longer paired-end read lengths (75 vs. 125 base pairs), total read counts (150.7 vs. 212.1 million paired-end reads) and a comprehensive combined de novo and ab initio assembly strategy as detailed above. Nevertheless, we used BLASTp to determine whether the newly identified genes from our study were also identified in this previous study. Of the 715 genes that did not map to the genome, 450 (~63%) found a match in the published Trichoplax sp. H1 transcriptome with an alignment length cut-off of 100 amino acids and a percent identity cut-off of 60% (Table S1). Increasing percent identity to ≥ 95% reduced this number to 335 (47%), and to 153 (21%) at 100% identity. Instead, of the 1,768 novel genes that did map to the genome, 1,312 (74%) found a match with ≥ 60% identity, 1,123 (64%) with ≥ 95% identity, and 460 (26%) with 100% identity. Altogether, the two transcriptomes corroborate the existence and expression of these novel genes, and our dataset likely adds several hundred additional genes to the known Trichoplax gene set. Using TBLASTN (filtering criteria: best hit per query where alignment length ≥100 bp), we also compared the recently published transcriptome of Trichoplax sp.

H2
[14] to our Trichoplax sp. H1 transcriptome. At a percent identity cut-off of 60%, 281 (39%) of the unmapped genes were also identified in the Trichoplax sp. H2 transcriptome, while these values decreased to 205 (29%) and 69 (10%) genes, at ≥ 95% and ≥ 100% identity cut-offs, respectively (Table S1). Of the novel genes that had a match in the genome scaffolds but not the genomepredicted gene set, 900 (51%), 762 (43%), and 218 (12%) genes were found in the Trichoplax sp. H2 transcriptome at ≥ 60%, ≥ 95%, and ≥ 100% identity cut-offs, respectively (Table S1). We similarly compared the novel Trichoplax genes to the 12,575 genes from the related placozoan Hoilungia hongkongensis [32] using BLASTp. Here, 178 (~25%) of the 715 un-mapped genes found a match with ≥ 60% identity, and only 3 with ≥ 95% (Table S1). Instead, 452 (26%) of the novel 1,768 mapped genes found a match with ≥ 60% identity, and only 2 with ≥ 95% (Table S1). Lastly, we sought to determine whether some of the 2,260 genome-predicted genes not present in our transcriptome were previously discarded during processing and filtering. The 2,260 genome-predicted genes were collapsed down to a 1,570 non-redundant gene set using CD-Hit (nucleotide sequence identity ≥ 90%) and aligned against a concatenated database of all 27 transcriptomes used for the EvidentialGene pipeline using BLASTp ( Figure 1A). Imposing alignment length and percent identity cut-offs of ≥ 33 amino acids and ≥ 95%, respectively, we recovered 615 previously discarded transcriptome sequences, leaving only 952 genome-predicted genes unmatched in the transcriptome. In total, our transcriptome expands the repertoire of genome-predicted genes that are expressed at the mRNA level to 13,548 ( Figure 3C). The resulting transcriptome was deposited to GenBank under accession SUB5274527.
Gene ontology highlights enriched expression of genes involved in ER trafficking and secreted/extracellular functions BLAST2GO [33]  processes. Furthermore, we conducted GO enrichment analysis to compare terms between the top 1,000 most highly expressed genes and the whole transcriptome (using Fisher's Exact Test in BLAST2GO; two-tailed analysis, false discovery rate = 0.05). Visualization of the top 90 enriched GO terms for all three gene ontology categories, selected according to lowest p-value, revealed enriched mRNA expression of genes with GO terms associated with protein localization to the endoplasmic reticulum, and hence the endomembrane/secretory pathway ( Figure S1). Consistent with this is the marked overrepresentation of GO terms for high TPM genes that are associated with the endomembrane, cell membrane and extracellular compartments of the cell, including exosomes, which are secreted vesicles with roles in cellular communication [36] (Figures 4B and S1). Notably, although small in number, it is interesting to find GO terms in the Cellular Component category associated with synapses and cell junctions ( Figure 4B), and mRNA enrichment of genes involved in vesicle function ( Figure S1), despite Trichoplax lacking synapses. Altogether, these data reveal that Trichoplax expresses a diverse set of genes associated with complex metazoan traits, including those involved in intra-and inter-cellular communication.
The Trichoplax secretome identifies genes involved in digestion, cell signaling, development and immune-related functions  (Table 1), immunity and defense (Table 2), development/reproduction (Table 3) and cell matrix/cell adhesion (Table 4). Notably, mRNAs for digestive enzymes are highly expressed at the whole-animal level, consistent with Trichoplax's external digestion of food algae [6], presumably through secretion of enzymes from lipophil cells located on its ventral epthelium [12,13]. Included among these are homologues for peptidases such as trypsin, kallikrein, cathepsin, lysozyme and chymotrypsinogen A, lipases such as phospholipase A1 and A2, and enzymes that hydrolyze sugars or amino acids such as alpha-amylase, chitobase and phospholipase B (Table 1). Hydrolytic enzymes are also important for immunity and defense.
Trichoplax exhibits high expression levels of the protective enzymes granzyme K and lysozyme, and others involved in immune signaling such as cathepsin ( Table 2). A putative orthologue for complement factor C3 was also identified (Table 2), however, domain prediction analysis via InterPro demonstrated that this protein more closely resembles the alpha-2-macroglobulin domain-containing CD109 antigen, previously identified in Trichoplax sp. H2 [42]. Thus, our work supports the previous report that Trichoplax lack a bona fide complement system and yet harbor a rich set of immunityrelated genes that likely figure in determining self from non-self [41]. Also highly expressed is the cytokine granulin, and other immunity-related genes involved in cellular stress responses and apoptotic signaling. In contrast, secretome genes involved in development/reproduction generally exhibited lower average TPM expression levels (Table 3). It is however notable that Trichoplax homologues for ovochymase-2 and vitellogenin-2, two genes involved in sexual reproduction [43,44], exhibit considerably high TPM values, given the uncertainty about whether placozoans are capable of sexual reproduction [10,[45][46][47]. Also identified in the secretome were numerous genes with putative extracellular matrix and cell adhesion functions, including stereocilin, involved in the organization of mechanosensitive stereocilia in vertebrate sensory hair cells [48], and adventurousgliding motility protein Z, involved in cellular gliding along external surfaces driven by intracellular molecular transport motors [49] ( Table 4).
Expected within the secretome are pre-pro-proteins for regulatory peptides and pro-hormones that are secreted after enzymatic cleavage into mature peptides in the ER. Previously, several such peptide precursors were identified from Trichoplax genome-predicted genes bearing repeated convertase cleavage sites predicted with the program NeuroPred [8,50]. Furthermore, several studies have shown that Trichoplax motile behavior can be dynamically modulated by exogenously applied peptides [9,11], indicating that Trichoplax likely uses secreted peptides for cellular communication and coordination. Given the completeness of protein-coding ORFs in our transcriptome, we reasoned that a similar analysis might identify additional regulatory peptides precursor genes. Indeed, we able to identify all 9 of the previously predicted short regulatory peptide precursor sequences, plus 3 novel ones bearing canonical convertase cleavage sites (WWamide, ITKL, and YPFFGN), and 4 bearing noncanonical cleavage sites (PERI, FALF, ERSA peptides, and DAYQamide; Supplementary File 2; Figure   5B). We also confirmed expression of Trichoplax insulin 1 and 2 pro-hormones, the recently discovered MFPF peptide, and granulin, the latter showing very high mRNA expression with an average TPM value of 965 ±118 SE ( Figure 5C). One of these novel peptides, YPFFGN (i.e. QDYPFFGN/S), has a striking resemblance to the vertebrate peptide endomorphin (YPFFamide), however, a C-terminal asparagine/serine residue makes it uncertain whether the glycine residue is amidated after cleavage. Nevertheless, ectopic application of an amidated version of this peptide, as well as vertebrate endomorphin, causes Trichoplax to pause ciliary locomotion as occurs during feeding [9]. Coincidentally, the YPFFGN gene is one of the 715 genes present in the transcriptome but completely absent from the genome ( Figure 3B), but its existence has been confirmed by cloning and sequencing [9] (NCBI accession number KY675296).
Our analysis also identified several proteins bearing repeating stretches of arginine residues, which coincidentally, failed to find BLASTp homology in the Swiss-Prot and RefSeq protein databases and thus likely represent genes that are unique to Trichoplax (Figure 2A, B). This prompted us to look more carefully within the secretome, where we manually identified several additional arginineenriched sequences (Supplementary File 2). Given the apparent non-canonical nature of these arginine stretches/repeats, it is uncertain whether they are cleaved by convertase, or instead, these novel proteins remain complete after removal of the signal peptide. However, it is notable that some exhibit extremely high mRNA expression levels, with TPM values of 7,252 (evg1172538), 5,180 (evg1228357) and 3,837 (evg452961) for the three most highly expressed transcripts. In contrast, we note relatively lower expression levels for all regulatory peptides in general, with FALF and FFNPamide precursors showing moderate average TPM values of 387 ±15 SE and 325 ±10 SE ( Figure   5C).

Trichoplax expresses a nearly complete set of genes involved in fast neural signaling
In vitro electrophysiological characterization of acid-sensing ion channel (ASIC) homologues cloned from gastropod molluscs revealed that select amidated regulatory peptides such as FMRFamide could activate them to produce depolarizing cation currents in cells [51][52][53]. Similarly, ASIC homologues cloned from the cnidarian Hydra magnipapillata activate in response to amidated regulatory peptides, where they might play a role in synaptic transmission across the neuromuscular junction [54]. Indeed, genomic expansion of ASIC channels and regulatory peptide genes in ctenophores and cnidarians has led to hypotheses that early-diverging animals employ secreted peptides and ASIC channel homologues for synaptic transmission [54,55]. In the Trichoplax transcriptome, we were able to identify 10 full length ASIC channel coding sequences, also predicted from the genome, which we named TadASIC1 to TadASIC10. An additional ASIC sequence, TadASIC11, was predicted from the genome but was absent in our transcriptome. A maximum likelihood phylogenetic tree of the Trichoplax ASIC channels, inferred from a MUSCLE [56] protein alignment with homologues from other animals, revealed that most of the TadASICs are tightly clustered with poor bootstrap support linking them to other clades, while TadASIC10 forms a relatively strong clade with vertebrate epithelial sodium channels (ENaCs), mechanosensitive channels Mec-4 and Mec-10 from C. elegans, the uncharacterized Hydra ASIC subunit HyNaC12, and the FMRFamide peptide-gated channels from gastropod molluscs Lymnaea stagnalis, Helix aspersa and Helisoma trivolvis [51-53] ( Figure 5D).
Whole animal mRNA TPM expression levels of the TadASICs revealed relatively low expression for TadASIC2 to TadASIC8, very low expression for TadASIC1, and moderate expression for TadASIC9 and TadASIC10. As part of an ongoing study to functionally characterize the Trichoplax ASIC channels in vitro, we sought to clone their cDNAs into the mammalian expression vector pIRES2-EGFP from wholeanimal total RNA (via RT-PCR using primers denoted in Table S2). TadASIC2 to TadASIC10 were successfully cloned in triplicate, producing consensus mRNA coding sequences for these genes  Using a database of different ion channel types as query, we identified numerous additional ion channel candidates in the Trichoplax transcriptome, which were subsequently carefully annotated using BLASTp protein alignment against the Swiss-Prot and NCBI non-redundant databases, combined with SMART-BLAST and InterPro [57] analyses, and in some cases, maximum-likelihood phylogenetic inference. Indeed, given the absence of bona fide neurons in Trichoplax, it is interesting that along with ASIC channels they express mRNAs for the vast majority of ion channels that are crucial for fast electrical signaling in the nervous system. For example, Trichoplax expresses a full complement of Pore-loop channels including fast voltage-gated sodium, calcium and potassium channels, a large conductance (BK) calcium-activated potassium channel, numerous AKDF (AMPA/Kainate-like) and epsilon subfamily ionotropic glutamate receptors, cation leak channels such as the sodium leak channel NALCN and a two-pore potassium channel, a cyclic nucleotide-gated channel, numerous inward rectifying potassium channels and numerous transient receptor potential (TRP) channels (Table S3).
Consistent with previous reports [55, 58], we failed to identify ion channels of the Cys-loop family which tend to play important roles in synaptic transmission, including ionotropic Glycine, GABA, acetylcholine, histidine and serotonin receptors, as well as innexins and connexins, which form gap junctions or hemi-channels [59]. Other identified channels included two ATP-gated P2X channels, one ryanodine receptor and three IP 3 receptors which operate in the ER, the mechanosensitive channel piezo, an orai calcium release-activated calcium channel and several CLC proton/chloride transporters that can act as voltage-gated chloride channels. Generally, the expression of these ionotropic genes at the whole animal mRNA level is low (Table S3). However, their presence in the genome is indicative of considerable potential for Trichoplax cells to undertake complex forms of electrical and calcium signaling. Unfortunately, to our knowledge, Placozoa is the only early-diverging phylum for which cellular electrical activity has not been successfully recorded using electrophysiology, and the role of fast electrical signaling in Trichoplax biology is completely unknown [60].
Trichoplax bears over 665 G protein-coupled receptor genes involved in slow neuromodulatory signalling In the nervous system, the machinery underlying fast electrical signaling is dynamically regulated by neuromodulatory input, which operates along much slower and longer timescales and hence serves to alter neural circuit function to exact changes in behavior and physiology [4]. The most prominent form of neuromodulation occurs through ligand-activated G protein coupled receptors (GPCRs). These membrane receptors respond to various secreted neurotransmitters and neuropeptides by activating intracellular G proteins (i.e. Gα and Gβγ), which subsequently activate a series of intracellular signaling molecules including kinases that can phosphorylate ion channel proteins to alter their function (e.g. such as protein kinase C and protein kinase A) [3]. In some cases, activated Gβγ subunits physically bind ion channels to alter their activity, as occurs during feedback auto-inhibition of pre-synaptic Ca v 2 calcium channels. In this context, ligand activation of peri-synaptic GPCRs by secreted neurotransmitters activates Gβγ proteins, which in turn bind to and inhibit pre-synaptic Ca v 2 channels to dampening their activity and hence neurotransmitter exocytosis [61]. Indeed, how these two forms of cell-cell signaling, fast ionotropic and slow neuromodulatory, became integrated during nervous system evolution to give rise to coherent neural output patterns underlying transitions in behavior is not known. Trichoplax, in its early-divergence and cellular simplicity, perhaps poses unique opportunities for exploring this question.
Here, we sought to identify Trichoplax GPCRs from the complete gene set using combined prediction with GPCRHMM [62] and InterProScan [63]  GPCRome, validated not only with InterPro but also with a hidden Markov model whose sensitivity for detecting bona fide GPCRs is approximately 15% higher than that of transmembrane predictors [62].
Of the 655 Trichoplax GPCRs for which TPM mRNA expression data was available (i.e. not from the genome-predicted gene set), TPM values ranged from 0.07 to 125.9 with a mean and standard deviation of 4.6 and 9.4, respectively.

Trichoplax expresses genes for glutamatergic and GABAergic biosynthesis and degradation
Interestingly, the soft-bodied sponge Tethya wilhelma contracts upon application of the neurotransmitters glutamate and gamma-aminobutyric acid (GABA), presumably through activation of metabotropic GPCRs [66]. Given the basal position of sponges within the Metazoa, it thus seems likely that these two ligand systems were adopted very early on during animal evolution for cellular communication, if not before animals emerged. A quick survey of Trichoplax GPCRs identified numerous putative glutamate and GABA metabotropic GPCRs in the transcriptome, as well as other transmitter-activated GPCRs such norepinephrine and opioid receptors (data not shown). We thus sought to update the annotation of Trichoplax genes for neurotransmitter biosynthesis and degradation, using BLASTp alignment against Swiss-Prot and NCBI non-redundant databases, combined with SMART-BLAST and InterPro analysis, and when required, phylogenetic analysis [10,58]. Not surprisingly, pathway components for glutamate and GABA appear to be complete in Trichoplax (Table S4). However, most other transmitter pathways were found to be incomplete, including those for the synthesis and/or degradation of catecholamines dopamine, epinephrine and norepinephrine, as well as those for serotonin, octopamine and histamine (Table S4). Interestingly, we did find two Trichoplax homologues for choline acetyltransferase, previously reported as absent [58], but not the acetylcholine degradation enzyme acetylcholinesterase. Finally, we identified 2 Trichoplax enzyme homologues for synthesis of the transmembrane diffusible ligand, nitric oxide, suggesting that one of the three inducible nitric oxide synthase (iNOS)-like genes previously reported in Trichoplax adhaerens [67] is not expressed, or, accordant with its namesake, is expressed only under select conditions yet to be identified.
Trichoplax points to an early establishment of the synaptic proteome and synaptic signaling machinery With our expanded Trichoplax gene set, we carried out an extensive annotation of synaptic gene homologues, using the same thorough annotation strategy as described for ion channels and neurotransmitter biosynthesis pathways above. Trichoplax was found to express all members of the exocytotic SNARE complex (i.e. synaptobrevin, SNAP25, and syntaxin-1), and Ca 2+ -sensitive elements of the exocytotic machinery synaptotagmin and complexin ( Figure 7A, B; Table S5). In addition to the pre-synaptic Ca v 2 calcium channel noted previously, Trichoplax also expresses an array of presynaptic scaffolding proteins that interact with Ca v 2 calcium channels to regulate their pre-synaptic localization and function, including RIM, RIM-BP, Mint, CASK and CAST/ELK [68][69][70][71][72] (Figure 7A, B; Table S6). Interestingly, Trichoplax possess two RIM (Rab3 Interacting Molecule) homologues, both bearing canonical Zn 2+ -finger motifs that mediate interactions with Munc-13, and two C2 domains for interactions with α-liprin and ELKs [70]. However, one homologue lacks a PDZ domain that in fruit fly and mouse is thought to bind the distal C-terminus of Ca v 2 channels promoting their localization at the synapse active zone [73,74]. RIM is particularly interesting being recently marked as one of only 25 genes that are both unique to animals and have resisted genetic loss across all major metazoan clades [75]. Some notable genes that we were unable to identify in the transcriptome were the related active zone scaffolding proteins Bassoon and Piccolo, and the invertebrate ELK-related protein Bruchpilot [76,77].
Overall, it is striking that the vast majority of Trichoplax homologues for the pre-synaptic assembly bear appropriate domain architectures required for specific interactions and functions at the synapse.
We note that the expression of these genes is quite variable at the whole-animal level ( Figure 7B; Tables S5, S6), and a conclusive determination about whether these genes are functionally coexpressed in Trichoplax cells will require deep single cell transcriptome profiling. Recently, a single cell transcriptome study on Trichoplax, identifying about 100 co-expressed genes per cell type, revealed that synaptobrevin and synaptotagmin are co-expressed in cells also expressing putative regulatory peptides [13]. Perhaps with deeper sequencing, it will become clearer whether cellular excitation and Ca 2+ -regulated exocytosis is taking place in select Trichoplax cell types. Of course, this would require co-expression with an array of ion channels that regulate membrane excitability and exocytosis. Interestingly, sponges appear to have lost core genes required for neural excitability including voltage-gated Na v and K v channels [60,[78][79][80]. In the Great Barrier Reef sponge Amphimedon queenslandica, which lacks synapses, the absence of temporal co-expression of synaptic genes suggests that synapse and neural evolution involved emergent co-expression of pre-  (Table S6).
Although not exclusive to the synapse, several types of protein-protein interaction domains are essential for the assembly of both pre-and post-synaptic protein complexes, such as PDZ, Src Homology 3 (SH3), C2, FYVE zinc-finger (Zn-ginger), guanylate kinase (GK), Rho-GTPase-activating protein (Rho-GAP), Unc13 homology and CaMKII domains (e.g. Figure 7A). Previously, comparison of the PDZ domain content and diversity in proteomes of several metazoans and related single-celled eukaryotes revealed both expansion and specialization of PDZ domains in animals associated with increased molecular complexity at the synapse [91]. To extend that analysis, we compared the total numbers of different synaptic domains within the Trichoplax proteome to those in other animals.
Domain prediction was carried out on the Trichoplax predicted proteome, plus 23 additional proteomes, using InterProScan. PDZ, SH3 and C2 domains were predicted using the Superfamily application (expect value ≤1E-6), while Zn-finger, GK, Rho-GAP Unc13 homology and CaMKII domains were identified with Pfam (expect value ≤1E-6), and counted to produce the plot shown in Figure 8.

Discussion
Here, we provide the first detailed annotation of the Trichoplax adhaerens whole animal mRNA transcriptome. We note that although a transcriptome shotgun assembly was recently released for Trichoplax sp . H1 [14], the focus of that study was on placozoan genomics, and not transcriptome annotation. Our deep sequencing and comprehensive assembly strategy produced a transcriptome data set that significantly expands upon the former, by identifying several hundred new genes, and with the majority of assembled transcripts bearing complete open reading frames which is useful for functional studies that require gene cloning. This was exemplified by our ability to successfully clone 9 Trichoplax acid-sensitive ion channel homologues using primers designed against their corresponding transcriptome sequences, without the need for classic and laborious gene cloning methods such as degenerate PCR and 5'/3' RACE (rapid amplification of cDNA ends). In separate ongoing studies, we have similarly used the transcriptome to clone other ion channel types, some of which bear protein-coding sequences longer than 5,000 base pairs. Clearly, the gene sequences that were generated provide a valuable resource for functional studies of Trichoplax genes.
Sequencing the Trichoplax whole animal mRNA transcriptome from four separate cDNA libraries permitted quantification of average gene expression levels (i.e. transcripts per million or TPM). Among the most highly expressed genes were those that encode putative secreted proteins, including hydrolytic enzymes such as proteases, lipases and amylases. Interestingly, a single cell transcriptomic study identified a specific Trichoplax cell type that co-expresses numerous hydrolytic enzymes, including those identified in this study (e.g. phospholipase A2, trypsin) [13]. However, whether these cells correspond with lipophil cells, proposed digestive cells located along the ventral epithelium that bear large acidophilic/lipophilic vesicles, is not clear [6,12]. The ability of Trichoplax to quickly break down living unicellular microorganisms along its ventral epithelium for feeding has been documented, and it is likely that regulated secretion of hydrolytic enzymes occurs in response to localized sensory feedback indicating the presence of food, since secretion/digestion seems to occur at discrete sites along the ventral epithelium, not throughout [6]. Perhaps instead, some of the identified enzymes are contained within lysosomal compartments in fiber cells, which are proposed to endocytose entire microalgae as a secondary form of food uptake [96].
Also identified within the Trichoplax secretome were genes important for multicellular processes such as development, reproduction and immunity. Of course, the mere presence of these genes is not enough to expect that their functions are necessarily similar, where at least a subset are likely to have pleiotropic and/or divergent functions. For example, the egg yolk protein vitellogenin, which was identified in the Trichoplax secretome (Table 3), has taken up non-reproductive functions in select arthropods, such as control of longevity [97] and social interactions [98]. Also, the protein Homer (Table S6), a key post-synaptic scaffolding protein, was found to localize to the nucleus in the choanoflagellate Salpingoeca rosetta, and to interact with the lipid raft protein flotillin [99], suggestive of roles in nuclear scaffolding and signaling. In this case, these observations do not represent functional divergence, but rather, an ancient pleiotropic function since both the interaction between Homer and flottilin, and the nuclear localization of both, were subsequently identified in vertebrates astrocytes [99].
The Trichoplax secretome was also found to contain several novel short regulatory peptide gene precursors, which is interesting because regulatory peptides appear to be important for Trichoplax motile behavior. For example, ectopic application of the YPFFGN endomorphin-like peptide identified in this study causes Trichoplax to suddenly pause ciliary locomotion [9], as occurs during feeding [6].
Also, the peptides FFNP, ELPE, MFPF and WPFF trigger animal flattening and internal movements; PWN and SIFGamide trigger detachment and folding, and LF and LFNE trigger rotation and flattening [11].
To our knowledge, the 7 novel peptides identified in this study have not yet been tested for their effects on Trichoplax motile behavior, including FALF and WWamide that exhibit fairly high mRNA expression at the whole animal level ( Figure 5C). Also interesting are the arginine rich secretome proteins (Supplementary File 2), in that they appear to be unique to Trichoplax and most exhibit very high expression levels. The function of these genes is completely unknown, but it is interesting to note that Trichoplax fiber cells possess rickettsial endosymbionts within the endoplasmic reticulum [12, 100, 101], a rare compartment for intracellular bacteria in animals, where perhaps, a subset of secretome proteins remain the ER to mediate symbiotic interactions. Alternatively, arginine-rich proteins are endowed with membrane permeating capabilities, where perhaps, these genes are involved in cellular signaling, or defense/immunity against pathogenic microorganisms [102].
It is striking that despite lacking a bona fide nervous system, Trichoplax expresses a nearly full complement ion channels, pumps and exchangers that permit fast electrical signaling in the nervous system. Sponges, which similarly lack nervous systems, are different from placozoans in that they appear to have lost key electrogenic genes including voltage-gated sodium and potassium channels [78]. Nevertheless, glass sponges exhibit electrical signals that propagate slowly between cells via cytoplasmic junctions, signaling arrest of choanocyte ciliary beating and the feeding current [103].
Furthermore, sponge contractile cells require transient influx of divalent cations such as Ca 2+ , Mg 2+ and/or Sr 2+ [103][104][105], and for some but not all species, contract in response to transient membrane depolarization caused by increasing extracellular [K + ] [104]. Not known is whether Trichoplax cells are also capable of electrical signaling in the form of graded or action potentials, or excitationinduced intracellular calcium signaling, such as exocytosis and contraction, since neither electrophysiological recording nor voltage/calcium imaging have been reported for any placozoan [60]. In a recent study, Trichoplax dorsal epithelial cells were shown to undergo ultrafast contractions, which interestingly, spread from cell to cell in waves suggestive of mechanical coupling. Contractions were proposed to be mediated by non-muscle actinomyosin filaments. Interestingly, application of ionomycin, which increases cytoplasmic [Ca 2+ ] via release from internal stores [106], caused all dorsal epithelial cells to contract, suggesting the contractile filaments are regulated by Ca 2+ . In this study, the authors did not test whether the contractions were also dependent on external Ca 2+ , which could flow into the cytoplasm through select ion channels. Should this be the case, the spread of contraction between cells might conceivably occur through the following mechanism: mechanical stress from adjacent cells activates stretch-sensitive ion channels at the cell membrane, which in turn either depolarize the cell to recruit voltage-gated calcium channels, or directly conduct Ca 2+ into the cell. The presence of ryanodine receptors in the transcriptome suggests that cells are capable of calcium-induced calcium release, where transient increases in cytoplasmic Ca 2+ activate ryanodine receptors in the ER, which in turn conduct ER Ca 2+ to further increase its cytoplasmic concentration.
Interestingly, the soluble cytoplasmic Ca 2+ sensor protein calmodulin was found to be highly expressed at the whole animal level. Localization of this protein at the cellular level might highlight cell types that rely on transient Ca 2+ signalling. Another cell type that might be contractile is the Trichoplax fiber cell, located between the dorsal and ventral epithelium and proposed to mediate localized churning movements observed during feeding [6]. Although direct evidence that fiber cells contract has not been reported, they do exhibit muscle-like filaments under electron microscopy [107]. Clearly, functional studies of distinct cell types, coupled with electrophysiology and/or optical recording, are a necessary next step in understanding the roles of electrical and calcium signaling in Trichoplax biology.
At a genetic level, Trichoplax cells seem poised for complex electrical and calcium signaling, expressing numerous ion channel genes including: voltage-gated potassium and sodium channels that drive fast action potentials; ion channels that modulate action potential shape and frequency such as Ca 2+ -activated potassium channels and cyclic-nucleotide gated channels; and channels that are activated by extracellular ligands such as ionotropic glutamate receptors, ASIC channels, P2X receptors and TRP channels (Table S3). Notably, Trichoplax is the most early-diverging animal to possess all three types of voltage-gated calcium channels found in animals, including a Ca v 1 channel homologue that in muscles drive excitation-contraction coupling, and a Ca v 2 channel homologue that in neurons drive excitation-secretion coupling (i.e. neurotransmitter exocytosis). Previously, we reported the electrophysiological properties of the cloned Trichoplax Ca v 3 calcium channel, expressed in a human cell line. In neurons and muscle, Ca v 3 channels help control action potential threshold, and play a major role in regulating cellular excitability. We found that despite its significant divergence at the amino acid sequence level, the Trichoplax Ca v 3 channel conducts low-voltage activated Ca 2+ currents with hallmark features for this channel type that are required for regulating excitability [108]. By extension, it is plausible that other and perhaps numerous Trichoplax ion channels share core functional attributes with their counterparts from other animals, which could be co-expressed in varying functional assemblages as occurs in neurons to produce alternate modalities of cellular excitation.
Bioinformatic prediction of 665 Trichoplax GPCRs from our expanded gene dataset significantly expands the repertoire previously identified from the genome [109,110]. Altogether, GPCRs encompass roughly ~5% of all Trichoplax protein-coding nuclear genes, which is remarkable given the apparent cellular and morphological simplicity of the animal. In contrast poriferans, which have more anatomically distinct cell types [111], are estimated to only have about 220-328 GPCRs [110,112], while Trichoplax more resembles cnidarians and ctenophores in its GPCR content [110]. In general, GPCRs are for integrating an array of extracellular signals, both intrinsic and sensory, into intracellular responses through various intracellular signaling programs. In a previous study, Trichoplax was shown to possess all of the core intracellular regulators of GPCRs [113], including various classes of G proteins, G protein-coupled receptor kinases (GRKs), arrestins, phosducins, Ric8, GoLoco and regulators of G protein signaling (RGS) [109]. Here, we identified key downstream GPCR signalling elements that in the nervous system form a critical link between neuromodulatory GPCR signaling and electrogenic/synaptic signaling [114]. For example, Trichoplax expresses homologues for adenylyl cyclase (increases cytoplasmic cAMP levels when bound to activated G proteins), and the cAMP-dependent protein kinase A (PKA). Activated PKA has many electrogenic downstream targets such as Ca v 1 channels, causing long-lasting changes in their biophysical properties and hence the state of cellular excitability [115,116]. Similarly, G-protein activation of phospholipase C leads to hydrolysis of membrane-bound phosphatidylinositol 4,5-bisphosphate (PIP 2 ), producing diacylglycerol (DAG) and inositol triphosphate (IP 3 ), the former of which activates PKC and the latter ER-localized IP 3 receptors [117,118]. Like PKA, PKC has various targets including ion channels and synaptic proteins [117,119], and hence similarly serves to alter neural excitability and synaptic signaling for extended periods. Additionally, like ryanodine receptors, IP 3 receptors conduct ER Ca 2+ into the cytoplasm upon activation, providing a functional link between extracellular GPCR ligands and intracellular Ca 2+ signaling [120]. And lastly, activated G proteins themselves can directly bind and modulate ion channels, including G protein activated inward rectifying K + (GIRK) channels, pre-synaptic Ca v 2 calcium channels, and Ca v 3.2 calcium channels [121]. In the vertebrate nervous system, the complexity of integrated fast and slow neural signaling is expected to be staggering, given the diversity of cell types and complexity of tissues, and the broad lexicon of neuromodulatory ligands and GPCRs [4]. Indeed, the cellular simplicity of Trichoplax provides a tantalizing prospect for the comprehensive determination of how fast and slow neural signaling genes co-segregate in their cellular expression across an entire animal.
Trichoplax expresses a rich complement of genes involved in synapse formation and function, most bearing appropriate domain architectures that are required for their integration into specific proteomic complexes at pre-and post-synaptic sites (e.g. Figure 7). Many studies have now shown that early-diverging animals, and single-celled choanoflagellates, possess a large complement of synaptic gene homologues [122]. However, largely unexplored is whether these proteins assemble into homologous complexes as those found in synapses. A major next step forward is therefore a determination of the complexing of these proteins, using classic methods such as yeast II hybrid, coimmunoprecipitation and mass spectrometry. We sought to evaluate whether the absence of synapses in Trichoplax and sponges is reflected by a reduced number of synapse-associated proteinprotein interaction domains (Figure 8). Sponges and choanoflagellates, which lack synapses, were found to generally possess more domains than Trichoplax and ctenophores, indicating that domain counts do reflect the presence/absence of synapses. What is clear from our analyses is that some lineages have considerably expanded domain counts, which for vertebrates and select arthropods, likely arose through independent genome duplication events. We note that our analyses did not explore the heterogeneity of the selected domains. For example, PDZ domains separate into distinct classes that bind different C-terminal sequence motifs on target proteins [123]. Previously, poriferans and choanoflagellates were found to have a reduced diversity in PDZ domain types [91], which might also be true for Trichoplax and ctenophores should they be analyzed in the same way. Interestingly, a recent study determined that a large proportion of metazoan proteins which interact with PDZ domains predate animals, and that their incorporation into PDZ-dependent protein complexes emerged later via de novo mutations that produced appropriate PDZ ligand motifs [124]. An interesting question is therefore: what are the ligands for different synaptic scaffolding proteins in early-diverging animals such as Trichoplax, ctenophores and poriferans, and how do their interactomes compare to those humans and other animals?

Methods
Due to the in silico nature of this study, the majority of methods used to generate the data are contained within the results section and corresponding figure legends. Transcriptome annotation for, among other research objectives, the purpose of inferring orthology between Trichoplax adhaerens genes and those found in other species, consisted of a combination of BLAST against UniProt/Swiss-Prot databases, reciprocal BLAST against NCBI, and InterPro protein domain analyses. In cases where orthology was uncertain, maximum likelihood phylogenetic inference was also used. Secretome transcripts were not subject to reciprocal BLAST analysis, however, BitScores and E-Values from our BLASTp annotation are provided in Tables 1-4, allowing the reader to critically assess whether grounds for inferring homology exist. Below, we only detail background methods used for obtaining and generating the proteomes used for Figure 8 Similarly, for Nematostella vectensis, single-end RNA-Seq reads were obtained from NCBI (accession numbers SRR5183917, SRR5183918, SRR5183919, SRR5183920, SRR5183921, SRR5183922, SRR5183923, SRR5183924, SRR5183925, SRR5183926, SRR5183927, SRR5183928, SRR5183929 and SRR5183930). These were assembled using a combined strategy as depicted in Figure 1A for the Trichoplax transcriptome, using trimmed RNA-Seq reads that were either normalized or un-normalized for de novo assembly with Trinity (i.e. 2 assemblies), or normalized for ab initio assembly with Cufflinks (1 assembly). A fourth assembly was obtained from FigShare [128], and the four assemblies were merged using the EvidentialGene pipeline. Prior to domain analysis, all of the obtained proteomes were processed for removal of redundant sequences at the protein level using CD-Hit  Tables   Table 1. Table denoting