Approximately 150 recognized species of comb jellies form a clade of pre-bilaterian animals1,2,3 (Fig. 1f) with an elusive genealogy, possibly tracing their ancestry to the Ediacaran biota4,5. We selected the Pacific sea gooseberry, Pleurobrachia bachei (A. Agassiz, 1860; Figs 1a, 2, Extended Data Fig. 1, Supplementary Data 1 and Supplementary Videos 1 & 2) as a model ctenophore because of preserved traits thought to be ancestral for this lineage (for example, cydippid larva and tentacles). Three next-generation sequencing platforms (454, Illumina and Ion Torrent) were used to obtain >700-fold coverage (Supplementary Tables 1 and 2) of Pleurobrachia’s genome, and about 2,000-fold coverage of the transcriptome representing all major organs and developmental stages (Supplementary Tables 3 and 4). Consequently, the draft assembly was 156,146,497 base pairs (bp) with 19,523 predicted protein-coding genes (Supplementary Tables 5–7). About 90% of these predicted genes are expressed in at least one tissue or developmental stage (Supplementary Methods) and 44% of Pleurobrachia genes have orthologues in other animals (Supplementary Tables 7 and 8). More than 300 families of transposable elements constitute at least 8.5% of the genome (Supplementary Table 9 and Supplementary Data 2) with numerous examples of diversification of some ancient transposable element classes (for example, transposases and reverse transcriptases). Approximately 1.0% of the genome is methylated. Pleurobrachia also uses DNA demethylation during development, with both 5-hydroxymethyl cytosine (5-hmC) and its synthetic enzyme TET6 (Extended Data Fig. 2). The obtained genome and transcriptome data provide rich resources ( for investigating both animal phylogeny and evolution of animal innovations including nervous systems2,3,7,8,9.

Figure 1: Ctenophores and their innovations.
figure 1

a, The sea gooseberry, Pleurobrachia bachei (Supplementary Fig. 1), was selected as a target for genome sequencing due to preservation of traits ancestral for this lineage and because in situ hybridization/immunolabelling is possible. be, Major ctenophore innovations. b, Nervous system revealed by tyrosinated α-tubulin immunolabelling (scale bar, 60 μm). c, Scanning electron microscopy (SEM) imaging of nerve net in a tentacle pocket (scale bar, 20 μm). d, Locomotory ciliated combs (SEM; scale bar, 100 μm). e, Glue-secreting cells (colloblasts) in tentacles (SEM; scale bar, 20 μm). f, Relationships among major animal clades with choanoflagellates sister to all Metazoa.

PowerPoint slide

Figure 2: Phylogenomic reconstruction among major ctenophore lineages.
figure 2

Cydippid (Euplokamis, Pleurobrachia, Dryodora and Mertensiidae) and lobate (Mnemiopsis and Bolinopsis) ctenophores were polyphyletic, suggesting independent loss of both cydippid larval stage and tentacle apparatus as well as independent development of bilateral symmetry in benthic ctenophores, Vallicula and Coeloplana (Supplementary Data 4).

PowerPoint slide

Ctenophore phylogeny

Although relationships among basal animal lineages are controversial1,10,11,12,13,14,15,16, our analyses (Supplementary Data 4) with Ctenophora represented by Pleurobrachia and Mnemiopsis suggest the placement of Ctenophora as the most basal animal lineage (Fig. 1 and Extended Data Fig. 3). Porifera was recovered as sister taxon to remaining metazoans (bootstrap support = 100%) with Cnidaria sister to Bilateria (bootstrap support = 100%, Fig. 1f). Shimodaira–Hasegawa tests17 (corresponding to Extended Data Fig. 3a–c with 586 gene matrix) rejected both Eumetazoa (sponges sister group to all other metazoans) and Coelenterata (Cnidaria+Ctenophora). Placement of Ctenophora at the base of Metazoa also provides the most parsimonious explanation for the pattern of global gene gain/loss seen across major animal clades (Fig. 3a and Supplementary Table 14a, b). Transcriptome data from ten additional ctenophores (Supplementary Table 13) with stricter criteria for orthology inference (Supplementary Methods 7) also placed ctenophores basal, albeit with less support (Extended Data Fig. 3d). When the most conserved set of genes was considered (Supplementary Methods 7.5 and Supplementary Data 4.3), the topology was unresolved. Weak support is probably due to under-representation of comparable transcriptomes from sponges and large protein divergence. Nevertheless, Shimodaira–Hasegawa tests based on expanded ctenophore sampling (with a reduced 114 gene matrix due to lack of other ctenophore and sponge genomes; Supplementary Methods 7.2) also rejected Coelenterata but not Eumetazoa. Notably, relationships within Ctenophora were strongly supported (Fig. 2). Both cydippid and lobate ctenophores, previously viewed as monophyletic clades, were recovered polyphyletic, indicating independent loss of both the cydippid larval stage and tentacle apparatus. Interestingly, Platyctenida was the second basal-most branch in the Ctenophore clade, suggesting that their benthic and bilaterial nature are secondarily derived.

Figure 3: Gene gain and gene loss in ctenophores.
figure 3

a, Predicted scope of gene loss (blue numbers; for example, −4,952 in Placozoa) from the common metazoan ancestor. Red and green numbers indicate genes shared between bilaterians and ctenophores (7,771), as well as between ctenophores and other eukaryotic lineages sister to animals, respectively. Text on tree indicates emergence of complex animal traits and gene families. Deu, deuterostomes; Pro, protostomes. b, Uniquely shared and lineage-specific genes among basal metazoans. Values under species names indicate number of genes (bold) without any recognized homologues (e-value is 10−4) versus the total number of predicted gene models in relevant species (Supplementary Table 14b).

PowerPoint slide

A highly reduced complement of animal-specific genes is a feature shared for the entire ctenophore lineage (Fig. 3 and Supplementary Table 15). HOX genes involved in anterio-posterior patterning of body axes and present in all metazoans are absent in ctenophores and sponges18 (Supplementary Tables 17 and 18). Likewise, canonical microRNA machinery (that is, Drosha and Pasha, Supplementary Table 19) is lacking in Pleurobrachia and other ctenophores. Using small RNA sequencing from Pleurobrachia, Bolinopsis and Beroe, we were unable to experimentally detect microRNAs (Supplementary Data 5.4). Pleurobrachia also lacks major elements that initiate innate immunity such as pattern recognition receptors (Toll-like, Nod-like, RIG-like, Ig-Toll-IL-1R domain) and immune mediators, MyD88 and RHD transcription factors, that are present in bilaterians, cnidarians and, in divergent forms, in sponges19,20 (Supplementary Table 20).

Key bilaterian myogenic/mesoderm-specification genes are absent in Pleurobrachia’s genome and transcriptomes of ten other ctenophores (Supplementary Table 35). These data indicate that muscles21 and, possibly, mesoderm evolved independently in Ctenophora to control the hydroskeleton, body shape and food capture. Thus, ctenophores might have independently developed complex phenotypes and tissue organization, raising questions about the nature of ctenophore-specific traits such as their unique development, combs, tentacles, aboral/apical organs and nervous systems.

Ctenophore innovations

To assess the genomic bases of ctenophore-specific innovations, we performed RNA-seq profiling of the major developmental stages (Fig. 4a, b) as well as adult organs and identified genes differentially expressed in these structures. Many Pleurobrachia genes that have no homologues in other species are specifically expressed and most abundant during the 4- to 32-cell cleavage stages as well as in tentacles, combs and the aboral organ (Fig. 4b and Extended Data Fig. 4). Thus, structures that are known as ctenophore innovations (Fig. 1d, e) have the largest complement of highly expressed Pleurobrachia/ctenophore-specific genes. These data suggest extensive gene gain in cell lineages associated with early segregation of developmental potential leading to ctenophore-specific traits in structures controlling feeding, locomotion and integrative functions; a finding consistent with proposed ‘orphan’ genes contributing to variation in early development and evolution of novelties22,23.

Figure 4: Nature of ctenophore innovations.
figure 4

a, Main developmental stages in Pleurobrachia from eggs to cleavage (2–64 cells), gastrulation (1–3 h) and formation of cydippid larvae (24 h). b, Hierarchical clustering of approximately 400 ctenophore-specific genes differentially expressed among different development stages and adult structures as revealed by RNA-seq experiments. Colour index: black indicates highest level of expression, followed by purple, red then down to white indicating no expression. Most of these ctenophore-specific genes are primarily expressed during 4–32-cell stages (asterisks). The red circle indicates a subset of novel genes uniquely expressed in combs, tentacles and the aboral organ (AO). These genes lack recognized homologues in other organisms. c, Diversity and differential expression of RNA-editing genes in Pleurobrachia development and adult tissues (RNA-seq). ADAR1 has the highest expression level in early cleavage stages whereas ADAR2 and ADAR3 and ADAT1 and ADAT2 are most abundant in the combs. d, Morphological appearance of neurons during the third day of development (the top inset, neuronal cell bodies are stained with tyrosinated α-tubulin antibodies, red arrows) correlates with abundant expression of multiple iGluR receptors, suggesting that glutamate has an important role as an intercellular messenger. Muscles formed well before neuronal differentiation at the end of the first day of development (the bottom insert, phalloidin staining, yellow arrow); white arrow points to the embryonic mouth with hundreds of cilia inside. In c and d expression levels of RNA-editing or iGluR genes are shown as a normalized frequency of sequence reads for a given transcript from all RNA-seq data for each developmental stage (Supplementary Methods). Scale bars in d: 15 μm (top); 40 μm (bottom).

PowerPoint slide

Examples of known metazoan gene families that are considerably expanded in Ctenophora (Supplementary Data 5 and Supplementary Table 16) include collagens, RNA-editing enzymes and RNA-binding proteins (Supplementary Data 5). Pleurobrachia’s genome encodes the highest number of RNA-editing enzymes (ADAR1–4, ADAT1–3, CDA1–2) reported so far among metazoans24,25 (Supplementary Data 5.5), possibly acting as the generalized mechanism generating post-transcriptional diversity and ctenophore-specific traits in locomotory and integrative structures (combs and aboral organ). Matching the expansion of RNA regulatory mechanisms, Pleurobrachia has more RNA-binding proteins (especially RRM/ELAV, KH and NOVAs26,27, Supplementary Table 21) than any basal metazoan or choanoflagellate examined. Dozens of RNA-binding proteins are selectively expressed and abundant during 8–64-cell stages (Supplementary Table 31), and might contribute to sequestration of RNAs and segregation in developmental potential leading to early cell-fate specification.

Phenotypic complexity positively correlates with presence and high differential expression of 92 homeodomain Pleurobrachia genes (Supplementary Data 5.2 and Supplementary Table 17)—76 genes have been reported in Mnemiopsis18, whereas the Amphimedon homeodomain complement consists of only 32 genes. In contrast, some developmental pathways are either absent (Hedgehog, JAK/STAT) or have reduced representation in ctenophores (TGF-β, Wnt, Notch). Surprisingly, most Wnt genes are weakly expressed during Pleurobrachia development, whereas the ctenophore-specific subtype WntX is primarily restricted to adult neuroid elements such as polar fields, aboral organ and tentacular conductive tracts (Extended Data Fig. 5e), suggesting a distinct molecular make-up of neural systems.

Parallel evolution of neural organization

Extensive parallel evolution of neural organization in ctenophores is the most evident. Compared to other animals with nervous systems, many genes controlling neuronal fate and patterning (for example, neurogenins, NeuroD, Achaete-scute, REST, HOX, Otx) are absent in the ctenophores that we sampled. Orthologues of pre- and postsynaptic genes also have limited representation (Supplementary Table 34), and they lack components (for example, neuroligin) critical for synaptic function in other eumetazoans.

Notably, our combined molecular, ultrasensitive metabolomic, immunohistochemical and pharmacological data strongly suggest that ctenophores do not use serotonin, acetylcholine, dopamine, noradrenaline, adrenaline, octopamine, histamine or glycine as intercellular messengers (Extended Data Figs 6 and 7g, Supplementary Data 5.8 and Supplementary Tables 22–26). Lack of ionotropic receptors for these molecules in ctenophores is consistent with this conclusion (Supplementary Table 26a). Most genes involved in the synthesis of neurotransmitters are absent in non-metazoan opisthokonts Monosiga and Capsaspora, suggesting that they are cnidarian/bilaterian innovations.

We next investigated the identity of the ctenophore transmitters. Physiological and pharmacological tests suggest that l-glutamate is a candidate neuromuscular transmitter in Pleurobrachia (Fig. 5b, Extended Data Fig. 7), able to induce rapid inward currents and raise intracellular Ca2+ concentration in muscle cells causing muscle contractions at nanomolar concentrations (10−7 M). In contrast, all other classical neurotransmitters were ineffective even in concentrations up to 5 × 10−3 M whereas d-glutamate as well as l-/d-aspartate have significantly less affinity in these assays (Fig. 5b).

Figure 5: Emergence of neural organization in Pleurobrachia.
figure 5

a, Two neural nets in Pleurobrachia as revealed by tyrosinated α-tubulin immunostaining. Top image shows subepithelial net with concentrations of neuronal elements in the polar fields and ciliated furrows, known as structures involved in sensory and motor functions, respectively (blue arrow in right insert indicates location of a neuronal somata with individual neurites marked by red arrows). The bottom image shows neurons of the mesogloeal net (arrows are neuronal somata; arrowheads are neuronal processes). Note that phalloidin (a muscle marker) did not stain these cells. Scale bar, 120 μm (top); 10 μm (bottom images). b, l-glutamate (10−7–10−3 M) induced action potentials in muscle cells, whereas other transmitter candidates were ineffective even at concentrations up to 5 mM. Typical responses of ctenophore muscle cells to local pulses of a transmitter application were recorded both as individual action potentials (whole-cell current-clamp mode) and video contractions from a single muscle cell. The graph shows normalized responses from the same muscle cell, indicating that l-glutamate is the most potential excitatory molecule compared to d-glutamate or l-/d-aspartate (Supplementary Methods). ACh, acetylcholine. c, Key molecular innovations underlying neural organization in ctenophores. 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, 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. d, The aboral organ has the greatest diversity and highest expression levels of 12 gap junction proteins, suggesting unmatched expansion of electrical signalling in this complex integrative organ—an analogue of an elementary brain in ctenophores. Expression of different innexins is shown as a summation of normalized frequencies of respective sequencing reads in RNA-seq data obtained from each developmental stage and adult tissues (y axis is ‘expression frequency’ or ‘expression’; see Supplementary Methods).

PowerPoint slide

The proposed role of glutamate as a signal molecule in ctenophores is supported by an unprecedented diversity of ionotropic glutamate receptors (iGluRs; Extended Data Fig. 7a, b and Supplementary Table 27)—far exceeding the number of genes encoding iGluRs in other basal metazoans28. iGluRs might have undergone a substantial adaptive radiation in Ctenophora as evidenced by unique exon/intron organization for many subtypes, and ctenophoran iGluRs form a distinct clade within the gene tree. Notably, during development, Pleurobrachia’s neurons are formed 2 days after the initial muscle formation, and first neurogenesis events correlate with co-expression of all iGlu receptors in hatching larvae (Fig. 4d). All cloned iGluRs also show remarkable cell-type-specific distribution with predominant expression in tentacles, followed by combs and the aboral organ, revealing well-developed glutamate signalling in adults (Extended Data Fig. 7b). Additionally, Pleurobrachia contains more genes for glutamate synthesis (8 glutaminases) and transport (8 sialins) than any other metazoan investigated29,30. Although we detected GABA (γ-aminobutyric acid, Supplementary Tables 22–24) and immunohistochemically identified its localization in muscles, lack of pharmacological effects of GABA on Pleurobrachia behaviour and major motor systems, such as cilia, muscle and colloblasts, suggest that GABA is a by-product of glutamate metabolism by l-glutamic acid decarboxylase.

The first nervous systems are suggested to be primarily peptidergic in nature7. Although we did not find any previously identified neuropeptide homologue, secretory peptide prohormone processing genes (Supplementary Table 31) are present. We predicted 72 novel putative prohormones in Pleurobrachia and found >50 homologues in other sequenced ctenophores (Extended Data Fig. 8 and Supplementary Tables 28 and 32). Functions of these prohormone-derived peptides could include cell-to-cell signalling, toxins, involvement in innate immunity, or a combination of these functions. Several ctenophore-specific precursors are expressed in polarized cells around the mouth, tentacles and polar fields, indicative of a signalling role (Extended Data Fig. 8b). Secreted products of these prohormones may be natural ligands for >100 orphan neuropeptide-like G-protein-coupled receptors31 identified in Pleurobrachia (Supplementary Table 26b). A second example of neuropeptide receptor candidates is amiloride-sensitive sodium channels (ASIC), which are also known to be regulated by different classes of short peptides and protons32. Pleurobrachia’s genome has 29 genes encoding ASICs—more than any organism sequenced so far—and expression of most correlated with developmental appearance of neurons (Supplementary Table 31). ASIC expression is most abundant in tentacles, combs and aboral organs—structures enriched in neural elements and under complex synaptic control.

Moreover, ctenophores evolved an enormous diversity of electrical synapses (absent in Nematostella, Amphimedon and Trichoplax) with 12 gap junction proteins (pannexins/innexins33 but not chordate-specific connexins) found in Pleurobrachia. All pannexins/innexins have their highest expression in the aboral organ followed by tentacles and combs (Fig. 5d). Non-metazoans lack pannexin orthologues, suggesting that these are metazoan innovations with profound expansion of this family in ctenophores. The aboral organ, combs and tentacles have a relatively large diversity of ion channels (Extended Data Fig. 9b), confirming complex regulation of excitability in these structures. However, the overall complement of voltage-gated ion channels in ctenophores is reduced compared to other eumetazoans34 (Extended Data Fig. 9a).

Our genome-wide survey also indicates that some bilaterian and cnidarian pan-neural markers are present (for example, three elav and musashi genes), but they are not expressed in neurons; a finding consistent with early divergence and extreme parallel evolution of neural systems in this lineage (Extended Data Figs 5 and 9b).


Figure 5c summarizes key molecular innovations underlying neural organization in ctenophores. Evidently, with a markedly different molecular and genomic make-up, ctenophores have achieved complex phenotypic plasticity and tissue organization. Thus, ctenophores might represent remarkable examples of convergent evolution including the emergence of neuro-muscular organization from the metazoan common ancestor without differentiated nervous system or bona fide neurons (Extended Data Fig. 10b and Supplementary Data 15). The alternative ‘single-origin hypothesis’, where the common ancestor of all metazoans had a nervous system with complex molecular and transmitter organization including all classical cnidarian/bilaterian transmitters and neurogenic genes (Extended Data Fig. 10a), is a less parsimonious scenario. This hypothesis implies that ctenophores, despite being active predators, underwent massive loss of neuronal and signalling toolkits and then replaced them with novel neurogenic and signalling molecules and receptors.

These findings are of relevance for regenerative and synthetic biology in designing novel signalling pathways and systems. In this case, ctenophores and their genomes present matchless examples of ‘experiments’ in nature and the possible preservation of ancient molecular toolkits lost in other animal lineages.

Methods Summary

Genome sequencing and transcriptome profiling used high-throughput methods. Gene prediction (Augustus/Fgenesh) and annotations were complemented by metabolome and microchemical assays using capillary electrophoresis with laser-induced fluorescence and mass spectrometry detectors to validate the presence/absence of selected transmitters. RNA-seq, PCR with reverse transcription (RT–PCR), in situ hybridization and immunohistochemistry were used to characterize and localize gene expression in adults and embryos. Pharmacological and electrophysiological tests were performed on semi-intact preparations using freshly isolated ctenophore cells.

Online Methods

Source material

Animals (Pleurobrachia bachei, Euplokamis dunlapae, Dryodora glandiformis, Beroe abyssicola, Bolinopsis infundibulum and Mertensiid) were collected at Friday Harbour Laboratories (Pacific North-Western Coast of USA) and maintained in running seawater for up to 2 weeks. Other species were collected at the Atlantic coast of Florida and around Woods Hole, Massachusetts (Pleurobrachia pileus, Pleurobrachia sp., Mnemiopsis leidyi) as well as central Pacific (Palau, Hawaii, Coeloplana astericola, Vallicula multiformis). Animals were anaesthetized in 60% (volume/body weight) isotonic MgCl2 (337 mM). Specific tissues were surgically removed with sterile fine forceps and scissors and processed for DNA/RNA isolations as well as metabolomics or pharmacological/electrophysiological tests. Whole animals were used for all in situ hybridization and immunohistochemical tests as described35. Genomic DNA (gDNA) was isolated using Genomic-tip (QIAGEN) and total RNA was extracted using RNAqueous-Micro (Ambion/Life Technology) or RNAqueous according to manufacturers’ recommendations. Quality and quantity of gDNA was analysed on a Qubit2.0 Fluorometer (Life Technologies) and for RNA we used a 2100 Bioanalyzer (Agilent Technologies). For all details see Supplementary Methods sections 1.1–1.3.

Genome sequencing

All genomic sequence data for de novo assembly were generated on Roche 454 Titanium and Illumina Genome Analyzer IIx, HiSeq2000 and MiSeq instruments using both shotgun pair-end and mate-pair sequencing libraries with 3–9 kb inserts as summarized in Supplementary Tables 1 and 2. Shotgun sequencing was performed from a single individual. Owing to a limited amount of starting gDNA, mate pair libraries were constructed from 10–12 individuals. In total, the genome sequencing is composed of 132,015,600,107 bp or 132 Gb of data, which corresponds to 733–825× physical coverage of the Pleurobrachia genome (the size of the P. bachei genome is estimated to be 160–180 Mb); see Supplementary Methods sections 1.4–2.1.2.

Genome assemblies

The Pleurobrachia bachei draft genome was assembled using a custom approach designed to leverage the individual strengths of three popular de novo assembly packages and strategies: Velvet36, SOAPdenovo37, and pseudo-454 hybrid assembly with ABySS38. First, using filtered and corrected data, we performed individual assemblies from 454 and Illumina reads by the Newbler (Roche, Inc.) software. Then the merged/hybrid assembly was achieved using three individual assemblies (SOAPdenovo, Velvet and ABySS/Newbler as described in Supplementary Methods 2.2). Three gene model predictions were performed by Augustus39 and Fgenesh predictions with the Softberry Inc. Fgenesh++ pipeline40,41 to incorporate information from full-length cDNA alignments and similar proteins from the eukaryotic section of the NCBI NR database42. After initial gene predictions in each of the three sets of genomic scaffolds, we screened each set of gene models for internal redundancy with the BLASTP program from NCBI’s BLAST+ software suite43. A model was considered redundant if it: had 90% identity to other model; the alignment between the two models had a bit score of at least 100; and the model was shorter than the other model.

Scaffolds producing these gene models were pooled and then screened for prokaryotic contamination using UCSC’s BLAT software package44 to produce the draft genome assembly version 1.0 (statistics can be found in Supplementary Table 5 and Supplementary Methods 2).

Genome annotation

For annotation, gene models were uploaded to the In-VIGO BLAST interface, a blastp alignment of gene models was performed against the entirety of NCBI’s non-redundant protein database and the Swiss-Prot protein database, and subsequently annotated in terms of Gene Ontology and KEGG pathways as well as Pfam domain identification. Transposable elements (TEs) were identified using not only WU-BLAST and its implementation in CENSOR but also databases for all known classes, superfamilies and clades of TEs described in the literature and/or collected in Repbase45. Detected sequences have been clustered based on their pairwise identities by using BLASTclust. All autonomous non-LTR retrotransposons have been classified based on RTclass1 (ref. 46). To merge partially predicted, non-redundant gene models with assembled transcriptome data, a custom Java tool was developed. This Java tool extended partial gene model predictions based on using transcriptome sequences to bridge 5′ and 3′ fragments of partially predicted genes. Using this Java tool, analysis of alignments of non-redundant gene models to assembled Pleurobrachia transcriptomes resulted to 19,523 (Supplementary Table 30) gene models. These gene models were used to also identify their possible homologues in assembled transcriptomes from 10 other ctenophore species sequenced (Supplementary Tables 10 and 11). All genomic sequences were submitted to NCBI on SRA accession number Project SRP001155 (Supplementary Methods 3.1–3.2).

Transcriptome sequencing and annotation

Three sequencing technology platforms were used for transcriptome profiling (RNA-seq): Roche 454 Titanium, Illumina HiSeq2000 and Ion Proton/PGM (Ion Torrent, Life Technologies). RNA-seq was performed from all major embryonic and developmental stages (1 cell, 2 cells, 4 cells, 8 cells, 16 cells, 32 cells, 64 cells, early and later gastrula, 1 day and 3 day larvae), major adult tissues and organs (combs, mouth, tentacles, stomach, the aboral organ, body walls), and whole body of Pleurobrachia bachei. We developed a reduced representation sequencing protocol for the 454 and Ion Torrent sequencing platforms that can detect low abundance transcripts47. The method reduces the amount of sequencing and gives more accurate quantification and additional details of the procedure are reported elsewhere47,48. In summary, we have generated 499,699,347 reads or 47.9 Gb to achieve approximately 2,000× coverage of the Pleurobrachia transcriptome.

In addition, Illumina HiSeq sequencing was also performed with RNA extracted from the following ctenophore species: Euplokamis dunlapae, Coeloplana astericola, Vallicula multiformis, Pleurobrachia pileus, Pleurobrachia sp. (collected from the Middle Atlantic and later identified as a subspecies of P. pileus), Dryodora glandiformis, Beroe abyssicola, Mnemiopsis leidyi, Bolinopsis infundibulum and an undescribed species which belongs to the family Mertensiidae (Supplementary Table 3). Each sequencing project was individually assembled using the Trinity de novo assembly package49 and in selected cases using MIRA. Reads from developmental stages were also assembled using the CLCBio Genomics Workbench. Before each assembly, reads were quality trimmed and had adaptor contamination removed with cutadapt50. Full summaries of the transcriptome assemblies are presented in Supplementary Tables 4 and 10. Each transcriptome was mapped to the Pleurobrachia genome, and aligned to both NCBI’s non-redundant protein database (NR) and the UniProtKB/Swiss-Prot (SP) protein database. Gene Ontology51 and Kyoto Encyclopedia of Genes and Genomes52,53 (KEGG) terms were associated with each transcript. By first translating transcripts in all six reading frames, Pfam/SMART domains54 were assigned to each reference transcriptome.

Each reference transcriptome and its full set of annotation and expression data was uploaded to our transcriptome database for downstream analysis and visualization55,56. The database is integrated with UCSC type genome browser. Via the genome project homepage ( all data sets have direct download options. Quantification of gene expression profiling was performed on all transcriptional data as described in Supplementary Methods 4.4. Hierarchical clustering was performed by Spotfire agglomerative algorithm. All primary transcriptome data was submitted to NCBI on SRA accession number Project SRP000992. (See Supplementary Methods 4.1–4.2.3 for details.)

Phylogenetic analyses

To reconstruct basal metazoan phylogeny (see controversies in10,11,12,13,14,15,57), we conducted two sets of phylogenomic analysis using tools described elsewhere58. All analyses included new data from Pleurobrachia bachei and the sponges Sycon (Calcarea) and Aphrocallistes (Hexactinellida). For the first set of analyses, Ctenophora was represented by two species of Pleurobrachia and Mnemiopsis leidyi. Initial analyses included the taxa in Supplementary Table 12. For a subsequent analysis, sampling within Ctenophora was expanded to include ten additional taxa, each represented by a relatively deeply sequenced Illumina transcriptome (Supplementary Table 13). In order to reduce noise in the phylogenetic signal, we used strict criteria to exclude paralogues, highly derived sequences, mistranslated sequence regions, and ambiguously aligned positions in sequence alignments. Analyses were conducted in RAxML 7.2.7 (refs 59) using maximum likelihood (ML) with the CAT +WAG + F model. Topological robustness (that is, nodal support) for all ML analyses was assessed with 100 replicates of nonparametric bootstrapping. Details of phylogenomic analyses are presented in Supplementary Methods 7. Shimodaira–Hasegawa test17 was implemented in RAxML with the PROTGAMMAWAGF model17.

In order to examine evolution of single genes or gene families, alignments were performed with either ClustalX260,61,62 or Muscle63 then, if appropriate, either trimmed manually or trimmed using GBlocks64 to exclude ambiguously aligned positions. Once alignments were obtained, gene trees were reconstructed in MEGA 565 using ML with the Whelan and Goldman (WAG) model. The bootstrap consensus tree was inferred from 100 replicates. All positions containing gaps and missing data were eliminated. Pfam composition54, Gene Ontology51 and KEGG52,53 were used to further validate P. bachei orthologues. Analyses of gene gain and gene loss were performed using custom scripts as described elsewhere66 and in Supplementary Methods 7.

Analysis of DNA methylation

ELIZA-based colorimetric assays (Epigenteck) were performed to quantify both global 5-mC and 5-hmC methylation in the P. bachei genome. A total of 6 individual P. bachei and three rat (positive control) were used (Supplementary Methods 1.2). Three biological and technical replicates were performed for every sample. Absolute quantification of 5-mC and 2-hmC were determined and date is reported as a mean ± s.e.m. (Supplementary Methods 8).

Molecular cloning, in situ hybridization and immunohistochemistry

Methods were similar as reported elsewhere35,47,48,67 with some modifications (Supplementary Methods 9–11).

Scanning electron microscopy

Animals were fixed in 2.5% glutaraldehyde in 0.2 M phosphate-buffered saline (pH 7.6) for 3–4 h at room temperature, and washed. For secondary fixation, we used 2% osmium tetroxide in 1.25% sodium bicarbonate for 2–3 h at room temperature. After dehydration in ethanol, samples were placed for drying in Samdri-790 Critical Point Drying. After drying the samples were coated on Sputter Coater. SEM observations and recordings were done on NeuScope JCM-5000 microscope (Supplementary Methods 12).

Electrophysiological methods, calcium imaging and pharmacological assays

Patch electrodes for extracellular and whole-cell recordings were pulled from borosilicate capillary (P-87, Sutter Instruments). All currents were recorded using an Axopatch or 200B amplifier controlled by a Digidata 1322A and pClamp 9.2. Action potentials (APs, spikes) were recorded in track mode using cell-attached loose-patch configuration. Whole-cell currents were recorded in voltage-clamp mode at a holding potential of −70 mV. Neurotransmitter candidate (see Supplementary Method 15) application for both extracellular AP and whole cell recordings were performed with a rapid solution changer, RSC-160 (Bio-Logic-Science Instruments, France). Data were analysed with Clampfit 9.0 (Molecular Devices) in combination with SigmaPlot 10.0. Videomicroscopy and time-lapse series were acquired with QImaging EXi CCD camera using DIC mode of Nikon Eclipse 2000 inverted microscope. Calcium imaging was performed on isolated ctenophore muscle cells using an Olympus IX-71inverted microscope equipped with a cooled CCD camera (ORCA R2, Hamamatsu). Cells were injected with calcium-sensitive probe (Fluo-4, 5 μM) through a patch pipette. Fluorescence imaging was performed under the control of Imaging Workbench 6 software. Stored time series image stacks were analysed off-line using Imaging Workbench 6, Clampfit 10.3, SigmaPlot 10/11 or exported as TIFF files into ImageJ 1.42. Pharmacological tests and behavioural assays with video recording were performed on intact animals in 5–40 l aquaria or on semi-intact preparations in a Sylgard-coated Petri dish with free cilia beating and muscle contractions. To monitor and quantify cilia movements we used glass microelectrodes filled with 2 M potassium acetate with resistances of 5–20 MΩ with electrical signals recorded by A-M System amplifiers (Neuroprobe 1600) and Gould Recorder (WindoGraf 980).

Determination of the presence of classical neurotransmitters by capillary electrophoresis (CE)

Two CE separation techniques were used to analyse tissue extracts for the presence of a number of neurotransmitters (Supplementary Tables 22 and 23; Supplementary Methods 17). While both methods used CE separations, complimentary detection methods, laser-induced native fluorescence (LINF)68 and electrospray ionization mass spectrometry (ESI-MS)69,70 were used to ensure broad coverage and low detection limits for the specific analytes of interest. Whole bodies of small animals as well as individual organs and tissues were removed, rinsed with ultrapure water and analytes were extracted using 49.5/49.5/1, methanol (LC-MS grade)/water/glacial acetic acid (99%) by volume, homogenized, centrifuged and supernatant was removed and frozen at −80 °C until analysis. The CE-LINF instrument used ultraviolet excitation at 264 nm and the native fluorescence emission was collected and recorded using a UV-enhanced CCD array (Spec-10; 2KBUV/LN; Princeton Instruments). CE separations were performed by hydrodynamic injection of 10 nl of sample and using 25 mM citric acid (pH 2.5, applied voltage +30 kV) or 50 mM borate (pH 9.5, applied voltage +21 kV). Analytes were identified based on comparison of both the migration time and fluorescence spectrum to that of standard mixtures of analytes. CE-ESI-MS analysis was performed using a Bruker Microtof or a Maxis (Bruker Daltonics) mass spectrometer for detection. All separations were performed using 1% formic acid in water as the electrolyte and applied voltage of +30 kV. Sheath liquid was 0.1% formic acid in 50/50 methanol/water. Samples were hydrodynamically injected for a total volume of 6 nl. Mass spectra were collected and recorded at a rate of 2 Hz with calibration performed using sodium formate clusters. Analytes were identified based on comparison of both the CE migration time and mass match to that of standard mixtures of analytes.