On the origin of appetite: GLWamide in jellyfish represents an ancestral satiety neuropeptide

Significance

Unlike popular animal models, jellyfish only have dispersed and decentralized neurons, and their nervous systems are thought to represent an ancestral state. Nevertheless, jellyfish display a rich repertoire of behaviors, such as sequential feeding rituals. Here, we identified a secreted peptide in a subset of jellyfish neurons and show that this peptide functions to inhibit food intake after a meal. Our comparative behavioral analyses further revealed that this jellyfish neuropeptide suppresses feeding in fruit flies. These results suggest deep origins of a conserved satiety system.

Abstract

Food intake is regulated by internal state. This function is mediated by hormones and neuropeptides, which are best characterized in popular model species. However, the evolutionary origins of such feeding-regulating neuropeptides are poorly understood. We used the jellyfish Cladonema to address this question. Our combined transcriptomic, behavioral, and anatomical approaches identified GLWamide as a feeding-suppressing peptide that selectively inhibits tentacle contraction in this jellyfish. In the fruit fly Drosophila, myoinhibitory peptide (MIP) is a related satiety peptide. Surprisingly, we found that GLWamide and MIP were fully interchangeable in these evolutionarily distant species for feeding suppression. Our results suggest that the satiety signaling systems of diverse animals share an ancient origin.

Animals must feed to gain nutrients for their growth and survival. Feeding behavior is therefore regulated by antagonistic internal systems that respond to an animal’s nutritional state. Upon satisfying their nutritional needs, animals display satiety: a decrease in their motivation to feed (i.e., appetite) (1). Molecular correlates of appetite include peptides secreted from the nervous system, gut, and other organs (1). Some peptides, such as neuropeptide Y and neuropeptide F, have similar structures and conserved feeding-regulating functions (2) across many bilaterian species (3). However, it is unclear how far we can date back the role of these peptides in regulating feeding behavior.

Cnidarians are the sister group of bilaterians and therefore powerful models to approach this question. Cnidarian nervous systems are often described as decentralized nerve nets (4), a trait thought to represent an ancestral state (5). Cnidarian feeding behavior has stereotyped sequences (6), and the underlying cellular pathways are characterized in some species (7). In contrast, both the mechanisms underlying appetite in cnidarians and the role of cnidarian peptides in feeding regulation remain largely unknown (8).

We used the jellyfish Cladonema pacificum (9) (Fig. 1A) to address molecular mechanisms of satiety in a cnidarian. Cladonema derives its name from highly branched and specialized tentacles decorated with bead-like sting cell batteries (10) (Fig. 1A). Our initial observations identified an elaborate, coordinated behavioral sequence of Cladonema feeding, from prey capture to ingestion (Movie S1). Satiety could influence each behavioral step of this sequence. Moreover, Cladonema stocks can be maintained in the laboratory using brine shrimp larvae as food, allowing strict control of its feeding state.

Fig. 1.

Starvation controls food consumption in the jellyfish Cladonema. (A) The jellyfish C. pacificum. (B) Starvation dependency of number of brine shrimps eaten for differentially starved jellyfish. Groups with statistically significant differences have no overlapping letters (Kruskal–Wallis test; Dunn’s posttest; P < 0.05). n = 12 jellyfish per group.

To identify genes regulating feeding in Cladonema, we carried out transcriptomics in starved and fed jellyfish. These analyses revealed that many differentially expressed genes (DEGs) upon feeding state change. By combined behavioral and anatomical characterization, we identified a secreted peptide that signals satiety. Strikingly, expression of this jellyfish peptide in Drosophila and the administration of a Drosophila homolog to the jellyfish revealed complete functional complementation in feeding regulation. Our results thus argue that feeding-regulating peptides have ancient origins dating back to the last cnidarian–bilaterian common ancestor.

Results

Feeding Regulation in Jellyfish and Associated Genes.

To understand Cladonema feeding regulation, we measured food intake by offering brine shrimps to differentially starved jellyfish across 10 rounds (SI Appendix, Fig. S1A). At the end of the experiment, consumption by starved animals was reduced to as much as for fed ones at the onset (SI Appendix, Fig. S1B), indicating that the protocol induced full suppression of appetite. As expected, total consumption increased with starvation time (Fig. 1B), and these parameters correlated well (logarithmic regression, R² = 0.92).

To identify molecules regulating appetite in jellyfish, we compared transcriptomes in animals that were starved or recently fed. We prepared mRNA separately from the "ring" and the manubrium (Fig. 2A). The ring included the margin of the bell that contains the neuron-rich ring nerve and the bases of the tentacles. The manubrium comprised the mouth and the gut. To eliminate the possibility of sample contamination by transcripts of brine shrimps, we separately analyzed their transcriptome and subtracted it from the jellyfish transcriptome in silico.

Fig. 2.

Identification of feeding-regulating peptides in Cladonema. (A) Schematic of samples prepared for transcriptomics. (B) Principal component analysis of gene expression. Fed and starved samples are indicated in black and white, and ring and manubrium samples are shown as circles and squares, respectively. (C) Circle chart summarizing families of gene ontology (GO) terms significantly altered upon feeding state change in ring samples. The percentages represented by each category are shown. (D) Scatterplot of nonredundant GO terms significantly altered upon feeding state change in ring samples. The P values are color-coded. Bubble size indicates GO term generality. GO terms related to neuronal processes are highlighted (black). (E) Peptide effect on Cladonema food consumption. Control, peptides that had no effect (Kruskal–Wallis test; Dunn’s posttest; P > 0.05), and peptides that caused a statistically significant reduction (**P < 0.01) are shown in white, gray, and black, respectively. n = 8 to 123 jellyfish per group.

Principal component analysis (PCA) of transcript expression levels clearly separated body parts (ring vs. manubrium) by the first principal component (Fig. 2B). Interestingly, the second principal component represented feeding states in the ring (fed vs. starved; Fig. 2B). Consistently, we found 4,696 and 444 DEGs in the ring and the manubrium, respectively. Gene ontology (GO) term enrichment analysis revealed that DEGs spanned diverse families, including metabolic processes (Fig. 2C and SI Appendix, Table S1). To obtain a more concise overview, we used reduce and visualize GO (REVIGO) to reduce semantic redundancy in GO terms. In the ring, this analysis highlighted several GO terms involved in regulation of ion, acetylcholine, action potential, and exocytosis (Fig. 2D), suggesting association of neuronal activity changes with feeding states. No GO terms were enriched in the manubrium, presumably due to a lower number of DEGs.

The majority of cnidarian neuropeptides and G protein–coupled receptors (GPCRs) are thought to lack bilaterian orthologs due to low sequence similarity (11, 12). Therefore, we ad hoc annotated putative peptide-encoding genes based on combinations of prepropeptide processing motifs. Indeed, we found 29 differentially expressed transcripts encoding putative amidated peptides (Dataset S1). Likewise, our computational predictions revealed 583 contigs encoding Family 1 GPCRs in the Cladonema transcriptome, but as previously reported (11), the majority of these GPCRs were unrelated to bilaterian GPCRs, with a few notable exceptions (SI Appendix, Fig. S2). A fraction of these Cladonema GPCRs, such as the one similar to human orexin receptors, were differentially expressed (SI Appendix, Fig. S2), suggesting involvement of neuropeptide systems in appetite regulation.

We synthesized 43 putative mature peptides whose expression is regulated by the feeding state and screened them for their ability to modulate food intake using a simplified consumption assay (Materials and Methods). Of 43 peptides, five significantly suppressed food intake (Fig. 2E). In line with this effect, four of five peptides were up-regulated in the fed state (SI Appendix, Fig. S3). Therefore, they may function as endogenous appetite-reducing signals in Cladonema. We chose to focus on (N)GPPGLWamide (referred to as GLWa hereafter) as it is a potent suppressor of feeding (Fig. 2E), has characteristic repeats in its precursor (Dataset S1), and is widely conserved in cnidarians (13).

GLWa Is a Feeding-Suppressing Peptide in Cladonema and Inhibits Tentacle Contraction.

GLWa is a group of secreted peptides that terminate with the sequence motif of Gly-Leu-amidated Trp. To better characterize its effect on feeding, we manipulated key properties of the mature peptide. Administration of 10 μM GLWa decreased overall feeding by at least 69%, and a significant reduction was also detectable at a further dilution at 1 μM (Fig. 3A). We found that peptide length was also important: N-terminal truncation to hepta- or pentapeptides still suppressed feeding, but the tetrapeptide had no significant effect (Fig. 3B). Moreover, changing the terminal amidated residue from tryptophan to alanine reduced or abolished feeding suppression (Fig. 3C).

Fig. 3.

Characterization of GLWamide-induced feeding suppression. (A) Dose–response profiles of the effect of GLWamides on Cladonema feeding (Kruskal–Wallis test; Dunn’s posttest; ***P < 0.001; **P < 0.01). From Left to Right, n = 10, 9, 9, 9, and 9 (GPPGLWa) or n = 15, 15, 14, 14, and 15 (NGPPGLWa) jellyfish per group. (B) Effect of peptide length on the GLWamide-induced feeding suppression (Kruskal–Wallis test; Dunn’s posttest; ***P < 0.0001; ns P > 0.99). From left to right, n = 54, 13, 23, 15, and 15 jellyfish per group. (C) Effect of the terminal amidated amino acid on the GLWamide-induced feeding suppression (Kruskal–Wallis test; Dunn’s posttest; ***P < 0.001; **P < 0.01; *P < 0.05; ns P > 0.05). From Left to Right, n = 39, 14, 16, 15, and 19 jellyfish per group. (D and E) Behavior schematics (Left) and Kaplan–Meier curves (Right) for cumulative fraction of prey paralysis (D) and tentacle contraction reflex (TCR, E) for starved (black), fed (black, dashed), and GLWamide-treated starved jellyfish (magenta) (log-rank (Mantel–Cox) test with Bonferroni correction; ***P < 0.001; ns P > 0.99). n = 156, 75, and 83 (C) or 151, 69, and 81 (D) brine shrimps for the ASW starved, ASW fed, and GLWa starved groups, respectively.

Cladonema feeding proceeds sequentially, with prey capture followed by a tentacle contraction reflex (TCR) and ingestion (Movie S1). To clarify the GLWa-regulated feeding step, we examined these behavioral steps to single brine shrimps. Fed jellyfish were less effective in paralyzing prey (Fig. 3D), potentially because prior capture had depleted sting cells, or satiety inhibited their discharge (14). GLWa-treated starved jellyfish were as effective as controls in prey capture (Fig. 3D). In stark contrast, we found that GLWa-treated jellyfish were impaired in TCR and to the same extent as fed animals (Fig. 3E).

GLWa could broadly inhibit tentacle movements. Therefore, we examined state dependency of GLWa by measuring the TCR onset. Stimulation of starved jellyfish with brine shrimp extract induced tentacle contraction and bending within 5 s, while the TCR onset was significantly delayed in fed animals (SI Appendix, Fig. S4). Application of GLWa abolished this feeding state–dependent delay (SI Appendix, Fig. S4B) with a significant interaction between feeding state and peptide treatment (two-way ANOVA; P = 0.0312).

To better characterize GLWa expression, we examined fluorescence immunohistochemistry using a monoclonal antibody (15) (Fig. 4 and SI Appendix, Fig. S5). Anti-GLWa signals were highly localized to subsets of neurons in various organs (Fig. 4A and SI Appendix, Fig. S5), confirming that Cladonema GLWa is a neuropeptide as in other cnidarians (15). We found a concentration of anti-GLWa signals in several dozens of large neurons at the tentacle bases, just distal to each ocellus (Fig. 4B). These connected with projections that extended to the ring nerve, radial channels, and manubrium (SI Appendix, Fig. S5). We confirmed that the cells in the tentacle bases were neurons by double-labeling with anti-tyrosinated tubulin (SI Appendix, Fig. S5 G–I), a common neuronal marker in cnidarians (8). Given the importance of GLWa in the TCR, we focused subsequent analyses on the tentacles.

Fig. 4.

Characterization of GLWamide expression. (A) Schematic showing anti-GLWamide signals in Cladonema (green). Dashed boxes include the tentacle base and ocellus (filled black circle) and represent the regions shown in panels B and F. (B) Anti-GLWamide signals in a tentacle base (green) with nuclear counterstaining (Hoechst 33258, magenta). Maximum intensity projection. (Scale bar, 100 μm.) (C) Feeding protocols tested for their effect on anti-GLWamide signal levels. (D) Representative images of anti-GLWamide signals in the tentacle bases under starvation (Upper) or 3 h after feeding (Lower). Maximum intensity projections. (Scale bar, 50 μm.) A heat map was used to color-code signal intensity. (E) Effect of feeding on the average voxel intensity of anti-GLWamide signals in the tentacle base (Kruskal–Wallis test; Dunn’s posttest; *P < 0.05). From Left to Right, n = 22, 23, 23, 20, and 14 jellyfish per group. (F and G) Double-labeling of anti-GLWamide (green) and muscle fibers (phalloidin, cyan) at the base of the tentacle. (G) is a higher-magnification image from F (white box). Arrowheads indicate processes of anti-GLWamide-positive neurons. Maximum intensity projection (F) or partial projection (G). (Scale bars, 50 μm and 15 μm, respectively.) (H) Schematic showing the relationship between GLWamide neurons (green) and longitudinal muscles (cyan).

To test whether the feeding state alters GLWa peptide levels, we measured anti-GLWa signals in the tentacle bases at different time points upon feeding (Fig. 4C). Quantification of the average voxel fluorescence intensity revealed significant increases 3 and 6 h after feeding initiation (Fig. 4 D and E). These results are consistent with our transcriptomic analyses (SI Appendix, Fig. S3D). We propose that feeding induces GLWa expression in the tentacles, and this in turn inhibits TCR.

High-magnification analysis of anti-GLWa signals at the tentacle base revealed direct and selective innervation to the longitudinal muscles of the tentacle (Fig. 4 F–H). In contrast, the circular muscles were devoid of anti-GLWa signals. Therefore, GLWa is likely expressed in a subset of motor neurons. As some members of the Wamide family regulate muscle activity (16), our anatomical and behavioral results strongly suggest that GLWa upon food consumption suppresses further feeding by directly inhibiting longitudinal muscle contraction for the TCR.

GLWamide and Myoinhibitory Peptide (MIP) Are Functionally Interchangeable.

The reported bilaterian homolog of GLWa is MIP (16), but their sequences differ significantly except for the amidated tryptophan in the mature peptides (Fig. 5A and SI Appendix, Fig. S6). MIP is best characterized in arthropods, where it is also known as Allatostatin B (17). Therefore, we examined the cross-reactivity of the anti-GLWa antibody to Drosophila MIP in the nervous system. Immunohistochemistry of the fly brain revealed a pattern strikingly similar to MIP (Fig. 5B) (18). Colocalization of immunoreactive cells with Mip-GAL4 corroborated the cross-reactivity (SI Appendix, Fig. S7). The amidated tryptophan alone is not a sufficient epitope for this antibody as we did not detect any characteristic patterns of other peptides with amidated tryptophan, such as short neuropeptide F (19). Moreover, signals were abolished in the brain of Mip^SK4, a null mutant (20), confirming the specificity of the cross-reaction to MIP in Drosophila (Fig. 5C). GLWa and MIP thus share sufficient structural similarity to be recognized by the same antibody, suggesting that GLWa represents an ancestral appetite-inhibiting signal that is conserved in the fruit fly.

Fig. 5.

Structural and functional relationship of GLWamide and MIP. (A) Diagrams of the preproneuropeptide sequence of GLWamide and MIP. Signal peptides are in black. Mature peptides are shown in magenta and yellow, respectively. (B and C) Anti-GLWamide staining in Drosophila brains of a wild-type animal (yw; B) and a Mip null mutant (Mip^SK4; C). Maximum intensity projections. (Scale bar, 100 μm.) A heat map was used to color-code signal intensity. (D) Schematic of the function of GLWamide and MIP in suppressing feeding in Cladonema and Drosophila, respectively. Hypothesized functional interchangeability is indicated by the diagonal inhibitory arrows. Figure panels showing the corresponding results are indicated. (E) Effect of Drosophila MIPs on Cladonema feeding (Kruskal–Wallis test; Dunn’s posttest; ***P < 0.001; *P < 0.05). ASW are artificial sea water–treated jellyfish (controls). From Left to Right, n = 90, 12, 12, 12, 12, 12, and 13 jellyfish per group.

To test this hypothesis, we examined the functional interchangeability of MIP and GLWa (Fig. 5D). We first tested if Drosophila MIPs can inhibit brine shrimp feeding in Cladonema. Despite the sequence differences including the conserved N-terminal tryptophan in MIP, two of five putative mature MIPs significantly suppressed jellyfish feeding to the same extent as Cladonema GLWa (Fig. 5E).

To test if the converse is true, we sought to rescue the hyperphagic phenotype of Drosophila Mip^SK4 mutants by functional complementation with GLWa. In fed wild-type flies, there was hardly any proboscis extension reflex (PER) to sucrose, but fed Mip mutants showed substantial PER (Fig. 6A) (18). For the Mip rescue, we generated three transgenic fly lines carrying variants of GLWa under the control of the Upstream Activation Sequence (UAS) (Fig. 6 B–D and Dataset S2). UAS-CpaGLWa is based on the jellyfish preproneuropeptide sequence to produce Cladonema GLWa (i.e., NGPPGLWa). DmelGLWa is a hybrid based on the MIP preproneuropeptide sequence but with all five mature MIPs replaced with GLWa (i.e., NGPPGLWa). Because MIPs contain a fully conserved N-terminal tryptophan (16) that likely conveys stability (21), we also constructed DmelW-GLWa, producing MIP–GLWa chimeric mature peptides from the Mip gene (e.g., AWNGPPGLWa). We also generated a control strain that drives expression of an “empty” MIP preproneuropeptide variant lacking all mature peptides (UAS-emptyMIP). Using Mip-GAL4 (SI Appendix, Fig. S7), we induced expression of these GLWa variants in Mip^SK4 mutants and confirmed successful expression in the brain using the anti-GLWa antibody (Fig. 6 B–D). As expected, no signals were detected with the expression of the empty Mip gene. Strikingly, driving Cladonema GLWa expression completely rescued the elevated PER in the Mip null mutant, restoring it fully to control levels (Fig. 6E). Similarly, the chimeric variant DmelW-GLWa was able to fully rescue the defect, whereas expressing DmelGLWa only showed a tendency (Fig. 6E). The behavioral results of these rescue experiments correlated well with GLWa expression levels in the brain (Fig. 6 B–D). Indeed, even DmelGLWa achieved full rescue when we boosted expression by raising the flies at a higher temperature (SI Appendix, Fig. S8). The most straightforward interpretation is that Cladonema GLWa is sufficiently similar to MIP to bind to its receptor(s). Taken together, our results (Figs. 5 and 6) clearly show that MIP and GLWa are functionally interchangeable, strongly arguing for a conserved role in feeding regulation in flies and jellyfish.

Fig. 6.

GLWamide rescues the hyperphagic phenotype of Drosophila Mip mutants. (A) Proboscis extension reflex (PER) probability of fed flies (controls and Mip mutants) in response to labellar stimulation with a 10% w/v sucrose solution (Fisher’s exact test; ***P < 0.0001). n = 30 flies per group. (B–D) Diagrams of preproneuropeptide sequences of generated GLWamide variants and their expression in brains of Drosophila Mip null mutants, as determined by anti-GLWamide staining. Maximum intensity projections. (Scale bar, 100 μm.) A heat map was used to color-code signal intensity. (E) Effect of GLWamide expression on the PER phenotype of the Mip null mutant. GLWamide groups are labeled with ‘a’ if they have statistically significant differences from the empty peptide group but not the wild type–like control group or ‘b’ if the converse is true (chi-square test with Bonferroni correction; P < 0.05). n = 50 flies per group. Fly genotypes are as follows: w; Mip-GAL4/+ (wild-type control), w; Mip-GAL4/+; Mip^SK4 (Mip mutant), w; Mip-GAL4/UAS-emptyMIP; Mip^SK4 (empty peptide), w; Mip-GAL4/UAS-CpaGLWa; Mip^SK4 (Cladonema GLWa), w; Mip-GAL4/UAS-DmelW-GLWa; Mip^SK4 (Drosophila W-GLWa), and w; Mip-GAL4/UAS-DmelGLWa; Mip^SK4 (Drosophila GLWa).

Wamides May Predate Cnidarians.

The structural similarities and functional complementation between cnidarian GLWa and arthropod MIP prompted us to search for similar peptides in other species. We therefore compiled published works and mined databases to identify or predict Wamides in a diverse set of organisms (SI Appendix, Table S2 and Fig. S9). With the exception of Octocorallia, a class of the phylum, all cnidarians encode GLWas or close variants. In addition to the well-conserved GLWa sequence, the N termini of these peptides also have strong similarities, such as proline and/or glutamine residues at the ultimate and/or penultimate position(s). Beyond cnidarians, recent studies show that placozoans (22) and ctenophores (23, 24) have Wamides, although their phylogenetic relationship to GLWa–MIP remains unclear (16). Interestingly, we also found putative precursors for RWamide in the sponges Amphimedon and Ephydatia, and even the choanoflagellate Salpingoeca may encode a GNWamide. These findings strongly suggest that Wamides existed prior to the divergence of Cnidaria, while these candidate peptides, beyond sequence motifs, need to be characterized and validated.

Discussion

Here, we show that feeding induces the expression of the neuropeptide GLWa in Cladonema (Fig. 4E and SI Appendix, Fig. S3D), leading to a reduction of ingestion (Figs. 2E and 3 A–C). Given the state-dependent TCR suppression (Fig. 3E and SI Appendix, Fig. S4B), the simplest interpretation of our results is that GLWa is a satiety signal in Cladonema. Internal states like satiety influence multiple behaviors and physiological processes (25). As GLWa controls selective steps of the feeding sequence (Fig. 3 D and E), satiety in Cladonema likely involves additional peptidergic systems as is the case for bilaterians (2). Similarly, in the jellyfish Clytia, the neuropeptide RFamide was reported to promote feeding by driving umbrella margin folding (8). Verification and further characterization of peptides, especially those responding to feeding states (Fig. 2E), would clarify the entire network of feeding-regulating molecules.

GLWa and similar peptides are widespread in cnidarians (SI Appendix, Fig. S9) (13). Beyond the GLWa sequence, the N termini of these peptides share additional similarities (SI Appendix, Fig. S9). While the precise N-terminal structure of Cladonema GLWa awaits biochemical determination, these motifs may be functionally critical, given the loss of feeding suppression with an N-terminally truncated GLWa variant (Fig. 4B). Consistently, those MIPs that suppressed jellyfish feeding share a glutamine near the N-terminal (Fig. 5E and SI Appendix, Fig. S6C) similar to other cnidarian GLWa peptides (SI Appendix, Fig. S9). Similarly, structural conservation of the N-terminal tryptophan is a hallmark of MIP (16) and may increase stability (21). Consistently, W-GLWa, a chimeric variant of MIP–GLWa was more effective in both peptide expression and the feeding rescue (Fig. 6E). Interestingly, one of the mature GLWa peptides in Hydra contains tryptophan residues at both termini, possibly representing a GLWa–MIP intermediate (16).

Although convergence should not be formally excluded, the most parsimonious explanation for the shared functions and the structural similarities of GLWa–MIP is a common evolutionary origin. Consistently, GLWa–MIP orthology was previously proposed mainly based on amino acid sequences of prepropeptides and mature peptides (16). The cnidarian GLWa receptor is unknown (16), and like others (11), we found that most candidate cnidarian neuropeptide receptors likely lack bilaterian orthologs (SI Appendix, Fig. S2). Nevertheless, identification of cognate receptor sequences could refine the evolutionary relationship of MIP–GLWa, especially since cnidarian neuropeptides can signal through atypical receptors (26).

Although no deuterostome GLWa–MIP homologs have been reported to date (16), anti-GLWa immunoreactivity was detected in the rat hypothalamus, a feeding center, suggesting functional counterparts in mammals (27). Sequence comparisons of Wamides revealed broad occurrence in metazoans (SI Appendix, Fig. S9). Even if the relationship to GLWa needs to be reinforced, a GWamide-encoding gene in Mnemiopsis is expressed in the mouth and pharynx (23). In line with this, recent peptidomic analysis showed that NPWamide is expressed in neurons in the oral region of Bolinopsis larvae and regulates mouth opening in this ctenophore (24). Moreover, the choanoflagellate Salpingoeca may encode a Wamide (SI Appendix, Fig. S9) in addition to other recently reported peptides (28). Collectively, these results imply early origins to the feeding-regulating functions of Wamides that could predate metazoans.

Are there similarities in the mechanisms underlying feeding inhibition by MIP and GLWa? In arthropods, MIP can directly inhibit sensory systems (29) and muscles (30). In Cladonema, GLWa-mediated sensory inhibition is unlikely given the lack of expression in the distal tentacles, where chemosensory pathways exist (Fig. 4 and SI Appendix, Fig. S5F) (31). Inhibition of gut muscles has been proposed to drive food accumulation in the foregut and in turn promote satiety by activation of stretch receptors (32). Considering muscle projections of the GLWa neurons and the myoinhibitory function of MIP in insects (30), the most straightforward mechanism is that the GLWa family inhibits feeding-related muscles (Fig. 4). Identification and localization of the GLWa receptor would clarify this hypothesis and define the cells capable of receiving GLWa signals. Satiety and locomotion arrest are closely associated, such as in postprandial quiescence. As cnidarians exhibit sleep-like states (33, 34), it would be interesting to test whether GLWa is involved in the satiety-induced quiescence by muscle relaxation.

In addition to feeding control, GLWa–MIP shares another common function: the regulation of life-phase transitions (35). In particular, both GLWa and MIP control larval settlement and/or metamorphosis. Similar pleiotropy in muscle, feeding, and life transition regulation is well documented for other peptides as well (17). It is therefore tempting to speculate that these functions were coupled from early on in animal evolution as food availability is crucial for determining growth vs. stasis across the entire animal kingdom (36).

Materials and Methods

Jellyfish Cultures.

The UN2 strain of C. pacificum (37) was used. Jellyfish were maintained throughout the year at 20 °C under a 12-h:12-h light–dark cycle in filtered artificial sea water (ASW) (Gex, Japan). Eggs of the brine shrimp Artemia salina (Tetra, Japan) were hatched in ASW at 30.5 °C overnight and used as jellyfish food for both stock maintenance and experiments. Jellyfish were fed 5 d a week. After at least 1 h of feeding, they were transferred to fresh ASW. Males aged 3 to 6 wk from budding were used for the behavioral and transcriptome experiments. For anatomy, males aged 2 wk were used in all cases, except for whole jellyfish imaging, where 1-d-old males were used.

Fly Strains and Cultures.

All flies were kept at 25 °C and 60% relative humidity under a 14-h:10-h light–dark cycle unless otherwise specified. Standard cornmeal medium was used as fly food. The following strains of Drosophila melanogaster were used for experiments: y¹ w¹¹¹⁸ (38), y¹ w¹¹¹⁸; Mip^SK4 (20), w¹¹¹⁸; Mip-GAL4 (18), yw; 20x-UAS-IVS-mCD8::GFP (RRID: BDSC_32194), and w¹¹¹⁸ (39). In addition, w¹¹¹⁸; UAS-emptyMIP, w¹¹¹⁸; UAS-CpaGLWa, w¹¹¹⁸; UAS-DmelGLWa and w¹¹¹⁸; UAS-DmelW-GLWa were generated in this study (see below).

Generation of UAS-Wamide Lines.

Four sequences were designed using Thermo Fisher’s GeneArt Instant Designer tool. emptyMIP, DmelGLWa, and DmelW-GLWa were based on the D. melanogaster sequence encoding MIP, but all mature peptides (MIP 1-5) and the C-terminal glycine were removed, replaced with the sequence NGPPGLWG, or replaced with MIP–GLWamide chimeric peptides (X_1-5WNGPPGLWG), respectively (Dataset S2). CpaGLWa was based on the C. pacificum sequence encoding GLWa, with a modification of the C-terminal cleavage sites to lysine–arginine (KR; Dataset S2). The GeneOptimizer algorithm (40) was used to maximize expression in Drosophila. The Drosophila Kozak sequence CAAA was included upstream of the start codon and restriction sites for the endonucleases AscI and NotI at the 5′ and 3′ ends, respectively, for vector integration. All sequences were synthesized using Invitrogen GeneArt Gene Synthesis and inserted into standard GeneArt delivery vector (pMA-RQ or pMK-RQ). The final constructs were sequence-verified.

Standard molecular biology procedures were used to generate UAS lines. Sequences were ligated into pBFV-UAS3, a custom UAS vector (38) and amplified. Fly transformation was carried out as described elsewhere (41), with vectors integrated into the attP40 landing site of the second chromosome (42).

RNA Extraction and Sequencing.

Fed (4 to 9 h postfeeding) and starved (48 to 58 h postfeeding) jellyfish were dissected in ASW prepared with RNase-free water (Thermo Fisher Scientific) with 40 U Protector RNase Inhibitor (Roche). First, the sting cell–rich tentacles were removed. Then, the bell margin, which included the tentacle bases (referred to as ring hereafter), and the manubrium were dissected and frozen on dry ice. Tissues from 10 to 20 jellyfish were pooled for each sample. Three samples for each condition (starved/fed ring/manubrium) were prepared. Three brine shrimp samples were separately prepared for "subtracting" the transcriptome of the prey (see below). To extract RNA, samples were defrosted; the dissection medium was quickly replaced with 500 μL ice-cold TRIzol™ Reagent (Thermo Fisher Scientific), and tissues were homogenized with BioMasher (Nippi, Japan) until debris was no longer observed. Homogenates were stored at −20 °C before RNA extraction. RNA extraction with TRIzol™ Reagent was conducted according to the manufacturer’s instructions. Total RNA from Artemia was extracted in the same way. The RNA concentration and quality were determined using the NanoDrop spectrophotometer (Thermo Fisher Scientific) and Agilent RNA 6000 Nano Kit on Agilent Bioanalyzer (Agilent Technologies). Total RNA samples were subject to library preparation using the Illumina^R TruSeq^R RNA Sample Preparation Kit v2 (Illumina, California, USA) according to the manufacturer’s protocol. These mRNA libraries were sequenced on Illumina HiSeq 4000 with 100-mer paired-end sequences at BGI Tech Solutions (Hong Kong) Co., Ltd.

Transcriptome Analysis.

128234026 sequence reads from a total of three Artemia libraries were obtained, trimmed, and filtered by the trimmomatic option of the Trinity program (43). De novo assembly using the Trinity program yielded 188,588 contigs. 504557388 sequence reads from 12 Cladonema libraries were mapped against Artemia transcript sequences using Bowtie 2 (44) (ver. 2.3.2) and resulted in mapped and unmapped data. The latter (228232332 reads), presumably consisting of Cladonema sequence data, were de novo assembled as above, yielding 250,980 contigs. Values of read count data of each gene were calculated using RNA-Seq by Expectation Maximization (RSEM) (45) with the default setting implemented in Trinity. Values of transcripts per kilobase million (TPM) (46) were calculated manually. To obtain a visual overview of the effect of feeding states and types of tissues (ring or manubrium) on global gene expression patterns, PCA was performed on the calculated TPM values using the function "prcomp" in R (47). Genes expressed in at least one sample were included in the PCA.

GO Term Enrichment Analysis.

GO term enrichment analysis was performed using the goseq package in R (48). To annotate each Cladonema gene based on orthology and the associated GO terms, Basic Local Alignment Search Tool (BLASTX) searches (49) against the database of Hydra vulgaris (Hydra 2.0 genome; https://research.nhgri.nih.gov/hydra/download/?dl=asl) as reference were performed with the E value cutoff 10⁻⁴. By taking the top hit of the BLASTX searches for all of the Cladonema transcript sequences used as query, 77,756 orthologs were obtained. The GO term analysis was summarized and visualized with the REVIGO web server (50). GO terms categorized as biological processes were used with default settings.

GPCR Prediction and Comparisons with Bilaterians.

Protein sequences of all Cladonema contigs were predicted using TransDecoder (51), and these were annotated using InterProScan 5.52-86.0 (European Bioinformatics Institute (EMBL-EBI)) and the Pfam database. Predicted proteins sharing a similarity of 95% or greater were combined into a representative sequence with Cd-hit (52). Of these, we selected those that were annotated as "GPCR Family 1" (PF00001) by the Protein families (Pfam) database. These sequences were added to a subset of a previously published dataset containing bilaterian GPCRs (53), and cluster analysis was carried out with CLANS (54).

Peptide Prediction.

To identify putative peptides encoded by differentially expressed transcripts, we selected predicted protein sequences that included a) a signal peptide, as determined by SignalP 4.1 (55) (cutoff D > 0.40), b) internal diresidue cleavage sites that were either dibasic (arginine/lysine) or basic–acidic (arginine or lysine followed by aspartic or glutamic acid), and c) a C-terminal glycine before the cleavage site as this is necessary for peptide amidation. As N-terminal cleavage sites are difficult to predict (56), we tested amidated peptides of seven amino acids or shorter in the behavioral experiments.

Behavioral Experiments—Jellyfish.

For all behavioral experiments, jellyfish were starved for 23 to 27 h unless otherwise specified. Jellyfish indicated as “fed” (Fig. 4 D and E) were tested 0 to 1 h after feeding. Peptides were synthesized by GenScript, dissolved in distilled water at 10⁻² M, stored at −20 °C, and diluted in ASW prior to usage. They were bath applied from at least 2 h prior to the experiment at a concentration of 10 μM unless otherwise specified. All experiments were carried out in 24-well plates at 24 °C under a Nikon SMZ745 microscope with dark-field illumination (Nikon P-DF LED Dark Field Unit).

Round-by-round feeding assay.

Ten wells, each containing 10 brine shrimps in approximately 1 mL ASW, were prepared for each jellyfish. Initially, an individual jellyfish was transferred into the first well using a pipette. Care was taken not to damage the fragile Cladonema tentacles. After 15 min, the noningested brine shrimps were counted to calculate the number of consumed brine shrimps. Subsequently, only the jellyfish was transferred into the next well. This procedure was repeated for 10 times (rounds). Finally, the total number of brine shrimps eaten was calculated for each group, and comparisons of group medians were performed.

Massed feeding assay.

For peptide screening, 100 brine shrimps were prepared in a single well, and a single jellyfish was allowed to feed for 3 h. These experiments were otherwise performed as the round-by-round assay above.

Video analysis of feeding behaviors.

Each jellyfish was offered an individual brine shrimp. Behavior was video-recorded from above through a binocular microscope equipped with a Nikon 1 J4 camera. Following prey capture, 2 min were offered for ingestion. The procedure was repeated for 5 times consecutively to each jellyfish. The time points of prey capture, prey paralysis, and tentacle contraction were determined by inspection of videos.

Stimulation with brine shrimp extract.

The extract was prepared by freezing a living brine shrimp suspension at −80 °C overnight. It was then thawed, ASW was added to adjust dead brine shrimp density to 200 shrimps/mL, and filtered. Jellyfish were prepared in 1 mL ASW and stimulated with 1 mL of extract (final concentration: 100 shrimps/mL). For the GLWa-treated groups, peptide (NGPPGLWa) was present in both ASW and stimulus. Experiments were video-recorded as above, and the time from stimulus application to tentacle responses (contraction and bending) was determined by inspection of videos.

PER assay.

Experiments were performed at 24 °C under a Nikon SMZ745 stereoscopic microscope, with 5- to 13-d-old female flies. Culture density was controlled as Mip mutant phenotypes are sensitive to this parameter (18). Specifically, 50 eggs per bottle (height 9 cm, diameter 5 cm, and food 20 mL) and 15 female adult progeny per vial (height 9 cm, diameter 2.4 cm, and food 6.5 mL) were collected. Flies were satiated (i.e., on food prior to the experiment) in all cases. PER was measured as described elsewhere (57). Briefly, flies were first offered water and allowed to drink ad libitum if necessary. They were then stimulated with a 10% w/v sucrose solution. PER was categorized as “none,” “partial,” or “full”.

Immunohistochemistry.

Jellyfish.

All steps were performed at room temperature except the incubation with the primary antiserum (4 °C). Jellyfish were anesthetized in 7% MgCl₂ in ASW for 10 min and fixed with 4% paraformaldehyde in ASW for 1 h. Samples were washed with phosphate-buffered saline (PBS) with 0.1% Triton X-100 (PBS-Tx) and blocked with 3% serum of goat in PBS for 1 h. Fixed animals were separated into Eppendorf tubes (2 jellyfish/tube) and stained using a mouse monoclonal antibody (produced from the hybridoma clone 4B10) to GLWamide (1:100) overnight. The antigen used for raising this antibody was keyhole limpet hemocyanin–succinimidyl m-maleimidebenzoate-cysteinyl–GLWamide (15). For detection of the primary antibody, we used Alexa 488–tagged goat anti-mouse IgG (1:500, Invitrogen, catalog #A-11001, RRID: AB_2534069) with Hoechst 33258 (0.1%, Dojindo) and Phalloidin-Atto 565 (0.5%, Sigma-Aldrich) for counterstaining of nuclei and actin filaments, respectively. For the double-labeling with anti-GLWamide and anti-tyrosinated tubulin, a similar procedure was used, but antibodies were applied sequentially (58). Specifically, mouse anti-GLWamide was added first (overnight) followed by the secondary antibody as above. Then, rat anti-tyrosinated tubulin (1:100, Bio-Rad, product code MCA77G, RRID: AB_325003) was added (overnight) and detected with Cy3-tagged goat anti-rat IgG (1:500, Jackson ImmunoResearch, code #112-166-003, RRID: AB_2338253) with Hoechst 33258.

Drosophila brains.

Brains of 1- to 4-d-old adult female Drosophila were dissected as described (59), fixed (2% formaldehyde in 0.1% PBS-Tx), washed with 0.1% PBS-Tx, and stained with the anti-GLWamide antibody (15) (1:200). For detection of the primary antibody, Cy3-tagged goat anti-mouse IgG (1:200, Jackson ImmunoResearch, product #115-166-003, RRID: AB_2338699) was used.

Imaging and processing.

Samples were mounted in 86% glycerol and imaged with an Olympus FV-1200 confocal microscope. A 10×/0.40 dry objective (UPLSAPO10X, Olympus) (SI Appendix, Fig. S5B), a 20×/0.85 oil immersion objective (UPLSAPO20XO, Olympus) (Figs. 5 and 6 and SI Appendix, Figs. S5E, S7, and S8 A and B), a 40×/1.3 oil immersion objective (UPLFLN40XO, Olympus) (Fig. 4B), or a 60×/1.42 oil immersion objective (PLAPON60XO, Olympus) (Fig. 4 D and F and SI Appendix, Figs S5 C, D, F, and G) was used for scanning samples. Scan settings were kept identical for the signal quantification experiments (Fig. 4E) and for the comparison of different fly genotypes (Figs. 5 and 6) or different temperatures (SI Appendix, Fig. S8 A and B). For the former, starved (48 h) and fed (Fig. 4C) jellyfish were used. A tentacle base for each jellyfish was fully scanned. An XY area of 212 μm × 212 μm centered around the jellyfish ocellus was selected for each sample. The step size along the Z plane was 1 μm. Images were processed using Fiji software (60). Signal quantification was performed by applying an appropriate threshold (350) on all slices of the 16-bit stacks and calculating the average voxel intensity using the Measure function in Fiji.

Compilation of Wamide Sequences.

Several sequences of mature Wamides (SI Appendix, Fig. S9) were compiled from the published literature as follows: bilaterians and Nematostella vectensis (16); Actinia equina, Anemonia sulcata, and Hydractinia echinata (61); Anthopleura elegantissima (62); Octocorallia (63); Staurozoa (63); Cubozoa (63, 64); Scyphozoa (63); Clytia hemisphaerica and C. pacificum (37); Hydra magnipapillata (65); Trichoplax adhaerens (22); and Mnemiopsis leidyi (23).

Wamides from Actinia tenebrosa, Acropora millepora, and Stylophora pistillata were obtained from Uniprot (accession numbers A0A6P8I7G4, G8HTA1, and A0A2B4STK2, respectively). Wamides from Pocillopora damicornis and Orbicella faveolata were identified with protein BLAST in these species by using the query sequence GLWG and identifying the most likely preproneuropeptide candidates among the hits (accession numbers XP_027060553 and XP_020623100, respectively). The remaining Wamides were identified from each species’ predicted proteins by sequential application of SignalP 4.1 (55) (selection criterion: D > 0.40), InterProScan 5.52-86.0 (EMBL-EBI) (selection criterion: no Pfam, Simple Modular Architecture Research Tool (SMART), and The Institute for Genomic Research Family (TIGRFAM) domain entries), selecting sequences containing the amino acid motifs WGK and/or WGR, and finally applying NeuroPID (66) (selection criterion: full or high confidence sequences). Species and accession numbers of corresponding Wamides are as follows: Exaiptasia diaphana (XP_020893119), Acropora digitifera (XP_015772272), Hoilungia hongkongensis (67) (Braker ID: braker1_g04334.t1), Pleurobrachia bachei (68) (sb|3477507|), Amphimedon queenslandica (XP_019850579 and XP_019862527), Xestospongia testudinaria (69) (maker-XT_scaffold95330-augustus-gene-0.9-mRNA-1 protein AED:0.12 eAED:0.12 QI:0|0|0|0.5|1|1|2|0|125), and Salpingoeca rosetta (XP_004988246). In cases where the mature peptide was uncertain, putative sequences were predicted by requiring them to contain an ultimate glutamine or penultimate proline at the N terminus (where possible), a common motif in cnidarian peptides (56).

Statistics.

All statistical tests were two tailed, where applicable. To define DEGs, the Benjamini–Hochberg method (70) was applied on p values produced by edgeR (71) to perform multiplicity correction for controlling the false discovery rates (FDRs) or q values. FDRs were calculated by edgeR implemented in the R package TCC (72) using 0.05 as a cutoff. For the GO term analysis (Fig. 2 C and D), overrepresented p values produced by GOseq (48) were adjusted using the Benjamini–Hochberg correction (70). GO terms with adjusted P or q values of less than 0.05 were defined as enriched.

For behavioral data, medians are presented as box-and-whisker plots unless otherwise noted. The box and whiskers represent the interquartile range and the minimum/maximum, respectively. Data were evaluated with Prism 6 (GraphPad, San Diego, CA). As most data were not normally distributed and/or had unequal variances, we used the Kruskal–Wallis test and Dunn-corrected pairwise comparisons. Data from the video recordings of feeding (Fig. 3 D and E) were evaluated with Kaplan–Meier survival analysis employing a log-rank (Mantel–Cox) test with Bonferroni correction for pairwise comparisons. The categorical proboscis extension data (Fig. 6 A–E and SI Appendix, Fig. S8) were tested with the chi-square test. Significance levels are as follows: ns P > 0.05, *P < 0.05, **P < 0.01, and ***P < 0.001.

Supplementary Material

Appendix 01 (PDF)

Dataset S01 (PDF)

Dataset S02 (PDF)

Movie S1.

Feeding behaviour of Cladonema jellyfish. Capture of a free-swimming brine shrimp by a starved jellyfish, followed by the full sequence of jellyfish feeding steps.

Data, Materials, and Software Availability

Original data underlying this study have been deposited to G-Node (German Neuroinformatics Node, doi: 10.12751/g-node.b7v65l). Sequence data have been deposited to the DNA Data Bank of Japan (DRA accession numbers: DRA013209 to DRA013223) (73, 74).