Control of water excretion in Hydra vulgaris by neural integration

Summary

Hydra vulgaris is one of the few cnidarian species that lives in freshwater environments. To understand this adaptation, we studied Hydra’s mechanisms of osmoregulation. Behavioral imaging showed that Hydra accumulates water in its gastric cavity over time and periodically excretes it through the mouth. Comparative genetic analysis revealed unique aquaporin water-channel expression in Hydra’s endodermal epithelium, where ultrastructural data demonstrated small clear vesicles near the gastric cavity, suggesting a potential water-release pathway. We further found that endodermal RP2 neurons increase their activity before water excretion, until a threshold activity level is reached, when the mouth opens, and their activity abruptly declines. The ramping of RP2 activity, well modeled by spike-count and leaky-integration algorithms with a 30 second integration time window, leads to the specific activation of epithelial muscle cells near the mouth region before mouth opening. Confirming a causal role of RP2 in the behavior, two-photon activation of RP2 neurons induces water excretion, while ablating RP2 neurons alters it. Consistent with this, GLWamide peptides, synthesized by RP2 neurons, induce water excretion. We conclude that activation of RP2 neurons and subsequent release of GLWamide family peptides promote water excretion, and propose a circuit model for the temporal integration and sequential unfolding of this osmoregulatory cycle. Our work demonstrates that neural integration algorithms and peptide-based signaling can be used by simple nervous systems to coordinate a behavioral and physiological program.

Graphical Abstract

eTOC Blurb:

Hydra is one of the few cnidarians adapted to freshwater. Yamamoto and Yuste show that, to survive, it periodically expels accumulated gastric water. An endodermal neural ensemble integrates its activity until it reaches a threshold that triggers excretion, via GLWamide signaling.

Introduction

Cnidarians are an ancient animal phylum that evolved during the Ediacaran, approximately 550M years ago.^1,2 The vast majority of the more than 11,000 known cnidarian species live in marine environments,^3,4 while some cnidarians, such as the anthozoan Nematostella vectensis and the hydrozoan Cordylophora, tolerate broad salinity ranges and can inhabit brackish or low-salinity environments.^5,6 In striking contrast, a small number of cnidarians,^7,8 represented primarily by the genus Hydra, have adapted to life in freshwater habitats. This transition is physiologically challenging because freshwater imposes a constant inward osmotic force, driving water into animal tissues and body cavities. Despite this challenge, Hydra has become one of the most represented freshwater cnidarians, colonizing ponds, lakes, and slow-moving streams across broad geographic regions in the globe.⁹ How this simple animal, lacking excretory organs, solves the problem of continuous water influx remains poorly understood.

Freshwater adaptation requires mechanisms that collect, redistribute, and expel excess water. In unicellular eukaryotes, contractile vacuoles act as osmoregulatory organelles that accumulate excess water and periodically discharge it from the cell. Recent work suggests that contractile vacuoles can respond to pressure or membrane tension, allowing water release to be coupled to the physical state of the organelle.^10,11 In freshwater fish, excess water is handled by specialized epithelia and excretory organs with nephrons, which produce large volumes of dilute urine while retaining essential ions. In more complex bilaterians, osmoregulation is not only achieved by epithelial transport, but also a problem of coordination, requiring sensory, neural, and muscular mechanisms that detect internal fluid state and drive appropriate water excretion behaviors.

Hydra provides an amenable system to study how such coordination emerged in one of the simplest nervous systems. The animal has a simple tubular body plan with a single gastric cavity surrounded by two epithelial layers, the ectoderm and endoderm. These epithelial layers are excitable tissues, with epitheliomuscular cells and neurons, that actively participate in behavior. Because the osmotic gradient in freshwater favors water entry into the body, excess water entering cells, tissues, and the gastric cavity must ultimately be eliminated. Previous physiological and ultrastructural work proposed that water reaches the gastric cavity through epithelial vacuoles and intercellular spaces.¹² A key unresolved question is whether this process is controlled primarily by epithelial transport, by neural activity, or by an integrated epithelial-neural mechanism.

As it monitors the environment and generates body responses, Hydra’s nervous system is likely involved in water regulation. Although Hydra lacks a centralized brain, it contains a distributed nerve net composed of functionally distinct neuronal ensembles.^13,14 Previous studies identified at least three major ensembles, contraction burst neurons (CB), rhythmic potential 1 neurons (RP1), and rhythmic potential 2 neurons (RP2), which show coordinated and rhythmic activity across the body.¹⁴ These neurons use neuropeptides as transmitters, allowing them to coordinate distant epithelial and muscular targets through volume transmission-like signaling.^15–17 Single-cell and imaging studies further revealed that Hydra contains several molecularly distinct neuronal populations, suggesting that even a simple nerve net can support specialized physiological functions.^14,18–20 However, whether this simple nervous system regulates the fundamental freshwater challenge of water balance remains unknown.

Here, we use Hydra vulgaris to investigate how a freshwater cnidarian solves the problem of osmotic water influx, by combining long-term behavioral imaging, bioinformatics, machine vision, electron microscopy, calcium imaging, computational modelling, two-photon stimulation, neuronal ablation, and pharmacology. We identify a periodic water-excretion behavior in which Hydra accumulates fluid in the gastric cavity and releases it through the mouth. We find that this accumulation is associated with specialized aquaporin expression and endodermal epithelial organelles that may provide a pathway for water movement into the gastric cavity. We further show that water excretion is associated with the endodermal RP2 neuronal ensemble, whose activity precedes excretion until it reaches a specific threshold, when water excretion occurs and RP2 neuronal activity decays back to baseline. This integrate-to-threshold algorithm is captured by a leaky integrator model to trigger the behavior. Consistent with this mechanism, GLWamide-family neuropeptides, synthesized by the RP2 ensemble and which share a conserved C-terminal GLWamide motif previously linked to sphincter-like muscle contraction,^21,22 can induce water excretion. Together, our results support a model in which freshwater adaptation in Hydra depends on a simple but elegant epithelial-neural mechanism: epithelial cells route excess water into the gastric cavity, and a dedicated RP2 neuropeptidergic circuit triggers its periodic expulsion through the mouth. This work reveals how a simple nerve net can control an essential homeostatic behavior and suggests that neural regulation of internal fluid balance and computational algorithms for neural integration may have deep evolutionary origins.

Results

Hydra has limited osmolarity tolerance

In contrast to most cnidarians, which live in seawater, Hydra vulgaris inhabits freshwater environments, which likely pose a constant osmotic challenge (Figure 1A). To understand this, we first investigated its salt tolerance. To determine the osmotic range that Hydra can endure, we exposed animals to artificial seawater at osmolarities ranging from 0 to 1,000 mOsmol/kg, a broad range that reflects potential cnidarian habitats and vernal pools annual cycle. As expected for a freshwater species, we found that most animals died within 15 minutes after exposure to 300 to 1,000 mOsmol/kg seawater (Figure 1B). In contrast, all animals survived for 72 hours at osmolarities below 50 mOsmol/kg. However, animals exposed to 50 mOsmol/kg displayed visible tentacle damage, which likely compromises prey capture and long-term survival (Figure 1B; n = 30 animals per condition, pooled from N = 3 experiments).

Figure 1. Hydra accumulates water in gastric cavity and periodically excretes it.

(A) Only around 20 cnidarian species, mostly Hydra, have adapted from marine to freshwater habitats. (B) Survival curves showing Hydra’s poor tolerance to seawater (n = 30 animals per condition, pooled from N = 3 experiments). (C) Representative images before and after water excretion. (D) Left: fold change in body area aligned to onset of water excretion. Right: fold change in body area comparing the last three measurements during water excretion (mean ± SEM; N = 6 animals, n = 10 to 11 water excretion events; ratio paired two-tailed t test, ****p < 0.0001). (E) Water accumulation in the body. Representative images at 0 h, immediately after water excretion, and after 3 h, with no water excretion events. Images selected at a similar body length during elongation. (F) Left: representative body area trace for one animal during elongation. Right: body area at a matched body length of 4 mm, immediately after water excretion and after 2–3 h with no water excretion events (mean ± SEM; N = 5 animals, n = 5 accumulation sessions; paired two-tailed t test, **p = 0.0047). (G) Left: representative trace of body length change, with overlaid water excretion and somersaulting events. Top right: enlarged view of boxed region in the left panel. Bottom right: pie chart summarizing 88 water excretion events and their relationship to other behaviors (N = 19 animals). (H) Raster plots of water excretion events under control and high-osmolarity conditions (50 mM sucrose). Five representative animals shown for each condition. (I) Intervals between water excretion events (mean ± SEM; N = 6 animals, n = 16 intervals). (J) Histogram of observed control inter-excretion intervals and randomized inter-event intervals generated computationally (F test to compare variances, **p = 0.0014). (K) Number of water excretion events under control and high-osmolarity conditions (50 mM sucrose) (mean ± SEM; control, N = 6 animals; 50 mM sucrose, N = 9 animals; unpaired two-tailed Student’s t test with Welch’s correction, ***p = 0.0002).

See also Figures S1, S2, and S3 and Videos S1 and S2.

We obtained similar results using sucrose, an uncharged osmolyte that allowed us to test the effect of osmolarity independently of seawater ion composition (Figure S1; n = 30 animals per condition, pooled from N = 3 experiments). Together, these results confirmed that Hydra has limited tolerance to increased environmental osmolarity, consistent with its specialization for low-osmolarity freshwater habitats.

Hydra periodically accumulates water in the gastric cavity and excretes it

While imaging Hydra specimens in physiological low osmolarity conditions resembling freshwater, we noticed that animals occasionally undergo a rapid decrease in body size. This behavior likely reflects the accumulation of water in the gastric cavity, the single internal organ of Hydra that functions as a gut, followed by the release of excess fluid by the opening of a mouth-like oral aperture, as previously reported.²³ To directly monitor fluid movement from the gastric cavity, we injected a fluorescent freshwater solution into the gastric cavity and imaged freely behaving animals. We found that Hydra periodically released fluorescent fluid from the mouth, and this was always accompanied by a significant decrease in body area (Figures 1C and D; Video S1; geometric mean ratio = 0.68 ± 0.02; N = 6 animals, n = 10 to 11 excretion events each; ****p < 0.0001). We refer to this behavior as water excretion. The duration of this excretion event was 63.18 ± 7.87 seconds, and body length decreased by a ratio of 0.77 ± 0.03 (Figure S2B; N = 6 animals, n = 10 to 11 water excretion events; **p = 0.0022). This body area decrease was significantly smaller than that observed during whole-body contraction bursts, a well described spontaneous motor behavior in Hydra, in which body length decreased by a ratio of 0.24 ± 0.03 (Figure S2C; N = 3 animals, n = 6 contractions; ****p < 0.0001).

To quantify water dynamics in the gastric cavity over time, we compared the two-dimensional body area of the same animals immediately after water excretion and 2 to 3 hours later, before the next water excretion event (Figure 1E). We found that body area gradually increased from 1.05 ± 0.07 mm² to 1.17 ± 0.08 mm² before an excretion event (Figure 1F; N = 5 animals, n = 5 accumulation sessions; **p = 0.0047), indicating that Hydra gradually accumulates water in the gastric cavity between excretion events.

We explored whether water excretion was accompanied by other behavioral events. We found that 95.5% of water excretion events occurred in isolation, without overlapping with other behaviors such as whole-body contractions or somersaulting (Figure 1G; N = 19 animals, n = 88 water excretion events). Unlike a previous model that proposed water excretion during each whole-body contraction,²⁴ this result suggests that water excretion is a distinct behavior, rather than a secondary consequence of other motor behaviors.

We next asked whether water excretion occurs periodically. We imaged freely behaving Hydra for 8 hours and scored the timing of each water excretion event (Figure 1H; Video S2). The average interval between water excretion events was 2.04 ± 0.14 hours (Figure 1I). The observed intervals were significantly less variable than randomized intervals (Figure 1J; coefficient of variation = 0.28 observed versus 0.88 randomized; randomized interval variability = 1.49 ± 0.32 hours; **p = 0.0014), indicating that water excretion occurs at periodic intervals rather than randomly.

These results suggested that water excretion may be triggered when the gastric cavity reaches a certain fluid volume or pressure. To test this idea indirectly, we first increased the medium osmolarity using 50 mM sucrose, which should reduce the osmotic driving force for water influx.²⁵ High osmolarity indeed significantly decreased the number of water excretion events over 8 hours, from 3.66 ± 0.21 events in control animals to 1.44 ± 0.38 events in sucrose-treated animals (Figures 1H and K; control, N = 6 animals; 50 mM sucrose, N = 9 animals; ***p = 0.0002). We next tested whether mechanical pressure signaling could affect water excretion by applying Jedi1, an activator of Piezo mechanosensitive channels originally characterized in mammalian systems and recently used in Hydra.^26,27 Jedi1 significantly increased water excretion compared with control conditions (Figure S3; control, 0.57 ± 0.30 events, N = 7 animals; Jedi1, N = 4 animals; ***p = 0.0004). Together, these results suggest that Hydra gradually accumulates water in the gastric cavity and periodically excretes it through a mechanism that may involve sensing of gastric volume or pressure, while leaving the specific Hydra Piezo target unresolved.

Endodermal epithelial cells express aquaporin channels

Water likely accumulates in Hydra’s gastric cavity through epithelial tissues rather than by oral drinking,¹² as Hydra’s mouth opening is usually closed except during feeding and waste expulsion (Figure 2A).^23,28 To test whether water enters through the mouth, we incubated animals in Alexa Fluor 488 dye and measured fluorescence released during water excretion. Fluorescence intensity after dye incubation was 94-fold lower compared with animals in which Alexa Fluor 488 was directly injected into the gastric cavity (Figure S4; incubation, 3.49 ± 0.81, N = 5 animals; injection, 329.3 ± 165.1, N = 6 animals; ***p = 0.0002). This suggests that little external water enters the gastric cavity through the mouth under normal conditions.

Figure 2. Potential mechanisms of water accumulation in the gastric cavity.

(A) Cartoon showing osmotic water movement from the freshwater environment into the gastric cavity. Ecto, ectodermal epithelium; Meso, mesoglea; Endo, endodermal epithelium. Blue arrows indicate the potential direction of water movement. (B) Dot plots showing mRNA expression of aquaporin variants in Hydra vulgaris. Several aquaporin variants show enriched expression in endodermal epithelial cells. (C) Cnidarian aquaporin/MIP phylogeny. Maximum-likelihood tree of 39 curated aquaporin/MIP protein sequences from the freshwater hydrozoans Hydra vulgaris and Craspedacusta sowerbii, the marine hydrozoan Hydractinia symbiolongicarpus, and the estuarine anthozoan Nematostella vectensis. Major clades are indicated: a, AQP11-like; b, AQP1-like; c, AQP4/5-like; d, AQP9/7-like; and e, divergent cnidarian AQP lineages. (D) Electron micrograph of an endodermal epithelial cell. LV, large vacuole; orange line, cell membrane; blue box, region enlarged in (E). (E) Enlarged view of the boxed region in (D). Light blue areas indicate small clear vesicles. Red line indicates an invagination at the gastric lining. Arrowheads indicate vesicles tethered near the membrane. (F) Proportion of small clear vesicles along the apical-basal axis, excluding large vacuole areas. Each cell was divided into 10 equally spaced sections. Section 10 showed higher proportion of small clear vesicles than sections 5, 6, and 7 (mean ± SEM; n = 6 EM sections from 3 endodermal epithelial cells; repeated-measures one-way ANOVA with Tukey’s multiple comparisons test; section 10 versus sections 5, 6, and 7, *p = 0.0149, 0.0332, and 0.0307). (G) Left: raster plots of water excretion events under control conditions and treatment with 10 μM Z433927330, an aquaporin blocker. Five representative animals are shown for each condition. Right: number of water excretion events under control and aquaporin blocker treatment (mean ± SEM; control, N = 6 animals; 10 μM Z433927330, N = 7 animals; unpaired two-tailed Student’s t test, **p = 0.0018). (H) Left: cartoon of survival analysis with aquaporin blocker treatment in 300 mOsm artificial seawater. Right: survival analysis under aquaporin blocker treatment (mean ± SEM; n = 15 animals per condition, pooled from N = 3 experiments, 10 μM Z433927330; unpaired two-tailed Student’s t test with Welch’s correction, *p = 0.0474). (I) Working model of water movement through endodermal epithelial cells in freshwater, seawater, and aquaporin blocker conditions.

See also Figures S4, S5, and S6.

Water can diffuse slowly across lipid bilayers, but efficient transcellular water transport is often mediated by intrinsic protein/aquaporin channels.²⁹ Aquaporins passively transport water or small neutral solutes across cellular and organellar membranes, depending on osmotic and concentration gradients. To test whether aquaporins contribute to water accumulation in Hydra vulgaris, we analyzed available single-cell RNA sequencing and genomic databases.²⁰ We found that H. vulgaris expressed seven aquaporin transcripts from seven genes. These aquaporins were particularly enriched in endodermal epithelial cells, but showed little or no expression in ectodermal epithelial cells or neurons (Figure 2B). In contrast, this endoderm-specific expression was not observed in the marine cnidarian Nematostella vectensis, based on the updated Nematostella single-cell reference atlas,³⁰ suggesting that endodermal aquaporin expression may be associated with freshwater osmoregulation in Hydra (Figure S5C).

To examine the evolutionary relationships of these aquaporins, we compared H. vulgaris aquaporin proteins with those from the marine hydrozoan Hydractinia symbiolongicarpus, the estuarine anthozoan Nematostella vectensis, and the freshwater hydrozoan Craspedacusta sowerbii (Figures S5A and S5B).³¹ Most Hydra aquaporins, including the isoforms highly expressed in endodermal epithelial cells, clustered within an AQP9-like lineage. In contrast, H. symbiolongicarpus and N. vectensis possessed aquaporins distributed across broader aquaporin lineages. Similarly, the putative C. sowerbii aquaporins were distributed across several major lineages (Figure 2C). These results show that the distribution of aquaporins among multiple lineages does not show a simple association with either marine or freshwater habitats. Hydra aquaporins may have undergone unique lineage-specific expansion or specialization, potentially related to freshwater adaptation.

Because endodermal epithelial cells showed high and cell-type-specific aquaporin expression, we next examined their ultrastructure using previously published electron microscopy datasets from Hydra.¹⁶ We found that endodermal epithelial cells contained large, clear vacuolar structures (Figure 2D). Smaller clear vesicles were preferentially localized toward the gastric cavity side of the cell (Figures 2E and F; n = 6 EM sections from 3 endodermal epithelial cells; section 10 versus sections 5, 6, and 7, *p = 0.014, 0.033, and 0.031; Figure S6; n = 6 EM sections from 3 endodermal epithelial cells; section 10 versus sections 4, 5, 6, and 7, *p = 0.029, **p = 0.0063, *p = 0.0195, and *p = 0.0227). We also observed membrane invaginations along the gastric-facing membrane and small clear vesicles tethered near this membrane. These ultrastructural features are consistent with the possibility that water accumulates through vacuolar or vesicular compartments in endodermal epithelial cells and is then released into the gastric cavity.

To test whether aquaporin-mediated water transport functionally contributes to water excretion, we treated Hydra with putative aquaporin/aquaglyceroporin inhibitor Z433927330. Z433927330 treatment reduced the frequency of water excretion events compared with control animals (Figure 2G; control, 4.67 ± 0.33, N = 6 animals; 10 μM Z433927330, 2.43 ± 1.13, N = 7 animals; **p = 0.0018). This decrease in excretion frequency indicates that Z433927330 may impair water accumulation in the gastric cavity by blocking aquaporin-dependent water transport.

If aquaporins facilitate water influx into the gastric cavity, blocking aquaporins might slow water accumulation and reduce the osmotic stress experienced by Hydra under high-osmolarity conditions, by promoting water retention for instance. Consistent with this idea, animals treated with Z433927330 showed increased survival in 300 mOsm artificial seawater compared with control animals (Figure 2H; control, 6.67 ± 6.67, N = 3 experiments; 10 μM Z433927330, 33.33 ± 6.67, N = 3 experiments; *p = 0.0474). These results support a model in which aquaporin-dependent water transport through Hydra’s endodermal epithelial cells may contribute to gastric water accumulation and periodic water excretion (Figure 2I). Because Z433927330 has not been validated against Hydra aquaporins, we interpret these pharmacological data as descriptive rather than definitive evidence for specific Hydra aquaporin function.

Together, these results show that Hydra endodermal epithelial cells have enriched aquaporin expression and specialized vacuolar morphology, suggesting a potential cellular pathway for water accumulation in the gastric cavity through Hydra’s endodermal epithelium.

Neurons may contribute to water excretion

We next asked whether water excretion, a periodic behavior that releases accumulated fluid from the gastric cavity, is controlled by neurons. Previous studies reported that nerve-free Hydra, made by colchicine treatment, fail to open their mouths normally for feeding, and become swollen from accumulated gastric fluid.^32,33 To further explore this, we chemically reduced the number of neurons in Hydra using colchicine and examined water excretion under this condition. Colchicine treatment can selectively reduce the interstitial cell lineage, which gives rise to animals with greatly reduced numbers of neurons.³⁴ In Hydra, interstitial stem cells differentiate into nematocytes, gland cells, and neurons so the phenotype from this approach could reflect combined effects of these cell types. This treatment indeed significantly decreased the number of neurons from 643.8 ± 92.67 to 20.4 ± 3.80 (Figure 3A; control, N = 5 animals; neuron-reduced, N = 5 animals; **p = 0.0025).

Figure 3. Reduced neuronal number suppresses water excretion.

(A) Left: representative images of pan-neuronal GFP Hydra under control conditions and after colchicine treatment (Reduced), which reduces neuron number. Right: number of neurons in control and colchicine-treated animals (mean ± SEM; Control, N = 5 animals; Reduced, N = 5 animals; unpaired two-tailed Student’s t test with Welch’s correction, **p = 0.0025). (B) Raster plots of water excretion events in control and colchicine-treated animals over 8 h. (C) Number of water excretion events over 8 h (mean ± SEM; Control, N = 5 animals; Reduced, N = 5 animals; unpaired two-tailed Student’s t test, ***p = 0.0005).

We then imaged control and neuron-reduced animals for 8 hours and quantified water excretion events. Neuron-reduced animals showed a significant decrease in water excretion, from 4.2 ± 0.49 events in control animals to 1.2 ± 0.2 events after colchicine treatment (Figures 3B and C; control, N = 5 animals; neuron-reduced, N = 5 animals; ***p = 0.0005). Some neuron-reduced animals also appeared bloated and showed restricted behaviors. These results indicate that Hydra’s nervous system likely contributes to periodic water excretion, while we cannot exclude the possibility that other interstitial-lineage cell types or broader physiological changes also contribute to the phenotype.

RP2 neurons increase their activity before water excretion

To identify the neuronal activity that drives water excretion, we focused on a neuronal population that is localized in the endodermal layer, close to the gastric cavity, and has known motor output. In this study, we refer to this molecularly and physiologically distinct neuronal population as rhythmical potential 2 neurons, or RP2 neurons. RP2 neurons generate all-or-none action potentials that are largely mediated by calcium influx, allowing calcium imaging to capture their spike timing.^14,35,36 RP2 neurons are sparsely distributed throughout the Hydra endoderm, highly localized near the mouth region, and absent from tentacles (Figure 4A and Figure 6E).

Figure 4. RP2 neuronal activity increases before water excretion.

(A) Left: cartoon showing the localization of RP2 neurons in the endodermal layer. Middle and right: representative fluorescence images of animals expressing GCaMP6s in RP2 neurons under the NDF-1 promoter, showing inactive and active RP2 neurons. (B) Changes in RP2 calcium signal amplitude before water excretion. Top: average fluorescence amplitude in the mouth region at −15 to −10 s and −5 to 0 s relative to water excretion onset (mean ± SEM; N = 5 pairs from 5 experiments; paired two-tailed t test, *p = 0.0249). Bottom: same analysis for the region around the foot, used as a control (mean ± SEM; N = 5 pairs from 5 experiments; paired two-tailed t test, p = 0.156). (C) Representative trace of body length change, with RP2 spike timing overlaid relative to water excretion and other behaviors. (D) Raster plots of RP2 spikes aligned to water excretion onset (±60 s). (E) Number of RP2 spikes before and after water excretion onset (±60 s) (mean ± SEM; N = 5 animals; paired two-tailed t test, ***p = 0.0002). (F) Distribution of RP2 interspike intervals shown on a log scale (mean = 172.2 s; N = 5 animals, n = 245 intervals). RP2 bursts were defined as three or more spikes with interspike intervals of ≤30 s. (G) Number of spikes per RP2 burst for bursts followed by water excretion and bursts not followed by water excretion (failed bursts) (mean ± SEM; failed bursts, n = 6; excretion bursts, n = 12; unpaired two-tailed Student’s t test with Welch’s correction, *p = 0.0147). (H) Association between RP2 burst events and water excretion (n = 18 RP2 bursts and 122 non-burst events; Fisher’s exact test, ****p < 0.0001).

See also Figure S7 and Video S3.

Figure 6. Photostimulation of RP2 neurons triggers water excretion, and ablation suppresses it.

(A) Cartoon of the experimental setup for simultaneous RP2 photostimulation and imaging. (B) Left: neuronal imaging of active RP2 neurons. Green arrowhead indicates photostimulated neuron. Right: whole-body behavioral imaging. Boxed region shows changes in head area after water excretion. (C) Number of RP2 neurons stimulated before the initiation of water excretion (mean ± SEM; N = 5 experiments in which stimulation induced water excretion). (D) Association between RP2 stimulation and water excretion. RP2 stimulation was significantly associated with water excretion compared with control stimulation conditions (n = 11 no-stimulation trials, n = 10 off-target stimulation trials, and n = 10 RP2 stimulation trials; Fisher’s exact test, **p = 0.003). (E) Left: cartoon of RP2 ablation in transgenic animals expressing GFP in RP2 neurons. Middle: representative image before ablation. Right: representative image after ablation. (F) Number of RP2 neurons before and after ablation (mean ± SEM; N = 9 animals; paired two-tailed t test, ****p < 0.0001). (G) Number of water excretion events before and after ablation (mean ± SEM; N = 9 animals; paired two-tailed t test, p = 0.072). (H) Correlation between number of remaining RP2 neurons and number of water excretion events after ablation (N = 9 animals; simple linear regression with test of non-zero slope, *p = 0.0321).

See also Video S4.

Using transgenic Hydra expressing GCaMP6s in RP2 neurons under the control of the NDF-1 promoter,^20,37 we recorded RP2 activity during water excretion (Figure 4A). As previously reported, RP2 neurons showed rhythmic synchronous ensemble activity.¹⁴ We first analyzed population calcium signals by placing regions of interest around the mouth region and around the foot region as a control. RP2 activity around the mouth increased significantly before water excretion and decreased abruptly, after the onset of water excretion. Specifically, calcium activity was significantly elevated from −15 to −10 seconds before excretion onset and from −5 to 0 seconds before onset in the mouth region (−15 to −10 seconds, 0.57 ± 0.10; −5 to 0 seconds, 0.27 ± 0.09), but not in the foot region (Figure 4B; −15 to −10 seconds, 0.045 ± 0.014; −5 to 0 seconds, 0.068 ± 0.015; N = 5 pairs from 5 experiments; *p = 0.0249). These results suggest that RP2 activity preferentially affects the mouth region before water excretion.

This increase in RP2 neuronal activity was followed by significantly increased activity of endodermal epithelial cells around the mouth, imaged in transgenic Hydra expressing GCaMP6s in the epithelial layer (Figure S7A; mouth, −15 to −10 seconds, 0.66 ± 0.21; −5 to 0 seconds, 1.31 ± 0.34; foot, −15 to −10 seconds, 0.49 ± 0.15; −5 to 0 seconds, 0.27 ± 0.10; N = 5 pairs from 5 experiments; *p = 0.0407). In contrast, ectodermal epithelial activity did not change significantly before water excretion (Figure S7B; mouth, −15 to −10 seconds, 0.11 ± 0.05; −5 to 0 seconds, 0.11 ± 0.04; foot, −15 to −10 seconds, 0.06 ± 0.02; −5 to 0 seconds, 0.054 ± 0.009; N = 5 pairs from 5 experiments; p = 0.54). These results suggest that RP2 neurons may act through endodermal epithelial cells around the mouth to drive mouth opening during water excretion.

We examined with more detail the ramping up of RP2 neuronal activity preceding water excretion. During spontaneous behavior, RP2 neurons fired sparsely, but their firing frequency increased specifically before the onset of water excretion (Figures 4C and D; Video S3). The number of RP2 spikes was significantly higher before water excretion than after its onset (Figure 4E; before, 5 ± 0.32 spikes; after, 0.2 ± 0.2 spikes; N = 5 animals; ***p = 0.0002). RP2 interspike intervals varied widely, ranging from 2 to 3,354 seconds (Figure 4F; mean = 172.2 ± 21.7 seconds; N = 5 animals, n = 245 RP2 spike intervals). To further analyze the relationship between RP2 firing and water excretion, we defined RP2 bursts as three or more spikes with interspike intervals of 30 seconds or less, based on the distribution of interspike intervals (Figure 4F). RP2 bursts that led to water excretion contained significantly more spikes than bursts that failed to trigger water excretion (Figure 4G; water excretion bursts, 4.67 ± 0.38 spikes; failed bursts, 3 ± 0.45 spikes). Moreover, only 1 of 122 non-burst events, defined as either a single spike or two spikes within 30 seconds, was followed by water excretion. In contrast, 12 of 18 RP2 bursts led to water excretion, showing a significant association between RP2 burst activity and the onset of water excretion (Figure 4H; n = 18 RP2 bursts and 122 non-burst events; ****p < 0.0001).

Together, these results demonstrate that RP2 neurons increase their activity before water excretion and that RP2 burst activity reliably predicts the onset of water excretion.

Spike-count and evidence-accumulation algorithms predict water excretion onset

The previous analysis showed that RP2 bursts were strongly associated with water excretion, suggesting that consecutive RP2 firing may trigger this behavior. We next asked what feature of RP2 spike timing best predicts water excretion onset. To test whether Hydra uses RP2 spikes in a decision-like process to initiate water excretion, we fit raw RP2 spike timing data with simple predictive models (Figure 5A). First, we used spike-count models, in which RP2 spikes are summed within a defined time window before each time point. As a second approach, we used leaky integrator models, in which each RP2 spike contribute evidence that decays exponentially over time, with a decay time constant tau determining how long previous spikes influence the accumulated signal.^38,39

Figure 5. Spike-count and evidence-accumulation models predict water excretion onset.

(A) Tested models and formulas. Count_T represents number of RP2 spikes within the previous T seconds. The leaky integrator represents exponentially weighted RP2 spike history with time constant τ. (B) Representative model traces generated from experimental RP2 spike timing data. Blue line, Count_30. Orange line, leaky integrator with τ = 27 s. Water excretion events are overlaid. Three representative recordings are shown. (C) Leaky integrator traces aligned to water excretion onset (τ = 27 s). Thirteen representative events are shown. (D) Model comparison for prediction of water excretion onset. Model performance is shown as AICc, where lower values indicate better fit. Count_30 and Leaky_τ=27 s showed the best performance among the tested models.

See also Table S1.

We tested a range of parameters for each model and identified the best-performing parameter within each class. The best spike-count model counted RP2 spikes within the previous 30 seconds, referred to as Count_30. The best leaky integrator model had a decay time constant of 27 seconds, referred to as Leaky_tau_27. These two models produced highly similar temporal profiles, with both signals increasing before water excretion onset (Figure 5B). The accumulated evidence also aligned with the timing of water excretion behavior, increasing before excretion and decreasing after the event began (Figure 5C).

We next compared these models with other simple spike-timing models, including models based on short interspike intervals and spike counts over different time windows. Count_30 and Leaky_tau_27 showed the best overall performance based on AICc comparison (Figure 5D and Table S1). Count_30 performed slightly better than Leaky_tau_27, with an AICc difference of 3.13 between the two models, likely reflecting the additional parameter in the leaky integrator model (Count_30, k = 2; Leaky_tau_27, k = 3). These results indicate that water excretion onset is best predicted by the accumulation of RP2 spikes over approximately 30 seconds.

Together, these results suggest that recent RP2 spike accumulation with a 30 second integration time window may contribute to the initiation of water excretion.

Photostimulation of RP2 neurons triggers water excretion

The previous experiments showed that RP2 activity precedes water excretion and that recent RP2 spike accumulation predicts water excretion onset. To test whether RP2 neurons causally regulate this behavior, we performed two-photon stimulation of individual RP2 neurons in transgenic animals expressing GCaMP6s under the control of the NDF-1 promoter.⁴⁰ During stimulation, we simultaneously imaged calcium activity in neighboring RP2 neurons and monitored behavioral responses (Figure 6A).

Two-photon stimulation successfully activated RP2 neurons and induced water excretion in a subset of trials (Figure 6B). Among successful trials, photostimulation activated 10.0 ± 1.7 RP2 neurons before water excretion onset (Figure 6C). Water excretion occurred in half of RP2 stimulation trials, whereas neither no-stimulation trials nor off-target stimulation trials induced water excretion. This showed a significant association between RP2 stimulation and water excretion (Figure 6D; Video S4; n = 11 no-stimulation trials, n = 10 off-target stimulation trials, and n = 10 RP2 stimulation trials; **p = 0.003). These results indicate that RP2 activation is sufficient to trigger water excretion.

Ablating RP2 neurons alters water excretion

To further test the contribution of RP2 neurons to water excretion, we used two-photon illumination to selectively ablate RP2 neurons in transgenic animals expressing GFP under the control of the NDF-1 promoter (Figure 6E).²⁰ This method reduced the number of RP2 neurons from 168.9 ± 18.09 to 20.11 ± 5.23 (Figure 6F; N = 9 animals; ****p < 0.0001). Initially, behavioral analysis showed no significant decrease in the number of water excretion events after RP2 ablation, although ablated animals showed higher variability (Figure 6G; control, 3.67 ± 0.21 events; ablated, 5.67 ± 0.83 events; N = 9 animals; p = 0.072).

Because RP2 ablation produced variable behavioral effects and the number of remaining RP2 neurons differed among animals, we examined in more detail whether the extent of RP2 neuron loss explained this variability. Indeed, we found a significant positive correlation between the percentage of remaining RP2 neurons and the number of water excretion events after ablation (Figure 6H; N = 9 animals; R² = 0.50; *p = 0.0321). Although compensatory changes may have occurred after ablation, this relationship suggests that animals retaining more RP2 neurons performed more water excretion events.

Together with the photostimulation experiments, these results show that RP2 neurons are correlated with water excretion and causally contribute to the initiation of this behavior.

GLWamide-family peptides stimulate water excretion

We next explored the molecular mechanisms that may mediate the causal role of RP2 neurons in triggering water excretion. In Hydra, bath application of neuropeptides can induce dose-dependent behavioral effects,⁴¹ suggesting that peptides applied externally through the media can reach their cellular targets. Single-cell transcriptomic data revealed that RP2 neurons express a gene encoding a GLWamide precursor protein. This precursor is processed into around 7 unique GLWamide-family peptides, including Hym-248 (Figure 7A).²¹ GLWamide-family peptides share a C-terminal GLWamide motif that can induce contraction of basal disc epitheliomuscular cells, leading to detachment of maturing buds from the parent animal.¹⁷ In addition, the N-terminal sequence of Hym-248, EPLPI, can stimulate somersaulting.⁴¹ We therefore tested whether Hym-248, its N-terminal sequence EPLPI, or the C-terminal consensus sequence GLWamide could regulate water excretion. To do so, we applied each peptide individually to freely behaving animals and quantified the number of water excretion events.

Figure 7. GLWamide stimulates water excretion.

(A) Cartoon showing processing of the GLWamide precursor into multiple mature peptide sequences. We tested Hym-248, as well as its N-terminal sequence (EPLPI) and C-terminal motif (GLWa). (B) Raster plots of water excretion events within 2 h under different peptide conditions. Five representative animals are shown for each condition. (C) Violin plots showing the number of water excretion events in each condition (mean ± SEM; control, N = 10; Hym-248, N = 11; EPLPI, N = 10; GLWa, N = 12; ordinary one-way ANOVA with Tukey’s multiple comparisons test; control versus Hym-248, **p = 0.0039; control versus GLWa, **p = 0.0047).

See also Video S5.

Both Hym-248 and GLWamide significantly increased the number of water excretion events, whereas EPLPI had no significant effect (Figure 7B and C; Video S5; control, 0.6 ± 0.27, N = 10 animals; Hym-248, 3.18 ± 0.66, N = 11 animals; EPLPI, 1 ± 0.39, N = 10 animals; GLWamide, 3.08 ± 0.50, N = 12 animals; control versus Hym-248, **p = 0.0039; control versus GLWamide, **p = 0.0047). These results indicate that the C-terminal GLWamide motif is sufficient to stimulate water excretion.

Together, these results suggest that GLWamide-family peptides act downstream of RP2 neuronal activity to promote water excretion. Because GLWamide-family peptides can also suppress whole-body contraction,⁴¹ this peptide signal could promote water excretion while reducing competing motor behaviors.

Discussion

Here we investigate the water excretion behavior of the cnidarian Hydra vulgaris and find that a specific neuronal circuit, formed by endodermal Rhythmic Potential 2 (RP2) neurons, is causally involved in it. RP2 activity ramps up before water excretion and contributes to the regulation and initiation of water excretion, an effect mediated by GLWamide-family peptides, synthesized by RP2 neurons.

To perform this study, we applied to freely behaving Hydra a combination of approaches, including media manipulation, bioinformatics, machine vision behavioral analysis, wide-field and two-photon fluorescent imaging, transgenic GCaMP lines, computational modelling, two-photon ablation and activation of neurons, and pharmacological application of neuropeptides. We first found that Hydra is intolerant to salty environments and continually accumulates water in its gastric cavity, until it releases it through a mouth opening in a stereotyped water excretion behavior, which occurs regularly (Figure 1). This freshwater adaptation in Hydra is likely mediated by the specific expression of aquaporins in Hydra’s endoderm epithelial cells, and by the directional trafficking of water vacuoles through them, from the body wall to the gastric cavity of the animal (Figure 2). The nervous system is necessary for the water excretion behavior to occur, as animals with a reduced number of neurons displayed reduced water excretion (Figure 3). Imaging neuronal activity during water excretion demonstrates that RP2 cells, a specific population of endoderm neurons, become increasingly active before mouth opening (Figure 4). Moreover, their activity abruptly decreases after water excretion occurs, a hallmark of an integrate to threshold algorithm in which neuronal activity resets after behavioral onset (Figure 4). Consistent with this, spike accumulation or leaky integrator mathematical models recapitulate the data, with best fits with an integrating window of time of approximately 30 seconds (Figure 5). The ramping up to threshold of RP2 neurons is closely followed by the activation of oral epitheliomuscular cells, consistent with their triggering mouth opening (Figure S7). A causal role of RP2 neurons in water excretion is further demonstrated by their photoactivation, which triggers water excretion, and their ablation, which reveals a significant relationship between the number of remaining RP2 neurons and water-excretion frequency (Figure 6). Interestingly, after ablating RP2 neurons, a significant amount of water excretion remained. This may reflect compensation by remaining RP2 neurons, or contributions from additional circuit elements. Indeed, we previously found that the frequency of RP1 neuronal activity increased when most RP1 neurons were ablated.⁴¹ A similar compensatory increase in the remaining RP2 neurons could help maintain water excretion events after significant ablation. Nevertheless, and consistent with a causal role of RP2 neurons in water excretion, animals with more remaining RP2 neurons exhibited a higher frequency of excretion events, resulting in a significant positive correlation between the number of remaining RP2 neurons and the frequency of water excretion (Figure 6H). Finally, bath application of GLWamide-family peptides, expressed in RP2 cells, can stimulate water excretion (Figure 7). Thus, RP2 activity, and its release of GLWamide-family peptides, appears necessary and sufficient to trigger water excretion, suggesting a spike-accumulation mechanism for initiating water excretion.

A sequential model for Hydra freshwater osmoregulation

Our results provide a framework for understanding how Hydra regulates internal water balance in freshwater. Based on our findings, we propose a model for the water excretion cycle:

1. In a low-osmolarity freshwater environment, Hydra experiences continuous osmotic water influx into its cells.

2. Excess water is trafficked across epithelial layers and accumulates in the gastric cavity. Specific aquaporin expression and clear vacuolar structures in endodermal epithelial cells facilitate this water transport into the gastric cavity. As more water accumulates, increased gastric volume or internal pressure could be detected by mechanosensitive channels such as Piezo.

3. RP2 neuronal activity increases as water accumulates, likely in response to Piezo activation in sensory neurons or in RP2 cells directly. RP2 spikes are integrated over approximately 30 seconds, and RP2 activity increases until it reaches a frequency threshold which initiates water excretion. The threshold could reflect an excitatory positive feedback loop mediated by the gap junctional coupling of RP2 cells, or by the self-activating release of their neurotransmitters. After the threshold is reached, cross-inhibition involving a second neuronal population could then inactivate RP2 cells, similarly to the reciprocal cross-inhibition circuit found in Hydra’s somersaulting.⁴¹ Alternatively, RP2 cells could have an intrinsic refractory period, by which neurons become inactive after reaching a maximum level of activity or of peptide release.

4. At their peak of their activity, RP2 neurons release GLWamide-family peptides, which activate endodermal epithelial cells around the mouth and promote mouth opening. At the same time, GLWamide signaling suppresses competing motor behaviors such as whole-body contraction, consistent with a cross-inhibitory circuit strategy.

5. Hydra would then release accumulated water through the mouth, reducing internal pressure and resetting the cycle.

Mechanisms of water homeostasis in Hydra

Our proposed model conceptually explains the ongoing periodic cycle of water accumulation and water excretion that operates in Hydra, and represents a significant departure from the past view that contraction bursts are used in Hydra to osmotically self-regulate by wringing itself clear of water.²⁴ At the same time, our work is but a first step in understanding the likely complex water homeostasis in Hydra. While it is clear from our data that RP2 and GLWamide-family peptides play a causal role in water excretion, the molecular mechanisms involved remain to be elucidated. Based on our experiments, we expect that both aquaporin channels and intracellular vacuoles will be involved in trafficking water through the epithelium and discharging it into the gastric cavity. Further evidence for this could involve targeted genetic manipulation of aquaporin channels and ultrastructural and functional tracing of water pathways in this animal. In addition, Piezo mechanosensory channels could provide pressure feedback that could activate RP2 neurons, either directly or indirectly, forming a signal that RP2 cells accumulate evidence to trigger water excretion. How this interoceptive feedback operates, and whether other neuronal or non-neuronal elements are at play, remains an open question and could also be studied further using cell-type specific expression or inactivation of mechanosensitive channels. Finally, it seems also necessary to endow our proposed osmoregulatory model with quantitative estimates of every one of the steps, ideally supported by the data, so that one could use the model to predict future behavioral responses, predictions that could be tested experimentally.

Integration of evidence for decision making: from cnidarians to bilaterians

A tantalizing finding in our experiments is related to the dynamics of RP2 neuronal activity during water excretion, that displays a slow buildup of spiking until it reaches a threshold level, which is immediately followed by an abrupt decrease in spiking. These integrate to threshold (ITT) neuronal dynamics are indicative of a decision making algorithm based on evidence accumulation and closely resemble results from bilaterian species. In fact, the ITT framework is currently the main hypothesis to explain the neural mechanisms of decision making in many animals. This hypothesis holds that the brain resolves uncertainty by accumulating noisy sensory evidence over time until an integrated total crosses a threshold, at which point a decision is triggered All major ITT models share this mechanism— with the drift-diffusion model (DDM) being the dominant mathematical formalization,⁴² and the threshold crossing determines both the behavioral choice and the reaction time. Salient experimental evidence for this framework comes from macaque LIP cortical neurons, where ramping activity correlates with task difficulty and response time, and recent electrical recordings have provided direct single-trial evidence for a drift-diffusion process controlling both choice and reaction time.⁴³

Besides this work on the mammalian cortex, there is an increasingly mechanistic body of work demonstrating ITT-based decision making in invertebrates. For example, in Drosophila trained to odor intensity discriminations, reaction times increased and accuracy declined with decreasing stimulus contrast, in quantitative agreement with a drift-diffusion model.⁴⁴ In C. elegans, temporal integration of sensory information for behavioral choice is analogous to evidence accumulation. As animals extract a long-term trend in sensory information by temporally integrating it, with AWB sensory neurons acting as leaky integrators whose intracellular calcium accumulates via voltage-gated channels until a behavioral threshold is crossed.⁴⁵

Our results on the temporal dynamics of RP2 activity during water excretion, where a spike counting or leaky integrator model accounts well for the experimental results, are consistent with this body of work, extending it to cnidarians. In fact, previous work on the neural mechanisms of somersaulting in Hydra could also be interpreted as further evidence of an ITT algorithm, as a neural circuit of ectodermal RP1 neurons ramps up its activity to a threshold firing rate, triggering the initial step of somersaulting, a fixed action pattern locomotion behavior.⁴¹ Thus, the Hydra nerve net could possess two complementary integrating neuronal ensembles, the ectodermic RP1 and the endodermic RP2, each implementing an evidence accumulation algorithm to mediate two different types of decision making. Together, these results indicate that decision making algorithms based on integrating accumulated evidence were already used by evolution in ancestral nervous systems. In closing, the relative simplicity and experimental accessibility of the nervous system of cnidarians could help illuminate neural computations, such as decision making, that are widely used through the metazoan lineage and have major repercussions in understanding the normal or diseased brains. By providing an evolutionary perspective, the study of the cnidarian nervous system could thus help advance neuroscience, perhaps in unexpected ways.

STAR METHODS

EXPERIMENTAL MODEL AND SUBJECT DETAILS

Hydra

Hydra vulgaris AEP strains were maintained in the dark at 18°C in Hydra culture medium (1 mM CaCl₂, 0.5 mM NaHCO₃, 0.33 mM MgSO₄, 0.03 mM KCl). Hydra were fed freshly hatched Artemia nauplii three times a week and were starved for 1 to 2 days before experiments.

Transgenic lines

In this study, we used transgenic Hydra expressing GCaMP6s or GFP in RP2 neurons, as well as previously generated transgenic Hydra expressing GCaMP6s in ectodermal or endodermal epithelial cells.⁴⁶ To generate the RP2 neuronal GFP line, DNA plasmids (0.5 mg/mL, provided by Dr. Celina Juliano’s laboratory) were microinjected into Hydra oocytes. The RP2 neuronal GCaMP6s line was kindly provided by Dr. Thomas Bosch’s laboratory.

METHOD DETAILS

Survival assay

Artificial seawater and sucrose solutions were prepared in Hydra culture medium at final osmolarities of 0, 50, 100, 300, 500, and 1,000 mOsm/kg. Artificial seawater was prepared using sea salts. Hydra were exposed to each solution for time windows ranging from 15 minutes to 72 hours. Because viability was sometimes difficult to determine immediately after treatment, animals were transferred back to standard Hydra culture medium after exposure and allowed to recover. Animals that resumed movement within 2 days were scored as surviving. Each condition included n = 30 animals per condition, pooled from N = 3 experiments (Figure 1B and Figure S1).

For the aquaporin blocker osmolarity tolerance assay, Hydra were treated with 10 μM Z433927330 in Hydra culture medium, exposed to 300 mOsm/kg artificial seawater for 15 minutes, and then transferred back to standard Hydra culture medium for recovery before survival was scored. Each condition included n = 15 animals per condition, pooled from N = 3 experiments.

Dye injection into the gastric cavity

20 μM Alexa Fluor 488 was delivered into the gastric cavity using aspirator tube assemblies. For behavioral imaging to quantify excretion frequency, Hydra were allowed to excrete once before imaging by gently poking the body column. This was done to compensate for the increased fluid volume in the gastric cavity caused by dye injection.

Water uptake assay

To test whether Hydra ingest water through the mouth, animals were incubated in Hydra culture medium containing 20 μM Alexa Fluor 488 for 2 hours without injection. This incubation allowed fluorescent dye to enter the gastric cavity if oral water uptake occurred. After incubation, animals were transferred to dye-free medium and gently rinsed several times to remove residual extracellular Alexa Fluor 488 while avoiding stimulation of water excretion. Animals were then imaged for 4 hours to detect fluorescence in water released from the mouth during spontaneous excretion events. As a positive control for fluorescence detection, 20 μM Alexa Fluor 488 was microinjected directly into the gastric cavity of separate animals. These animals were imaged under the same conditions to measure fluorescence in excreted water originating from the gastric cavity.

Fluorescence intensity of excreted water was quantified from image sequences by measuring the mean fluorescence intensity within the region containing the expelled water plume and subtracting local background fluorescence. The fluorescence signal was expressed as background-subtracted intensity, ΔF = b − a, where b indicates the mean fluorescence intensity of the expelled water plume and a indicates the local background intensity. Statistical comparisons between groups were performed using Welch’s t test on log-transformed values to account for unequal variance.

Behavioral recording and tracking

Wide-field imaging of Hydra behavior was conducted in bright field at 1 Hz using a fluorescence dissecting microscope (Leica M165) equipped with a 0.63× Plan Apo objective and an sCMOS camera (Hamamatsu ORCA-Flash 4.0). Hydra were placed in wells of glass-bottom dishes for freely behaving recordings at a depth of 700 to 750 μm. In this preparation, Hydra exhibited spontaneous behaviors, including body-column contractions, water excretion, and somersaulting, with a full range of motion.

Using DeepLabCut, a machine-learning-based tracking method, we tracked changes in body area (Figures 1D and 1F) and longitudinal body length (Figure 1G and Figure 4C) by placing labels around the body outline or along the longitudinal body axis. Water excretion events were manually scored by visualizing the release of fluorescent fluid from the mouth. All imaging was conducted at room temperature, approximately 23°C.

Generation of neuron-reduced animals

Because Hydra become motionless when all neurons are removed, we tested how a partial reduction of neuronal number affects water excretion behavior. Neuron-reduced animals were generated by incubating Hydra in 0.4% colchicine, an inhibitor of cell division that affects interstitial cells, which are precursors to neurons, overnight at 18°C. Hydra were then maintained at 18°C for approximately 15 days and allowed to recover before experiments. The medium was replaced twice daily and kept as clean as possible.

Wide-field calcium imaging

Wide-field calcium imaging of Hydra was conducted at 2 Hz for population analysis using a fluorescence dissecting microscope (Leica M165) equipped with a long-pass GFP filter set (Leica filter set ET GFP M205FA/M165FC), a 1.6× Plan Apo objective, and an sCMOS camera (Hamamatsu ORCA-Flash 4.0). A mercury arc lamp was used to illuminate the sample. For all imaging experiments, Hydra were mounted between coverslips with an approximately 200-μm spacer. This preparation restricted movement along the z axis while allowing free movement along the x and y axes, which allowed water excretion behavior to occur without disturbance while enabling visualization of RP2 population activity. In Figure S7, either ectodermal epithelial or endodermal epithelial activity was recorded. All imaging was conducted at room temperature, approximately 23°C.

Two-photon stimulation and ablation of RP2 neurons

Imaging experiments were performed by mounting Hydra expressing GCaMP6s in RP2 neurons between glass coverslips with an approximately 200-μm spacer. The imaging setup and objective lenses were completely enclosed with light-shielding cloth and black electrical tape to prevent light from leaking into the photomultiplier tubes. Calcium imaging was used to monitor RP2 neuronal activity.

Two-photon imaging and photostimulation were performed with two different femtosecond-pulsed lasers attached to a commercial microscope. An imaging laser, Ti, λ = 940 nm, was used to excite GCaMP6s, while a photostimulation laser, low-repetition-rate pulse-amplified laser, λ = 1040 nm, was used to stimulate RP2 neurons. The power of both lasers was controlled by two independent Pockels cells. The two laser beams on the sample were individually controlled by two independent sets of galvanometric scanning mirrors. Short movies, approximately 270 seconds, were collected at a sampling rate of 3.64 Hz with imaging laser power below 50 mW, dwell time of 2 milliseconds per pixel, and 256 × 256 pixels in the whole field of view.

The photostimulation power for each neuron was adjusted to the lowest value that elicited a response while minimizing cell damage. Single-cell photostimulation was performed with a spiral pattern scanned by a pair of galvanometric mirrors and delivered to the center of the cell body, with a 3-μm diameter, 3 spiral revolutions, and 30-millisecond duration. Simultaneous imaging and photostimulation were controlled by Prairie View and custom-made software running in MATLAB. To overcome difficulties in targeting neurons in constantly moving animals, we used the Live/Ablation feature to stimulate cells on the image during live scanning. Behavior was monitored during calcium imaging and two-photon stimulation using a USB camera placed below the sample.

Ablation of RP2 neurons was performed using Hydra expressing GFP in RP2 neurons. The photostimulation laser power was increased to ablate neurons, typically to above 50 mW. As with photostimulation, neurons were targeted one by one to address the challenges caused by Hydra movement and the need for precise laser targeting. Neuronal ablation was confirmed by observing the release of GFP from the soma into the extracellular space. Laser ablation took approximately 3 hours.

QUANTIFICATION AND STATISTICAL ANALYSIS

Analysis of periodicity

To test whether water excretion occurred periodically rather than randomly, we compared observed inter-excretion intervals with intervals generated from randomized excretion timing. For each simulation, the same number of excretion events was randomly sampled without replacement from possible time points between 0 and 5 hours at 0.5-hour resolution. Randomized event times were sorted, and inter-event intervals were calculated from adjacent randomized times. Interval variability was quantified using the coefficient of variation. Because periodic events are expected to show lower interval variability than random events, we tested whether observed intervals had significantly lower variance than randomized intervals using a one-sided F test.

Analysis of aquaporin expression

Aquaporin-family gene expression was analyzed from published single-cell RNA sequencing datasets for Hydra vulgaris and Nematostella vectensis. For Hydra, we used a Seurat object derived from the whole-transcriptome Hydra vulgaris AEP single-cell dataset from Siebert et al. The legacy Seurat object was updated using UpdateSeuratObject, and aquaporin-related features were identified by exact transcript feature names. Seven Hydra MIP-family features were analyzed: t12047aep-AQP9-HUMAN, t582aep-AQP9-HUMAN, t8313aep-AQP9-HUMAN, t28188aep-AQP9-HUMAN, t32664aep-AQP9-HUMAN, t36529aep-AQP4-MOUSE, and t8863aep. These features were relabeled for figure presentation as AQP9-like 1 to 5, AQPcic, and AQP11-like. Fine cell-type annotations from the cluster.manuscript metadata column were collapsed into broad cell classes, including ectodermal neurons, endodermal neurons, gland cells, interstitial stem cells, ectodermal epithelial cells, endodermal epithelial cells, germline cells, and nematocytes.

For Nematostella, we analyzed a published Nematostella vectensis single-cell RNA sequencing atlas distributed in UCSC Cell Browser format. The expression matrix, cell metadata, and UMAP coordinates were loaded into Seurat v5. Cell barcodes were matched between the expression matrix and metadata, UMAP coordinates were attached to the Seurat object, and expression values were log-normalized using NormalizeData with the LogNormalize method and a scale factor of 10,000. Candidate AQP/MIP genes were identified from atlas feature names by searching for “AQP,” “aquaporin,” or “MIP,” yielding 14 annotated AQP/MIP-family genes. To enable comparison with Hydra, atlas-provided cell identity labels were mapped into conservative HydraMatched_min groups, including ectodermal epithelial cells, endodermal epithelial cells, neurons, cnidocytes, gland/secretory cells, interstitial progenitors, and other categories without clear Hydra equivalents.

For both datasets, grouped dot plots were generated in R using Seurat and ggplot2. For each gene and broad cell group, we calculated the fraction of cells with detectable expression and the mean log-normalized expression among expressing cells. Dot size represents the percentage of cells expressing each transcript, and dot color represents z-scored mean expression across groups within each gene. Summary tables were exported to CSV, and final figures were exported as PNG and SVG files.

Analysis of aquaporin molecular evolution across cnidarians

Protein sequences of cnidarian major intrinsic proteins, including aquaporin-family proteins, were compiled from the NCBI RefSeq proteomes of Hydra vulgaris (GCF_038396675.1), Nematostella vectensis (GCF_932526225.1), and Hydractinia symbiolongicarpus (GCF_029227915.1), together with eight putative MIP-family protein sequences identified from the Craspedacusta sowerbii genome assembly GCA_003687565.2 (ASM368756v2), generating a curated sequence set containing 39 proteins. RefSeq proteins were retained under their XP_ accession numbers, whereas the C. sowerbii candidates were designated CsMIP01–CsMIP08; CsMIP02 and CsMIP06 were partial sequences. Sequences were aligned using MAFFT with the --auto option, generating a final curated alignment with 39 sequences and 474 aligned amino acid columns. Maximum-likelihood phylogeny was inferred using IQ-TREE2, version 2.0.7, with ModelFinder enabled. Branch support was calculated using 1,000 ultrafast bootstrap replicates and 1,000 SH-like aLRT replicates. ModelFinder selected LG+R4 as the best-fit model according to BIC. The resulting Newick-formatted maximum-likelihood tree was visualized and annotated in iTOL.

Analysis of endodermal epithelium by EM

High-resolution electron microscopy images of Hydra vulgaris endodermal epithelial cells were analyzed using the Volume Annotation and Segmentation Tool, VAST. Three endodermal epithelial cells from the body column were selected based on their orientation along the mesoglea-to-gastric-cavity axis. Two sections from each cell were manually annotated using a Wacom pen tablet, resulting in six analyzed sections from three cells. In each section, the cell body, large clear vacuoles, and small clear vesicles were manually segmented. Large vacuoles were defined as membrane-bound clear compartments connected to the central clear cavity, whereas small vesicles were defined as membrane-bound clear compartments that remained separate from the central vacuole.

To quantify the spatial distribution of small vesicles, each endodermal epithelial cell section was divided into 10 equally spaced bins along the mesoglea-to-gastric-cavity axis. Small vesicle area was quantified as the percentage of the analyzed endodermal area in each bin, excluding large vacuoles from the denominator. Quantification was performed from n = 6 sections from 3 endodermal epithelial cells.

Analysis of computational models

RP2 spike trains were modeled using spike-count and leaky evidence-accumulation frameworks. For each imaging recording, RP2 spike times were extracted from cumulative spike-time measurements. In the leaky evidence-accumulation model, each RP2 spike contributed one unit of evidence, which decayed exponentially over time with time constant τ. Accumulated evidence at time t was calculated as E(t) = Σ exp[-(t − ti)/τ], where ti indicates each RP2 spike time before time t. The probability of water excretion was then modeled using logistic regression as a function of accumulated RP2 evidence. The time constant τ was selected by maximum likelihood across candidate values ranging from 5 to 300 seconds.

Model performance was compared using AICc and ROC AUC. The leaky evidence-accumulation model was compared with simpler models based on RP2 spike count within the previous 30, 60, or 120 seconds, as well as a short-interspike-interval model. Water-excretion-associated spikes were defined as the last RP2 spike occurring within 20 seconds before water excretion initiation. Two visually ambiguous water excretion events in Imaging 5 were excluded from the primary analysis and included in a sensitivity analysis.

Statistics

Statistics for all experiments were performed using Prism 10 (GraphPad), except for the interval variability analysis in Figure 1J, which was performed using a one-sided F test for variance. Unless otherwise stated, values are shown as mean ± standard error of the mean (SEM). Fisher’s exact tests were used for categorical association analyses in Figures 4H and 6D. Adobe Illustrator was used for final adjustment of fonts, colors, and figure positions. Drawings were also created in Adobe Illustrator.

Supplementary Material

1

Document S1. Figures S1-S7 and Table S1.

2

Video S1. Water excretion in Hydra. Related to Figure 1. 5x speed, 53-second video, scale bar = 1 mm.

3

Video S2. Freely behaving Hydra with water excretion visualized. Related to Figure 1. 2,000x speed, 6-hour video, scale bar = 2 mm.

4

Video S3. GCaMP imaging of RP2 activity, during water excretion in a mounted animal. Related to Figure 4. 5x speed, 31.5-second video, scale bar = 500 μm. The arrowhead indicates the timing of water excretion (±2.5 seconds from the onset of water excretion).

5

Video S4. Two-photon laser stimulation of RP2 neurons, and evoked behavior. Related to Figure 6. Left: calcium imaging, scale bar = 50 μm. Right: behavior imaging, scale bar = 1 mm. 20x speed, ~5-minute video.

6

Video S5. Effect of GLWamide on free-behaving Hydra. Related to Figure 7. 1,000x speed, 2-hour video each, scale bar = 2 mm.

KEY RESOURCES TABLE

REAGENT or RESOURCE	SOURCE	IDENTIFIER
Experimental models: organisms/strains
Hydra vulgaris strain AEP, SS1	Dr. Robert Steele, UC Irvine	N/A
Hydra vulgaris strain AEP	Dr. Hiroshi Shimizu, National Institute of Genetics, Japan	N/A
Hydra vulgaris strain AEP, transgenic line expressing GCaMP6s in RP2 neurons under the control of the NDF-1 promoter	Dr. Thomas Bosch, University of Kiel, Germany	N/A
Recombinant DNA
NDF-1 promoter with GFP in pHyVec1	Dr. Celina Juliano, UC Irvine	N/A
Synthesized peptides
Hym-248 (EPLPIGLWamide)	Bachem	Custom
GLWamide (CGLWamide)	Bachem	Custom
EPLPI (CGGQSSQNSEPLPI)	Bachem	Custom
Chemicals
Sea salts	MilliporeSigma	S9883
Sucrose	MilliporeSigma	573113
Colchicine	MilliporeSigma	C9754
Jedi1	MilliporeSigma	SML2533
Z433927330	MedChem Express	HY-126074
Alexa Fluor 488	Thermo Fisher Scientific	A10436
Data and code
Code used for the paper	GitHub and Zenodo	https://github.com/wayamamo/Water-excretion-paper DOI: 10.5281/zenodo.21696228
Original raw videos and datasets upon publication	Zenodo	DOI: 10.5281/zenodo.21696228

Highlights:

• Hydra accumulates gastric water and periodically excretes it through the mouth

• Endodermal RP2 neurons integrate their activity to a threshold before water excretion

• Stimulation of endodermal RP2 neurons triggers water excretion

• GLWamide peptides produced by RP2 neurons induce water excretion

RESOURCE AVAILABILITY

Lead contact

Further information and requests for resources and reagents should be directed to and will be fulfilled by the Lead Contact, Wataru Yamamoto (wy2303@columbia.edu).

Materials availability

Transgenic Hydra are available upon request.

Data and code availability

• The data generated or analyzed during this study will be available through Zenodo (DOI: 10.5281/zenodo.21696228).

• All the codes will be available on GitHub (https://github.com/wayamamo/Water-excretion-paper) and Zenodo.

• Any additional information is available from the Lead Contact upon request.