The cnidarian Hydra vulgaris has a simple nervous system with a few hundred neurons in distributed networks. Yet Hydra can perform somersaults, a complex acrobatic locomotion. To understand the neural mechanisms of somersaulting we used calcium imaging and found that Rhythmical Potential 1 (RP1) neurons activate before somersaulting. Decreasing RP1 activity or ablating RP1 neurons reduced somersaulting, while two-photon activation of RP1 neurons induced somersaulting. Hym-248, a peptide synthesized by RP1 cells, selectively generated somersaulting. We conclude that RP1 activity, via release of Hym-248, is necessary and sufficient for somersaulting. We propose a circuit model to explain the sequential unfolding of this locomotion, using integrate-to-threshold decision-making and cross-inhibition. Our work demonstrates that peptide-based signaling is used by simple nervous systems to generate behavioral fixed action patterns. 

Detailed methods are provided in the online version of this paper.

Lead contact

Further information and requests for resources and reagents should be directed to and will be fulfilled by the Lead Contact, Rafael Yuste (rmy5@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 Mendeley as well as this repository. All the codes are available at GitHub (https://github.com/NTCColumbia). All data are archived at the NeuroTechnology Center at Columbia University.

EXPERIMENTAL MODEL AND SUBJECT DETAILS

Hydra

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

Transgenic lines

In the study, we used newly created transgenic Hydra expressing GCaMP6s or GFP in RP1 neurons, and previously-created transgenic Hydra expressing GCaMP6s in neurons⁶ or in ectoderm epitheliomuscular cells.²⁰ For RP1 neuronal GFP or GCaMP6s lines, we conducted microinjection of DNA plasmids (0.5 mg/mL) to Hydra oocytes. The DNA plasmid injected had a modified version of the pHyVec1 plasmid (Addgene cat#34789) where we replaced the Actin-promoter and -terminator sequence with Hym-355 promoter (997 bp of 5-genomic region) and terminator sequence (500 bp of 3-genomic region).

METHOD DETAILS

Behavioral characterization

Wide-field imaging of Hydra behavior was conducted in bright field at 1Hz using a fluorescence dissecting microscope (Leica M165) equipped with 0.63x Plan Apo objective, and a sCMOS camera (Hamamatsu ORCA-Flash 4.0). Hydra was either placed in a well of a glass-bottom dish (free behavior: 700–750-µm depth) or between glass cover slips with a spacer (mounted: 150–200-µm depth). In either preparation, Hydra exhibited spontaneous free behavior such as body-column contractions, tentacle contractions, and somersaulting with full or limited range of motion. Using DeepLabCut, a machine-learning neural network-based method, we tracked four different body parts of Hydra over time (Figure 1B). The derived values were used to compute parameters such as body length, basal disc speed, and foot angle. The parameters for each step (S0-S5 or contraction) were computed using custom code after manually annotating the initiation time of each step. All custom-made code was written in Python, with the exception of peak detection of contraction, which was written in Matlab. 

Detection of somersaulting and contraction

Using the data from DeepLabCut, a method to identify somersaulting or contraction automatically were designed. We focused on basal disc velocity, i.e. the velocity at which basal disc apex moves, and analyzed its changes as a function of time using a custom-made code, finding that rapid increases in basal disc speed were always correlated with somersaulting (Figure 1E; threshold = 1 to 5 pixels/(seconds)²). Although foot detachment can lead to Hydra floating in the media, we never observed foot detachment that did not result in somersaulting. To automatically detect body contraction, we analyzed changes in body length as a function of time using a custom code, and found that rapid decreases in body length were always correlated with body contractions (Figure 1E; threshold = -0.1 to -0.5 pixels/(seconds)²). Visual inspection of the movies confirmed that the significant decreases in body length occurred during the onset and subsequent steps of contraction burst.¹⁸ Thus, fast foot release (S2) is a key initial step in of somersaulting, as it is found only during somersaulting, and, conversely, somersaulting always involves fast foot detachment.

Head preparation and analysis of neuronal and nematocyte activity

The head, including of the hypostome, the dome containing the mouth, and tentacles, was isolated by slicing with a scalpel. The isolated head was placed in culture medium at 18°C for 4 hours to heal the wound prior to imaging. As a head has smaller movements during behavior, it enabled single-cell tracking, and has nematocytes and all neuronal cell types, including CB and RP1 neurons. We generated an in vitro preparation consisting of the isolated head, after sectioning the body wall near the tentacle’s insertion point (Figure S2A). The position of GCAMPs fluorescent cells were tracked using a machine learning approach (Figure S2B) and analyzed their calcium dynamics in this in vitro preparation (Figure S2C). CB and RP1 neurons were identified by their characteristic activity patterns, as well as many nematocytes, which were identified by their larger size and brighter basal fluorescence (Figure S2B).

Detection of RP1 bursts

An RP1 burst is defined as a sharp increase in RP1 frequency that precedes step S3 of somersaulting. An RP1 burst initiates when the slope of the plot is greater than 0.05, and it was detected by custom Python code.

Generation of neuron-reduced animal

Hydra will become motionless once all neurons are removed, so we tested how reducing number of neurons by colchicine, instead of removing all, affect Hydra’s behavior. Neuron-reduced animal was made by incubating Hydra in 0.4% colchicine, an inhibitor of cell division that affects interstitial cells, which are precursors to neurons, at 18°C overnight. Hydra was kept in at 18 °C for approximately 20 days and allowed to fully recover before used in experiments. The medium was replaced twice daily and kept as clean as possible.

Grafting

A foot of Hydra expressing GFP in RP1 neurons was sliced and replaced with a foot of Hydra expressing GCaMP6s in RP1 neurons (Figure S6A), We thus generated hybrid Hydra which had fewer RP1 neurons than normal ones, but had GCAMP6s-expressing RP1 neurons in the foot, and imaged these grafted animals in mounted preparations (Figure S6A-C). Hydra was sliced with a scalpel and skewed on a fishing line. The two body pieces were pressed against each other facing wounded areas. The Hydra was allowed to heal and connect overnight and were removed from the fishing line. The resulting Hydra (Foot RP1 neurons express GCaMP6s, and the rest of RP1 neurons express GFP) was further allowed to recover for 4 hours before using for the experiment.

Induction of somersaulting by photostimulation of RP1 neurons

Imaging experiments were performed by mounting Hydra expressing GCaMP6s in RP1 neurons between glass cover slips with a spacer (mounted: 150–200-µm depth). The imaging setup and the objective lenses were completely enclosed with light-shielding cloth and black electrical tape to prevent light from leaking into the photomultiplier tubes. We used calcium imaging to monitor the activity of RP1 neurons. Two-photon imaging and photostimulation were performed with two different femtosecond-pulsed lasers attached to a commercial microscope. An imaging laser (Ti:sapphire; λ = 940 nm) was used to excite a GCaMP6s while a photostimulation laser (low repetition rate pulse-amplified laser; λ = 1040 nm) was used to excite RP1 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 (~270 s) with a sample rate of 3.64 Hz were collected (Imaging laser power < 50 mW; dwell time 2 ms/pixel; 256 x 256 pixels in the whole field of view). We adjusted the power of photostimulation on each neuron (Photostimulation laser power ~20 mW) to the lowest value that elicited a response (> 2SD) and minimized cell damage. These experiments confirmed the cell-specificity of two-photon photoactivation without damaging the surrounding tissue, as tissue damage would have led to generalize activity of all neurons in the vicinity. Single-cell photostimulation was performed with a spiral pattern scanned by a pair of galvanometric mirrors delivered to the center of the cell (3 µm diameter; 3 spiral revolutions) for 30 ms. The experiments in Figure S7 confirmed the cell-specificity of two-photon photoactivation of neurons without damaging the surrounding tissue, as tissue damage would have led to generalize activity of all neurons in the vicinity. This selectivity strengthened the validity of two-photon activation to functional probe the network. Simultaneous imaging and photostimulation were controlled by Prairie View and custom-made software running in MATLAB. To overcome difficulties targeting neurons in constant motion, we utilized the Live/Ablation feature to stimulate cells on the image during live scanning.

Induction of contraction by photostimulation of CB neurons

Similar to the experiment of the photostimulation of RP1 neurons, two-photon imaging and photostimulation were performed using Hydra expressing GCaMP6s in all neurons. The CB neurons were selectively activated because some CB neurons were still fluorescent from the activity that occurred during the previous contraction.

Ablation of RP1 neurons

The ablation of RP1 neurons were achieved by using Hydra expressing GFP only in RP1 neurons. We adjusted the power of photostimulation to ablate neurons (Photostimulation laser power > 50 mW). As with photostimulation, to address the issue of Hydra’s constant movement and precise targeting of laser, we stimulated neurons one by one. Neuronal ablation was confirmed by observing the release of GFP from soma to extracellular space (Figure 3B). The imaging of Hydra behavior was conducted in bright field at 1Hz for various time points (-2h, 0h, 2h, and 4h of ablation). The -2h indicates two hours before the start of laser ablation. The laser ablation itself took about 2 hours. In addition, still images of whole Hydra body were taken at each time point to detect the number of neurons. Neurons were manually annotated based on their size and location, and the number of somersaulting and contractions were detected using the pipeline developed for behavior characterization. In the grafted animal we ablated only GFP-positive neurons in the upper body, leaving GCaMP6s-positive neurons in the foot intact. Neuronal activity was recorded from the GCaMP6s-positive neurons in the foot.

Wide-field calcium imaging

Wide-field calcium imaging of Hydra was conducted at 2 Hz for population analysis and 5-10 Hz for using a fluorescence dissecting microscope (Leica M165) equipped with a long-pass GFP filter set (Leica filter set ET GFP M205FA/M165FC), 1.6x Plan Apo objective, and a sCMOS camera (Hamamatsu ORCA-Flash 4.0). A mercury arc lamp was used to illuminate the sample. For all imaging, Hydra were mounted between coverslips (100- to 200-µm depth) except for the Figure S3. In Figure S3, Hydra was placed in a well of a glass-bottom dish (700–750-µm depth) to record the neuronal or epitheliomuscular activity during free behavior, by focusing into the basal disc. Ectodermal epitheliomuscular activity was recorded using transgenic Hydra expressing GCaMP6s in ectodermal epitheliomuscular cells. The correlation between foot detachment and somersaulting in free and mounted preparations (Figures 1 and S1) enabled us to study neuronal and epitheliomuscular activity during somersaulting. When imaging pan-neuronal GCAMP transgenics, we classified as RP1 or CB by using their characteristic spontaneous calcium dynamics, and used the slower calcium dynamics and distinct shape to identify nematocytes (The same applied to the two-photon imaging in Figure S7). All imaging was conducted at a room temperature ~23°C.

QUANTIFICATION AND STATISTICAL ANALYSIS

Markov Chain analysis

We constructed a Markov Chain directed graph based on the obtained experimental data to describe a sequence of possible steps in somersaulting. Each number represents the probability of the Markov process from one step to another, and each arrow indicates the direction. The process only went in one direction and not in both directions. To assess how similar Markov Chain is in between free-behavior and mounted preparation, we constructed transition matrix showing last move and next move. A two-way logistic regression analysis was performed using the values and sample size of each transition.

Neuronal population analysis

Using Fiji, we first created a ROI covering the whole-body to obtain fluorescent intensity in each frame. We then used a semi-automated program in MATLAB to detect the temporal location of CB and RP1 spikes. Basal disc RP1 neuronal activity and ectodermal epitheliomuscular activity (Figure S3) was analyzed in the same manner. Spike frequency was obtained by calculating the inverse of interspike interval. The temporal location of RP1 burst was obtained by detecting the slope of the spike frequency above 0.05 for each spike.

Single cell tracking and neuron identification

In this experiment, we used the head of the Actin-GCaMP6s transgenic Hydra, in which GCaMP6s is expressed in all cells of the interstitial stem cell lineage, including neurons and nematocytes. We first performed a population analysis of neural activity in the head to detect the time location of RP1 spikes. This allowed us to detect the RP1 burst, which identified the time in the video where the Hydra exhibited somersaulting. Using these videos, the location of the active cells was tracked using the open-source cell tracking software ICY, using parameters tuned to the unique requirements of experiments. 

We next analyzed obtained single-cell intensity profiles to identify RP1 neurons and CB neurons based on the population activity. Most of the remaining single cells that were not RP1 or CB neurons were identified as nematocytes based on their larger, rounder cell morphology and wide monophasic calcium transients. The last remaining cells were designated "others" because their spikes were not synchronized with any of the cells. These cells could be sensory neurons or independently active CB or RP1 neurons.  

Statistics

Statistics for all experiments, except for a two-way logistic regression analysis, were performed using Prism 9 (GraphPad). Adobe Illustrator was used for final adjustment of fonts, colors, and positions of figures. Drawings were also created in Adobe Illustrator.