We used clonal cultures of Trichoplax adhaerens (Grell’s strain Н1, from the Red Sea). We maintained individuals in closed Petri dishes with artificial seawater (ASW, 35 ppm, pH 3–5 days), which was changed (70% of the total volume) every 3–5 days. At the same time, a suspension of the green alga Tetraselmis marina (WoRMS Aphia, ID 376158) was added to the culture dishes. We maintain placozoans at the constant temperature of 24 ⁰С and natural light in environmental chambers (Romanova et al., 2022).

We have routinely observed the relatively stereotyped feeding cycles with the green algae Tetraselmis or cyanobacteria Oscillatoria as food (Figure 1C).

However, on cyanobacteria Oscillatoria, animals can unexpectedly stop feeding and begin to glide around for several hours (Supplementary Figure S2). We could not reproduce these long gliding episodes or elucidate their factors. Therefore, only Tetraselmis biofilm has been used in pharmacological tests on feeding behavior (below). Nevertheless, we occasionally employed cyanobacterial biofilm for illustrative purposes of stereotyped patterns (Supplementary Videos S2, S9, S10) because of its uniform color and absence of carbonate sediments.

We have recorded 861 h of time-lapse video, usually recording 8–12 animals at once (over 8,000 animal hours). This dataset included 280 hours (h) of normal behavior (118 h on clean glass and 162 h on algal biofilms), 32 h of experiments with altered salt composition, 96 h with glutamate and aspartate, 278 h with GABA, 75 h with glycine, and 100 h with ATP.

Animal behavior was observed in small (52 mm) glass Petri dishes. Canon PowerShot SX100 camera in macro mode and Gphoto2 software was used for making time-lapse photo series of entire Petri dish with 10 or 30 s intervals. To increase contrast, the experimental Petri dish was installed on a wet black background and illuminated by a thin (3 mm) layer of light from all sides using a custom-made illumination device.

Experimental animals were transferred from a cultural dish in the late exponential phase to a small clean Petri dish for 1 h. They transferred to a glass Petri dish for behavior ecording. 8–12 animals were filmed simultaneously in each experiment.

Experimental glass Petri dishes with algal biofilms were prepared for some experiments. Growing an algal biofilm from small inoculate invariably introduces numerous bright calcareous inclusions which interfere with time-lapse recording and video processing. For cyanobacteria, we placed 5–6 pieces of Oscillatoria biofilm (size ~ 1 mm) from cultural to experimental dish with clean ASW, waited 2–3 days for cyanobacterial filaments to spread across the dish, and gently removed residual pieces of transferred thick biofilm. After another 24 h, cyanobacteria spread to clean patches left after removing dense biofilm remnants, and fresh, uniformly thin biofilm was ready for experiments. For green algae biofilm, we scraped some Tetraselmis cells from a cultural dish with a pipette, transferred 200–300 μL of the cell suspension to a microcentrifuge tube, and waited for 1 min. After that, we gently moved most of the suspension (discarding the bottom 50 μL with calcareous sediment) to the experimental dish and waited for 1–2 h for algal cells to attach to the glass. All amino acids and ATP were from Sigma-Aldrich. Stock solutions were prepared for each experimental day.

Ca-free and high-Mg artificial seawater (ASW) recipes were based on the ASW recipe from Cold Springs Harbor Protocols (Seawater, 2012). Ca-free artificial seawater was prepared as follows: NaCl: 436 mM; NaHCO₃: 2 mM; KCl: 9 mM; MgCl₂x6H₂O: 23 mM; MgSO₄x7H₂O: 25.5 mM; dissolved in distilled water. For behavior experiments with low Ca^2+, animals were initially recorded in ~1 ml drops of normal ASW (10 mM CaCl₂) and then 6X volume of Ca-free seawater was added to the final concentration of CaCl₂ 1.5 mM.

For behavior experiments with 200 mM MgCl₂ animals were initially recorded in normal ASW (48 mM MgCl₂) and then an equal volume of high-Mg seawater was added to a final concentration of MgCl₂ 200 mM. We have tested the effects of ASW with 200 mM MgCl₂ and appropriately decreased Na⁺ (standard ASW contains 48 mM MgCl₂).

Time-lapse sequences of Trichoplax behavior were imported to the Fiji ImageJ program (Schindelin et al., 2012). Animal positions and paths were determined using the WrMtrck plugin (Nussbaum-Krammer et al., 2015). Illustrative videos were created from the same time-lapse photo series using Avidemux video editor¹.

Track data was analyzed in OpenOffice Calc (graphs, Welch t-test, straightness and turn angle calculation). There are several measures of animal path tortuosity described in the literature: Intensity of Habitat use (IU), Fractal D, MSD (Mean Squared Distance), Straightness (ST), and Sinuosity (SI; Benhamou, 2004; Almeida et al., 2010). In our data, path shape changes in different conditions were best captured by the simplest of these measures, the straightness index, which is just displacement (linear distance between start and finish) divided by path length. It takes values from 0 (closed loop) to 1 (straight path). We used 5-min and 60-min straightness in this work.

Feeding cycles were counted manually. Power spectral density was calculated using the Signal package from SciPy library (Virtanen et al., 2020) using a custom Python script.

We used the data from the sequenced genomes of animals deposited in GenBank and elsewhere (listed in Supplementary Table S2). The search for possible homologs was performed using sequence similarity methods (BLAST/DELTA BLAST) algorithm using all human iGluRs, mGluRs, GABA-Brs, and vGluTs from SwissProt database as initial queries. Protein sequences were aligned in MAFFT Online (Katoh et al., 2005, 2019).² Phylogenetic trees were inferred using either the Maximum Likelihood algorithm implemented in IQTREE 1.6.12 (Nguyen et al., 2015) with automatic choice of evolutionary models or the Bayesian algorithm in MrBayes 3.2.6 (Ronquist et al., 2012). Maximum Likelihood tree robustness was tested with 2000 ultrafast bootstrap replicates (Hoang et al., 2018). Trees were visualized and annotated in iTol WEB (Letunic and Bork, 2021).

Here, we observed spontaneous oscillations in the locomotion of Trichoplax adhaerens. We also revealed the induction and modulation of endogenous locomotory and feeding cycles by transmitter candidates such as enantiomers of glutamate and aspartate, GABA, glycine, and ATP. The administration of each of these molecules resulted in induction or inhibition of well-coordinated behavioral responses suggesting that these are endogenous signal molecules. They might represent the cohort of signaling molecules, enabling behavioral integration in the common metazoan ancestor (Moroz et al., 2021b).

Integrative systems in biology express themselves in global electrical, chemical, and mechanical oscillations on the scale of the entire brain or organism (Hanson, 2021). These low-frequency electrical oscillations are often spontaneous (not caused by external stimuli), ubiquitously found in neural systems from Hydra to humans, and are also shown for higher plants, fungi, and bacterial biofilms (Hanson, 2021). Spontaneous mechanical and chemical oscillations are well-known as growth and development regulation factors from hydrozoans to mammals (Tsiairis and Aulehla, 2016; Holmes et al., 2018; Marfenin and Dementyev, 2019).

In simpler animals like Trichoplax, analyses of mechanisms and integrative functions of such behavioral oscillations are important because placozoans lack chemical or electric synapses, gap junctions, pannexin, and connexin genes.

For example, the feeding cycle in placozoans includes coordinated changes in cilia beating, substrate adhesion, and body shape changes. Under the volume transmission hypothesis (Moroz et al., 2021b), we can expect the integrative function of established but dynamic chemical gradients (e.g., amino acids) in placozoans. Predicted chemical oscillations can be coupled to electrical signals and play a role in information processing (Romanova et al., 2020b). Mechanical integration is also possible as ultrafast contraction waves propagating across the placozoan body were observed (Armon et al., 2018).

Glutamate and ATP are among the most ancient signaling molecules in all life domains. They are abundant in cytoplasm, released to the extracellular medium upon cell rupture, and used as injury signals in a wide range of organisms, including plants and bacteria (Moroz et al., 2021b). GABA is a metabolite of glutamate pathways and is broadly utilized as a signaling molecule even in unicellular eukaryotes (Delmonte Corrado et al., 2002; Bucci et al., 2005). Here, we show that both pathways are involved in controlling feeding and locomotory patterns without overlapping functions, although the sources and cellular targets of glutamate/GABA release and action are currently unknown. The predicted sources could be primary extracellular such as products of extracellular digestion following its diffusion or selected uptake. However, specific glutamatergic or GABAergic cells might also exist. Such cells can selectively accumulate these molecules and release them endogenously or following selective stimuli. The co-existence of both scenarios is to be determined in future studies.

The genome of Trichoplax adhaerens encodes numerous components of predicted peptidergic, glutamatergic, GABAergic, purinergic (ATP), and nitric oxide-mediated transmission components (Srivastava et al., 2008; Moroz and Kohn, 2015; Moroz et al., 2020a,b, 2021b) but their spatial cellular distribution is also unknown. It was shown that secreted peptides could regulate some parts of the feeding cycle, some homologous to vertebrate endomorphins (Senatore et al., 2017; Varoqueaux et al., 2018) and peptidergic cells might also use certain amino acids as co-transmitters with canonical or novel regulatory release mechanisms. Some peptide-secreting cells are flask-shaped, have primary sensory cilia, and express secretory (exocytosis/presynaptic-like) components such as SNAP-25 and synaptobrevin (Smith et al., 2014).

We conclude that Trichoplax utilizes peptides (Senatore et al., 2017; Varoqueaux et al., 2018) and amino acids (our data) as endogenous signaling molecules. In the nervous systems of vertebrates and many invertebrates peptides exert slower effects via metabotropic receptors, while amino acids usually have faster postsynaptic actions mediated by ionotropic receptors (Greengard, 2001; Jing and Weiss, 2001; Nusbaum et al., 2001). This distinction is less evident in cnidarians and other invertebrates with a diverse set of peptide-gated ion channels of the DEG/ENAC family, which mediate fast signaling (Lingueglia et al., 1995; Grunder and Assmann, 2015; Schmidt et al., 2018). Trichoplax behavioral reactions are relatively slow, on the order of minutes, and ionotropic and metabotropic receptor-mediated effects are within the physiological range.

There are several descriptions of Trichoplax behavior in the existing literature (Okshtein, 1987; Seravin, 1989; Ueda et al., 1999; Heyland et al., 2014; Smith et al., 2015; Senatore et al., 2017; Fortunato and Aktipis, 2019; Smith et al., 2019). However, all papers before the 1990s used manual drawings or short photo sequences based on observations of a few animals and lacked statistical analysis. More recent articles (Smith and Mayorova, 2019) focused on shorter and smaller scales—several mm, compared to the 52 mm dish in our work. Only one paper reported video recording and statistical analysis of placozoan locomotion on a timescale of hours and spatial scale of over 10 mm (Ueda et al., 1999), but it does not present any tracks. It is based on an unknown (presumably small) number of observed animals.

We have recorded and digitized tracks for over 2,500 animal hours of normal placozoan behavior, testing at least 50 animals for each type of behavior and using long record times of up to 24 h. This is the most comprehensive analysis of normal placozoan behaviors performed so far (see also a recent preprint by Zhong et al. (2022) about a very detailed analysis of Trichoplax thermotaxis). Also, we made an effort to standardize animal conditions before testing by using only animals from the late exponential phase of culture growth. This approach helped to mitigate the poor reproducibility of placozoan behavior.

These efforts allowed us to observe the long-term dynamics of endogenous placozoan locomotion patterns: gradual acceleration of locomotion and its long-term oscillations. Gradual acceleration was observed when video recording was longer than 2 h and must be considered in all prolonged experiments with Trichoplax. These datasets also provide unique opportunities to study integrative systems of placozoans underlying long-term modulation of behaviors in the simplest free-living animals.

Trichoplax, an animal without neurons, synapses, and muscles, exhibits the complex feeding cycle—a coordinated and stereotyped sequence of patterns leading to changes in cilia beating, body shape, cell adhesion, and secretion (Smith et al., 2015). The feeding cycle is activated by external stimuli (food), but it might also occur spontaneously due to endogenous center pattern generator activity phenomenologically similar to bilaterians and cnidarians. We also report endogenous oscillations in Trichoplax’s locomotory rhythm for directional motion and rotations.

High [Mg²⁺] or low [Ca²⁺] have abolished endogenous locomotory and feeding oscillations, suggesting the requirement for intercellular chemical communications for these behaviors. A high magnesium concentration causes stable hyperpolarization of cell membranes and suppresses action potentials. Low [Ca²⁺] also decreases the evolutionarily conserved machinery of calcium-dependent exocytosis. These observations are consistent with the hypothesis of volume transmission as the ancestral integrative systems using electrochemical vesicular secretion (Moroz et al., 2021b).

L-glutamate is a versatile signal molecule in various organisms – from prokaryotes to bilaterians (Moroz et al., 2021a). D-aspartate is a co-agonist of NMDA-type glutamate receptors and was found in Trichoplax in sub-millimolar concentrations (Moroz et al., 2020b). D-glutamate also induced action potentials and muscle contraction in ctenophores (Moroz et al., 2014).

We have demonstrated that glutamate, aspartate, glycine, and GABA profoundly affect the locomotory and feeding patterns of the Trichoplax. L-glutamate (and, to less extent, aspartate isomers and D-glutamate) turn on the feeding-like behavior sequence even in the absence of food. Glycine and GABA have more subtle modulating effects. Glycine suppresses feeding on algae and increases curvature tracks on clean dishes. GABA suppresses feeding and locomotory rhythms and acts on orders for several hours.

Thus, GABA and glycine might act as antagonists of glutamate in regulating Trichoplax feeding behaviors. In mammalian CNS, glutamate is a predominant excitatory transmitter, while GABA and glycine—are primary inhibitory transmitters. The functional ‘antagonism’ of GABA and glutamate can be an ancient trait related to their metabolic relationships (Moroz et al., 2021a). GABA is synthesized from glutamate by glutamate decarboxylase. When the time-limited response to glutamate is desired, an antagonistic response to GABA is the simplest way to coordinate the overall outcome. The concerted action of these two transmitters forms a feed-forward loop (Alon, 2007).

L-Glutamate is a universal animal ‘food’ signal (Torii et al., 2013; Moroz et al., 2021a). It is not surprising that in Trichoplax, glutamate also invokes components of feeding-like behavior. It is interesting that a high concentration of L-glutamate (3 mM) caused irreversible body expansion in addition to other components of the feeding cycles (ciliary beating arrest, adhesion, and deadhesion). Enantiomers of aspartate do not show this effect in all tested concentrations. Such a situation may point to the dual role of L-glutamate in the placozoan feeding sequence: (1) initiation of feeding sequence, also caused by aspartate, and (2) regulation of body expansion and contraction, more associated with L-glutamate.

Sponges (phylum Porifera) are sedentary filtrators. Their coordinated locomotory responses are limited to contractions of the osculum (water canal system opening), rhythmic contractions of the entire canal system, and “sneezing”—biphasic expansion and contraction of the whole body to remove the unwanted particles from canals (Leys, 2015). Most sponges are challenging to maintain in the lab, and most experimental data come from freshwater Ephydatia muelleri and marine Tethya wilhelma. Tethya responds with “sneeze” to glutamate, GABA, glycine, acetylcholine, serotonin, dopamine, cyclic AMP, caffeine, and nicotine (Ellwanger and Nickel, 2006; Ellwanger et al., 2007). In Ephydatia, glutamate stimulates “sneeze,” while 0.1 mM GABA suppresses “sneeze” responses both to glutamate and mineral particles (Elliott and Leys, 2010). Mammalian mGluR antagonists AP3 and kynurenic acid inhibit ‘sneeze’ of Ephydatia in response to glutamate and mineral particles (Elliott and Leys, 2010), suggesting the involvement of mGluRs in the coordination of this reaction.

The role of action potentials and other electric processes in sponge coordination is unclear. An early study by Prosser (1967) has shown that contractions in marine sponges are regular in high extracellular potassium (100 mM) and even increase in amplitude in seawater with a complete exchange of sodium to potassium (450 mM). These high [K⁺] concentrations cause stable membrane depolarization and complete blockade of action potentials. Therefore, desmosponges seem to coordinate body contraction without classical Na-dependent action potentials or have action potentials based on another set of ions.

In summary, we emphasize that glutamate causes coordinated behaviors in both sponges and placozoans. Effects of GABA are likely antagonistic to glutamate in Trichoplax and Ephydatia but synergistic in Tethya. Our data on the effects of high [Mg²⁺] and low [Ca²⁺] are compatible with the role of calcium-dependent exocytosis of signal molecules in integrating Trichoplax behaviors. However, the role of the Ca-dependent secretion of signal molecules in the coordination of sponge behavior is unknown.

Our experiments provide initial information on the roles of glutamate, GABA, glycine, and ATP in integrating placozoan behaviors. This situation might represent the ancestral models of chemical integration, via volume transmission, in early metazoans without canonical neurons and synapses.

However, future experiments must be conducted by deciphering cellular bases of multiple behaviors to demonstrate the specific roles of these amino acids (and other signal molecules) in Placozoa conclusively. The first priority is the unbiased identification of endogenous sources of glutamate, GABA, glycine, and ATP: the identification of specific cells and release mechanisms. In other words, we need to know the localization, quantitative secretion of these molecules, and the distribution of such hypothetical chemoconnectomes in the simplest free-living metazoans (Moroz et al., 2021b). The strategy can employ conventional approaches (e.g., single-cell RNA-seq, in situ hybridization to putative glutamate vesicular transporter mRNAs, immunocytochemistry, etc.). However, the predicted dynamic nature of numerous chemical gradients requires innovative components for real-time physiological measurements of intercellular communications in placozoans, complemented with microchemical quantitative assays.