Distinctive mechanisms underlie the emission of social electric signals of submission in Gymnotus omarorum

South American freshwater weakly electric fish of the order Gymnotiformes rely on a common and relatively fixed electromotor circuit to produce species-specific electric organ discharges (EODs). This stereotyped EOD carries information not only about species but also about sex, individual identity, maturity, season, time of day and motivational state by both rhythm and waveform modulations (Bennett, 1971; Caputi et al., 2005). A spontaneously firing hindbrain structure, the pacemaker nucleus (PN), commands the timing of the EOD. The PN contains two neuronal types: actual pacemaker cells (PM-cells) and bulbospinal projection neurons or relay cells (R-cells), which activate electromotor neurons in the spinal cord, which in turn activate the peripheral electric organ. The command for each EOD is initiated at the PM-cells and then transmitted 1:1 to R-cells. Available evidence suggests that PM-cells are electrotonically coupled with each other and with R-cells (Bennett et al., 1967; Bennett, 1971) that form a rather simple circuit with exclusive feedforward connections (Quintana et al., 2011). Although the PN itself produces and maintains the regular EOD, it receives multiple descending inputs that modulate its firing, resulting in EOD rate modulations that adjust electric signaling to ongoing environmental demands (Kawasaki and Heiligenberg, 1989, 1990; Spiro, 1997; Zupanc and Maler, 1997; Caputi et al., 2005; Comas and Borde, 2010). Despite the conserved rigid wiring of the electromotor circuit, it allows a high versatility of behaviorally relevant rate modulations such as gradual raises and falls of EOD rate without waveform distortions, transient high rate barrages of low amplitude and distorted waveform discharges termed chirps, and the cessation of EOD emission termed offs and sudden interruptions, all of which can be generally observed across gymnotiform species (Dye et al., 1989; Kawasaki and Heiligenberg, 1989, 1990; Spiro, 1997; Quintana et al., 2014).

Within pulse Gymnotiformes, early electrophysiological studies unraveled the neural mechanisms underlying social EOD rate modulations in the genus Brachyhypopomus (Kawasaki and Heiligenberg, 1989, 1990; Spiro, 1997; Quintana et al., 2014). Distinct types of modulations depend not only on the specific prepacemaker structures being activated but also on the cellular target of prepacemaker inputs within the PN and neurotransmitter receptors involved. Specifically, GABAergic prepacemaker inputs to PM-cells induce them to silence their discharge (EOD offs), whereas glutamatergic prepacemaker inputs increase their discharge rate, and hence the rate of EOD, via activation of NMDA receptors (NMDARs). In R-cells, the activation of glutamatergic prepacemaker inputs provokes chirps by causing the repetitive discharge of R-cells during AMPA receptor (AMPAR)-mediated depolarization, or sudden EOD interruptions via a large NMDAR-mediated sustained depolarization that inactivates R-cells (Kawasaki and Heiligenberg, 1989, 1990; Spiro, 1997). The robustness of these early demonstrated mechanisms, as well as the good fit between pharmacologically induced rate modulations and behavioral ones in Brachyhypopomus, led to the unproven assumption that chirps and EOD interruptions were produced in a similar way at least among pulse-type gymnotiform species.

The agonistic behavior of Gymnotus omarorum is the best understood example among teleosts of non-breeding territorial aggression (Batista et al., 2012; Jalabert et al., 2015; Quintana et al., 2016). In dyadic contests, the dominant fish defends its territory, while the subordinate fish interrupts its EOD to hide from the dominant one, emits bouts of chirps, and adopts a lower post-resolution EOD basal rate than the dominant fish (Batista et al., 2012; Perrone and Silva, 2018). Two important issues emerge from the above studies. First, as no glutamatergic actions have been ever demonstrated on the R-cells of the G. omarorum PN (Curti et al., 1999, 2006), mechanisms underlying subordinate chirps in this species are unlikely to be similar to those demonstrated in Brachyhypopomus. Second, chirps are only emitted by subordinates as the most unambiguous signal of submission (Batista et al., 2012; Quintana et al., 2016). This electric behavior is status dependent (i.e. is emitted by the subordinate fish once its condition emerges from the contest) and is modulated by the hypothalamic neuropeptide 8-arginine vasotocin (AVT), the administration of which induces an increase in chirp emission in subordinates but does not induce any change in the electric behavior of dominants (Perrone and Silva, 2018). Taken together, to discern whether or not Brachyhypopomus and Gymnotus use the same strategies to emit electric signals with similar characteristics (chirps and interruptions), electrophysiological experiments in G. omarorum should be carried out in behaviorally tested subordinate individuals. If glutamate has no effect on R-cells of subordinates, this would indicate that R-cells in G. omarorum do not possess ionotropic glutamate receptors, and thus cannot produce either chirps or sudden interruptions. The electrical signals of submission should consequently imply an alternative non-glutamatergic mechanism.

In this study, we aimed to contribute to the understanding of the versatility in the mechanisms underlying the emission of social electric signals among the highly conserved electromotor circuitry of pulse gymnotiform fish. In subordinate G. omarorum individuals, 30 min after their submissive status had been established in the behavioral arena, we explored the mechanisms of submissive electric signals by analyzing the effect of local administration of endogenous neurotransmitters and of the modulation of R-cell excitability on the activity of the PN. Our data suggest that while offs are most probably driven by GABAergic inputs on PM-cells, the emission of chirps does not involve direct glutamatergic activation of R-cells – as described in most Gymnotiformes – and probably depends on a previously unknown mechanism based on the enhancement of excitability of these cells. Together with previously described neural underpinnings of electric signaling in weakly electric fish, modulation of electrophysiological properties of PN neurons emerges as a putative novel mechanism that grants a higher degree of functional versatility to electromotor networks.

We used non-breeding adult Gymnotus omarorum (Richer-de-Forges et al., 2009) in three sets of experiments: (1) dyadic agonistic encounters (n=12), in which the dominant–subordinate status was attained in less than 25 min after social engagement; (2) in vivo simultaneous recordings of the PN field potentials and the head-to-tail EODs of subordinates (n=12) immediately after each agonistic encounter, in which the effects of neurotransmitters on PN activity and the modulation of electrophysiological intrinsic properties of its cellular components were tested; and (3) in vitro intracellular recordings of neurons of the PN (n=6), in which R-cell electrophysiological intrinsic properties and their modulation were explored.

Fish ranged from 15 to 19 cm in body length and 8.8 to 22.3 g in body mass. As sex in G. omarorum is not externally apparent (either morphologically or electrophysiologically), it was determined by gonadal inspection either before (≥1 months) or immediately after the experiments. All experiments were performed during the non-breeding season (May–August 2017).

Fish were collected as described elsewhere (Silva et al., 2003) in Laguna del Sauce (34°51′S, 55°07′W, Department of Maldonado, Uruguay), and housed in individual mesh compartments in 500 l outdoor tanks for at least 10 days before the experiments. All environmental variables were kept within the normal range exhibited in the natural habitat in the non-breeding season. Water temperature ranged from 8 to 21°C, and natural photoperiod ranged from 10 h:14 h to 11 h:13 h light:dark. Water conductivity was adjusted and always maintained below 250 µS cm⁻¹ by the addition of deionized water. Aquatic plants (Eichhornia crassipes, Pistia stratiotes, Salvinia sp.) covered the surface of the water and provided shelter for the fish. Fish were fed with Tubifex tubifex once a week.

Electric fish collection for experimental purposes was authorized by DINARA (National Direction of Aquatic Resources) and MGAP (Ministry of Agriculture and Fisheries), resolution no. 065/2004. All experimental procedures complied with ASAP/ABS Guidelines for the Use of Animals in Research and were approved by our institutional ethical committees (Comisión Bioética, Instituto Clemente Estable, MEC, 007/02/2010, and Comisión Nacional de Experimentación Animal, exp. 071140-000092-13 and 071140-000105-13).

Fish were placed (2–3 h before the experiments) in an experimental setup (30 l glass aquarium, 55×40×25 cm) that allowed simultaneous video and electric recordings following Silva et al. (2007). Water temperature, conductivity and pH of the recording tank matched those of the outdoor housing tanks. The experimental tank was fitted with one pair of electrodes attached to two parallel tank walls. The electric signals of freely moving fish recorded by these electrodes were connected to a high-input impedance amplifier (DAM 50, World Precision Instruments, Sarasota, FL, USA). Signals were low-pass filtered at 3.0 kHz, sampled at 20.0 kHz through a Digidata 1200 (Molecular Devices, San José, CA, USA) and stored on a PC for further analysis with the aid of the pCLAMP programs (Molecular Devices). All the behavioral experiments were performed in total darkness, with illumination from an array of infrared LEDs included in an infrared-sensitive video camera (TP-link cloud NC200, TP-link Technologies Co. Ltd, Nanshan, Shenzhen, China) that was focused on the top of the tank. Images were WiFi transmitted to the computer and stored for further analysis.

Following Batista et al. (2012) and Perrone and Silva (2018), we tested the non-breeding territorial aggression of G. omarorum under experimental conditions in which territory is the only resource that individuals fight for, by providing symmetric resources and resource values for the two contestants. As the non-breeding territorial aggression of G. omarorum is sex independent (Batista et al., 2012), we used either inter-sexual or intra-sexual dyads. In all experiments, a removable glass gate was raised 5 min after sunset, and fish were separated 10 min following conflict resolution. Conflict resolution was established as the moment we observed the third consecutive retreat of one fish without attacking back. As body mass difference is a proxy for dominance (Batista et al., 2012), we used dyads with a body mass difference that ranged from 9.5% to 42.5% (n=12) to predict the contest outcome, and we only proceeded to the second experiment (in vivo recordings, see below) if the smaller fish of the dyad subordinated in less than 25 min. To potentiate subordination, as AVT is known to increase the electrical signaling of submission in G. omarorum (Perrone and Silva, 2018), we administered AVT (1 µg g⁻¹ body mass of a 1 µg µl⁻¹ saline solution; American Peptide Company, Sunnyvale, CA, USA) to expected subordinates by intraperitoneal administration 30 min before the agonistic encounter. In all the experiments, the other animal of the dyad was also intraperitoneally injected with the same volume of a physiological saline solution. With the same purpose of potentiating subordination, we performed a second agonistic encounter using the same subordinate against a different larger contender (60 min after the first fight) in cases in which the electrical signaling of submission in the first fight was scarce (n=4). However, as experience may influence the outcome, dynamics and level of aggression of the agonistic encounter (Hsu et al., 2006), we only used the first encounter (in cases in which we had two) in the behavioral description presented in Results to favor statistical comparisons.

We analyzed the locomotor displays in each recorded dyad to identify the three phases of the agonistic encounter following Batista et al. (2012): (1) evaluation phase (pre-contest): from time 0 (gate removal) to the occurrence of the first attack; (2) contest phase: from the first attack to conflict resolution (resolution time); and (3) post-resolution phase (post-contest): 10 min after conflict resolution. As reported elsewhere (Batista et al., 2012), contest resolution was established when we observed the third consecutive retreat of one fish without attacking back.

We measured the following parameters in all the experiments: latency to the first attack (including nips, nudges, bites), contest duration, post-resolution attack rate (number of attacks/contest duration in minutes) of dominants and subordinates, and the time of occurrence of post-resolution dominants' attacks. With respect to electric displays, we measured the occurrence of offs (interruptions of EOD emission) and chirps (transient high-rate barrages of low amplitude and distorted waveform), post-resolution chirp rate (number of chirps per 10 min), and the time of occurrence of post-contest chirps.

Immediately after each contest, subordinate fish were prepared for simultaneous PN field potential and head-to-tail EOD recordings. Individuals were anesthetized as described elsewhere (Comas and Borde, 2010) and were placed in a plastic box with the abdomen lying on a wet sponge. All surgical surfaces and fixation points were heavily infiltrated with 2% lidocaine hydrochloride. During surgical procedures, the head was maintained in a horizontal position by a pair of plastic-tipped metal bars attached to the box and the gills were perfused with tricaine methanesulfonate (MS-222, Sigma, St Louis, MO, USA) dissolved in iced tap water (0.3 mg l⁻¹). The dorsal surface of the brain and of the left anterior lateral line nerve (ALLn) – a large-diameter nerve trunk that contains mechanosensory and electrosensory afferent fibers innervating the fish head (Carr et al., 1986; Castelló et al., 1998) – near its entrance to the brainstem were exposed through an opening in the skull to provide access to micropipettes used for recording from and drug application to the PN and for nerve stimulation. Following these procedures, the animals were injected with d-Tubocurarine (1–3 μg g⁻¹ i.m.) at doses that produced paralysis but did not completely eliminate the EOD. After surgical preparation and curarization, the gills were continuously perfused with aerated dechlorinated tap water (aquarium water) at room temperature (20–25°C). Conductivity was controlled regularly and was usually under 250 µS cm⁻¹. Immediately following electrophysiological experiments, animal were killed with an overdose of MS-222.

In these experiments, the same micropipette was used for field potential recordings and drug application. Electrical recordings were obtained using micropipettes (10–15 MΩ) filled with NaCl (154 mmol l⁻¹)-based solutions of different compounds (see below) connected to a homemade high input impedance DC amplifier. Drug-containing solutions were applied by pressure (Picospritzer II, General Valve Corporation, Cleveland, OH, USA). Pulses of 10–20 psi and 10–100 ms duration were used. In order to calibrate drug applications, before each experiment, pressure and pulse duration parameters were adjusted while visualizing the formation of microdroplets under a microscope. The volumes of these microdroplets, calculated from their diameters, were relatively small (between 5 and 15 pl). It was thus supposed that the injection affected a restricted volume of brain tissue. Taking advantage of the spatial segregation of PM- and R-cell populations within the PN, extracellular electrophysiological recordings can be obtained from each cell population independently, and drug-containing solutions (see below) can also be administered locally, separately to each cell population (Bennett et al., 1967; Ellis and Szabo, 1980; Trujillo-Cenóz et al., 1993; Curti et al., 2006). In most experiments, similar microvolumes of drug-containing solutions were injected at two different depths within the nucleus along the same micropipette vertical track passing near (±100 µm) the rostro-caudal center of the PN (called 0 µm in Curti et al., 2006). Dorsal injections were performed to apply drugs to PM-cells whereas injections performed at 200–300 µm from this location in the ventral direction allowed us to apply drugs to R-cells. Although a detailed cytoarchitectural description of the PN of G. omarorum is not yet available in the literature, specific waveforms and depth profiles of field potentials obtained along different vertical electrode tracks in the rostro-caudal axis differed from those obtained in Brachyhypopomus (Quintana et al., 2011), suggesting structural differences between the two genera. Although an exhaustive analysis was not performed, during pilot experiments significant differences in the responses to drugs (particularly glutamate) applied at different levels in the rostro-caudal axis (±100 µm from the center) were not detected. Control experiments included the injection of similar volumes of the vehicle solution (154 mmol l⁻¹ NaCl).

A Grass Technologies (Quincy, MA, USA) P15 preamplifier was used to monitor the EOD (head to tail) with a pair of metal electrodes placed next to the fish and in contact with the supporting wet sponge, with a gain of 100× and a low-pass filter (cut-off at 3 kHz).

Clampfit software was also used to fit the smooth onset and offset of EOD offs observed during behavior and of EOD interruptions evoked by GABA and glutamate injection to a single exponential function (Levenberg–Marquardt and sum of squared errors as search and minimization methods, respectively). The time constant of the onset (τ_on) and of the offset (τ_off) of the best exponential fit were determined.

The effects of the following substances (dissolved in 154 mmol l⁻¹ NaCl) were assessed: l-glutamic acid (glutamate, 10 mmol l⁻¹), GABA (1 mmol l⁻¹), 4-aminopyridine (4-AP, 10-25 mmol l⁻¹) (all purchased from Sigma-Aldrich) and AVT (1 µg µl⁻¹ saline solution). The concentrations refer to the micropipette filling solutions. During pilot experiments, dose–response relationships were evaluated and concentrations of drugs were selected to produce maximal effects with minimum volumes.

Electrical stimuli were applied to the ALLn by way of a bipolar stimulating electrode. Stimuli consisted of single rectangular current pulses (0.15–0.3 mA, 0.01–0.02 ms) that were applied using an S88 stimulator and a PSIU5 isolation unit (Grass Technologies). Stimulus strength was maximal as further increases of intensity did not evoke further increases in the amplitude of responses. A series of 10–30 stimuli were delivered either manually (minimum interval of 20 s) or at a rate of 0.05 Hz. The timing between each stimulus and the preceding EOD varied randomly. For each experiment, the electrical stimulation threshold value was determined as the intensity evoking any detectable change in EOD rate.

We followed similar surgical procedures to those for in vivo experiments except that the dorsal surface of the brain was totally exposed while bathed with cold Na-free artificial cerebrospinal fluid (ACSF)–sucrose solution (described in ‘Recording and stimulation’, below). The brain and part of the spinal cord were rapidly removed from the skull and submerged in cold ACSF–sucrose solution. Transverse sections of the brainstem (400 µm thick) containing the PN were obtained under cold ACSF–sucrose solution using a Vibratome 1000 plus (The Vibratome Company, St Louis, MO, USA), and were incubated (>1 h, at room temperature, 21–23°C) in a 1:1 solution of ASCF–sucrose and control ACSF solution (see ‘Recording and stimulation’, below). Slices were transferred to a 2 ml recording chamber fixed to an upright microscope stage (Eclipse FN1, Nikon Company, Minato, Tokyo, Japan) equipped with infrared differential interference contrast (DIC) video microscopy and a 40× water immersion objective. Slices were perfused with carbogen-bubbled ACSF (1.5–3 ml min⁻¹) and maintained at room temperature (20–23°C). Under these experimental conditions, the PN maintains its spontaneous synchronized activity with a stable firing rate for at least 6 h.

The electric activity was monitored using an Axoclamp 2B amplifier (Molecular Devices) by recording intracellularly from R-cells using patch pipettes in the whole-cell configuration. The patch pipette (5–10 MΩ) was filled with a potassium gluconate-based intracellular solution (see below). Microelectrodes were placed under visual control with a hydraulic micromanipulator (Narishige, Setagaya, Tokyo, Japan). Signals were low-pass filtered at 3.0 kHz, sampled at 10.0 kHz through a Digidata 1322A (Molecular Devices) and stored on a PC for further analysis with the aid of the pCLAMP programs. Repetitive firing behavior of R-cells was explored by injecting slow current ramps (−2 to 2 nA, 4 s) in silenced PN, or long-lasting depolarizing pulses (800–1000 ms, 1–4 nA) in spontaneously discharging neurons. When injecting current ramps, PN spontaneous rhythmic discharge was inhibited by adding a GABA_A receptor agonist (muscimol, 50 µmol l⁻¹) to the perfusion solution (Spiro, 1997; Curti, 2007). In the presence of muscimol, in the absence of PM-cell input, R-cells do not fire action potentials spontaneously. Spikes evoked by current injection or the input resistance of these neurons are not significantly affected by muscimol. In some experiments (n=6), the consequences of blockade of slow I_A-like potassium currents of PN neurons on R-cell spontaneous firing behavior were analyzed by adding 4-AP (1 mmol l⁻¹) to the ACSF solution.

ACSF contained (in mmol l⁻¹): 124 NaCl, 3 KCl, 0.75 KH₂PO₄, 1.2 MgSO₄, 24 NaHCO₃, 10 d-glucose, 1.6 CaCl₂, pH 7.2–7.4 after saturation with carbogen (Spiro, 1997), while NaCl was replaced with 213 sucrose in ACSF–sucrose solution. Intracellular solution contained (in mmol l⁻¹): 140 potassium gluconate, 0.2 EGTA, 4 ATP-Mg, 10 Hepes, pH 7.3. All substances were purchased from Sigma-Aldrich.

All data were analyzed by non-parametric tests. Unless otherwise indicated, Mann–Whitney U-test was used for independent variables with sets of data from different fish, or for comparing dominants versus subordinates. For paired comparison, Wilcoxon test was used. Accordingly, data are expressed as median±median absolute deviation (MAD) throughout. Differences were considered statistically significant when P-values were less than 0.05.

All dyads of non-breeding G. omarorum tested (n=12) displayed agonistic behavior immediately after the gate was removed and the dominant–subordinate status was established within a few minutes with the expected contest outcome (i.e. the larger fish of the dyad always won the fight). In accordance with previous reports (Batista et al., 2012; Quintana et al., 2016), all the agonistic encounters also showed similar temporal profiles and followed the typical three phases: (1) a short pre-contest of 46.5±43 s; (2) the contest, characterized by highly aggressive displays by both contenders with a duration of 228.5±100 s; and (3) the 10 min post-contest phase, in which dominants, but not subordinates, persisted in attacking while subordinates attempted to flee and emitted submissive electric signals.

EOD offs and chirps were profusely emitted by subordinates (8/12 and 6/12 produced chirps and offs, respectively) after contest resolution, while three subordinates also displayed offs and/or chirps during the contest phase. As shown in Fig. 1B, offs are complete EOD cessations preceded by a smooth decrease in EOD rate with a time constant (τ_οn) of −147.65±105.91 ms (n=12) and followed by a progressive increase in EOD rate before reaching the regular previous EOD rate with a time constant (τ_off) of 156.78±52.29 ms (n=12). Chirps are transient high-rate (170.95±51.91 Hz, n=20) barrages of low amplitude and distorted waveform discharges (Fig. 1C, left). During chirps, the full-size EOD or its remnant at the same prior rate cannot be recognized. However, the time course of this chirp-related EOD cessation does not resemble that observed during the EOD off, as the chirp occurs suddenly, interrupting the normal regular discharge with no anticipated smooth decrease in EOD rate. In contrast, chirps are usually preceded by a smooth increase in EOD rate that slowly returns to baseline after the chirp (Fig. 1C, right). Because of the relatively high frequency of irregular discharges during the chirp, the variability of the recording was comparatively much higher (3.199±0.183%) than that observed during the inter-EOD interval of the regular full-size EOD baseline discharge (0.653±0.219%).

During the post-resolution phase, the dominant fish’s attack rate was significantly higher than that of the subordinate fish (Fig. 2A). Moreover, 63% of the post-resolution attacks performed by dominants to chirping subordinates (145 out of the 231 attacks, n=8 dyads) occurred in a time window of 10 s around the emission of chirps, and 88 out of these 145 dominant attacks occurred immediately before the occurrence of one chirp. As shown in Fig. 2B, the probability of occurrence of a dominant attack was maximal during the 1 s preceding the subordinate chirp, when 38.1% of the total post-resolution dominant attacks were observed. Therefore, a common feature of the post-resolution phase was the observation of the sequence: dominant approach–dominant attack–subordinate chirp–subordinate retreat.

To explore whether submissive electric signals, offs and chirps, result from activation of NMDAR and AMPAR of R-cells, respectively, microvolumes of glutamate (10 mmol l⁻¹) were applied at different locations within the PN (Fig. 3). Injections at the level of R-cell somata in subordinate fish were ineffective in producing either chirps or EOD interruptions (Fig. 3B). We observed a small (26.87±11.77% of basal pre-injection rate) and slow (time constant of frequency rise of 189.78±60.48 ms) transient increase in EOD rate with no changes in EOD amplitude (Fig. 3B, right). In contrast, when similar microvolumes were applied near PM-cells, transient EOD accelerations of 130.76±44.11% increase with a significantly shorter time constant of frequency rise (87.36±18.09 ms; P=0.04, n=7) were systematically evoked (Fig. 3C). The sensitivity of R-cells to glutamate was further explored by comparing baseline inter-EOD variability observed immediately after glutamate injection (0.089±0.025% of EOD amplitude), which was significantly smaller than that observed during the natural chirp (3.199±0.183%; P=0.02, n=7; Fig. 3D) and similar to that observed during the regular pre-injection discharge (0.565±0.155% of EOD amplitude, P=0.81, n=7). In addition, no changes in EOD amplitude were observed after glutamate injection at R-cells; pre-injection values did not differ from post-injection ones (Wilcoxon signed-rank test, pairing pre- and post-injection values, P=0.69, n=7).

GABA applied at the level of PM-cell somata evoked interruptions of the EOD (Fig. 4A, left) that were preceded by a slow decrease in EOD rate (Fig. 4A, right) with a τ_on of 174.27±113.13 ms (n=7). Immediately after interruptions, EOD rate slowly returned to pre-injection values with a τ_off of 148.97±32.29 ms (n=7). Smaller volumes caused transient decreases in EOD rate instead of interruptions (data not shown). Similar volumes applied near R-cells were almost ineffective in changing EOD rate or amplitude (Fig. 4B). Usually, the injected volume was adjusted to provoke interruptions lasting the same length of time as behavioral interruptions (∼2500 ms). The median τ_on and τ_off of the behavioral offs (−147.65±105.91 ms and 156.78±52.29 ms; n=12) did not differ from these GABA-induced interruptions (Fig. 4C; P=0.89 for τ_on and P=0.47 for τ_off).

As the production of chirps by the subordinate fish during the agonist encounter was tightly correlated with the occurrence of dominant attacks (Fig. 2B), we tested whether an intense sensory stimulation was a prerequisite for the production of chirps. For this purpose, we mimicked the sensory pattern of activation presumably induced in the subordinate after the dominant attack by the electrical stimulation of the ALLn in immobilized subordinate fish (n=5, white arrowhead in Fig. 5A,B). Electrical stimulation of the ALLn evoked transient increases of the EOD rate with a median peak amplitude of 31.99±3.33 Hz, a 289.63±32.86% increase from basal pre-stimulus rate (Fig. 5A). Typically, after ALLn stimulation, EOD rate increased abruptly (time to peak of 76.80±16.15 ms) and slowly returned to baseline with a time constant of 351.14±75.16 ms (Fig. 5A, left). In spite of the range of intensities explored, electrical stimulation of the ALLn was ineffective in eliciting chirps by itself (Fig. 5A). In some experiments (n=3), electrical stimulation of the ALLn was combined with glutamate injection at R-cells using different timings between the two challenges. This combined stimulation pattern was also ineffective in eliciting chirps (data not shown). However, when associated with an increase in R-cell excitability by local application of 4-AP, a blocker of I_A-type voltage-activated K⁺ conductance, electrical stimulation of the ALLn systematically elicited chirp-like discharges (Fig. 5B). These signals consisted of transient (340.08±70.43 ms duration), high-rate (178.55±43.04 Hz) barrages of low-amplitude discharges of the electric organ (inset in Fig. 5B, right). The stimulation of the ALLn by itself was also ineffective in increasing baseline inter-EOD variability, which did not differ from pre-stimulation baseline inter-EOD variability (P=0.84) and was significantly lower than that observed during chirp-like discharges (P<0.01; Fig. 5C). Although some spontaneous chirp-like discharges were observed, 88.4% occurred during the first 200 ms after ALLn stimulation (Fig. 5D). An example of the effects of 4-AP applied to R-cells on PN extracellular electrical activity evoked by activation of ALLn afferents is given in Fig. 6. Before 4-AP application, field potentials evoked by ALLn stimuli recorded at the level of R-cell somata consisted of a short-latency (latency to peak ∼4 ms), brief (20–30 ms), positive potential (Fig. 6A), suggestive of a passive depolarization of R-cells. Shortly (1–2 min) after the injection of 4-AP, the ALLn stimulus evoked a relatively long-lasting (300 ms) negative field potential that followed the positive potential (Fig. 6B). Coincidently with this negative potential in the extracellular recording at R-cells, an incipient chirping activity characterized by the occurrence of isolated extra-EOD small irregular discharges was observed in the EOD recording.

Chirp-like discharges closely resembled natural behavioral chirps in many critical aspects such as the intra-chirp discharge rate and the changes in baseline inter-EOD variability (Fig. 7). Both electrical signals consisted of transient high-rate barrages of low amplitude and distorted waveform discharges of the electric organ (Fig. 7A,B, left). The discharges of the behavioral and pharmacologically induced chirps were similar and exhibited similar modal values in frequency histograms (∼180 Hz; Fig. 7A,B, right). Overall, the intra-chirp rate of chirp-like discharges (178.55±43.04 Hz) was marginally higher than that of natural chirps (170.95±51.91 Hz) (Fig. 7C; P=0.06, n=20). Baseline inter-EOD variability of chirp-like discharges and behavioral chirps was also marginally different (3.92±0.37% versus 3.20±0.18%, respectively, P=0.05). Finally, the duration of the two electromotor displays was not statistically different (340.08±70.43 ms and 233±11 ms for chirp-like discharges and natural chirps, respectively, P=0.55).

As AVT has been reported to be a status-dependent modulator of G. omarorum agonistic behavior (Perrone and Silva, 2018), we explored the effect of AVT applied at different locations within the PN in subordinate immobilized fish (n=6). In 4 out of 6 experiments, AVT was ineffective in eliciting any change of the electromotor pattern. In the two remaining experiments, AVT evoked small-amplitude chirp-like discharges with different latencies (Fig. 8). When applied to R-cells, a burst of high-rate discharges was observed immediately after AVT injection (Fig. 8B, trace 1) followed by small-amplitude repetitive discharges (Fig. 8B, trace 2). Injection of AVT at the level of PM-cells evoked rhythmic barrages of small-amplitude discharges at a high rate (Fig. 8C,D), similar to those observed after application of AVT to R-cells but with a longer delay (13 s). In both cases, during chirp-like discharges, the EOD kept its regular rate. Putative effects of AVT were reversible and bursts of discharges were no longer observed 20 min after injection.

We conducted a series of experiments to explore, in a CNS in vitro preparation containing the PN, the capacity of R-cells to sustain a high-rate repetitive discharge of action potentials when depolarized (Fig. 9) as well as to investigate changes of R-cell excitability produced by 4-AP (Fig. 10). Under perfusion of control solution, R-cells showed a spontaneous rhythmic (<15 Hz) activity similar to that observed in in vivo preparations (e.g. Curti et al., 2006). R-cells exhibited a relatively hyperpolarized basal membrane potential (−73.9±1.8 mV) and action potentials arising abruptly from the baseline (Fig. 9B). Pulse-evoked depolarization elicited the discharge of small-amplitude action potentials at a high rate (130–190 Hz; Fig. 9C). The occurrence of spontaneous rhythmic discharges that followed 1:1 the regular pacemaker command minimally altered repetitive R-cell discharge evoked by the depolarizing pulse. In the silenced PN by perfusion of GABA agonists (Fig. 9D), depolarization-induced repetitive firing of R-cells was further characterized by injecting slow ramp depolarizing currents. Characteristically, discharge rate increased almost linearly with depolarization (16.39±4.87 Hz mV⁻¹) until reaching a maximum rate (Fig. 9D, inset) and neurons usually inactivated at membrane potentials more positive than −45 mV.

The perfusion of 1 mmol l⁻¹ 4-AP provoked the cessation of spontaneous PN rhythmic activity after 6±1 min of perfusion, which was recovered after 62±32 min of washout with normal solution (n=6). Before the complete cessation of PN activity, R-cells showed spontaneous rhythmic (900±200 ms mean interval) abrupt transient plateau depolarizations that elicited bursts of small-amplitude action potentials at a high rate (136±68 Hz; Fig. 10A, inset). During the bursts there was an apparent interruption of rhythmic PN full activation. At this stage, R-cells were depolarized (basal membrane potential −60.1±9.0 mV) and full regular spikes were followed by depolarizing afterpotentials and several spikelets (Fig. 10A). In the silent PN, plateaus were still observed as rhythmic (Fig. 10B; ∼700 ms interval), abrupt and long-lasting depolarizations that usually elicited the discharge of action potentials at a high rate (Fig. 10B), although on some occasions, subthreshold plateaus were also observed in intracellular recordings of other cells in the same slice (not shown).

Since the pioneering work of Heiligenberg (reviewed in Heiligenberg, 1991), based on extensive work on both pulse- and wave-type gymnotiform fish (Dye et al., 1989; Kawasaki and Heiligenberg, 1989, 1990; Keller et al., 1991; Kennedy and Heiligenberg, 1994; Spiro et al., 1994; Spiro, 1997; Curti et al., 1999; Caputi et al., 2005; Quintana et al., 2011, 2014), it has been accepted that the EOD rate modulations depend on electromotor behavior-specific innervation patterns combining the neurotransmitter receptor subtype and the target cell type within the PN. Specifically, in G. omarorum, the absence of chirps and sudden interruptions among the electromotor behaviors displayed by this species (Black-Cleworth, 1970; Westby, 1974, 1975; Kramer et al., 1981; Barrio et al., 1991) fitted with the lack of evidence supporting the activation of the R-cells of their PN through AMPAR and NMDAR (Curti et al., 1999). New evidence obtained in this species, however, challenged the current conception regarding the neural underpinnings of the electromotor repertoire in gymnotiform fish. As an unequivocal electrical social signal of submission after the resolution of agonist encounters, only subordinate fish emit chirps and interrupt their EOD emission (Batista et al., 2012). Given that the adaptation of the electric repertoire to different environmental demands has been demonstrated to rely on context-dependent modification of the PN innervation pattern (Quintana et al., 2011, 2014), in the present study the presence of glutamate receptors in PN neurons was re-examined in subordinate fish that were emitting EOD offs and/or chirps 30 min (or less) before the experiment. In spite of this, we failed to obtain any evidence indicating direct glutamatergic activation of R-cells via AMPA or NMDA receptors in these fish. It is unlikely that the extracellular concentration of glutamate applied locally in this work was insufficient to activate R-cell glutamate receptors. Local application of microvolumes of glutamate is a common experimental technique that has been extensively used in the CNS of gymnotiform fish species to mimic the action of this neurotransmitter (Dye et al., 1989; Kawasaki and Heiligenberg, 1989, 1990; Keller et al., 1991; Curti et al., 1999; Metzner, 1999; Zupanc, 2002; Comas and Borde, 2010; Quintana et al., 2011, 2014). Here, we used similar parameters of glutamate application to those utilized in many of these studies. Moreover, the amplitude of rate responses provoked by glutamate applied to PM-cells in our experiments was similar to that of responses provoked by activation of the Mauthner cells in this species (Falconi et al., 1997; Curti et al., 1999, 2006; Comas and Borde, 2010), suggesting that the amount of glutamate applied to PN neurons was in the physiological range. The possibility of uneven (and probably asynchronous) activation of a few scattered R-cells by glutamate injections is also unlikely. This type of R-cell activation through AMPARs would increase the baseline inter-EOD variability, equivalent to subtle chirping activity, whereas if NMDARs were involved, a decrease in EOD amplitude would be expected as the result of NMDAR-mediated inactivation of a small group of R-cells. None of these effects were evoked by glutamate when applied to R-cells. In spite of the above arguments against the possibility of activation of glutamate receptors of R-cells, the occurrence of glutamate-evoked subthreshold depolarization of these cells without any detectable consequence in the EOD recordings (even during strong ALLn stimulation) cannot be definitely ruled out in our experiments. Taken together, the above considerations indicate that alternative non-glutamatergic mechanisms should operate at the PN to produce submissive electrical signals in G. omarorum.

Cessations in the emission of electric signals have been recognized as general submissive displays in several species (Black-Cleworth, 1970; Westby, 1975; Batista et al., 2012; Zubizarreta et al., 2012). The characteristic time course of onset and offset of interruptions together with a stable potential in head-to-tail recordings during the absence of regular discharges allowed us to identify this submissive signal as being distinct from sudden interruptions provoked by NMDAR-mediated sustained depolarization of R-cells. Glutamatergic sudden interruptions usually occur without significant leading or following modulation of EOD rate (Kawasaki and Heiligenberg, 1989; Spiro, 1997; Caputi et al., 2005; Quintana et al., 2011). Moreover, as the onset and offset of EOD interruptions evoked by injection of microvolumes of GABA at PM-cells exhibited almost the same time course to that observed during behavioral offs, our results suggest that offs most likely result from the activation of GABA receptors on PM-cells. Although this is the first report of the effects of GABA on PN discharge in G. omarorum, similar results were obtained in Brachyhypopomus, in which interruptions were also evoked by stimulation of specific diencephalic structures containing inhibitory prepacemaker neurons and blocked by GABA antagonists applied to the PN (Kawasaki and Heiligenberg, 1989, 1990). Although these prepacemaker structures have not been described yet in G. omarorum, GABAergic prepacemaker neurons could be selectively activated by descending inputs to produce offs in the subordinate fish. However, this speculation is difficult to experimentally test as it requires, for example, evaluation of the effects of the blockade of GABA receptors of PM-cells in the behaving animal.

In gymnotiform fish, chirps consist of transient (50–400 ms) high-rate barrages of low-amplitude, distorted-waveform discharges that are produced by repetitive discharge of R-cells at a high rate (up to 180–200 Hz) in response to a transient suprathreshold AMPAR-mediated depolarization (Kawasaki and Heiligenberg, 1989, 1990; Spiro, 1997; Quintana et al., 2014). In G. omarorum, submissive chirps exhibited similar high-rate, low-amplitude discharges of the electric organ and probably also rely on the repetitive high-rate discharge of R-cells, although, unlike Brachyhypopomus, the suprathreshold depolarization of these cells cannot be explained by ionotropic glutamate receptor activation. One alternative mechanism for R-cell depolarization in this species involves their passive depolarization secondary to PM-cell depolarization evoked by excitatory inputs as the two groups of cells are electrotonically coupled (Bennett et al., 1967; Bennett, 1971). Several lines of evidence suggest that this might occur in the subordinate fish. First, previous work indicates that PM-cells are the exclusive direct cellular target of prepacemaker glutamatergic inputs to the PN in G. omarorum (Falconi et al., 1997; Curti et al., 1999, 2006). Second, the close correlation of submissive chirps with dominant attacks indicates that chirp emission is probably associated with a strong excitatory drive upon the PN derived from the massive recruitment of sensory afferent inputs elicited by the assault. Finally, the fact that submissive chirps are usually preceded by a gradual increase and followed by a slow decline in EOD rate suggests that the depolarization of PM-cells is critical for chirp production. In this study, the activation of sensory inputs provoked by the dominant attack to the subordinate fish was mimicked by electrical stimulation of the ALLn at a high strength. The short-latency positive extracellular potential recorded at the level of R-cells provoked by the activation of these sensory afferents suggests that depolarizing inputs to PM-cells can passively depolarize the R-cells. However, although strong ALLn stimulation induced an abrupt and long-lasting acceleration of the EOD, indicative of a large depolarization of PM-cells, it was ineffective in eliciting chirps. Probably, during the contest in the behaving fish, concurrent activation of inputs of other sensory modalities (Moller, 2002) as well as of descending neuromodulatory influences upon the electromotor system (see below) is required to stimulate the emission of chirps in the subordinate fish.

As in Brachyhypopomus (Spiro, 1997), we demonstrated that G. omarorum R-cells can indeed discharge action potentials at a high rate (up to 200 Hz) when depolarized. According to the level of membrane depolarization, R-cells can discharge repetitively at rates in the range of intra-chirp rates observed during submissive chirps and chirp-like discharges. Consequently, a concurrent increase in R-cell excitability boosting the response of these cells to passive depolarization is probably needed to bring these cells to fire repetitively during chirps. Here, an increase in R-cell excitability was induced by 4-AP, a K⁺ channel blocker that has been widely used as a pharmacological tool for both experimental and therapeutic purposes (Kita et al., 1985; Jahnsen, 1986; Bekkers and Delaney, 2001; Hayes, 2004; Avoli and Jefferys, 2016). In a concentration-dependent manner, 4-AP is particularly effective as a blocker of K⁺ channels of the broad I_A-type family, increasing neuronal excitability and facilitating the discharge of transient depolarizing plateau potentials (Kita et al., 1985; Voskuyl and Albus, 1985; Szente and Baranyi, 1989). In wave-type gymnotiform fish, pharmacological evidence obtained in vitro suggests that a 4-AP-sensitive potassium current is involved in the control of excitability of PN neurons although plateau potentials have not been reported (Dye, 1991; Smith, 1999; Smith and Zakon, 2000). Low concentrations of 4-AP initially increased pacemaker firing frequency and, shortly after, stopped pacemaker firing. In G. omarorum, in addition to stopping pacemaker firing, bath application of 4-AP increased excitability of R-cells, leading to the discharge of a burst of spikelets at a high rate (150–250 Hz) following each rhythmic action potential and during spontaneous transient depolarizing plateau potentials. Under blockade of 4-AP-sensitive K⁺ channels of R-cells in vivo, subordinate fish emit spontaneous chirp-like discharges, the occurrence of which was greatly facilitated by strong synaptic depolarization of PM-cells. These results suggest that 4-AP-sensitive K⁺ channels may participate in the control of excitability of PM- and R-cells in G. omarorum and signals probable differences in the functional role of this membrane conductance in the PN of wave-type gymnotiform fish.

Boosting the responses of R-cells to passive depolarization by blocking K⁺ channels of the broad I_A-type family may be the basis of submissive chirp emission. Moreover, an indirect increase in the strength of electrotonic coupling between PN neurons because of K⁺ channel blockade may also be involved in chirp mechanisms (Pereda et al., 2013). The occurrence of a slow negative extracellular potential (indicative of active depolarization) at R-cells correlated with chirping activity evoked by ALLn stimulation after 4-AP application to R-cells in the subordinate fish supports this notion. Even though during chirps, boosted slow depolarizing responses or spontaneous plateau potentials could involve most R-cells, the magnitude of the baseline inter-EOD variability (<3% of EOD amplitude), a parameter that depends directly on the amplitude of small high-rate discharges, suggests that synchronization of action potentials of R-cells triggered by slow depolarizing electrical events is poor. This is consistent with the well-known low-pass filter properties of electrical synapses (Galarreta and Hestrin, 1999; Gibson et al., 1999, 2005), although electrotonic coupling between PN neurons in G. omarorum and its frequency-dependent behavior remain to be explored. The above-proposed model of chirping may account for the apparent loss of regular rhythmic PN discharges during chirps, a phenomenon that has been described in Brachyhypopomus as the functional uncoupling of R- and PM-cell discharge (Kawasaki and Heiligenberg, 1989). Further investigation is needed to obtain evidence about the electrophysiological properties of the PN neurons and their connectivity within the nucleus in pulse gymnotiform fish to support the proposed model.

Nonapeptide regulation of social behavior is a conserved feature in vertebrate evolution, particularly in teleosts (Godwin and Thompson, 2012; Stoop, 2012). In gymnotid fish, AVTergic fibers project to areas related to the control of social behavior and electromotor displays (Pouso et al., 2017), and a role of AVT in the modulation of emission of social electric signals via V1a receptors in PN neurons has been documented (Perrone et al., 2010, 2014). More recently, Perrone and Silva (2018) demonstrated that AVT enhances the electric signaling of submission in subordinates and modulates the intensity of aggression and the readiness to attack in dominants in G. omarorum. Inhibition of several potassium currents (including 4-AP-sensitive I_A-type K⁺ currents) has been described as a mechanism of neuromodulatory effects of AVT in several neural systems (Raggenbass, 2008; Breton et al., 2009; Stoop, 2012), raising the intriguing possibility that AVT may act as an endogenous modulator of R-cell excitability during contests. Despite the above data, an effect of AVT compatible with an increase in R-cell excitability was detected in only two out of six experiments. Differences in the latency and magnitude of effects suggest that AVT is acting directly on a specific group of AVT-sensitive R-cells that were probably not affected by AVT in most experiments. The development of immunolabeling methods of specific AVT receptors to identify the cellular targets of AVT within the PN in gymnotid fish may provide critical evidence to explain the above results and will contribute to understanding more precisely the modulatory effects of AVT on the electromotor system.

Our results indicate that submissive electric signals in G. omarorum do not involve direct glutamatergic mechanisms. Whereas EOD interruptions are likely to result from activation of GABAergic receptors of PM-cells, emission of chirps most probably does not involve direct glutamatergic activation of R-cells and involves a mechanism based on the enhancement of the excitability of these cells. Modulation of electrophysiological properties of R-cells probably triggered by an as-yet unknown neuromodulatory input activated during contests represents a previously unknown mechanism for the production of electric social signals in gymnotiform fish. Whereas Brachyhypopomus and Gymnotus most probably share a cellular mechanism for the generation of chirps based on the synchronous repetitive high-rate discharge of action potentials of a group of R-cells, the two genera most likely differ in the mechanisms underlying the depolarization capable of triggering high-rate repetitive firing of these cells. In Brachyhypopomus, depolarization and firing of the R-cell rely on the activation of AMPAR by specific glutamatergic prepacemaker structures. In Gymnotus, in contrast, high-rate repetitive spiking of these cells requires an increase of R-cell excitability leading to the boosting of responses to depolarization of these cells, most likely originating in PM-cells with subsequent passive R-cell depolarization. We propose that modulation of electrophysiological intrinsic properties of neurons that comprise the PN constitutes an additional neural strategy that has evolved in gymnotid fish to produce specific electromotor outputs. This novel mechanism confers a higher degree of functional versatility to the PN and a high wired command nucleus of the electromotor system, and allows dynamic adaptation of the electromotor repertoire of gymnotid fish to different social and environmental demands.

Kim Langevin participated in the behavioral experiments and data processing during her internship in Facultad de Ciencias, Universidad de la República, Montevideo, Uruguay, as part of her undergraduate studies in Biological Sciences at the Université Laval, Québec, Canada. We thank Magdalena Vitar and Carolina Acordagoitia for their assistance during electrophysiological experiments. Special thanks to Alberto Pereda for critical reading and helpful suggestions on the manuscript.