The sensory effects of light on the electric organ discharge rate of Gymnotus omarorum

Light is a major determinant of life. It affects various aspects of animal behavior, including the control of body composition (Tucker et al., 1984; Teschke et al., 2008) and structure (Webb and Holick, 1988), circadian (Roenneberg and Foster, 1997) and seasonal (Chen et al., 2020) rhythms of growth (Cowan et al., 2017) and activity (Godsil and Fanselow, 2004), and also serves as a carrier of sensory signals (Lythgoe, 1988) used for migration (Rowan, 1925), communication between con-specifics (Théry et al., 2008) and visual evaluation of the nearby environment.

Some animals prefer to remain in the dark and others in the light (De Sousa et al., 2022). Among those showing a preference for the dark, i.e. ‘tenebrotaxis’, various species have evolved peculiar senses to cope with dark environments. Most of these senses rely on the self-generation of the carrier of sensory signals (i.e. homeo-active sensing; Zweifel and Hartmann, 2020), e.g. the whisker system in nocturnal Rodentia (Adibi et al., 2012; Diamond and Arabzadeh, 2013; Adibi, 2019; Harrell et al., 2020), the echolocation system of Chiroptera (Spallanzani, 1798; Griffin, 1958) and the electric sense of Gymnotiformes and Mormyridae (Bullock and Heiligenberg, 1986; Moller, 1995; Bullock et al., 2005), although there are examples of both good diurnal vision with allo-active hearing, as found in Strigiformes (Knudsen and Konishi, 1979; Carr and Konishi, 1990; Peña and DeBello, 2010), and diurnal vision with homeo-active electroreception (Schuster and Amtsfeld, 2002; Schumacher et al., 2017a,b).

Vision is well developed in Mormyroidea. These fish live in clear waters and, moreover, have photonic crystal light collectors that improve vision in turbid waters (Kreysing et al., 2012). Behavioral studies show that vision and active electroreception have similar importance in determining behavior (Rojas and Moller, 2002; Schuster and Amtsfeld, 2002; Schumacher et al., 2017a,b). Moreover, Mormyroidea are able to cognitively integrate both modalities and construct abstract representations (Schumacher et al., 2017a,b).

However, Gymnotiformes are nocturnal fish living in muddy waters under floating water plants that reduce the available amount of light during the day. Moreover, while the preference for living in darkness (i.e. tenebrotaxis) in Gymnotus omarorum (O. Macadar, personal communication) and the congeneric species Gymnotus carapo was behaviorally reported (Maximino et al., 2007), the sensory effects of light on G. omarorum behavior have scarcely been studied (Capurro et al., 1994), with the exception of those induced by the endogenous circadian cycle (Stoddard et al., 2007; Migliaro González, 2018; Silva et al., 2007; Migliaro and Silva, 2016; Migliaro et al., 2018; Gascue et al., 2019). Here, we focused on the less explored light-induced sensory-electromotor behaviors.

Gymnotiformes is an order of Actinopterygii that utilizes the electrosensory system to live among the root mass of floating macrophytes characteristic of muddy lakes and flood plains of South America (Waddell et al., 2019; Waddell and Crampton, 2020). A self-emitted electric field generated by an electric organ discharge (EOD) serves as a carrier for a haptic omnidirectional sensory system in which the objects polarized by the field behave as virtual electric sources in the same way that objects reflecting light function as light sources (e.g. the moon illuminated by the sun acts as a light source; Pereira and Caputi, 2010; Caputi et al., 2003). Floating plants provide a dark environment, which limits vision, and the mesh of thin roots attenuate lateral line signals but do not affect self-generated electric fields, which allows electric fish to take advantage of this niche in order to avoid macroptic predators and prey on small invertebrates. The EOD additionally serves as a cryptic communication channel (Aguilera et al., 2001) for identifying the sex and species of a nearby electric fish (Waddell and Caputi, 2020a,b, 2021).

In G. omarorum, the EOD constitutes a series of stereotyped brief pulses emitted approximately every 30 ms with a small coefficient of variation (about 0.01–0.02; Capurro et al., 1994). The constancy of the waveform (which is equivalent to the color of the illuminating light for the visual system) allows the fish to evaluate the impedance of the object relative to water (Aguilera and Caputi, 2003; Rodríguez-Cattáneo et al., 2017; Caputi and Aguilera, 2019). The repetition rate of the EOD is used for adapting the temporal resolution of the system to the activity state. Novel electrosensory (Caputi et al., 2003; Aguilera and Caputi, 2003) and mechanosensory stimuli (Barrio et al., 1991) evoke ‘novelty responses’ consisting of an abrupt decrease of a few intervals after stimulus onset, followed by a slow relaxation. As the sensory re-afference is processed frame to frame (i.e. EOD to EOD), this responses may be used by the fish for increasing the sampling rate and therefore the time resolution. More sustained increases in EOD rate are triggered by skeletomotor activity (Falconi et al., 1995; 1997) and cognitive states (Jun et al., 2014; 2016; Fotowat et al., 2019). Therefore, the EOD rate in Gymnotus is a faithful measure of activity level and attention (Jun et al., 2014, 2016). In addition, it exhibits seasonal and daily rhythms with increases in mean EOD rate associated with periods of greater activity (Silva et al., 2007; Migliaro and Silva, 2016; Migliaro et al., 2018; Gascue et al., 2019). Light has been shown to induce changes in the EOD rate in pulse Gymnotiformes (Maximino et al., 2007; Stoddard et al., 2007; Migliaro and Silva, 2016; Migliaro et al., 2018) and in Mormyroidea (Cobert, 1984; Moller, 2002).

Gymnotus omarorum shows a characteristic modulation of the mean EOD rate with the circadian cycle, with an increase in the early night and a decrease at dawn (Silva et al., 2007; Migliaro and Silva, 2016; Migliaro et al., 2018; Gascue et al., 2019). The analysis of inter-EOD interval sequences showed endogenously generated high-frequency bouts intercalated between periods of lower and regular frequency bouts (Forlim, 2008; Forlim and Pinto, 2014). This finding indicates that the mean or median values commonly used to describe the EOD rate circadian cycle are insufficient on their own for evaluation of the structure of the time sequence of intervals, and therefore their sole use discards important information about fish activity on a short-term time scale. For example, bimodal distributions may evidence significant changes of the interval and instantaneous frequency, which are not detected when the mean rate is measured over long intervals.

Here, we report the effects of light on the inter-EOD interval series in distinct experimental conditions either in steady-state illumination or under transient illumination pulses. To rule out the circadian effect, experiments were repeated at opposite phases of the circadian cycle. Our results show tonic responses to light which are added to the indirect effects on the EOD rhythm, caused by the endogenous circadian cycle (Stoddard et al., 2007). Fish also respond phasically to light pulses but their slower time course and the differences in adaptation differentiate them from electrosensory ‘novelty responses’. Finally, we confirmed the tenebrotaxic behavior in an experimental series. We postulate that tonic and phasic accelerations caused by the direct sensory effects of light are part of a ‘light-avoidance response’ that allows the fish to follow a mobile habitat.

Thirty Gymnotus omarorum (Richer de Forges et al., 2009) were captured in the summer season using dip nets in the Laguna del Cisne (34.9848 S, −55.1146 W) Uruguay, under permission of the Ministerio de Agricultura y Pesca. All experiments were performed under the regulations of the animal care and use committees of the Instituto de Investigaciones Biológicas Clemente Estable (protocol no. 001/03/2011). None of the procedures caused harm to the specimens. All specimens were returned to the aquarium after completion of the behavioral trials. During the captivity period, fish were kept in individual tanks with conditions matching those found in the wild (water conductivity of 70 µS cm⁻¹, 25°C water temperature) and were fed ad libitum with insect larvae. Experiments were performed in 30 fish (length range 12–15 cm, with <2 weeks in captivity). The light cycle in the aquarium was natural, provided by large windows, and matched light levels found in wild conditions.

The experiments were carried out in a chamber with dim light (1 lx measured underwater at the depth of the fish's eye) and sound isolation. The fish were placed in a fusiform pen consisting of a bamboo stick frame and nylon stocking mesh. This pen (18×2.5×8 cm, length×width×height) was placed inside a tank (24×36 cm base) filled with water to a depth of 5 cm with a conductivity of 70 μS cm⁻¹.

The EOD was recorded with a differential amplifier (AM Systems 3000, gain ×1000, bandwidth 10–10,000 Hz), visualized using a digital oscilloscope (Model GDS-1962A, GW-Instek, New Taipei City, Taiwan). Its timing was recorded by passing the raw signal through a homemade Schmitt trigger circuit. The output of this circuit was monitored by a microcontroller (Arduino Uno R3 ATmega328 microcontroller with up to 16 million instructions per second throughput at 16 MHz, Arduino LLC) which was also used to record the EOD timing and to control the stimuli generated by a white LED. We used a single LED (3.5 mm diameter) with a wavelength emission band bounded about 460 and 645 nm directly connected to the Arduino (Fig. 1A). The interval between pulses was evaluated by configuring an internal timer that keeps track of time. This timer was configured to generate an interrupt working as an internal clock. The time evaluation was performed by counting the interruptions of said timer. For example, between the two edges that measure the EOD, the number of counts was counted and scaled to time. The basic codes used for controlling the stimulus and acquisition are given in the Supplementary Materials and Methods. The current through the LED was determined by a series resistor. To change the intensity within the physiological range (i.e. up to 450 lx), we either placed the LED at different distances or changed the resistor controlling the current. In one experiment, we used a very bright intensity generated by four LEDs powered simultaneously to generate an illumination of 75,000 lx, well above the physiological range. The stimulus intensity used in each experiment was measured underwater at the depth of the fish's eye position using a light-dependent resistor (LDR, code GL5516; details as per the manual, https://www.kth.se/social/files/54ef17dbf27654753f437c56/GL5537.pdf) connected to a multimeter (Fluke 77, Fluke Co.). This device was calibrated using two methods at the temperature used during the experiments (24±1°C): (i) using a Minolta Autometer IV F (Minolta Camera Co. Ltd 1991) in illuminance measurement mode; (ii) with a Sper Scientific 840020 (Sper Scientific, Scottsdale, AZ, USA). The details of the LDR sensitivity to wavelength and intensity, the calibration curves and the method of construction of this device are given in the Supplementary Materials and Methods and Fig. S1. This device has the advantage over the Minolta and Sper Scientific meters that is waterproof and therefore can yield online measurements during the experiment in the field. This allowed us to measure underwater illumination at different depths in several conditions including open water, under the water plants (a large island Eichornia crassipes covering about 100 m²), and under a light spot caused by removing water plants in an area of 30 cm diameter.

These experiments were carried out on 7 fish. We recorded 22,000 EOD time stamps continuously in 7 fish under four constant illumination conditions. Because of the regularity of the discharge (Bullock, 1969; Forlim, 2013) when considering long evaluation periods, the experiment duration can be equivalently defined by the number of EODs or the total time. The first had a procedural advantage. Trials lasted between 10 and 11 min, and varied according to the fish and experimental conditions. The four experimental conditions were: 1 lx (comparable to the light measured in the fish's natural environment under aquatic plants during the day) in the morning (between 08:00 and 10:00 h), 1 lx during the early night (between 20:00 and 22:00 h), 420 lx (comparable to the light measured in the fish's natural environment without plants) in the late morning (between 10:00 and 12:00 h), and 420 lx during the early night (between 22:00 and 24:00 h). The procedure was as follows (Fig. 1B). (i) Fish were kept in darkness all night, and were recorded in the 1 lx light condition in the morning (recording a). (ii) Next, 420 lx illumination was applied with a tungsten filament lamp positioned 77 cm above the fish's head for 1 h, after which another recording trial was taken (recording b). (iii) This 420 lx illumination stimulus was kept on until early night, when the third recording trial was taken (recording c). (iv) Lights were then switched off, and after 1 h of darkness (1 lx), the final recording was performed in these dark conditions (recording d) (Fig. 1B). Each recording was preceded by an adaptation period of 1 h (Gascue et al., 2019) at the stimulus intensity for a given condition. Recordings within a given time period (e.g. the early night) were carried out in the same fish order (individuals 1–7).

The time course of the responses to brief pulses of light were studied in 22 fish. Stimuli consisted of light pulses emitted by a LED controlled with homemade Arduino software. The intensity and duration of the pulse varied according to experimental conditions. All experiments consisted of 22 concatenated trials, without pause between them. Each experiment lasted 22,000 inter-EOD intervals corresponding to about 10 min depending on fish rate. The stimulus began at a given EOD ordinal number (referred to as EOD₀) and lasted for a fixed number of EODs (this value depended on the experiment). We performed three types of experiments in which the stimulus regime was varied in order to address the time course, the effect of stimulus intensity, and response adaptation during the day (all phase response trials were recorded between 10:00 h and 16:00 h). To evaluate the effects of the endogenous circadian cycle we repeated one of the experiments between 22:00 h and 24:00 h.

In 22 fish, we evaluated the time course of the transient effect of a light stimulus on the EOD interval sequence. In 16 fish, we stimulated stepping light intensity from 1 lx (light intensity observed under the floating plants) to 150 lx (half of the light intensity without plants at noon in summer). In the other 6 fish, the light intensity effect was studied systematically (see below). The duration of the steps (100 EOD time stamps) was enough to encompass the transient response. Twenty-two stimulus steps were applied separated by periods of 900 EODs which were several times longer than necessary to reach a stable EOD rate (Fig. 1C).

In 10 fish, we stepped light intensity between 1 lx, and several values of illumination. We performed two series of experiments. In the first series made with 4 fish, we stepped the illumination from 1 to 30, 150 and 75,000 lx (this last, well above natural conditions) and vice versa. The dark and light conditions lasted 1000 EOD time stamps, repeated sequentially without pause (Fig. 1D). Steps down caused ‘off’ phasic responses only for the largest step well above natural conditions (between 75,000 and 1 lx). In the second series, we used 6 other fish in which we systematically explored the amplitude of the phasic response as a function of illumination. We stepped the illumination from 1 lx to six different light intensities compatible with values obtained in the field (19, 27, 52, 84, 262 and 411 lx) (Fig. 1E).

To evaluate whether light-evoked phasic responses exhibit paired-pulse adaptation, we used two series of experiments. In 5 fish, we used light pulses of intermediate magnitude (1–150 lx) and 100 EOD duration (about 3–5 s depending on the fish) separated by different dark periods. Every fish was explored with four series of 44 pulses. Each series of paired stimuli consisted of 22 identical cycles of 1000 EODs. In each cycle, the pulses were separated by 100 and 700 EODs, 200 and 600 EODs, 300 and 500 EODs, or 400 EODs (Fig. 1F). In six other fish, we compared the adaptation of the response to light and to changes in conductance of a neighbor object using paired pulse stimulation. We used two protocols. In the first protocol, light pulses lasted 100 EOD and were separated by 100 EOD (similar to one of the experiments in a previous series – see above); in the second protocol, each stimulus lasted 50 EODs and they were separated by 50 EODs (Fig. 1G).

To test phototaxis versus tenebrotaxis, we used two different experiments. First, we observed the behavior after placing the fish in a small narrow tank (30 cm long and 3 cm width) filled with aquarium water. This tank had a dark and an illuminated compartment. Seven fish were placed either in the bright or in the shaded compartment and fish behavior was observed for 10 min (see Movie 1). Second, we used the Y-maze method that was introduced by Waddell and Caputi (2021) to evaluate waveform-specific electrotaxis. To evoke electrotaxis, we applied the same electrical stimulus to the two arms of the Y-maze and evaluated the fish choice. In a preliminary experiment on 10 fish, we found that 5 fish chose the right side and the other 5 the left side in three out of three trials. Taking into account that fish lateralization (see Results) may bias the results, our experimental design focused on exploring the number of trials in which the difference in illumination between the two arms changed the natural choice. Then, we explored whether the fish reverse their natural choice side as a function of the difference in illumination between arms. A Y-maze with opaque walls was used. The arm not usually chosen was covered with thick cardboard and the other arm was illuminated in each series of experiments with 70, 1300 and 4745 lx using a tungsten filament light bulb placed at different distances. We counted the number of reversals of the natural tendency for each fish in three trials at each illumination intensity.

In all cases, data from the Arduino was displayed on the screen as a text file, copied to a text editor, and transformed into binary to be processed off-line using Octave. The inter-EOD intervals were calculated as the difference in timing between subsequent EODs.

To evaluate tonic effects, we constructed first-order interval histograms and calculated and compared key statistical parameters. To evaluate transient effects, we normalized the sequence of inter-EOD intervals to the median of the 20 intervals before EOD₀ (basal interval) and expressed the time course of the inter-EOD interval sequence as a percentage of the control.

To evaluate the amplitude of the phasic response, we used an index (introduced in Pereira et al., 2012; also implemented in Jun et al., 2016). The response consists of an inter-EOD interval reduction reaching a minimum at about 20 EODs after stimulus onset. Therefore, we constructed a filter to detect a peak occurring at about 20 EODs on the normalized series of intervals. The cornerstone of this filter is to calculate the series of normalized intervals encompassing 20 EODs after every EOD for each trial. These series have a sharp minimum when the interval extends from EOD₀ to EOD₂₀ and is about null in the control condition. Receiver operating characteristic curves (ROC curves) were constructed in the following way to verify the statistical significance of the response. We defined 250 equally spaced threshold values between −25% and 25%, and measured the number of trials that crossed every threshold within two windows: one including the intervals starting at the first 100 EODs after the stimulus (true positives) and the other including the intervals starting between EOD₄₀₀ and EOD₅₀₀ (false positives). The ROC curve results from plotting the fraction of trials yielding true positive versus the fraction of trials yielding false positive responses for the thresholds explored. When the true and false positive distributions do not differ, the ROC plot yields a bisecting line (area under the curve=0.5). The degree of significance of the difference between two responses can be measured by the departure of the ROC curve from the bisecting line. When the area limited by the ROC curve and the bisecting line was larger than 0.25 (area under the curve=0.75), the difference was considered statistically significant (Green and Swets, 1966).

To evaluate the effect of intensity, we made a regression analysis between the amplitude of the response calculated as expressed above and the logarithm of the illumination. To evaluate adaptation, we used non-parametric tests between the responses obtained with the first and second paired stimuli.

The hypothesis evaluated in this study was that light has a direct sensory effect on fish behavior, in addition to the well-known effects on setting a circadian cycle. In order to use a light stimulus within the natural range, we explored the illumination intensity levels in nature. Fig. 2 compares the decay of illuminance with depth in three conditions: open water, under a light spot caused by removing the water plants, and under the plants; the corresponding values at which tonic, phasic and behavioral experiments were conducted are also indicated.

Taking into account that light may have multiple effects conveyed through different neural paths, we separately explored the effects of continuous illumination and step increments of light intensity at two opposite phases of the circadian cycle. Classically, sensory responses have two types of components: those depending on the present intensity of the stimulus, i.e. tonic responses, and those depending on the history of stimulation, i.e. phasic responses. To explore these components separately, a typical strategy is to apply abrupt steps between two constant values of stimulus intensity. Such abrupt increments caused a transient response (phasic) that increased as a function of the stimulus increment. Long-lasting periods of adaptation allowed us to explore steady-state components (tonic) signaling the present stimulus intensity.

We evaluated the tonic response to light at opposite phases of the circadian cycle, comparing interval series obtained under light intensities similar to those observed in open water and beneath aquatic plants during the morning and early night under steady-state conditions (each stimulus intensity value was kept constant for at least 1 h prior to the start of the recordings).

 Fig. 3 illustrates the results obtained from one fish. Under both dark experimental conditions, we found a tendency to switch between two activation states (previously described by Forlim and Pinto, 2014), reflected as a main (longer intervals, low frequency states) and a secondary (shorter intervals, high frequency states) mode in the inter-EOD histograms (compare histograms in Fig. 3, right).

Kruskal–Wallis one-way ANOVA test indicated that the mean (P=0.0059), median (P=0.0088) and coefficient of variation (P=0.016) of the EOD rate were not the same across the four conditions. All distributions exhibited a negative skewness due to the tendency of the fish to accelerate the EOD pacemaker, but no significant differences were observed across skewness values (Kruskal–Wallis test, P=0.22). However, the frequency of high-frequency bouts was larger and their average duration was longer during the night and dark conditions (Fig. 3C,D). This change in the inter-EOD interval pattern is expressed in the inter-EOD interval as a secondary peak with a lower mode (arrow Fig. 3D).

Two-way ANOVA analyses (Friedman tests with repetition) confirmed previous observations that the EOD rate in G. omarorum exhibits an endogenous circadian cycle (Silva et al., 2007; Migliaro and Silva, 2016; Migliaro et al., 2018) and also indicated that light has an independent sensory-electromotor effect on the pacemaker (Fig. 4). We used the Friedman test with repetitions to compensate for the effects of light and dark, and found that median intervals at night were significantly smaller than median intervals during the day (P=3×10⁻⁴; Fig. 4A). No significant differences in median interval between light and dark were found when compensating for the effect of the endogenous rhythm (P=0.7664). However, the median interval in morning recordings was significantly shorter in light than in dark conditions (sign rank test P=0.0391; Fig. 4A), suggesting a tonic increase of the EOD rate due to photo-stimulation.

Similarly, two way ANOVA analysis indicated that light also reduced the interval variability after compensating for the effects of the circadian cycle (P=0.038; Fig. 4B). This was mainly due to an increase in the coefficient of variation during the night and dark conditions (P=0.0156 when compared with the other three conditions). During the night, median values of the coefficient of variation (0.037, 0.078 for light and dark, respectively; Fig. 4B) were larger than those during the day in the same condition (median values, 0.034 and 0.05 for light and dark, respectively; Fig. 4B). However, after compensating for the effects of light and darkness, there were no significant differences due to the endogenous rhythm (P=0.278; Fig. 4B).

This increase in variability under dark and night conditions led us to examine in detail the first-order histograms. We found that the inter-EOD interval distribution in dark conditions was bimodal for all fish at night and only for two fish during the day. The mean and median values frequently used are adequate for comparing distributions with a single mode. However, our results show skewed and bimodal distribution. Thus, such location parameters reflect a trade-off between the probability of intervals around each of the two modes and the separation of the modes. Therefore, to gain additional information, we complemented the previous analysis with a comparison of each mode of the distributions and the peak values for each mode.

When compensating for light and dark conditions, the mode corresponding to the histogram's largest peaks (the so-called main mode, reflecting the basal frequency state) is smaller during the night, (P=0.0294; Fig. 4C). This reflects a nocturnal EOD acceleration caused by the endogenous cycle. When compensating for the endogenous effect, the main modes in light conditions are shorter than in dark conditions (P=0.0057; Fig. 4C). This indicates that light caused per se an increase of the basal pacemaker frequency, independently of the endogenous rhythm effect.

The secondary peak mainly appears in dark conditions (Fig. 3). This corresponds to the emergence of high-frequency bouts that increase in duration and probability during the night in dark conditions (all 7 fish versus only 2 during the day). This nocturnal rise of the peak corresponding to the secondary mode is reflected in the reduction of the peak corresponding to the main mode (Fig. 4D; sign rank test, P=0.0156). The two effects, the increase in the basal rate and the increase in probability of the high-frequency bouts, combine to cause the highest EOD rate observed in G. omarorum during the night.

The responses evoked by short pulses of light were studied in 22 fish. They consisted of a successive reduction of the first 20 inter-EOD intervals after the onset of the light pulse, followed by a slower relaxation, which lasted approximately 60–100 EODs. For intensities within the natural daytime range (30 and 150 lx), pacemaker acceleration responses occurred at the stimulus onset only. At high stimulus intensities (75,000 lx, well above natural stimuli) the light-evoked response occurred at both the onset and cessation of the stimulus. In this case, the response at the ‘off’ was slower and of lower amplitude.

To characterize response time course, we recorded the sequence of intervals in contiguous trials of 1000 EODs initiated by a brief pulse of light (150 lx) lasting 100 EODs in 16 fish. Results obtained in a complete experiment on the same fish are plotted as a function of the EOD ordinal number in polar coordinates in Fig. 5A. As stimulation was repetitive, to show the consistency of the response, every circle turn corresponds to each of the 22 trials and the distance from the center corresponds to the inter EOD interval. The white sector corresponds to ‘light-on’ and the gray one to ‘light-off’ periods. For the sake of clarity, averaged responses from 7 fish are superimposed in Fig. 5B. These responses peak at about the 20th interval following stimulus onset. Taking this observation into account, we calculated a series of 20th order intervals for each trial (i.e. intervals between every 20 EODs; between EOD₀ and EOD₂₀, EOD₁ and EOD₂₁, EOD₂ and EOD₂₂, and so forth, with the subscript indicating the ordinal EOD number, with respect to the stimulus onset). This filtering procedure allowed us to define a clear peak in the 7 fish (Fig. 5C). This signal transformation also allowed us to verify the statistical significance of the response by contrasting, for each fish, the amplitude of the response obtained after the stimulus onset with the amplitude of any incidental accelerations that occur in the absence of stimulus in the period between EOD₄₀₁ and EOD₅₀₀. To perform this comparison, we defined a series of threshold values and compared the fraction of traces crossing each threshold after the stimulus onset (true positives) with the fraction of traces crossing the same threshold in the absence of stimulus (false positives). This allowed us to construct ROC curves (Fig. 5D), which showed significant differences in all cases (the area under the curve ranged from 0.79 to 1). Thus, in the rest of this study, the peak value of the series of the 20th order intervals was considered as a faithful indicator of response amplitude.

In a first series of experiments, we found that the amplitude of the response increased with stimulus strength (Fig. 6A). However, one of these intensities was beyond the natural range. Thus, we designed a second series of experiments on 6 fish at selected intensities within the natural range in order to perform a regression analysis. This experiment showed that the amplitude of the phasic response increased linearly with the logarithm of the illumination intensity (Fig. 6B).

Finally, we explored the adaptation of the response. In a first series or experiments, the amplitude of the light-evoked phasic response was not consistently altered when using pairs of similar stimuli (150 lx, 100 EOD) separated by inter-stimulus intervals of 100, 200, 300, 400, 500, 600 and 700 EODs (Kruskal–Wallis, P=0.53; Fig. 7A). In a second series of experiments, we paired a test pulse of 50 EODs with a conditioning pulse of 50 EODs separated by 50 EODs (about 1.5–2 s, depending on the fish). This experiment had the minimum possible separation between stimuli as the test stimulus was applied at the end of the response evoked by the conditioning. The results were visually similar, although when comparing the median amplitude of the responses to the conditioning stimulus across trials (Fig. 7B) with the median amplitude of the response to the test stimulus (Fig. 7C), we found a significant reduction (sign rank test, P=0.0068, N=11 fish; Fig. 7D).

An interesting observation during these experiments was that one small fish (9 cm) exhibited a tendency to turn around after stimulus onset, which was clearly seen by the reversal of EOD waveform polarity monitored via an oscilloscope. While we were not able to use this fish in our analyses because of its intolerance to the experiment, this behavior suggested that the phasic response to light was part of a ‘light-avoidance skeleton electromotor response’. This led us to examine whether fish prefer to stay in the shade or the bright side of a small tank. When fish were initially placed in the bright region, they immediately swam to the shaded region, and when placed in the shaded region, they remained there during 10 min of observation (sign test: P=0.0156, N=7).

To confirm this observation, we performed forced-choice experiments during electrotaxis behavior in a Y-maze. To provoke electrotaxis, we applied the same stimulus between the ends of the two parallel arms of the Y-maze and the base point. As the maze pen has a constant section, the current density and electric field were similar at each branch. In this condition, we expected that the fish would make a choice by chance between the two arms. However, we found that half of the fish systematically chose the right arm and the other half systematically chose the left one. This suggests that fish may have either a right or a left swimming preference (i.e. a rudimentary form of lateralization). Then, to test choices in the Y-maze, we evaluated whether the tenebrotaxis/photophobia has enough influence to revert this natural lateralization. The distribution of reversals in the three trials for the three experimental conditions is shown in Table 1. These data indicate that the average number of reversals in the fish's habitual choice side increases with the difference in illumination between the two arms.

This article shows that photo-stimulation influences electromotor behavior in various complex ways. We found that besides its previously described influence on the endogenous circadian rhythm (Migliaro and Silva, 2016; Migliaro et al., 2018), light within the physiological stimulus range tends to increase pacemaker rate per se. Steady and transient stimuli evoke tonic and phasic responses, respectively, that may be relevant to fish life. A tentative hypothesis is that the complex effects on the EOD rate and variability previously observed along the circadian cycle (Silva et al., 2007) occur because the pacemaker interval receives multiple descending drives. Among these, we focus on two main light-dependent drives: one endogenous, associated to the circadian rhythm, and the exogenous, the illumination cycle. Although they have different and, in some cases, opposed effects, the light stimuli and the circadian cycle are associated in nature in a season-dependent way.

To reveal the sensory effects of light on pacemaker activity per se, we had to separate the effects of the endogenous circadian cycle, which are associated with the light cycle. Therefore, the first part of the study compared the histograms under light and dark conditions at two opposite phases of the circadian cycle. We confirmed that during the circadian cycle, the mean EOD rate and the EOD variability increase during the night (Silva et al., 2007; Migliaro and Silva, 2016; Migliaro et al., 2018; compare Fig. 3A and D). However, to make a proper analysis of pacemaker activity in steady-state conditions, we studied the inter-EOD interval histogram. As it showed a skewed distribution for long periods, the commonly used parameter mean rate would be influenced by the degree of asymmetry of the distribution (Pereira et al., 2012). We interpret that the main mode more likely corresponds to the tonic activity of the pacemaker, and the skewness of the histogram corresponds to transient effects driven by sensory stimuli, internal collaterals of skeletomotor commands and other endogenous signals. We found that during the night phase of the circadian cycle, the inter-EOD interval distribution histogram showed two modes, one larger, corresponding to the absolute maximum of the histogram (here referred to as the main mode) and another smaller one, corresponding to a local smaller maximum (here referred to as the secondary mode). The main source of this secondary mode is the emergence of spontaneous bouts of high frequency that stand out, intercalated between periods of lower basal frequency (Forlim, 2013; Forlim and Pinto, 2014). This suggests that the accelerating effects on the pacemaker of the night phase of the endogenous circadian cycle (Perrone et al., 2009; Migliaro and Silva, 2016; Migliaro et al., 2018) have two main causes: (i) an increase in pacemaker tonic activity, reflected in the reduction of the main modes of nocturnal with respect to diurnal histograms in the same light conditions, and (ii) an increase in the probability of high-frequency bouts during the night (Fig. 3C), contributing additionally to the increase in mean rate and also explaining the larger rate variability of the EOD.

The first set of evidence in favor of the hypothesis that light has a direct sensory effect on the pacemaker rhythm is that sustained illumination provoked a reduction of the mean EOD rate during the day (Fig. 4A) and also a reduction of the coefficient of variability of the EOD rate independently of the time of the day (Fig. 4B). The main mode of the inter-EOD interval histogram in the dark was larger than under illumination in both phases of the endogenous circadian cycle (Fig. 4C). As the main mode is likely a good indicator of the pacemaker tonus in a given condition, its reduction by illumination implies that tonic photo-stimulation per se causes an increase of the excitatory inputs on the pacemaker. In addition, light stimuli similar in amplitude to those found in open waters suppress high-frequency bouts, both during the day and during the night, indicating a powerful suppressing effect on this behavior. High-frequency bouts alter the interval distribution from mono- to bi-modal, with the net result being an increase in the EOD rate. However, there is a reduction of the median interval, an increase in skewness and also an increase of the peak of the secondary mode at expense of the peak of the main mode. This is outstanding during the night (Fig. 4D). Thus, the direct tonic effect of photo-stimulation on EOD rate is twofold: it reduces the basal interval and it also reduces the probability of high-frequency bouts.

The second set of evidence in favor of the hypothesis is that light pulses evoke transient accelerations of the pacemaker. These phasic responses show three characteristics: (i) they are slow, (ii) their amplitude increases with stimulus intensity amplitude, and (iii) the effect of a phasic response on a subsequent one is present only when the interstimulus interval is very short.

Light-evoked phasic responses share with novelty responses the graded amplitude with stimulus intensity. However, their rising phases (about 20 EODs or about 500–600 ms) were significantly slower than those evoked by active electrosensory (2 EODs or about 60 ms; Aguilera and Caputi, 2003; Caputi et al., 2003) or mechanosensory (7–8 EODs or about 200 ms; Barrio et al., 1991) stimuli reported in the literature in similar experimental conditions. This slower time course suggests that these EOD rate accelerations are different from fast novelty responses observed with other sensory modalities. Another consistent difference is that the effect of one response on a subsequent one can be shown at short intervals only, implying that the memory trace decays much faster than in the case of novelty responses elicited by a change in a nearby object conductance (Caputi et al., 2003).

Phasic responses appear to be associated with skeletomotor activities, which have been observed in other circumstances in these fish. For example, Mauthner neuron antidromic activation causes a strong and sudden pacemaker acceleration response via the activation of multiple paths (Falconi et al., 1995, 1997; Comas and Borde, 2010). This response, referred to as a ‘Mauthner initiated abrupt increase in rate’ (Falconi et al., 1995), is associated with the escape reflex triggered by intense mechanosensory stimuli, and may have the role of increasing the active electroreception system sampling frequency during a fast movement that involves not only linear motion but critically a drastic change in swimming direction (Borde et al., 2004). Other pacemaker acceleration responses induced by skeletomotor activities are associated with back and forth movements during object electrosensory exploration (referred to as B-scans; Jun et al., 2016) and during object-following movements (Uyanik et al., 2019; Comertler and Uyanik, 2021). The latter is notable because in the wave fish Eigenmannia, the main feedback used by the fish for object-following behavior is visual stimuli (Comertler and Uyanik, 2021) and this may have a role in the adaptation of fish to their habitat.

The observed differences from novelty responses, including the slow time course, the shortness of the memory trace during the response, and the frequent association of accelerations and skeletomotor activity in these fish suggest that in our experiments the phasic electromotor response to light is the first stage of a global light-avoidance response in which the fish accelerates the EOD rate to increase the sampling sensory frequency just before moving away from the light. This was confirmed by the tenebrotaxic behavior observed in two independent light-avoidance behavioral experimental series.

These fish are solitary animals whose natural habitats are islands of aquatic floating plants, primarily composed of the species Eichornia crassipes. These plants are clumped together by an entangled complex of fine and relatively long roots in which these fish hide during the day. Data obtained in experimental light conditions similar to those in nature suggest that in open water, the fish will have a regular and faster rate than expected by the effect of the endogenous cycle alone, while under the plants, the lack of a light acceleration effect is somewhat compensated for by the transient increases in rate caused by the appearance of high-frequency bouts.

The wide leaves and the root network of E. crassipes form a light barrier that protects these fish from their natural macroptic predators (genera Australoheros and Hoplias). The tonic effects on the pacemaker rate described herein may also have an important role in animal life. When the fish swims in open water during the day, light stimuli accelerate and regularize the rhythm, which increases the sensory sampling rate, facilitating the evaluation of heightened predation risks after losing the protection of floating plants. In the dark, fish display high-frequency bouts which are more frequent and last longer during the active phase of the circadian cycle; these may occur even though the fish is quiet, but also may be associated with skeletomotor exploratory behaviors. The light-avoidance response can be considered a subset of an object-following behavior (Uyanik et al., 2019; Comertler and Uyanik, 2021) that allows the fish to follow the shade of the floating aquatic plants (in which these fish find shelter during the day), and therefore allows them to stay within the island-like congregations of these plants as they move as a result of water currents and winds. In addition, light penetration between the leaves may generate bright ‘sunspots’, which slowly move according to the sun's position, and may reveal hiding fish to predators.

Our present view on how illumination and circadian cycle interact is schematized in Fig. 8. We postulate that the pacemaker interval is under two main drives, one tonic and the other transient, causing either high-frequency bouts or other sensory-driven accelerations. The endogenous rhythm tends to increase the rate by both mechanisms (reducing the main mode and facilitating high-frequency bouts during the night), but light has opposed effects on the mean EOD rate (increases the basal rate but blocks the appearance of bouts). In nature, the endogenous rhythm and the light cycle are associated. Therefore, during the night, the mean pacemaker rate increases because of the effect of the endogenous cycle only (reducing the basal intervals and increasing the probability of high-frequency bouts; Fig. 8A). The presence of high-frequency bouts also explains the higher nocturnal variability.

During the day, the endogenous circadian rhythm tends to reduce the pacemaker rate. This endogenous effect is attenuated by daylight in an intensity graded fashion (Fig. 8B,C). The observed differences in illumination between experimental conditions are similar to those in open water and hiding under aquatic plants, and suggest that in the first condition the fish would have a regular and faster rate (Fig. 8B), whereas under the second there was an increase in rate caused by the appearance of high-frequency bouts that somewhat compensated for the shift of the main mode to larger values (Fig. 8C).

Additional evidence supporting the nature of these behaviors as a global light-avoidance response is the absence of a direct anatomical projection from the optic tectum to the prepacemaker nucleus. The modulations of the EOD rate induced by light on the pacemaker nucleus require the activation of a connecting path originating from the optic tectum. No direct projections have yet been described between the tectum and the prepacemaker. For example, in Eigenmannia virescens, anterograde transport of byocitin injected into the prepacemaker nucleus showed no projections on the optic tectum (Wong, 1997). In G. omarorum, the injection of byocitin near the pacemaker cells resulted in two groups of retrogradely marked neurons at the level of the diencephalon and the brainstem (Comas and Borde, 2010). The activation of these groups of cells by the injection of glutamate resulted in increases of the frequency of the EOD.

In pulse fish – as in wave fish (Sas and Maler, 1986) – a direct projection from the retina to the suprachiasmatic nucleus has been found (Lázár et al.,,1987; Wong, 1997). This suggests a possible role of the suprachiasmatic nucleus in the generation of the phasic response while tonic responses could be mediated by the tectum and pretectum.

A final and more general consideration is that peculiar homeoactive senses (i.e. those that use a self-generated carrier; Zweifel and Hartmann, 2020), such as the active electric sense, active whiskering and echolocation, are most often found in animals adapted to live in dark environments. It appears that light influences on these homeoactive sensory modalities may modulate species-specific sensory channels that enable swimming, running and flying, and also cryptic communication in relatively crowded dark environments, endowing these taxa with sensory advantages that allow them to capture prey and reproduce while avoiding predators with a sensory predominant visual system.

We thank Dr Joseph Waddell for critical comments and English editing of the manuscript, and Drs Leonardo Barboni and Federico Pedraja for their help with Arduino software.