Zebrafish breeding and maintenance was performed under standard conditions (Westerfield, 2000). Embryos were collected in the morning and raised on a 12:12 light/dark cycle in E2 medium (Westerfield, 2000). All experiments were carried out with wild-type (cross between AB and TL strains) 6 dpf larvae. Zebrafish experimental procedures were performed according to the guidelines of the German animal welfare law and approved by the local government.

Larvae were incubated for 10 min in steady state E2, E2 + 50 mM NaCl (NaCl_50mM) or E2 + 100 mM NaCl (NaCl_100mM) media, at 28°C under white light illumination. They were then washed three times with E2 medium and kept in a small container for oxygen consumption measurements (6–8 min later), or transferred to a custom made swimming chamber for behavioral testing. The wash and transfer period took 3 min (± 10 s) and was performed at room temperature. Behavioral recordings began 5 min after stressor exposure. Cortisol detection was carried out using a home-made cortisol ELISA protocol, as described in (Yeh et al., 2013).

Larvae were presented with fast hydrodynamic flows caused by the rapid lateral displacements (LDs) of a partially submerged inflexible silica capillary tube (Polymicro Technologies, AZ, 360 µm OD, Optronis GmbH; Kehl, Germany) fixed to a multilayer piezo bender actuator (PICMA® PL140.10, Physik Instrumente (PI) GmbH and Co. KG; Karlsruhe, Germany). The actuator had an operating voltage of 0–60V, a maximum displacement of ±1000 µm, and an unloaded resonant frequency of 160 Hz. The bender was connected to a dual-piezo-amplifier (maximum voltage: 10 V), a pulse generator and a TTL control system allowing for computer control. The tip of the silica capillary tube was placed, partially submerged (2 mm), at the center of a 35 mm petri dish half filled (1.8 ml) with E2 medium (orientation relative to water surface: 90°). The voltage applied to the bender (V_act) determined the speed of the capillary’s LDs, or stimulus strength (in % relative to maximum voltage). Groups of 30 larvae were exposed to 6 stimulation units delivered with an inter-stimulation-interval of 250 ms. Each unit consisted of 99 40 ms LDs. We used a V_act of 3, 4.5 or 6 V. Stimulations were carried out under white illumination at 25°C. After stimulation, larvae were kept in the petri dishes for cortisol measurement (9.5 min later) as described in (Yeh et al., 2013), or transferred to a custom made swimming chamber for behavioral testing, where they remained unperturbed for 10 min before recordings.

Oxygen levels inside a single custom made flow chamber with 8 freely swimming larvae were monitored (as % air saturated) every 5 s for 30 min using fibre optic sensing technology (PreSens Precision Sensing GmbH; Regensburg, Germany). To minimize perturbations during recordings, the chamber was placed inside a light-proof box. Its inlet and outlet channels had silicon tubes (750 µm OD) and a 5 ml Luer-Lok™ Tip Syringe (BD, Heidelberg, Germany) was connected to the inlet to fill the chamber with solution; both inlet and outlet remained closed during measurements. To calibrate the oxygen-sensitive sensor, we used oxygen-free (Cal 0) and air-saturated water (Cal 100). For Cal 0, Na₂SO₃ (0.2 g) was dissolved in distilled water (20 ml). For Cal 100, air was blown in distilled water (100 ml) for 20 min while stirring; air flow was then stopped and the solution stirred for another 10 min to avoid hyper-saturation. We used a two-point calibration with temperature and atmospheric pressure compensation. For each measurement, the oxygen consumption rate (OCR) was approximated as the slope of a linear fit to the oxygen level for 10 min < time < 30 min. Prior to measuring the OCRs of larvae, the OCR of E2 alone was measured twice and averaged. Net OCRs of the larvae were calculated by subtracting the average OCR of E2 of the actual day from the OCRs of the larvae.

Experiments on feeding were conducted under both white and infrared (IR) light, delivered through a custom-made array of white- and IR-LEDs mounted inside a custom-made light-proof enclosure placed on a vibration-free platform (Newport, Irvine, CA, USA). Avoidance and individual locomotion tests were conducted under IR light only. In all cases, larvae were imaged at 25 frames s⁻¹ through an infrared-sensitive camera (ICD-49E B/W; Ikegami Tsushinki Co., Ltd. Japan) with its lens (TV Lens, Computer VARI FOCAL H3Z4512 CS-IR, CBC; Commak, NY, USA) positioned above a swimming chamber. We used EthoVision XT software (Noldus Information Technology; Wageningen, The Netherlands) and algorithms written in MATLAB 7.8 (MathWorks, Inc.; Natick, MA, USA) to monitor the movements of larvae swimming either individually (avoidance and individual locomotion tests) or in groups (feeding) inside custom-made chambers, respectively. In all experiments, larvae were allowed to adjust to the chamber conditions for several minutes prior to recording. Experiments were conducted at 28 ± 1°C. A thermocouple (npi electronics GmbH; Tamm, Germany) connected to a temperature control system (PTC 20, npi electronics GmbH; Tamm, Germany; Exos-2 V2 liquid cooling system, Koolance; Auburn, WA, USA) monitored the temperature inside the chambers. All the experiments were performed in a blind fashion as to group identity. Control animals for each group were handled in the same fashion, but omitting stimulus presentation. Tests were conducted between 9:00 and 18:00 and different experimental groups intermixed throughout the day.

All data are shown as single measurement points, mean and standard error of the mean (S.E.M.) or box-and-whisker plots. We used paired Student’s t-tests (two-tailed) for two-group comparisons as well as ANOVAs, followed by Bonferroni’s post-hoc tests, or Kruskal-wallis tests (if the data did not fulfill the assumptions of the ANOVA), followed by Dunn’s tests, for multiple group comparisons. We also used linear and non-linear regression analysis. Analyses were carried out using MS-Excel (Microsoft; Redmond, WA, USA), MATLAB 2009b (MathWorks, Inc.; Natick, MA, USA), Prism 5, (Graphpad Software Inc.; San Diego, CA, USA), Sigma Plot (Systat Software Inc.; San Jose, CA, USA), R (Freeware) and Virtual Dub (Freeware).

Twenty larvae were placed in a custom made swimming chamber where two transparent walls created three contiguous compartments: S1, Centre and S2 (Figure 1A). The Centre compartment, twice as large as the side compartments, acted as a transition zone between S1 and S2. The transparent walls divided only the upper part of the chamber’s cross section, leaving space at the bottom for the larvae to move freely across compartments while a camera placed above recorded their movements under white light illumination. The walls contained rows of small openings allowing the medium to diffuse and acted as barriers to confine moving prey (paramecia), or food items, to a single side compartment, either S1 or S2 (see Methods). Each recording session lasted 600 s. Larvae were allowed 10 min of acclimation before recordings started. Following acclimation, video-recordings were made first in the absence of food items (initial 600 s) and, then, after having added a known quantity of paramecia to one of the chamber’s side compartments, either S1 or S2 (final 600 s). In the absence of prey and external stimuli (other than visual and hydrodynamic cues caused by their own movements), larvae spent equal amounts of time in S1 and S2 (Figure 1B, left: Paired t-test, t₍₁₀₎ = 0.1, p = 0.92). On average, the percentage of larvae found in either S1 or S2 was ~25%, as expected from a symmetric distribution of individuals within the chamber (Figure 1B, right). This changed when paramecia were added to one of the side compartments, either S1 or S2 (Figure 1C). The presence of prey reliably triggered a differential space use (DSU) by the larvae. They spent more time in the compartment with paramecia (Figure 1D, left: Paired t-test, t₍₁₀₎ = 5.8, p = 0.0002), increasing the number of individuals in it after 600 s (Figure 1D, right). To quantify the overall motion and the number of larvae swimming in specific areas of the chamber, we used a tracking algorithm which computes the pixel-by-pixel m.s.e. between images from consecutive video-frames (Figures 1E,F, see also Methods). First, we calculated the area under the curve (a.u.c.) from ensued m.s.e. values (in 600 s) in order to estimate the amount of larvae that swam concurrently within each of the chamber’s compartments. Using these values, we then calculated a global measure of DSU, as the difference (in %) between the a.u.c. values from S1 and S2, or [(S1_a.u.c. – S2 _a.u.c.)/(S1 _a.u.c. + S2_a.u.c.)]*100, always using data from the side with prey as S1. A comparison of DSU values recorded with and without paramecia validated the approach as an accurate method for quantifying the relation between the spatial distributions of both prey and larvae (Figure 1G, Paired t-test, t₍₉₎ = 10.8, p < 0.0001). Importantly, by counting the number of paramecia consumed by the group of larvae (Figure 1H, Paired t-test, t₍₉₎ = 17.2, p < 0.0001), we observed that DSU variability accounted for 97% of the variance of prey consumption (Figure 1H, Pearson’s correlation, p < 0.0001). Further, a comparison of total a.u.c. values (i.e., summed values from S1, Centre and S2) showed that prey-dependent DSU changes could not be accounted for by changes in the overall level of motion displayed by larvae throughout the recording periods (Paired t-test, t₍₉₎ = 0.21, p = 0.84).

Salt exposure is a well-established stressor in teleosts and has been shown to increase whole-body cortisol in larval zebrafish (Yeh et al., 2013). We observed that freely swimming larvae avoided a sudden increase of salt concentration, as they rapidly moved away from a spot where 5 µl of 50 mM NaCl solution was added to their medium and continued to avoid the vicinity of the addition spot over a 180 s video-recording period; such a series of avoidance reactions did not occur when 5 µl of E2 were added to the medium (Figure 2A). In separate experiments, we next measured whole-body cortisol as a function of (1) NaCl concentration in a steady-state hyperosmotic medium, (2) NaCl exposure duration; and (3) time after exposure (Figures 2B–D). The results showed that NaCl exposure increased cortisol in a concentration-dependent manner (Figure 2B, Kruskal-Wallis test, H = 48.6, p < 0.0001, followed by Dunn’s multiple comparison tests and linear regression), that maximum cortisol levels were detectable 5–10 min after NaCl exposure (Figure 2C, One-Way ANOVA, F_(4,29) = 22.8, p < 0.0001, followed by Bonferroni post-tests for pair comparisons and non-linear regression), and that basal levels were re-established 30 min later (Figure 2D, One-Way ANOVA, F_(3,23) = 25.4, p < 0.0001, followed by Bonferroni post-tests for pair comparisons and non-linear regression).

To investigate how HPI-axis activation influences feeding in freely swimming larvae, we measured prey-dependent DSU changes in larvae that had been previously exposed (for 10 min) to hyperosmotic E2 media containing additional NaCl (NaCl_50mM or NaCl_100mM). Both incubations (NaCl_50mM and NaCl_100mM) suppressed prey-dependent DSU changes 15–25 min after the offset of NaCl exposure (Figures 2E,F, left, Paired t-tests, NaCl_50mM: t₍₉₎ = 0.51, p = 0.62, NaCl_100mM: t₍₉₎ = 0.38, p = 0.72). Later, 40–50 min after exposure, prey-dependent DSU changes were detectable in larvae pre-incubated with NaCl_50mM but not with NaCl_100mM (Figures 2E,F, right, Paired t-tests, NaCl_50mM: t₍₉₎ = 4.02, p = 0.003, NaCl_100mM: t₍₉₎ = 0.51, p = 0.63). Larvae pre-incubated with NaCl_100mM showed normal prey-dependent DSU changes, i.e., indistinguishable from those of control larvae, three hours after NaCl exposure (Paired t-test, t₍₉₎ = 4.3, p = 0.002). Control larvae (pre-incubated with normal E2 medium) showed prey-dependent DSU changes 15–25, 40–50 and 180–190 min after incubation (Paired t-tests, 15–25 min: t₍₉₎ = 8.72, p < 0.0001, 40–50 min: t₍₉₎ = 7.1, p < 0.0001, 180–190 min: t₍₉₎ = 4.6, p = 0.001). A comparison of the total a.u.c. values from all groups indicated that motion variability could not account for the observed prey-dependent changes in DSU (Two-Way Repeated-Measures ANOVA, Group factor: F_(7,72) = 0.56, p = 0.78, Prey factor: F_(1,72) = 0.21, p = 0.65, Group X Prey factor: F_(7,72) = 0.12, p = 0.99).

We next asked whether altered locomotion, oxygen consumption or visual reactions could account for the differences in prey-dependent space use observed after NaCl exposure. We first monitored post-exposure levels of locomotor activity in treated and control larvae within a stimulus-deprived environment. For this, larvae pre-incubated in either control or hyperosmotic mediums were placed in a custom-made, vibration-free swimming chamber (10 mm ID, volume: 400 µl) and allowed to swim individually in complete darkness for 600 s under constant temperature (28 ± 0.1°C). We found that NaCl exposure failed to modify post-exposure locomotion (Figure 3A, One-Way ANOVA, F_(2,71) = 2.2, p = 0.13); as a result, peak cortisol levels after salt exposure did not account for locomotion variability (Figure 3B, Pearson’s correlation, p = 0.42). To expand the analysis of NaCl effects, we monitored oxygen levels every 5 s for 30 min (as % of air saturated) in a custom-made chamber containing groups of eight freely swimming larvae that had, or had not, been pre-incubated in a hyperosmotic medium (of either NaCl_50mM or NaCl_100mM). For each measurement, the OCR was approximated as the slope of a linear fit to the oxygen level for 10 min < time < 30 min. Prior to measuring the OCRs of larvae, we measured (twice) the OCR of E2 medium alone. Net OCRs were calculated by subtracting the average OCR of E2 from the OCRs of the larvae. We found that pre-incubation with either NaCl_50mM or NaCl_100mM did not change net OCRs, as compared to pre-incubation with E2 (Figure 3C, One-Way ANOVA, F_(2,23) = 0.26, p = 0.77). Next, because larvae rely on vision for detecting and capturing prey, we set up to detect possible effects of NaCl exposure on their reactions to illumination change and moving visual stimuli. First, we examined motion reactions of control and treated larvae to a sudden illumination change and observed that dark-adapted larvae of either group reacted similarly to a 10 s square pulse of white light (0.1 mW*cm⁻²). They reduced their locomotor activity similarly when exposed to light (Figure 3D, Two-Way Repeated-Measures ANOVA, Group factor: F_(1,36) = 0.41, p = 0.53, Light factor: F_(2,36) = 391.7, p < 0.0001, Group X Light factor: F_(2,36) = 0.69, p = 0.51). Next, we tested the larvae’s optomotor response, i.e., spontaneous swimming in the direction of large-field displacements in the visual field. For this we presented control and treated larvae with ventrally moving dots of variable contrast, diameter, velocity and number (Figure 3E, see Methods). The results showed that NaCl exposure did not impair the larvae’s optomotor response. Figure 3F shows how dot contrast improved the larvae’s response, and how the response of larvae pre-incubated in either NaCl_50mM or NaCl_100mM did not differ from that of control larvae (Two-Way ANOVA, Group factor: F_(2,54) = 2.49, p = 0.09, Contrast factor: F_(1,54) = 69.57, p < 0.0001, Group X Contrast factor: F_(2,54) = 0.59, p = 0.56). Similar results arose from varying the diameter, velocity and number of moving dots (data not shown). In sum, locomotion, oxygen consumption and responses to visual inputs appeared unaltered under conditions of elevated HPI-axis activity caused by exposure to hyperosmotic mediums.

We next examined the extent to which stressor identity accounted for the effect of NaCl exposure on food intake. We therefore measured prey-dependent DSU changes in larvae pre-exposed to a novel stress protocol based exclusively on mechanosensory stimuli. To evoke mechanosensory stress, we used fast water movements caused by the rapid LDs of an inflexible silica capillary (360 µm OD) submerged partially (2 mm) in the larvae’s surrounding medium. The capillary was fixed to a computer-controled piezo bender actuator and the voltage applied to the actuator (or stimulus strength, in % relative to maximum voltage) determined the speed of the capillary’s LDs (maximum displacement: ±1000 µm). To examine how larvae reacted to the capillary’s LDs, we placed them individually in a rectangular swimming chamber and LDs of known frequency and duration were triggered only when they swam within the half of the chamber containing the tip of the capillary (stimulus source). Larvae responded to LDs by increasing rapidly their distance to the centre point of the moving capillary (Figure 4A). In line with this, their overall locomotion during continuous stimulation (i.e., series of 10 2-ms LDs delivered at 1 Hz, irrespective of the larva’s position relative to the source) increased together with stimulus strength (Figure 4B). We thus assumed that LD-borne stimuli resembled, at least partially, motion and pressure waves (cues) encoding predation threat, such as those derived from the movements of larger fish and approaching predators. Notably, such a form of repeated mechanosensory stimulation increased whole-body cortisol in a stimulus strength-dependent manner (Figure 4C, Kruskal-Wallis test, H = 22.4, p < 0.0001, followed by Dunn’s multiple comparison tests and linear regression), and post-peak cortisol levels were undistinguishable from basal levels 30 min after stimulation (Figure 4D, Kruskal-Wallis test, H = 10.8, p = 0.0045, followed by Dunn’s multiple comparison tests and non-linear regression). An analysis of space use after mechanosensory stimulation showed that low (30%) and moderate (60%) stimulus strength levels abolished prey-dependent DSU changes 15–25 min after exposure to LDs (Figures 4E,F, left, Paired t-tests, 30%: t₍₂₃₎ = 0.48, p = 0.63, 60%: t₍₂₃₎ = 0.004, p = 0.99). Normal DSU changes were detectable 40–50 min later (Figures 4E,F, right, Paired t-tests, 30%: t₍₂₃₎ = 4.2, p = 0.0004, 60%: t₍₂₃₎ = 4.6, p = 0.0001). As before, across group comparisons of total a.u.c. values indicated that locomotion variability could not account for prey-dependent DSU changes (Two-Way Repeated-Measures ANOVA, Group factor: F_(5,54) = 0.38, p = 0.86, Prey factor: F_(1,54) = 0.40, p = 0.53, Group X Prey factor: F_(5,54) = 0.20, p = 0.96).