Experiments took place at the Department of Neuroscience, located at the Biomedical Center, Uppsala University in Sweden in September and October 2017. Ethical approval for the use of animals was given by the Uppsala Regional Animal Ethical Committee (permit C55/13), following the guidelines of the Swedish Legislation on Animal Experimentation (Animal Welfare Act SFS1998:56) and the European Union Directive on the Protection of Animals Used for Scientific Purposes (Directive 2010/63/EU).

In the current study, a total of 73 zebrafish were involved in behavioral testing; 13 females and 17 males of the AB strain, and 21 females and 22 males of the “wild” strain, i.e., offspring of wild-caught fish. Adult AB zebrafish (d.o.b. October 2015) were obtained from SciLifeLab (Evolutionary Biology Centre, Uppsala University) and transferred to the Department of Neuroscience. The wild-caught strain was collected from the river Ichamati (approximately 70 km from Calcutta), bred in ponds, and the resulting offspring were transferred as adults to the Department of Neuroscience where they were kept in two large, aerated aquaria (200 L). The lab-raised F1 offspring (d.o.b. April 2016) from the wild-caught individuals were included in this experiment, and will hereafter be referred to as “wild zebrafish.” Hence all animals were adults at the time of testing; the AB zebrafish were 23 months and the wild strain was 18 months of age.

Experimental animals were kept in 9.5L tanks in a stand-alone rack system (Aquaneering, San Diego, United States) that was maintained at 27 ± 1.5°C with a photoperiod of 14L:10D (lights on at 07:00 AM). The aquarium system contained a particular filter pad (exchanged weekly), a fluidized bed biological filter, carbon filters and an ultraviolet sterilizer. Fish tanks were supplied with recirculating copper-free Uppsala municipal tap water of which 10% was exchanged daily. Alkalinity (66–119, median 87 mmol L^–1), conductivity (34–48, median 40 mS m^–1) and pH (8.2–8.5, median 8.4) were monitored daily. Animals were fed twice daily with flakes (tropical energy food, Aquatic Nature, Roeslare, Belgium) and Artemia brine shrimp (Argentemia Platinum Grade 0, Argent Aquaculture, Redmond, United States).

AB zebrafish were tested twice in the zMCSF test (on experimental days 1 and 7; Run 1 and 2, respectively), and twice in the NTDT (on days 2 and 8; Run 1 and 2, respectively). Wild zebrafish were tested once in each test (zMCSF on day 1 and NTDT on day 2). The morning feed was provided at least 30 min before behavioral tests were performed. None of the zebrafish used in this experiment had any previous experience of behavioral testing. All behavioral tests took place in a separate room located inside the aquarium room. The experimenter was not present or visible during video recordings. Male groups were tested before female groups, to minimize confounding effects of pheromones from ovulating females. Between trials, testing arenas were sprayed with ethanol (96%), rinsed twice with water and refilled.

All statistical analyses were carried out in R statistical computing software version 4.0.2 (R_Core_Team, 2020) with added packages “lmer” (Bates et al., 2015), “emmeans” (Lenth, 2020), “bestNormalize” (Peterson and Cavanaugh, 2019), “ggplot2” (Wickham, 2016), “pals” (Wright, 2019) and “ggalluvial” (Brunson and Read, 2020). The data was split up into two datasets, since the experimental design was not fully factorial (i.e., there was no Run 2 for the wild zebrafish). The “retested dataset” contained the data from all tests on the AB zebrafish, while the “strain dataset” contained the data from both strains on the first testing occasion. We first explored the data by conducting a principal component analysis (PCA) on each dataset, using the “prcomp” function with scaling and centering of variables.

Our initial PCA on the variables from the zMCSF revealed largely overlapping distributions of the two sexes within the same strain, while AB fish only partially overlapped with the wild strain (Figure 2A). The loading plot showed a clear separation of test variables (Figure 2B). The DCR, the CENT/CIRC and the RAMP variables all fell on the PC1 axis yet at distinct values, suggesting that these represent different magnitudes of exploration/avoidance. The corridor variables (CORR1-CORN-CORR2, hereafter CORRS) were located at a distinct position on the PC2 axis, which appeared to be related to locomotory activity.

There was a tendency for increased locomotory activity from Run 1 to 2, although this effect only reached statistical significance for total activity [GLMM, F_{(1, 28)} = 6.648, p = 0.011; Supplementary Table 4]. The sex difference in locomotory activity became more pronounced after taking into account both runs, with AB males traveling longer distances [LMM, F_{(1, 28)} = 7.452, p = 0.011] at higher speed [LMM, F_{(1, 28)} = 7.466, p = 0.011] and exhibiting higher total activity [GLMM, F_{(1, 28)} = 5.098, p = 0.024; Supplementary Table 4].

On the second testing occasion, AB zebrafish again spent the longest duration in the DCR and shortest in RAMP4, in the order: DCR > RAMP1 > REST = CENT = START = CORR1 > CORN = CORR2 > RAMP2 = CIRC > RAMP3 > RAMP4 (LMM, F_{(11, 308)} = 10.516, p < 0.001; Supplementary Table 2 and Figure 3D). For duration and frequency in zone, there was a main effect of Run [LMM, F_{(1, 644)} = 22.866, p < 0.001 and F_{(1, 512)} = 8.150, p = 0.006, respectively; Supplementary Table 4]. However, the lack of a Zone by Run interaction for all zone-related variables indicated that AB zebrafish did not allocate their time differently over zones on the second testing occasion (Supplementary Table 4). Only a single pair-wise comparison between the runs was statistically significant: AB males had a shorter average duration per visit to DCR in Run 2 compared to Run 1 [LMM contrast, t_{(10, 505)} = 3.215, p = 0.008; Supplementary Table 2].

The six areas of correlated zones outlined in section “Behavior of Female and Male AB and Wild Zebrafish in the zebrafish Multivariate Concentric Square Field” could also be distinguished in the second run, but some relationships between areas changed (Supplementary Table 3A). Upon repeated testing, activity was no longer related to duration in RAMP and CIRC zones. Animals that stayed longer in the DCR now spent less time in the CORN and CORR2 zones. Also, the negative correlation between the REST and RAMP zones was no longer present (Supplementary Table 3A).

Consistency repeatability was significant for distance moved and velocity (r = 0.432, p = 0.014 and r = 0.446, p = 0.013, respectively; Supplementary Table 5). RAMP1 had high repeatability both in terms of duration (r = 0.508, p = 0.038), frequency (r = 0.376, p = 0.041) and frequency (%) (r = 0.533, p = 0.002; Supplementary Table 5). Significant repeatabilities were also seen for the neighboring zones, average duration per visit to CORR2 (r = 0.556, p = 0.001) and frequency (%) of entering RAMP2 (r = 0.379, p = 0.039). Finally, the frequency of visits to the DCR was repeatable (r = 0.497, p = 0.005) and the repeatability of frequency to CIRC was approaching the significance level (r = 0.326, p = 0.079; Supplementary Table 5).

In the first run, the DCR was the home base for 60% of AB zebrafish (Figure 3I) and this preference was not altered by repeated testing [Poisson GLM, χ²_{(1, 57)} = 0.000, p = 1.000]. There was no significant effect of Run, Sex or their interaction on the number of zones explored [Poisson GLM, Run: χ²_{(1, 58)} = 2.251, p = 0.134; Sex: χ²_{(1, 57)} = 0.041, p = 0.840; Run by Sex: χ²_{(1, 56)} = 0.100, p = 0.752], nor whether or not all zones had been entered [Binomial GLM, Run: χ²_{(1, 58)} = 0.000, p = 1.000; Sex: χ²_{(1, 58)} = 1.248, p = 0.264; Run by Sex: χ²_{(1, 58)} = 0.000, p = 1.000]. In both runs, 8 AB females and 8 AB males (53%) left one or more zones unexplored (Figure 4).

In the NTDT, there was a main effect of Strain on all activity variables and a main effect of Sex on distance moved and velocity (Supplementary Table 1). Wild zebrafish had higher locomotory activity than AB as indicated by a longer distance moved [ANOVA, F_{(1, 69)} = 53.091, p < 0.001; Table 2 and Supplementary Table 6], higher velocity [ANOVA, F_{(1, 69)} = 18.563, p < 0.001; Figure 5A] and higher total activity [GLM, χ²_{(1, 71)} = 11.422, p < 0.001; Supplementary Table 6]. In both strains, females moved longer distances [ANOVA, F_{(1, 69)} = 9.501, p < 0.003) and at higher speed than males [ANOVA, F_{(1, 69)} = 4.214, p = 0.044], although there was no sex difference in total activity [GLM, χ²_{(1, 70)} = 0.446, p = 0.504; Supplementary Table 6].

Both zebrafish strains spent most time in the bottom and least time in the top third of the NTDT [LMM contrast, AB TOP vs. BOT, t_{(2, 179)} = –4.053, p = 0.001; wild Top vs. Bottom, t_{(2, 179)} = –7.066, p < 0.001; Supplementary Table 6 and Figure 5B]. There was a main effect of Strain and a Zone by Strain interaction for some zone-related variables (Supplementary Table 1), indicating that there were some differences in how much time each strain spent in the top, middle and bottom third of the arena. Post-hoc tests showed that wild females entered the top and middle third more often than AB females [GLMM contrast, z_{(1, 205)} = –4.830, p < 0.001; z_{(1, 205)} = –3.805, p < 0.001], while wild and AB males did not differ [GLMM contrast, z_{(1, 205)} = –0.628, p = 1.000; z_{(1, 205)} = –2.005, p = 0.270; Table 2 and Figure 5C].

Repetition of the NTDT for AB fish one week later revealed a tendency for an increase in velocity [LMM, F_{(1, 28)} = 3.791, p = 0.061] and total activity from Run 1 to 2 [GLMM, F_{(1, 28)} = 3.311, p = 0.096, Supplementary Tables 4, 6]. We did not detect a main effect of Run nor a Zone by Run interaction for any of the zone-related response variables (Supplementary Table 4).

In AB, correlations between duration spent in each zone of the zMCSF and NTDT revealed that in Run 2, duration in bottom of the NTDT correlated negatively with duration in RAMP2, 3, and 4 and duration in the top correlated positively with RAMP 2 and 3 (Supplementary Table 3A). Furthermore, duration in bottom correlated positively with DCR (Supplementary Table 3A).

We detected considerable differences in explorative behavior between laboratory and wild zebrafish. Notably, wild zebrafish swam faster and explored all zones, while AB zebrafish often left one or more zones unexplored. Wild zebrafish showed stronger avoidance of the open area in the center, but entered the shallowest zone in the arena, RAMP4, earlier. The strains did not differ in the time spent in the sheltered area (DCR). In nature, open areas are associated with increased predation risk for individual prey fish (Ruxton and Johnsen, 2016) and wild zebrafish occur in small vegetated streams feeding on insects, moving into shallow flooded areas for spawning at the start of the rainy season (Engeszer et al., 2007; Sundin et al., 2019). By contrast, the environment of laboratory zebrafish has for many generations consisted of deep tanks with no shallow areas and no vegetation cover (Sessa et al., 2008). It is possible that laboratory zebrafish over generations have become less selective with regards to the exact housing conditions, while in nature selection pressures for avoiding open water and exploring shallow water continued to act. This could explain the greater avoidance of the center zone and greater exploration of the shallow zones of the ramp by the wild strain. Indeed, the strain differences detected in the zMCSF imply that the behavioral adaption to the laboratory environment is characterized by relaxed aversion to risky areas with no potential gain (open area) and reduced exploration of risky areas with potential gain (the inclined ramp), while attraction to safe areas seems to be unaltered. These findings are congruent with the idea that domesticated animals are characterized by a low propensity to perform active behaviors (Van Reenen et al., 2005), and low stress reactivity, in other words, are “docile” rather than “shy” (Koolhaas et al., 2010; Koolhaas and Van Reenen, 2016; Rauw et al., 2017).”

The results from the zMCSF correspond well to the behavior of zebrafish in classical tests. The lower velocity and avoidance of the center of an open field by domesticated zebrafish has been reported before (Baker et al., 2018) and extends to other anti-predator behaviors, including reduced diving responses and altered responses to conspecific alarm pheromone (Mustafa et al., 2019; Vossen et al., 2020b). Moreover, the absence of large strain differences in DCR duration is in line with the absence of a strain difference in the shelter test in the same population of wild zebrafish (Mustafa et al., 2019). Thus, the strain differences we observed in the zMCSF appear to be comparable to those reported using single tests. However, the zMCSF adds the advantage of reducing the need for repeated testing, which makes the zMCSF more time-efficient, reduces handling stress and minimizes possible carry-over effects.

Comparing these results to wild and inbred lines of laboratory mice, it becomes apparent that the strong avoidance of the central zone in wild animals is a particular consistent finding across species (Augustsson and Meyerson, 2004; Augustsson et al., 2005). A higher duration and frequency of visits to the slope (corresponding to RAMP1) has also been observed in wild mice in comparison to the BALB/c laboratory line, although no differences were observed with regard to the bridge (Augustsson and Meyerson, 2004; Augustsson et al., 2005). Similar to zebrafish, the duration in DCR was not different across wild and laboratory lines of mice (Augustsson and Meyerson, 2004; Augustsson et al., 2005). Hence, actively seeking shelter does not appear to be a major factor differentiating wild and laboratory bred animals, at least not in the current test environment in which animals can freely choose where to reside. It should be noted, however, that some wild zebrafish may have considered the RAMP1 zone an (additional) home base, in which case wild zebrafish may have spent more time in the safety of their home base than AB zebrafish. We found only one discrepancy in MCSF behavior between the species; wild zebrafish showed an increased duration in the corridors, while wild mice showed a lower use of this area compared to the laboratory lines. The corridors present a semi-sheltered area where movement is possible (Roman and Colombo, 2009), therefore we suggest that this difference is driven by the higher velocity of wild zebrafish. Wild mice did not have a higher velocity than the laboratory lines, explaining the absence of a difference in the corridor duration for this species. Mice and zebrafish may indeed differ at the species level in the propensity to respond reactively (freeze) or proactively (swimming) when entering a novel environment.

Overall, we found only minor changes in explorative behavior when the AB zebrafish were tested in the zMCSF 1 week later, and the few differences we found were confined to males. AB males were more active than females in the second run, an effect that has also been reported for the open field using the same interest interval (Thomson et al., 2020). In rodents, activity during the second test occasion is dependent on developmental stage. Adult rats were less active in the MCSF upon retesting (Meyerson et al., 2006; Momeni and Roman, 2014), while adolescent rats increased the total number of visits to zones (Lundberg et al., 2019).

Male zebrafish decreased their visit duration to the DCR, an effect that has been reported in rats by some studies (Momeni and Roman, 2014; Lundberg et al., 2019), whereas others found no effect (Meyerson et al., 2006; Roman and Colombo, 2009). A consistent finding in rats is a decrease in duration and/or frequency of visits to the hurdle, slope and bridge in the second trial (Meyerson et al., 2006; Roman and Colombo, 2009; Momeni and Roman, 2014; Lundberg et al., 2019), and in most of these studies, an increase in visits to the corridors. We did not detect these same effects in zebrafish, which may reflect underlying differences in episodic/spatial memory between the zebrafish and rodents. Zebrafish can be taught to associate a color with a mild electric shock within a single training session but fail to show avoidance of this color when tested again 1 week later (Vossen et al., 2020). By contrast, rats can spatially locate exposure to an aversive stimulus when repeatedly tested 2 weeks later (Karlsson et al., 2009).

Significant repeatabilities of r∼0.5 were found for activity variables as well as for the DCR and RAMP1 zones, which is considered to be a high repeatability (Bell et al., 2009). High consistency of velocity has also been reported for zebrafish repeatedly tested in the open field (Toms and Echevarria, 2014; Baker et al., 2018; Fangmeier et al., 2018; Thomson et al., 2020). Regarding responses to novelty, high consistency of inspection duration has also been reported for zebrafish repeatedly tested in the mirror (Toms and Echevarria, 2014), novel object and predator tests (Toms and Echevarria, 2014; Fangmeier et al., 2018). The high consistency of RAMP1 is interesting since rodents frequently display stretched attend posture in this zone (the slope in the rodent MCSF), which is a key part of risk assessment behavior (Macintosh and Grant, 1963; Rodgers et al., 1999; Augustsson and Meyerson, 2004). Under the assumption that the RAMP1 (zMCSF) and the slope (rodent MCSF) are homologous, this would imply that zebrafish show equal risk assessment in repeated testing while risk taking differs.

In line with the usual pattern reported for the NTDT (Levin et al., 2007), both AB and wild zebrafish avoided the top zone. However, wild zebrafish paid more visits to the middle and top third of the tank compared to AB zebrafish. High levels of bottom dwelling in the AB strain have been reported before (Gerlai et al., 2008), although it is not clear whether this accurately reflects increased anxiety-like behavior of this strain (Vossen et al., 2020b). The positive correlation between duration in bottom (NTDT) and DCR (zMCSF) suggests that these zones contain elements of safety. The positive correlation between the number of visits to the TOP (NTDT) and RAMP (zMCSF) was expected, since both zones are located close to the water surface, an area associated with a risk for avian predation (Levin et al., 2007).

At a first glance, the overall conclusions from the zMCSF and NTDT seem to agree that the strongest differences were those between the strains and no or minor effects were seen between the sexes and between repetitions of the zMCSF. The differences between the strains were furthermore in the same direction in both tests, with wild zebrafish exhibiting higher activity than AB and moving more in shallow water. However, upon a closer look it becomes apparent that a much smaller number of effects was detected in the NTDT, even after correcting for the lower number of zones in this arena. Apart from the strain difference in activity, the NTDT revealed only two differences (wild females paid more visits to the middle and top), whereas in the zMCSF 21 strain differences were detected (17 in males and 4 in females), related to 7 zones. Indeed, this suggests that the zMCSF, like the rodent MCSF, constitutes a more nuanced behavioral test.

Yet the distinction between the zMCSF and NTDT goes deeper than just a quantitative difference in the number of effects detected. The zMCSF was able to distinguish between two independent risk-related behaviors, which are intertwined in the NTDT. Paradoxically, in the zMCSF the wild strain displayed more elaborate exploration and entered the shallowest zone more readily, while it simultaneously avoided the open areas more than AB zebrafish. In the NTDT, only a reduced diving response of wild zebrafish was detected, which (when viewed in isolation) could lead to the erroneous conclusion that wild zebrafish “showed reduced anxiety-like behavior” or “were more bold/risk taking,” which is only one side of the coin. Taking into account both tests, it rather seems that while wild zebrafish are more active explorers that avoid shallow areas less, they also display more elaborate risk assessment and are more sensitive to certain kinds of risk (i.e., open areas). Indeed, the results from the zMCSF suggest that risk taking should be considered in relation to the nature of the challenge.

Our findings indicate that the interpretation of the MCSF areas in terms of risk and safety is largely conserved between rodents and zebrafish. The DCR may be considered the safest zone of the zMCSF, since it often was the zone that was visited first, with the longest cumulative duration, indicative of a home base (Eilam and Golani, 1989; Stewart et al., 2010). The NTDT provided cross-validation for this interpretation since the duration in the DCR was positively correlated with the duration in the bottom zone. We suggest that the corridors (CORR1, CORN, CORR2) also contains aspects of safety, but in contrast to the DCR the fish can swim larger distances in this relatively sheltered area. In the corridors motor restlessness can be expressed without affecting measures of active exploration, which has also been reported for rats (Roman and Colombo, 2009). The RAMP is an area of gradually decreasing water depth, zebrafish spent little time here and were most hesitant to enter this area (especially the shallowest part, RAMP4), AB more so than wild zebrafish. The RAMP has an inverse relationship to the DCR especially in AB zebrafish, providing further evidence for an interpretation as a high-risk zone. Also the center of the arena (CIRC and CENT) may be a high-risk zone, which in particular wild zebrafish are hesitant to enter and spend little time in, while AB zebrafish more readily move through it much like movement in the corridors, suggesting the interpretation of this area is more strain dependent. Hence the RAMP and arena center may each reflect different “qualities” in terms of risk assessment and risk taking (Roman et al., 2012; Meyerson et al., 2013). Further experiments using pharmacological pre-treatments with for instance anxiolytic substances such as benzodiazepines (Bencan et al., 2009) are needed to further validate the here proposed interpretation of the zMCSF zones.