Analysis of behavioral variables across domains and strains in zebrafish: Role of brain monoamines

Important neurochemical variations between strains or linages which correlate with behavioral differences have been identified in different species. Here, we report neurochemical and behavioral differences in four common zebrafish wild-type strains (blue shortfin, longfin stripped, leopard and albino). leo zebrafish have been shown to display increased scototaxis in relation to the other strains, while both nacre and leo zebrafish show increased geotaxis. Moreover, leo displayed increased nocifensive behavior, while nacre zebrafish showed increased neo-phobia in the novel object task. lof zebrafish showed decreased turn frequency in both the novel tank and light/dark tests, and habituated faster in the novel tank, as well as displaying increased 5-HT levels. leo zebrafish showed decreased brain 5-HT levels and increased 5-HT turnover than other strains, and nacre had increased brain DA levels. Finally, specific behavioral endpoints co-varied in terms of the behavioral and neurochemical differences between strains, identifying cross-test domains which included response to novelty, exploration-avoid-ance, general arousal, and activity.


Introduction
Important genetic variations between strains or lineages which correlate with behavioral differences have been identified in different species. The degree with which these variations explain the behavioral differences is not fully understood, but the use of behaviorally distinct strains might represent an important model to understand human behavioral disorders [1][2][3][4].
In this direction, mouse and rat inbred strains have been shown to differ in anxiety-like behavior and impulsivity [5][6][7][8][9]. The use of genetically tractable organisms, including invertebrate models and non-mammalian vertebrates, could generate important information regarding the genetic architecture underlying these disorders [10].
In this sense, zebrafish (Danio rerio Hamilton 1822) represent important additions to this arsenal, presenting technical advancements such as optogenetics and transgenesis which facili-strains have been shown to differ in many important traits which are relevant for their behavior, such as brain transcriptome [14] and neurochemistry [15]. Some of these strains have also been subjected to behavioral testing, such as the novel tank test, and shown to differ in terms of anxiety-like behavior [14,[16][17][18][19] and associated functions, such as habituation to novelty [20,21] and boldness [22,23]. While these behavioral endpoints have been assessed mostly in animals from inbred and outbred strains such as AB, WIK, SH, Tü and TL (http://zfin.org), commonly found mutant phenotypes were also used, including skin mutant phenotypes such as leopard, albino and longfin [24]. For example, Egan et al. [24] demonstrated that, in relation to wild-type shortfin, albino and leopard zebrafish show increased bottom-dwelling; Kiesel et al. [16] demonstrated a similar profile for longfin mutants.
In the present work, we analyze the behavioral and neurochemical differences between the common WT phenotypes shortfin, longfin, leopard and albino.

Animals and housing
40 animals from the blue shortfin phenotype (bsf), 40 from the longfin stripped phenotype (lof), 40 from the albino phenotype (nacre) and 40 from the leopard phenotype (leo) were used in the present study. Animals were group-housed in mixed-phenotype 40 L tanks, with a maximum density of 25 fish per tank. Tanks were filled with deionized and reconstituted water at room temperature (28 °C) and a pH of 7.0-8.0. Lighting was provided by fluorescent lamps in a cycle of 14-10 hours (LD), according to the standard of care zebrafish [25]. All manipulations minimized their potential suffering, as per the recommendations of the Canadian Council on Animal Care [26]. All procedures complied with the Brazilian Society for Neuroscience and Behavior's (SBNeC) guidelines for the care and use of animals in research.

Novel tank test
The protocol for the novel tank diving test used was modified from Cachat et al. [27]. Briefly, animals were transferred to the test apparatus, which consisted of a 15 x 25 x 20 cm (width x length x height) tank lighted from above by two 25 W fluorescent lamps, producing an average of 120 lumens above the tank. As soon as the animals were transferred to the apparatus, a webcam was activated and behavioral recording begun. The webcam filmed the apparatus from the front, thus recording the animal's lateral and vertical distribution. Animals were allowed to freely explore the novel tank for 6 minutes, after which they were removed from it and exposed to the scototaxis tank. Video files were later analyzed by experimenters blind to the treatment using X-Plo-Rat 2005 (http://scotty.ffclrp.usp.br), and the images were divided in a 3 x 3 grid composed of 10 cm² squares. The following variables were recorded: erratic swimming: the number of "erratic swimming" events, defined as a zig-zag, fast, unpredictable course of swimming of short duration; and freezing: the total duration of freezing events, defined as complete cessation of movements with the exception of eye and operculae movements.
homebase: For the establishment of homebases, the number of visits and time spent in each 10 cm 2 square were calculated and expressed as % of total; a zone qualified as a homebase based on the maximal percentages for individual animals.

Light/dark test
Determination of scototaxis was carried as described elsewhere [28,29]. Briefly, animals were transferred to the central compartment of a black and white tank (15 cm X 10 cm X 45 cm h X d X l) for a 3-min. acclimation period, after which the doors which delimit this compartment were removed and the animal was allowed to freely explore the apparatus for 15 min. The following variables were recorded:  risk assessment: the number of "risk assessment" events, defined as a fast (<1 s) entry in the white compartment followed by re-entry in the black compartment, or as a partial entry in the white compartment (i.e., the pectoral fin does not cross the midline).

Novel object exploration test
The novel object task was adapted from Sneddon [30]. Animals were transferred to a 15 x 25 x 20 (width x length x height) tank and allowed to acclimate for 5 minutes. After that period, a novel object (made up of a combination of red, yellow, green, blue and black Lego® Duplo bricks such that the object was no longer than 9 cm in length and 6 cm in height) was slowly lowered into the tank (so as not to startle the fish) and placed at the center of a (previously defined) 10 cm diameter circle at the middle of the tank. A webcam filmed the apparatus from above, and the time spent within that circle and the number of squares crossed were recorded for 10 minutes.

Nocifensive behavior
To assess behavioral responses to a chemical, inescapable nociceptive stimulus, animals were acclimated to the test tanks (10 cm length X 10 cm width X 20 cm height Plexiglas tanks containing water from the home tank) for 30 min and then individual baseline (pre-treatment) locomotor responses (number of 3 x 3 squares crossed during the session) were monitored for 5 min. Each fish was then individually cold-anesthetized and injected in the anal fin with a 1% solution of acetic acid. Afterwards, animals were returned to the original test tanks to recover from anesthesia, after which behavioral recording took place. The frequency of tailbeating events, in which the animal vigorously moves its tail but do not propel itself in the water [31], and the change in total locomotion in relation to the baseline [32], were recorded as variables pertaining nocifensive behavior.

Cluster analysis
Raw data was first transformed into Maximum Predictive Values (MPV), following the approach of Linker et al. [33]. Briefly, taking the data from wild-type shortfin animals as reference, for each variable the MPV was calculated as the ratio of the mean difference between two groups and their pooled standard deviations as follows: , where . CC-BY-NC 4.0 International license a certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made available under The copyright holder for this preprint (which was not this version posted May 27, 2016. . Given the mathematical simplicity of these measures, MPV scores were automatically calculated by LibreOffice Calc 3.6.6.2. These scores represent the intensity (positive or negative) that fish from a given strain displayed a given behavioral endpoint in relation to shortfin fish. Resulting scores were normalized by centering each endpoint around the mean. Hierarchical clustering was then performed across behavioral endpoints and strains ('arrays') with Cluster 3.0 (University of Tokyo, Japan) using uncorrected correlation as clustering method, and single linkage as similarity metric.
Clustering results were visualized as a dendrogram and colored "array" in Java TreeView (University of Glasgow, UK).

Novel tank test
nacre and leo fish spent less time in the top third of the novel tank than bsf zebrafish (F [3,39] Figure 1C). lof and nacre froze more than bsf, and leo froze less than lof (F[3, 39] = 6.506, p = 0.0012; Figure 1D). Homebase behavior did not differ across strains (F [3,39] = 0.9261, NS; Figure 1E). Habituation scores were higher in lof than bsf, smaller in nacre than bsf, and smaller in nacre and leo than in lof (F [3,39] = 24.63, p < 0.0001; Figure 1F). With the exception of nacre, all strains spent more time on the top in the last 3 min than in the first 3 min (F[3, 72] = 2.82, p = 0.0449; Figure 1G).

Novel object test
nacre zebrafish spent less time near the novel object than bsf (F [3,39]

Clustering
Cluster analysis using lof, nacre and leo zebrafish against a bsf "reference" produced four . CC-BY-NC 4.0 International license a certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made available under The copyright holder for this preprint (which was not this version posted May 27, 2016. ; https://doi.org/10.1101/055657 doi: bioRxiv preprint identifiable behavior clusters (Figure 6), the first including tail-beating, time on top, habituation score, time near novel object and NE levels (r² = 0.579); the second ("exploration-avoidance") including freezing in the light/dark test, time on white, change in activity after acid injection, number of entries on white and DA and 5-HT levels (r² = 0.741); the third ("general arousal") including turn frequency in both the novel tank test and the light/dark test, locomotion in the novel tank test, thigmotaxis, risk assessment, latency to white, DOPAC and MHPG levels, and the turnover of all monoamines (r² = 0.767); and the last ("activity") including time spent on the homebase, locomotion in the novel object test, freezing in the novel tank test, and 5-HIAA levels (r² = 0.930). nacre and leo clustered together (r² = -0.243), with lof as outgroup.

Discussion
In the present work, leo zebrafish have been shown to display increased scototaxis in relation to the other strains, while both nacre and leo zebrafish show increased geotaxis. Moreover, leo displayed increased nocifensive behavior, while nacre zebrafish showed increased neophobia in the novel object task. lof zebrafish showed decreased turn frequency in both the novel tank and light/dark tests, and habituated faster in the novel tank, as well as displaying increased 5-HT levels. leo zebrafish showed decreased brain 5-HT levels and increased 5-HT turnover than other strains, and nacre had increased brain DA levels. Finally, specific behavioral endpoints co-varied in terms of the behavioral and neurochemical differences between strains, identifying cross-test domains which included response to novelty, exploration-avoidance, general arousal, and activity.
In both the novel tank test and in the light/dark test, nacre zebrafish showed increased freezing. While this behavior is poorly understood in zebrafish, freezing behavior does seem to vary with genetic background. Cachat et al. (2011) observed a small difference between bsf and leo zebrafish in freezing in the novel tank test, while no differences between lof and leo zebrafish were observed both in the NTT and the light/dark test [34]. AB zebrafish selected for high freezing in the open field test show increased bottom-dwelling, increased alarm reaction, increased scototaxis and increased latency to feed in both disturbed and undisturbed conditions [19]. In another study, Blaser et al. (2010) demonstrated that zebrafish which consistently avoid the white compartment also freeze more after being confined to the white com-. CC-BY-NC 4.0 International license a certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made available under The copyright holder for this preprint (which was not this version posted May 27, 2016. ; https://doi.org/10.1101/055657 doi: bioRxiv preprint partment. Thus, freezing seems to reflect either a fear response or a response to stressful manipulations, and therefore strains which show prominent freezing in the novel tank and light/dark tests could represent 'reactive' strains. Interestingly, the response of adult zebrafish with a mutation in the glucocorticoid receptor (gr s357 ) after transference to a novel environment is to freeze instead of explore, an effect which is reversed by acute diazepam or subchronic fluoxetine treatment [36]. Likewise, transient knockdown of tyrosine hydroxylase 1 during development decreases freezing in adult zebrafish exposed to a novel tank [37], suggesting an important role for catecholamines in this response.
Lending credibility to such interpretations is the observation that nacre zebrafish also show decreased exploration of novel objects, as well as decreased habituation in the novel tank test and increased risk assessment in the light/dark test. In another study, differences between nacre zebrafish and bsf were observed in bottom-dwelling [24]. Thus, these animals show exaggerated stress responses to novelty, similarly to gr s357 mutants [36,38]. This response does not necessarily result from increased anxiety, as decreased 'novelty-seeking' could also be responsible for these results (Hughes, 1997;Hughes, 2007). A non-selective exaggerated responsiveness to stressors is discarded by the observation that nacre display normal nociceptive behavior after acetic acid injection. Thus, this common mutant may represent an important addition in behavioral genetics in the sense that it shows selective responsiveness to novelty, but not to nociceptive or simple anxiogenic stimuli.
In the literature, erratic or burst swimming has been defined as sharp changes in direction or velocity and repeated darting [39] which, in the novel tank test, are increased by 'anxiogenic' manipulations such as morphine withdrawal, alarm substance presentation and caffeine administration [18]; and decreased by acute fluoxetine and 5-HT1B receptor antagonists [40].
These measures commonly, but not necessarily, include the fast turns quantified in the present article as 'turn frequency'. While not necessarily being equivalent to erratic swimming measures reported elsewhere in the literature, turn frequencies are of ecological relevance, because zebrafish can turn against the water current only until the current speed equals their routine maximum swimming speed [41,42]. In the present manuscript, turn frequency was higher in leo than in other zebrafish strains in both the novel tank and the light/dark test.
Nonetheless, turn frequencies did not differ between lof and bsf zebrafish, which should be expected if this variable was controlled solely by metabolic and/or biomechanic constraints, . CC-BY-NC 4.0 International license a certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made available under The copyright holder for this preprint (which was not this version posted May 27, 2016. ; https://doi.org/10.1101/055657 doi: bioRxiv preprint as observed in routine swimming [41]. These results are also consistent with observations that erratic swimming does not differ between bsf and lof zebrafish [16], between bsf and leo [18] or between lof and leo [34].
From the neurochemical point of view, some observations call attention. First, all strains had higher norepinephrine levels than the reference bsf. In the multivariate analysis, NE levels grouped in the first cluster, which included tail-beating in the nocifensive behavior assay, time spent near the novel object, and time on top and habituation in the novel tank test. The third cluster extracted in our analysis shows behavioral endpoints more consistent with generalized arousal, such as turn frequency in both the novel tank test and the light/dark test, locomotion in the novel tank test, thigmotaxis, risk assessment and latency to white; moreover, the metabolites of dopamine and norepinephrine, DOPAC and MHPG, as well as the turnover rates of all neurotransmitters analyzed, clustered in this group. While NE has been proposed to mediate many different behaviors, in zebrafish noradrenergic drugs so far have been shown to modulate arousal [43]. Along with increased responsiveness to sensory stimuli and voluntary motor activity, increased arousal leads to increased emotional reactivity [44,45], and other neurotransmitter systems, including 5-HT [46], have been implicated in zebrafish arousal.
While NErgic neurotransmission was higher in lof, nacre and leo, 5-HT levels were lower in leo, which also show increased anxiety-like behavior in the light/dark test and in the novel tank test, as well as increased nocifensive behavior. leo also showed increased 5-HT turnover and increased MPHG levels, suggesting increased monoamine oxidase activity. Dopamine levels were altered only in nacre, reinforcing the hypothesis of elevated reactivity to novelty in these mutants. DA and 5-HT levels clustered together with change in activity in the nocifensive behavior assay, as well as freezing, time on white, and number of entries on white in the light/dark test. A role for serotonin in this assay has been proposed in zebrafish [34,40,47,48], but so far little is known about the role of this neurotransmitter in fish nociception, nor on the role of dopamine in scototaxis.
The heterogeneous nature of behavioral variation in this paper supports our anterior notion that behavioral tests of 'anxiety' in zebrafish do not necessarily measure the same dimensions [49]. The present results suggest that these tests fall under the aegis of 'domain interplay' [50], with different behavioral endpoints mapping to different behavioral domains. Using a similar approach to cluster analysis presented in this paper,  demonstrated . CC-BY-NC 4.0 International license a certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made available under The copyright holder for this preprint (which was not this version posted May 27, 2016. ; https://doi.org/10.1101/055657 doi: bioRxiv preprint the existence of two major clusters in the novel tank test, the first including (among others) latency to upper half of the tank, freezing and erratic swimming, and the second including time spent in the upper half, distance traveled and average velocity. Importantly, these clusters grouped in relation to the effects of 'anxiolytic' manipulations (which decrease behaviors from the first cluster and increase behavior from the second) and 'anxiogenic' manipulations (with the opposite effect); the latter include animals from the leo strain. Moreover, clustering based on habituation rates, instead of anxiety level, produces different results in the same assay [21], suggesting that anxiety and habituation are independent in the NTT. In this latter work, leo zebrafish were also shown to habituate freezing faster than bsf, but this effect was inversely affected by exposure to an alarm substance or to acute caffeine treatment, and freezing habituation was actually increased by anxiolytic treatments (chronic ethanol, chronic fluoxetine, acute nicotine, chronic morphine).
In conclusion, the present paper demonstrated that common wild-type zebrafish strains differ in their behavior in multiple behavioral assays, suggesting a genetic basis for conflict-and novelty-stress induced behavior, as well as in nocifensive behavior. Moreover, a monoaminergic substrate for these differences has also been described. In general, the identification of the genes and neural substrates underlying the behavioral variation of these common zebrafish mutants could represent important additions to the arsenal of tools to understand the neurogenetics of anxiety disorders. Bars represent mean ± standard error. Boxplots represent median ± interquartile range, with Tukey whiskers. ***, p < 0.001; **, p < 0.01; *, p < 0.05.
. CC-BY-NC 4.0 International license a certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made available under The copyright holder for this preprint (which was not this version posted May 27, 2016. ; https://doi.org/10.1101/055657 doi: bioRxiv preprint . CC-BY-NC 4.0 International license a certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made available under  . CC-BY-NC 4.0 International license a certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made available under The copyright holder for this preprint (which was not this version posted May 27, 2016. ; https://doi.org/10.1101/055657 doi: bioRxiv preprint . CC-BY-NC 4.0 International license a certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made available under The copyright holder for this preprint (which was not this version posted May 27, 2016. ; https://doi.org/10.1101/055657 doi: bioRxiv preprint Figure 3 -Behavioral differences between zebrafish from the blue shortfin (bsf), longfin (lof), albino (nacre), and leopard (leo) phenotypes in the novel object exploration test (NOET). (A) Time spent near the object in the whole 10-min session; (B) Total number of squares crossed in the 10-min session. Bars represent mean ± standard error. Boxplots represent median ± interquartile range, with Tukey whiskers. *, p < 0.05.
. CC-BY-NC 4.0 International license a certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made available under The copyright holder for this preprint (which was not this version posted May 27, 2016. ; https://doi.org/10.1101/055657 doi: bioRxiv preprint . CC-BY-NC 4.0 International license a certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made available under The copyright holder for this preprint (which was not this version posted May 27, 2016. ; https://doi.org/10.1101/055657 doi: bioRxiv preprint Figure 4 -Behavioral differences between zebrafish from the blue shortfin (bsf), longfin (lof), albino (nacre), and leopard (leo) phenotypes in the nocifensive behavior assay. (A) Frequency of tail-beating events; (B) Change in baseline activity in relation to pre-injection levels. Bars represent mean ± standard error. Boxplots represent median ± interquartile range, with Tukey whiskers. **, p < 0.01; *, p < 0.05.
. CC-BY-NC 4.0 International license a certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made available under The copyright holder for this preprint (which was not this version posted May 27, 2016. ; https://doi.org/10.1101/055657 doi: bioRxiv preprint . CC-BY-NC 4.0 International license a certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made available under The copyright holder for this preprint (which was not this version posted May 27, 2016. ; https://doi.org/10.1101/055657 doi: bioRxiv preprint Bars represent mean ± standard error. ***, p < 0.001; **, p < 0.01; *, p < 0.05. . CC-BY-NC 4.0 International license a certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made available under The copyright holder for this preprint (which was not this version posted May 27, 2016. ; https://doi.org/10.1101/055657 doi: bioRxiv preprint Figure 6 -Hierarchical clustering of behavioral and neurochemical variables (rows) vs. phenotypes (columns). Clustering was made by calculating Maximum Predictive Values in relation to a reference phenotype (blue shortfin).
. CC-BY-NC 4.0 International license a certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made available under The copyright holder for this preprint (which was not this version posted May 27, 2016. ; https://doi.org/10.1101/055657 doi: bioRxiv preprint