The canonical HPA axis facilitates and maintains light adaptive behavior

The vertebrate stress response (SR) is mediated by the hypothalamic-pituitary-adrenal (HPA) axis and contributes to generating context appropriate physiological and behavioral changes. Although the HPA axis plays vital roles both in stressful and basal conditions, research has focused on the response under stress. To understand broader roles of the HPA axis in a changing environment, we characterized an adaptive behavior of larval zebrafish during ambient illumination changes. The glucocorticoid receptor (nr3c1) was necessary to maintain basal locomotor activity in light and darkness. The HPA axis was required to adapt to light more efficiently but became dispensable when longer illumination was provided. Light adaptation was more efficient in dimmer light and did not require the mineralocorticoid receptor (nr3c2). Our findings show that the HPA axis contributes to the SR at various stages, facilitating the phasic response and maintaining an adapted basal state, and that certain adaptations occur without HPA axis activity.


Introduction
The stress response (SR) is de ned as the body's response to actual or perceived threats.The SR is mediated by the hypothalamic-pituitary-adrenal (HPA) axis and has evolved in vertebrates.When a stimulus reaches the paraventricular nucleus (PVN) of the hypothalamus, the SR is initiated in a hormonal cascade starting with the secretion of corticotropin releasing hormone by the PVN.Subsequently, the anterior pituitary releases adrenocorticotropic hormone (ACTH) that binds its receptor, MC2R (melanocortin receptor type 2) on the adrenal gland.The adrenal gland in turn secretes the effector molecule glucocorticoids (GCs; cortisol for humans and sh; corticosterone for rodents).GCs travel throughout the body and effectuate the SR by binding to their cognate receptors (glucocorticoid receptor [GR; nr3c1] and mineralocorticoid receptor [MR; nr3c2]). 1,2Thus, the SR is composed of two parts: the central perception of the stimulus and production of the effector molecule (GCs), and the physiological changes occurring in diverse peripheral tissues by the function of the GCs and their receptor binding. 3havioral changes occur together with these physiological changes.The responses from the brain and peripheral tissues are intertwined and modulate each other's states. 4spite the connotation of the name, the SR machinery operates in both stressful (phasic) and basal (tonic) states of the body. 3,5During the phasic response after encountering potent stressors, high levels of GCs are secreted and drive a broad range of physiological and behavioral changes.Increased lipolysis and blood glucose levels, increased attention to the stressor, and increased or halted locomotion are some examples. 6,70][11][12] In this manner, the SR machinery and GCs regulate basal wakefulness and metabolism. 13[16][17][18] Such broad-scale actions are made possible by the biphasic (slow or rapid) signaling of the GCs and its receptor (GR).While glucocorticoid receptor (GR) is a transcription factor that modulates gene expression levels (slow, genomic response), GR also functions as a signaling molecule enabling a rapid, non-genomic GC response within seconds to minutes of a stimulus exposure. 3,19The long-lasting modulation of gene expression and rapid signaling in both the stressful (phasic) and basal (tonic) states make GCs the key molecule with which our perception in the brain is translated into tangible physiological and behavioral changes.Together with the nervous system, GCs impart our body's physiological responses and coordinate appropriate adaptations in accordance with the changing environment, both gradually and rapidly. 7,20Thus, it is not a surprise that altered HPA axis activity is one of the most common ndings in people with psychiatric disorders.Peripheral physiological symptoms accompany mental burdens in mood disorders and the ways in which psychological and somatic symptoms in uence each other continues to be of importance. 21,22Thus, we sought to deepen our understanding of the dynamics and role of the HPA axis during an adaptive process to changing environmental conditions.Such knowledge may bring insight into how the HPA axis operates and how maladaptation of our body to the environmental challenges may arise.
The zebra sh is well-suited for such investigations involving the SR and corresponding changes in physiology and behavior.The hypothalamic-pituitary-interrenal (HPI) axis in zebra sh is functionally homologous to the mammalian HPA axis. 23,245][26][27][28] Zebra sh are a diurnal species with the same circadian pattern as humans. 29Larval zebra sh develop rapidly, establishing most organ systems by 5 dpf, including the visual system. 25,30Larval zebra sh exhibit a repertoire of phototaxis and prototypical swim behavior. 31,323][34] Using a larval zebra sh model, we speci cally asked what the role of the HPA axis is during the adaptation to changing ambient light conditions.We did not presume that the change itself is a stressful encounter, rather inquired the role of the HPA axis during the period of changing illumination.We leveraged a photo adaptive behavior in larval zebra sh that we previously showed to be HPA axis dependent.Dark-acclimated zebra sh larvae moved signi cantly more after a 1-min illumination of white light in the post-illumination darkness, which was dependent on the key receptors in the HPA axis. 23With zebra sh that carry a mutation in mc2r (ACTH receptor), nr3c1 (glucocorticoid receptor), or nr3c2 (mineralocorticoid receptor), we had shown that mc2r and nr3c1 homozygous (HM) mutants had decreased locomotion in the post-illumination darkness after 1-min illumination, whereas nr3c2 mutants did not. 23Using the same mutant zebra sh lines, we expanded our investigation into locomotor responses to changing ambient illumination.
We hypothesized that sh with HPA axis perturbation show compromised light adaptation.In the expanded paradigm, we provided alternating illumination and dark conditions.Fish were acclimated in darkness, and dark and light conditions were provided for 7.5 minutes in four cycles nishing the experiment in 25 min of darkness: dark acclimation (30 min) -4x [dark (7.5 min) -light (7.5 min)] -dark (25 min) (Fig. 1).We observed changes in locomotor responses after changing the duration of illumination to 2, 4, or 6 minutes (i.e., dark acclimation (30 min) -4x [dark (7.5 min) -light (2 min)] -dark (25 min).When we varied the illumination in duration and intensity, HPA axis receptors were required to mount increased locomotor response only after shorter durations of light (2 and 4 min), but not after longer illumination (7.5 min).At baseline without any changes in illumination, nr3c1 mutant larvae moved less than their WT siblings in constantly lit and dark conditions.Beyond a certain threshold of brightness, dimmer and brighter light elicited the same pattern of response while dimmer light did not require HPA axis signaling to do so even during shorter durations of light.Using the adaptive behavior of zebra sh, we report that a functioning HPA axis is required to facilitate photoadaptation, enabling adaptation even after a short 1 or 2-min illumination, and to sustain basal locomotor activity.Rejecting our initial hypothesis, HPA axis perturbation did not lead to a uniform disruption of light adaptation process.When the stimulus was long enough in duration (≥ 6 min) or was low intensity (300 lux), the adaptive behavior could be achieved without HPA axis activity, indicating multiple pathways that lead to the adaptation.The HPA axis contributes to characteristic locomotor responses during light adaptation at various stages, facilitating initiation and maintaining higher levels of locomotion following basal adaptation.

Materials and equipment
We listed the materials and equipment used in this study (Supplementary Table S1).

Zebra sh husbandry
Wild-type (WT) zebra sh (Danio rerio) originally purchased from Segrest Farm in Florida were outbred to keep a genetically diverse, healthy stock.Fish were handled and cared for following standard practices. 35The Institutional Animal Care and Use Committee (IACUC) in the Mayo Clinic (A345-13-R16, A8815-15) approved the animal husbandry and study protocol.All experiments and methodologies were performed in accordance with relevant guidelines and regulations.The Methods and Results are reported in accordance with ARRIVE guidelines. 36Adult sh were kept in a 9 L (25-30 sh) or 3 L (10-15) housing tank at 28.5°C with a light:dark (14:10) cycle.All experiments in this study were conducted between 4 and 7 dpf before the zebra sh sex is determined at about 15 dpf, 30,[37][38][39] and thus sex determination was not made.

Mutant zebra sh lines
The same WT and mutant zebra sh lines that we previously reported were used. 23All sh were maintained through outbreeding.There were three mc2r mutant lines that each carried a frameshift mutation in exon 1 (two 4-and one 5base pair deletions; mn 57 , mn 58 , and mn 59 , respectively).The annotation on the gene mc2r in the National Center for Biotechnology Information (NCBI) has changed from having 2 exons to having 1 exon since our previous report.Four nr3c1 frameshift mutants were used, each of which carried a 7-or a 17-bp deletion in exon 2 (mn 61 , mn 62 ) or a 4-or a 5-bp deletion in exon 5 (mn 63 , mn 65 ).A nr3c2 frameshift mutant was used that carried a 55-bp deletion in exon 2 (mn 67 ).For detailed information on the mutant lines, refer to our previous paper. 23

Behavioral assay preparation
Adult mating pairs were placed in mating tanks separated by dividers (-1 dpf).On the following day, the divider was pulled between 8:30 and 9:00 am and embryos were obtained via natural spawning (0 dpf).Unfertilized and unhealthy embryos were cleaned up on 0, 1, and 3 dpf in the petri dish.Morphological defects (deformity, death) were the only exclusion criteria for an embryo to be excluded from an experiment.Larvae from different parent pairs were mixed in each petri dish to randomize the animals.On 3-4 dpf, a single larva was placed in each well of a 48-well plate.Larvae were caught from varying regions of the dish (i.e., center of the dish, areas close to the wall), from various depths (i.e., surface of the water, close to bottom of the dish), and throughout various swim behavior (i.e., fast swimmer, quiescent sitter) to ensure each 48-well plate contains larvae with diverse behavioral patterns.On 5 dpf, behavioral assays were performed in a custom-built assay chamber.From 0-5 dpf, plates were stored in an incubator with a light:dark cycle (14:10 hrs) at 28.5°C.

Custom light boxes
Light boxes were custom produced by the Mayo Clinic Division of Engineering.The light box has a control panel with two knobs that enable light intensity adjustment.Both infrared (IR) and white light have a low, medium, and high intensity.The illumination was provided from the bottom of the box through an additional translucent white acrylic board to evenly diffuse the light.
The spectral range of white light was approximately between 420 and 780 nm (STS-VIS; Ocean Optics Inc.; Supplementary Figs.S1-S3) and its light power (irradiance; wattage/unit area) was 20.5, 240.0, and 469.4 µW•cm − 2 for the low, med, and high, respectively (Benchtop optical power meter; 1936-R; Newport Corp.).When translated into brightness (illuminance; lux) measured by a light meter application on a cell phone, the power measurements were equivalent to ~ 300, 4000, and 8000 lx, respectively.
The spectral range of infrared (IR) light was approximately between 780 and 880 nm (STS-NIR; Ocean Optics Inc.; Supplementary Figs.S4-S6) and its light power was 6.3, 58.1, and 116.0 µW•cm − 2 for the low, med, and high, respectively.The IR light produces 0 lx in brightness measurement.For detailed information on the custom light boxes (e.g., dimensions), refer to our previous paper. 23sal locomotor activity assays On the assay day, larvae were acclimated in either dark or light for 30 minutes before videorecording started (HDR-CX560V; Sony Corp.; Supplementary Table S1).Videorecording (30 frames per second) started around 9 am and the recording continued for 13 hours without any changes in the light condition that the sh were acclimated (Fig. 1B).

Dark-light repeat assays
All assays were performed at the high intensity for both IR (116.0 µW•cm − 2 ; 0 lx) and white (469.4µW•cm − 2 ; 8000 lx) light, except the dim light assays.The dim light assays were done at the high intensity for infrared and at the low (20.5 µW•cm − 2 ; 300 lx) for white light.
The dark period was recorded in IR.Although zebra sh are thought to be unable to detect infrared and the acclimation to IR led to quiescence in locomotion, negative phototaxis to near-IR was reported. 40The larval zebra sh on 5 dpf were acclimated in IR for 30 minutes and underwent the dark (7.5 min) and light (7.5 min) periods four times.In the assays with shorter durations of illumination, 2, 4, or 6-min illumination was provided while keeping the length of the dark phase constant at 7.5 minutes.Assays ended with a 25-min dark period.The regimens were abbreviated with the repeat element (i.e., [7.5 + 2-min]) in the text.Regimen: 30-min dark acclimation + 4x [7.5-min dark + 7.5-min light] + 25-min dark (Fig. 1C).In some assays that were performed earlier in the project, the nal 25-min dark period was omitted since the full experimental protocol was not established.

Dim light assays
In the dim light assays, the same protocol was followed except that the lowest of three intensity settings was used for white light (20.5 µW•cm − 2 ; 300 lx). 23
Response (dependent) variable.Locomotor activity was videorecorded (30 fps), and one data point was used for every second.Total distance moved for one minute is summed at every second in a sliding window of a minute (rolling sum; mm/min).The mean of the total distance moved per minute at each second was calculated for each experimental condition of the explanatory variable for each assay (i.e., WT, HT, and HM in assay 1).This rolling sum in each assay was the primary analysis unit.Since the variability in the dark-light assay is not established, we used our previous study to estimate the appropriate number of animals necessary.In our previous paper, 23 we used 500-1,000 sh for a set of an experiment (i.e., 1-min light assay with mc2r lineage siblings).We used a pair of 48-well plates and thus 500-1,000 sh translate to approximately 5-10 assays.Our aim was to t the number of assays between 5 and 10.However, since we intended to perform a set of experiment on multiple days to account for unexpected effects of a particular day and used natural spawning, the nal number of assays in each set of experiment varied between 4 and 11, with an exception of dim light assays using WT sh.The dim light assay with WT sh was concluded with 3 assays per experiment.
Statistical analysis work ow.Using raw movement data obtained in csv les for each assay (Fig. 1D): (1) The total distance moved for a min (rolling sum) at each second was computed for each explanatory variable group, (2) a Generalized Additive Model (GAM) 41 was tted to the data, (3) pairwise comparison on the predicted response variable from the model was conducted at each second, (4) proportion of the signi cance during a photo period was computed, and (5) comparison between the proportions of the signi cance was performed among different assay paradigms.All analyses were performed using the R language (4.3.0;Already Tomorrow) 42 and all gures except the schematics (Fig. 1) were produced with R. All primary data and supplementary materials were deposited in the open access data repository ( gshare.com;10.6084/m9.gshare.23734398).The core R scripts for GAM modeling were deposited to GitHub.com(https://github.com/moonlarkalto/HPA_gam)and the data sets for test-running the R scripts were deposited to FigShare at the same link).
Model.A generalized additive model (GAM) was developed to describe the locomotion of larval zebra sh in response to illumination changes.The generic formula for GAMs can be conceptually represented as: where is a link function, is mean (expected) values of the response variable, and are a design matrix and its parameter vector, respectively, of the linear term of , is a group of smooth functions to the term of , is a set of smooth functions to another explanatory variable of , and is an error term.The response variable can come from any exponential family and some non-exponential family distributions.
A GAM with the explanatory variables of genotype and time was tted to the dark-light repeat assay data in R 42 using the mgcv package (v1.8-42): 43where is the response variable of the rolling sum, denotes the relationship between the response and explanatory variables, is a linear explanatory term as a categorical variable (WT, HT, and HM; the main effect of this term is estimated), constructs a set of smooth functions for the smooth term , (basis dimension) is the number of base functions that constitute the smooth functions for the term (sets the upper limit on the degrees of freedom for the smoother of the term), and means a separate set of smooth functions are estimated for the term for each condition in .The smooth term for (representing each assay) is added as a random effect among the different assays where is an option to choose the type of spline (the thin plate spline [ ] is used as the default technique to produce smooth curves) and is a spline option of random effect for the term .Refer to the mgcv package manual for information. 44Similarly, a GAM with the explanatory variables of illumination condition and developmental stage was tted to the baseline assay data.There was no interaction detected among the terms, so the interaction was not modeled.
3 When the experimental period is longer (about an hour in the dark-light repeat assay vs. about 12 hours in the baseline assay), the number of base functions necessary to t the model increased ( in the equation [2] vs. in [3]).In GAMs, increased numbers of base functions produce more wiggly lines, more closely re ecting the changes in the actual data, while a straight line is produced when .To prevent over tting, unnecessarily high values were penalized during the tting process as well as when competing models were compared for tness.
The experimenter only sets the upper limit ( ) and the optimal value is determined by the underlying optimization algorithm in the mgcv package, which can be further validated for its tness statistics.
The model was rst heuristically evaluated by the proportion of the variation explained, the distribution of the residuals, and the parameters produced by the function in mgcv.The model was considered inadequate when (1) the predicted values do not reasonably trace the actual data upon visual inspection, (2) the proportion of variation explained is too low, (3) the p-value (different from the p-value for inference on main effects) for each smooth term in the summary is too small (has to be comfortably not signi cant).Among the valid models with differing parameters, the tness of those models was compared using Akaike information criterion (AIC).
Inference for the locomotion.Once a GAM model is tted, the overall effects and signi cance of each linear and smooth term on the response variable were evaluated, visualized, and recorded.Based on the tted model, the predicted values of the response variables were extracted using the function.With the predicted values of the rolling sum, post hoc pairwise comparisons were performed using the emmeans package (v1.8.5). 45The estimated marginal means (EMMs; least-squares means) for each explanatory variable condition was computed (i.e., equation [4]), the difference between a pair of conditions was analyzed, and the signi cance of the difference was produced at every second (i.e., equation [5]): The pairwise comparison ( ) produced the Tukey-adjusted p-values based on the t ratios drawn from the Studentized Range distribution, adjusted for multiple comparisons.The p-values were used to determine whether there is a signi cant difference in the distance moved between the two conditions (i.e., WT vs. HM) at each second.
Proportion of signi cance during a photo period.Based on the inference (signi cance) drawn at every second, the proportion that a group moved signi cantly more than the comparison group during a given photo period was computed (i.e., mc2r WT larvae signi cantly moved more than their HM siblings in 89.3% of the "3rd Dark" photo period in the 7.5 + 2-min illumination assay; Supplementary Fig. S12; Supplementary Table S67; Fig. 4Ac).
Inference for the proportion.The proportions of signi cance during a photo period were compared using the twoproportions test with Yates' continuity correction to see whether the durations of illumination led to difference in the proportions of signi cant difference in locomotion [46][47][48] (i.e., nr3c1 ex5 WT larvae moved signi cantly more than their HM siblings in 100% of the "5th Dark" photo period in the 7.5 + 2-min illumination assay, whereas the proportion of signi cant difference in the 7.5 + 6-min illumination assay was 90.1%.It was asked if these two proportions were signi cantly different; Supplementary Data S32; Supplementary Table S67; Supplementary Figs.S21 and S23).

Generalized additive model describes locomotor response during environmental changes
To better understand the dynamics of locomotor response over time, a generalized additive model (GAM) was developed (Fig. 1D; Model, Statistical Analysis in the Methods).A GAM for each assay consisted of the linear (parametric) and smooth terms that described the response variable, locomotion.The effect of each term on locomotion was evaluated for its main effect.With all the assays, the overall variation in locomotion explained by the model ranged from 68.3-97.5% (mean = 86.6%),demonstrating a satisfactory explanatory power (Supplementary Data S1-S23).The residuals, unexplained variations, and tted values were visualized to assess their appropriateness (Supplementary Figs.S63-S85).Post hoc pairwise analysis was performed to assess the signi cance of the difference in locomotion between the groups every second (Supplementary Tables S2-S66).The proportions of the signi cance between different assay regimens were compared to understand the effect of different durations of illumination (i.e., 4x[7.5 dark + 2-min light] regimen vs. 4x[7.5+ 7.5-min] regimen; Supplementary Figs.S8-S62; Supplementary Data.S24-S59; Supplementary Tables S67-S70).
To make the analysis process more rigorous, we used the mean of each experimental group in each assay, rather than the raw movement measurement of individual sh conventionally utilized in zebra sh behavioral studies (Response variable, Statistical Analysis in the Methods).The aggregation of the locomotor response mitigated the severe skewedness in the data.
WT larvae move more in light than in darkness during the day To understand the changes in locomotion following illumination changes, we rst established the pattern of basal locomotor activity during the day without any stimuli.In addition, we were curious whether zebra sh would move more in a lit or dark condition in the daytime.As a diurnal species, zebra sh have a circadian cycle that has evolved for high activity during the day and rest at night. 49It was not clearly established whether the larvae would move following the circadian rhythm regardless of the presence of illumination or the condition of ambient light (lit or dark) would dictate the levels of locomotion overriding the circadian cycle (called "masking").We recorded wildtype (WT) larvae (4-, 5-, 6-, and 7-days post-fertilization [dpf]) from 10 am to 10 pm in constant illumination or darkness.
We found that illumination was a critical determinant for the larvae's locomotion (Fig. 2Ab; Supplementary Data S1).
Light had a signi cant positive effect on locomotion compared to the null hypothesis of no effect (light, main effect, p = 1.31e-12;Supplementary Data S1) and led to a signi cantly different trend over time (light spline; F = 4.937, edf = 171.881,p < 2e-16).Darkness did not result in a different trend in locomotion (dark spline; F = 0.05, edf = 0.844, p = 0.951).Larvae in light moved signi cantly more than their age-matched counterpart in darkness across all developmental stages over 12 hours (Fig. 2Ac).That is, 4-, 5-, 6-, and 7-dpf larvae in light moved signi cantly more than 4-, 5-, 6-, and 7-dpf larvae in dark at all time points (100%), respectively (Supplementary Figs.S8-S9; all p-values provided in Supplementary Tables S3-S4 and all proportions provided in Supplementary Tables S67-S68).A dark environment decreased baseline activity, masking the basal circadian activity levels in light during the day in larvae aged between 4 and 7 days.
nr3c1 mutant larvae move less than WT siblings in light and darkness during the day Following the study of basal locomotion in WT sh, we investigated the basal locomotion of the nr3c1 ex5 sh (5 dpf) in which glucocorticoid receptor (nr3c1) is knocked out in homozygous (HM) larvae. 23Similar to the WT stock sh, illumination was a critical determinant of locomotor activity of nr3c1 ex5 sh.Compared to darkness, a lit condition increased locomotion (light; main effect, p = 3.27e-6; Supplementary Data S2; Fig. 3Ab).Genotype was another determinant.nr3c1 WT and heterozygous (HT) siblings showed comparable levels of locomotion in darkness or light (HT, main effect, p = 0.683).Homozygosity decreased overall locomotion (HM, main effect, p = 0.014).WT larvae moved signi cantly more than HM siblings in 34.84% of the time both in darkness and illumination and the signi cance was concentrated between 10 am and 6 pm (Supplementary Figs.S10-S11; Supplementary Tables S6-S7 and S67-S68).Homozygosity had a sizable impact on locomotion such that HM larvae in light moved more than WT siblings in darkness only in 1.39% for the 12-h period (Fig. 3Ac; Supplementary Fig. S11; Supplementary Table S68).
mc2r mutant larvae show less activity than WT siblings only when illumination duration is shorter After establishing that both darkness and blocking HPA axis activity decrease basal locomotion in larvae, we investigated the dynamic changes in locomotion in response to changing illumination.We previously showed that, to increase locomotor activity in post-light darkness after a brief 1-min illumination, a functioning HPA axis is essential as loss of mc2r and nr3c1 signi cantly decreased the locomotor response in these sh. 23Our initial hypothesis on the larval response in the dark-light repeat assay was that sh with a mutation in one of the key HPA axis receptor genes would have a de ciency in mounting an appropriate locomotor response to changing light.However, the duration of illumination is a determinant of the ensuing locomotor activity in darkness, as well as the HPA axis genes.
In another lineage of nr3c1 mutant sh where the frameshift mutation was introduced in exon 2, the effect of the duration of illumination was clearer.Homozygous mutant larvae in nr3c1 ex2 moved signi cantly less in darkness when repeated illumination was 4 min (HM, main effect, p = 0.0071), but not when illumination was 6 min (p = 0.197) and 7.5 min (p = 0.615; Supplementary Fig. 7; Supplementary Data S21-S23).The proportion of locomotion where the WT sh moved signi cantly more than their HM siblings in darkness was 94.65, 71.58, and 0% when the repeated illumination was 4, 6, and 7.5 min (Supplementary Figs.S57-S59; Supplementary Table S69).The difference in overall activity levels in darkness between the WT and HM siblings decreased in both nr3c1 ex5 and nr3c1 ex2 mutant sh (100, 87.78, 68.1, and 60.88%; ND, 94.65, 71.58, and 0% at 2, 4, 6, and 7.5 min), respectively, as the repeated illumination increased [50][51][52] , showing that the nr3c1 HM mutants could mount increasingly similar locomotor responses in darkness as the duration of illumination increased.

Mutations in nr3c2 do not appear to have consistent effects on locomotion
Mineralocorticoid receptor (nr3c2) is another nuclear receptor that binds to glucocorticoids with higher a nity.
However, we did not nd a clear role in locomotor response to light changes in our previous paper. 23Similarly, again we could not identify any pattern of difference in locomotion between the WT and HM siblings (Fig. 6).The HM mutant larvae in nr3c2 ex2 apparently moved less in 2-min repeated illumination at some time points (Fig. 6Ac), but there was no main effect of homozygosity compared to WT siblings (HM, main effect, p = 0.104; Supplementary Data S10).Likewise, in 4-and 7.5-min repeated illumination, the HM larvae apparently moved more at some time points, but there was no main effect of homozygosity compared to WT siblings (p = 0.0728 and p = 0.758), respectively (Supplementary Data S11-S12).Since there was no main effect of homozygosity on locomotion, the meaning of the apparent decrease or increase of locomotion amongst HM siblings at some time points is unlikely to be biologically meaningful.
Locomotor patterns over dark-light repeats are reproduced with much dimmer light All light experiments were conducted with consistent illumination intensities: IR (116.0 µW•cm − 2 ; 0 lx) and white light (469.4µW•cm − 2 ; 8000 lx).Since the duration of illumination (quantity of light by time) was a determinant of locomotor response in darkness, we asked whether lower intensities of illumination (quantity of light by intensity) would elicit the same behavioral adaptation and, if so, whether it would take longer.In the dim light assays, we used the same dark-light repeat regimen but used a much lower intensity of white light (20.5 µW•cm − 2 ; 300 lx).With dim light repeat assays, WT sh showed the same pattern of locomotor response as the assays with higher intensities of light (Fig. 7), increasing locomotor response in darkness while having lower locomotion in light.However, this increase did not reliably occur until the duration of illumination was 2 min or more (Fig. 8).With the nr3c1 ex5 lineage, the HM larvae showed equivalent levels of locomotion to WT sh in all 2-, 4-, and 7.5-min repeated dim illumination and there was no main effect of homozygosity on locomotion compared to WT siblings (HM, main effect, p = 0.224, p = 0.263, and p = 0.485), respectively, (Supplementary Data S18-S20).Since there was no main effect, further analysis was not pursued.The tile graphs (Figs. 8 Bc, Cc, Dc) are provided to show the patterns in difference.
Unexpectedly, dimmer illumination effectively facilitates the locomotor response in darkness after repeated illumination without a requirement for HPA axis signaling.

Discussion
We found that HPA axis activity is required to facilitate light adaptive behavior and maintain baseline activity.Larvae with mutations in a key HPA axis receptor could not increase their locomotor activity levels in darkness after a brief (1-2 min) illumination.The mutants required longer durations of illumination (≥ 6 min) to increase locomotor response in the ensuing darkness (Figs. 4, 5, and Supplementary Fig. S7).In addition, abrogation of HPA axis function (nr3c1 knockout) led to decreased basal activity under either constant illumination or darkness (Fig. 3).These ndings show that the canonical HPA axis contributes to the e ciency of adaptation to changing light conditions, but the photoadaptation can be slowly achieved in HPA axis mutants when light is provided for a long enough duration (≥ 6 min).Thus, the phasic locomotor response can be achieved even without HPA axis activity, while maintaining the tonic (basal) locomotion appropriate for the given illumination required the canonical HPA axis (Fig. 3).Our ndings contrast a previous study that found an increased circadian locomotor activity in nr3c1 mutants (gr s357 ) that carried a missense mutation. 53This discrepancy may be due to the difference in mutant types.While our nr3c1 mutants in exon 2 or 5 carried a frameshift mutation that was expected to yield a truncated protein, the DNAbinding function was abrogated in the gr s357 variant, possibly leaving the signaling capacity intact. 54e same pattern of phasic locomotor response was reproduced in much dimmer light only with higher e ciency in HPA axis mutant animals (Figs. 7 and 8).Both the dim (20.5 µW•cm − 2 ; ~300 lx) and high (469.4µW•cm − 2 ; ~8,000 lx) intensities of white light increased locomotion in darkness post-illumination.Yet, nr3c1 ex5 homozygous (HM) mutant larvae achieved WT-level activity after 2 min in the dim illumination, compared to 6-7.5 minutes they needed in brighter light.Thus, light adaptation was facilitated not only by the number of photons, but also by the appropriate intensity of light.In most animals, the illumination during twilight (dawn, dusk) is a salient zeitgeber (signals that entrain the circadian rhythm). 55,56The electroretinogram (ERG) wave patterns change in birds during this period, 57 and there may be distinct photopigments detecting twilight illumination. 58During the twilight period, animals detect changes in the amount and composition (a shift toward more blue spectrum) of illumination and the angle to the sun.Since the amount (intensity) of illumination is the only variable that changed in this experiment, it is unclear whether the increased e ciency in the light adaptive behavior is related to the twilight-time effect.Further investigation is required for the relationship between light adaptation and illumination intensity.nr3c1 WT and HM siblings showed a marginal increase in locomotion (< 4 mm/min) after 1-min illumination in dim light (Fig. 8Ab).That was likely to be noise since those low levels of locomotion (effect size) were within the range of basal locomotion, rather than a response to light changes.The likely reason why general WT stock sh showed a more robust response (Fig. 7A) than the WT siblings of the nr3c1 sh (Fig. 8A) after 1-min dim illumination may be familial difference in background genetics.[61][62][63] The characterization of the locomotor response in this study reveals complex behavior over time.In the earlier studies using ultrahigh-speed cameras, larvae showed a sharp increase in locomotion immediately after transitioning from dark to light or from light to dark in millisecond scales. 64These re exive movements were classi ed into different categories based on the nature of stimuli, the characteristics of the movement repertoire, and the underlying neural circuits responsible for the movements.6][67][68][69][70][71][72][73][74] Our study characterized the rapid non-genomic locomotor responses in the timescale of hormonal response [75][76][77] that followed the immediate re exive responses.The adaptive response to changing illumination arose with or without HPA axis activity, demonstrating that vital adaptations essential to the organism would occur by means of multiple redundant pathways.Nevertheless, the canonical HPA axis, as the backbone of the stress response (SR), played a critical facilitative role, making such locomotor adaptation more e cient.
Importantly, a functioning HPA axis was necessary to maintain basal locomotor activity, which implies that a hyperor hypo-preparedness of the baseline may precede maladaptive phasic responses.11]18 The circadian regulation is more complex in teleost sh as a range of tissues including the skin, brain, pineal gland, and heart are directly responsive to light and autonomously maintain the circadian clock. 78-82A broad spectrum of light (ultraviolet, visible, and infrared light) could directly induce clock gene expression in zebra sh cell culture while infrared light could not phase-shift the circadian clock. 78espite such complex interaction between the light and clock systems in zebra sh, the genetic abrogation of DNAbinding in gr s357 did not change clock-related gene expression levels while increasing basal locomotor activity. 53locking glucocorticoid (GC) signaling did not change the circadian uctuation of crh (corticotropin releasing hormone) in the neurosecretory preoptic area 29 (zebra sh homolog to the mammalian hypothalamus). 83GC signaling was instead necessary to maintain the level of overall basal crh production, contrasting the negative feedback of the phasic GC signaling against crh production. 29Thus, the role of GC/GR signaling and HPA axis dynamics in basal physiology needs further investigation and may add insight into maladaptive HPA axis function during the phasic responses.
Mutations in nr3c2 ex2 did not show any consistent direction of in uence in behavior.After 2-min illumination, WT sh appeared to move more than their HM siblings whereas, after 4-min illumination, HM siblings moved more (Fig. 6).After 7.5-min illumination, there was no difference in locomotion between the WT and HM siblings.We found no main effect of homozygosity on locomotion in any of these apparent increases and decreases of locomotion among nr3c2 sh.Thus, the differential effect of the HM genotype on locomotion is unlikely to have biological signi cance.It appears that nr3c2 does not play a critical role in the photo-adaptive behavior.
The locomotor response of zebra sh larvae to exogenous stimuli is not normally distributed (non-normal).A large proportion of sh do not move at all or move very little (many zero and negligible values) while a small minority show excessive movement (outliers).The variation in the response is not homogeneous among experimental groups (heteroscedasticity).5][86][87] By comparing the single values averaged over a stretch of time (i.e., total movement during the experimental period), we miss the dynamics of the response over time. 88,891][92] Non-normality and violation of homoscedasticity can be addressed, to some degree, by increasing the sample size, 87,93 using the sample means (sampling distribution rather than raw data from individual sh), [93][94][95] and including the variation in each experiment as random effects in mixed model approaches. 86The size of the samples that yield reliably accurate hypothesis tests in non-normal distribution is often less than 100.Even with extremely non-normal and heteroscedastic data sets, a less than 500 sample can perform statistical tests based on linear regression models. 93However, it is reported that whereas increasing sample size could address nonnormality issues more easily, other types of the assumption violation including homoscedasticity, extreme outliers, and independence of errors may persist despite the increased sample size. 87Moreover, outcome transformations change the result estimates and may bias the result. 8786]96 However, with nonparametric tests using rank comparison, we forgo quantitative estimates of the response although the inference may be more justi able. 97When a generalized linear model analyzes the proportion of the response vs. nonresponse, we do not show the effect size but just summarize the presence or absence of the response.While nonlinear models effectively characterize the properties of the response (i.e., the rate of increase), the rigidity of the underlying equation limits the applicability of a model to a few speci c phenomena.In our case, it was di cult to apply the same nonlinear model to the two distinct response pro les during the dark or light photo period.Thus, it is unlikely that a single solution will be able to address all the assumption violations in the zebra sh locomotor data and produce the right inference.][100][101][102] For example, what to report should be determined by the research interest, when a group of zebra sh consistently moved 10 mm per minute making a 100-mm displacement for 10 min and another group moved 90 mm for a minute and stayed without movement for 9 min.Whether comparing the total distance of 100 mm vs. 90 mm or the pattern of evenness vs. dynamic change is more relevant should be determined by the scienti c question in hand.The statistical assumptions violated or satis ed should be chosen by the same standard.
To describe zebra sh locomotor response to changing light, we chose to adopt a statistical model, not commonly used in zebra sh behavioral studies.A generalized additive model (GAM) provided a practical and expandable framework, 103,104 successfully describing locomotor response throughout the entire experimental process.A GAM is an extended generalized linear model (GLM) with a linear predictor that is composed of a sum of smooth functions of the explanatory variable (smooth terms) in addition to the parametric explanatory variables. 41,43,105A range of link functions can be used to linearly relate the predictors to the response variable of an exponential or some nonexponential family distribution.However, we focused on estimating the effect size and did not use additional link function (used the identity link function).The variance among the individual assays was modeled as a random effect.The simplicity and exibility of the GAMs allow describing nonlinear relationships in the linear additive framework.In our GAM analysis pipeline, the main effect of the explanatory variable for the whole experiment was rst evaluated, followed by post hoc pairwise analyses at each second that reported the signi cance of the difference of the means between the compared pair.Then, the proportion of signi cant difference between each pair was compared among different assay regimens to see if different durations of illumination led to distinct locomotor response compared to another.The analysis pipeline allowed quantitative estimation of the effect of the explanatory variable at each second without losing the dynamics of changes in locomotion and the effect of different assay regimen on locomotion.The body's homeostasis represents a point in the continuum of physiological and behavioral states in an organism.Allostasis describes how a homeostatic point is determined in response to the changing environment.The redundancy in the SR ensures an adequate adaptation even if a defective pathway is present.However, the compensatory pathways in achieving allostasis may be biologically expensive over time, eventually leading to maladaptation.In our investigation, the canonical HPA axis (backbone of the stress response) was dispensable to mounting light adaptive phasic responses when the illumination duration was long enough and the intensity was optimal.However, maintaining light adaptive tonic states at the baseline required HPA axis activity.Our ndings imply that both phasic and tonic responses of the HPA axis need to be investigated to understand the stress response.Stress-aggravated psychiatric and metabolic disorders may arise not only from the aftereffect of HPA axis activity (hyper-or hypo-cortisolemia) but also from the de cient primary pathways in relevant organ systems where cortisol and its cognate receptors fail to support them in generating optimal adaptive responses.Mutation in nr3c2 ex2 do have main effect on locomotion.Aa, Ba, Ca Locomotor activity (mean predicted value (mm/min) ± 95%CI) for each experimental condition predicted by the GAM for each photo period (gray: dark, white: light period).Ab, Bb, Cb Locomotor response of nr3c2 ex2 larvae during dark-light repeat assays.The scatterplot (points) shows actual mean locomotor activity (mm/min) for each experimental condition of each assay.The line graph shows predicted locomotor activity for each experimental condition by the GAM (predicted value ± 95%CI; gray: dark, white: light period).Ac, Bc, Cc Time points where an experimental condition showed signi cantly high locomotor activity compared to the other in a pairwise comparison.Despite such apparent signi cant difference at some time points, there was no main effect of being HM on locomotor response compared to WT siblings (Refer to the Results).The tile graphs are provided to show the patterns in difference.Ad, Bd, Cd Density distribution of actual mean locomotor activity shows severely right skewed distribution.The integration of the curve equals 100%.(D: dark, L: light, WT: wildtype, HT: heterozygous, HM: homozygous, n.s: not signi cant) Figure 7 Much dimmer illumination reproduces the same pattern dark-light responses in WT larvae.Dimmer illumination (20.5 µW•cm -2 ; 300 lx) was used compared to that of all other experiments (469.4 µW•cm -2 ; 8000 lx).IR illumination was the same (116.0µW•cm -2 ; 0 lx).Aa, Ba, Ca, Da Locomotor activity (mean predicted value (mm/min) ± 95%CI) predicted by the GAM for each photo period (gray: dark, white: light period).A brief illumination assay (1-min light) without the repeat components was included to understand behavior in dim light (A).Ab, Bb, Cb, Db Locomotor response of WT larvae during dark-light repeat assays.The scatterplot (points) shows actual mean locomotor activity (mm/min) of each assay.The line graph shows predicted locomotor activity by the GAM (predicted value ± 95%CI; gray: dark, white: light period).Ac, Bc, Cc, Dc Time points where an experimental condition showed signi cantly high locomotor activity compared to the other in a pairwise comparison.Ad, Bd, Cd, Dd Density distribution of actual mean locomotor activity shows severely right skewed distribution.The integration of the curve equals 100%.(D: dark, L: light, n.s: not signi cant) Figure 8