NeuroToxicology 52 (2016) 23–33

Contents lists available at ScienceDirect

NeuroToxicology

Persistent behavioral effects following early life exposure to retinoic

acid or valproic acid in zebrafish
Jordan M. Bailey a, Anthony N. Oliveri b, Nishika Karbhari a, Roy A.J. Brooks a,
Amberlene J. De La Rocha a, Sheila Janardhan a, Edward D. Levin a,b,*
a
Department of Psychiatry and Behavioral Sciences, Duke University School of Medicine, Durham, NC 27710, USA
b
Department of Pharmacology and Cancer Biology, Duke University School of Medicine, Durham, NC 27710, USA

A R T I C L E I N F O A B S T R A C T

Article history: Background: Moderate to severe dysregulation in retinoid signaling during early development is
Received 11 March 2015 associated with a constellation of physical malformations and/or neural tube defects, including spina
Received in revised form 30 September 2015 bifida. It is thought that more subtle dysregulation of this system, which might be achievable via dietary
Accepted 1 October 2015
(i.e. hypervitaminosis A) or pharmacological (i.e. valproic acid) exposure in humans, will manifest on
Available online 9 October 2015
behavioral domains including sociability, without overt physical abnormalities.
Methods: During early life, zebrafish were exposed to low doses of two chemicals that disrupt retinoid
Keywords:
signaling. From 0 to 5 dpf, larvae were reared in aqueous solutions containing retinoic acid (0, 0.02, 0.2 or
Zebrafish
Retinoic acid
2 nM) or valproic acid (0, 0.5, 5.0 or 50 mM). One cohort of zebrafish was assessed using a locomotor
Vitamin A valproic acid activity screen at 6-dpf; another was reared to adulthood and assessed using a neurobehavioral test
Behavior battery (startle habituation, novel tank exploration, shoaling, and predator escape/avoidance).
Cognition Results: There was no significant increase in the incidence of physical malformation among exposed fish
Social behavior compared to controls. Both retinoic acid and valproic acid exposures during development disrupted
Development larval activity with persisting behavioral alterations later in life, primarily manifesting as decreased
social affiliation.
Conclusions: Social behavior and some aspects of motor function were altered in exposed fish; the
importance of examining emotional or psychological consequences of early life exposure to retinoid
acting chemicals is discussed.
ß 2015 Elsevier Inc. All rights reserved.

1. Introduction
Retinoic acid is a small, lipophilic ligand for nuclear RA
receptors (RARs) and is derived from vitamin A. In embryonic
tissue, RA concentrations are tightly regulated via both synthesis
and degradation processes (Rhinn and Dollé, 2012). Functionally,
RA is involved in gene expression and thought to control the
differentiation, proliferation, and migration of numerous stem cell
populations (Rhinn and Dollé, 2012). As such, optimal mammalian
development requires functional retinoic acid (RA) signaling, and
chemicals that disrupt this process have teratogenic consequences.
For instance, when maternal vitamin A is increased, whether from
dietary sources (e.g. the liver of certain animals),various supple-
ments (which contain preformed vitamin A or cod liver oil), or
* Corresponding author at: Department of Psychiatry and Behavioral Sciences,
Box 104790, Duke University Medical Center, Durham, NC 27710, USA.
E-mail address: edlevin@duke.edu (E.D. Levin).
retinoic acid-acting drugs (e.g. valproic acid, 13-cis retinoic acid
(accutane) RA signaling in the embryo can become disrupted
(Lammer et al., 1985; Koch et al., 1983; Hathcock et al., 1990;
Coberly et al., 1996; Hayman and Dalziel, 2012; Duerbeck and
Dowling, 2012)).
When administered during gestation, substantially elevated
doses of vitamin A and related retinoic acid receptor acting
compounds have long been associated with severe CNS structural
malformations and organ damage. In particular, excess vitamin A
during early development results in gross neural malformations,
including neural tube defects, newborn lethality, growth reduc-
tion, and behavioral dysfunction in experimental animals (Shaw
et al., 1997; Steele et al., 1983; Shenefelt, 1972; for a review see
Adams, 1993 or Adams, 2010). Early work in this field character-
ized the persisting effects of disrupted retinoid signaling on
behavior and cognition, which can occur in the absence of overt
toxicity or physical malformation (see Vorhees et al., 1978, 1979;
for a review see Adams, 2010). In the late 1990s, the theory that
autism spectrum disorder may arise from gestational events at the

http://dx.doi.org/10.1016/j.neuro.2015.10.001
0161-813X/ß 2015 Elsevier Inc. All rights reserved.
24 J.M. Bailey et al. / NeuroToxicology 52 (2016) 23–33
time of neural tube closure (e.g. Rodier et al., 1996) reignited an
interest in chemicals that affect retinoid signaling. More recently,
valproic acid (VPA), a drug used as an anti-epileptic medication and
mood-stabilizer, but that has known teratogenic effects, has been
linked to the developmental effects of RA (Kawanishi et al., 2003;
Faiella et al., 2000). VPA acts as a histone deacetylase (HDAC)
inhibitor, and this inhibition has been posited to facilitate the
effects of RA and its receptor on the expression of a number of
genes (Li et al., 2014; Boudadi et al., 2013). A gene family notably
affected by this inhibitory mechanism is the Hox family of genes,
which is crucial to the development of the embryonic body plan
and rostro-caudal axis (Menegola et al., 2006).
While classically studied via mammalian models (primarily
rodents), retinoid signaling and its role in development is greatly
facilitated via genetic and molecular applications in the zebrafish
(For a review see Samarut et al., 2015) and numerous reports
describe the effects on development and specific developmental
processes following exposure to RA or VPA in zebrafish (e.g. Wang
et al., 2014; Aluru et al., 2013; Farooq et al., 2012). In fact, a more
general emphasis on intermediate models, such as the zebrafish, has
encouraged the use of this species in behavioral pharmacology and
toxicology studies. Zebrafish have highly conserved neural archi-
tecture that facilitates comparisons to other, more commonly used,
species. Accordingly, zebrafish engage in complex behavior while
accommodating high throughput data collection (Gerlai et al., 2000;
Kimura-Kuroda et al., 2012). This model is widely considered to offer
numerous logistical and economic advantages over mammalian
models (Levin and Cerutti, 2008) as zebrafish spawn overnight,
hatch in 2–3 days post fertilization (dpf), develop complex behavior
within the first week of life, and reach sexual maturity in 2–3
months. With the genetic tools that are already in place to examine
retinoid signaling and genetics in the zebrafish, complementary
behavioral/functional characterization of this pathway is needed in
order to appreciate the functional implications of molecular data.
Clearly, there are clinical implications for understanding how
chemicals might alter retinoid signaling (particularly in maternal
nutrition and medication use), and such investigations may also
offer insight into the role of RARs in downstream processes (e.g.
dopaminergic or serotonergic systems) that affect behavior.
Currently, the role of developmental retinoid dysfunction in
behavioral impairments, such as cognitive dysfunction and emo-
tional dysregulation, remains unclear. Disrupted RAR signaling may
be important in the etiology of a number of developmental
neurobehavioral disabilities, including autism spectrum disorder
and disruptive mood dysregulation disorder. Early life retinoid
dysfunction may in fact account for some manifestations of
behavioral dysfunction that are otherwise unexplained. Therefore,
it is the goal of the present study to characterize, in zebrafish, the
impact of early life disruptions in RA signaling on later-life behavior,
specifically the effects of exposure on cognitive domains that are
clinically significant. This work provides the functional/behavioral
phenotype for subsequent studies of the cellular and molecular
mechanisms by which disruptions of RA signaling during develop-
ment can produce long-term behavioral dysfunction in zebrafish.
2. Methods
2.1. Subjects
2.1.1. Breeding
Zebrafish from the AB strain were bred onsite from progenitors
originally obtained from the zebrafish international resource
center (ZIRC, Eugene, OR, USA). Zebrafish embryos were collected
at the beginning of the 14-h light cycle on the morning following
the pairing of adult breeders. Random pair-wise mating of
zebrafish breeders was conducted and all breeders were kept in
large (N = 15) spawning groups at a male to female ratio of 2:1.
Collected eggs were inspected under a dissecting microscope at 2-
hours post fertilization (hpf), and those unfertilized or showing
obvious malformations were excluded. Larvae were then randomly
distributed among all aqueous exposure conditions (described
below), housed in glass dishes, and placed inside incubators
maintained at a constant temperature of 28.5 8C and on a 14:10 h
light/dark cycle until 5-days post fertilization (dpf).
2.1.2. Housing and husbandry
Zebrafish > 6-dpf were housed in 3-L tanks on a circulating rack
system in a colony room arranged on a 14:10 h light/dark cycle.
Water temperature was maintained at approximately 28.5 8C and
salinity and pH were monitored bi-weekly. Aquarium water was
mixed using de-ionized H2O, sea salt (Instant Ocean, 9.0 g/5 gal
H2O), and neutral regulator (Seachem, 2.5 g/5 gal H2O). Young fish
were fed twice daily: brine shrimp (Brine Shrimp Direct, Ogden, UT,
USA) in the morning and solid food (at increasingly large particle size
as they grew) (Brine Shrimp Direct Golden Pearl; TetraMin1 Tropical
Flakes, Blacksburg, VA, USA) in the evening. On days when adult fish
were subjected to behavioral tests, the evening feeding was given
after the completion of testing. Adult behavioral testing spanned the
hours of 11:00 AM and 5:00 PM, with exact testing time counter-
balanced among exposure groups. All behavioral testing of larvae
was completed between the hours of 3:00 PM and 5:00 PM. All fish
care and experimental procedures were approved by the Duke
University Institutional Animal Care & Use Committee.
2.2. Chemical exposures
Beginning at 4-hpf, larvae were exposed to retinoic acid (0.02,
0.2 or 2-nM), valproic acid (0.5, 5.0 or 50-mM) or vehicle control in
glass Petri dishes. Exposure was conducted at a density of 60 eggs/
40-mL aqueous solution (Easton and Goulson, 2013), which was
done in duplicate twice (for a total of four dishes per exposure
group, see below), and with solutions prepared and changed daily
for 5 days (0–5 dpf). A follow-up study was conducted to provide a
finer-grained dose effect function with RA at doses of 0.02, 0.1, 0.2,
0.33, 1, 2 and 3.3 nM and VPA at doses of 15 and 30 mM to fill in
between the ineffective and effective doses as well as unexposed
vehicle controls. Before each solution change, eggs in each plate
were inspected under a dissecting microscope and eggs with
arrested development or obvious malformations were excluded.
Retinoic acid (Sigma–Aldrich, St. Louis, MO, USA) and valproic
acid (Sigma–Aldrich) solutions were made using aquaria water to
ensure proper salinity. A million-fold dilution of DMSO (Sigma–
Aldrich) was required for RA solution, which also served as vehicle
control for this group. VPA dissolves readily in H2O. A total of
4 dishes (replicates) per exposure group provided a cohort of
duplicate exposures for larval testing and, separately, adult testing.
The stress associated with post-experimental transfer of larvae out
of the 96-well plate, a procedure required for larval activity
assessment (described below) prohibits the later use of those same
fish on adult behavioral assays. Regardless of cohort assignment
(i.e. larval or adult testing), all zebrafish were transferred to fresh
aquarium water at 5-dpf, ending the exposure period. A total of 60
(7) zebrafish larvae per VPA exposure group and 70 (5) zebrafish
larvae per RA exposure group were used for the larval motility assay.
For the adult assays, 30 (2) zebrafish per VPA exposure group and 30
(5) zebrafish per RA exposure group were used.
2.3. Behavioral assessments
2.3.1. Larval motility assay
On 5 dpf all fish designated for larval testing were distributed
pseudo-randomly across 96-well plates containing .45 mL of
 J.M. Bailey et al. / NeuroToxicology 52 (2016) 23–33
un-dosed aquarium water per well. In this way, each exposure
group was represented within a 96-well plate. The 96-well plates
housed the fish for 24-h, until 6-dpf , at which time they were
subjected to behavioral testing. We utilized a common behavioral
test for assessing larval swimming activity (i.e. distance traveled)
and the capacity to adapt to changing environmental stimuli (i.e.
alternating periods of light and dark) (Ahmad et al., 2012).
DanioVisionTM (Noldus, Wageningen, The Netherlands) hardware
was used, which is a zebrafish larval tracking device that measures
movement during alternating periods of white light (‘‘100%
illumination’’, 5000 lux) and dark (‘‘0% illumination’’, <1 lux). An
initial acclimation dark period of 10 min was followed by 2 phases
of a 10 min light/10 min dark period, and larval motion was
tracked 30 times/s over the course of the 50 min trial. Video data
was then analyzed by computer tracking software, EthoVision
XT1 (Noldus, Wageningen, The Netherlands), to calculate total
distance traveled for each individual larvae during every min of
the trial.
2.3.2. Adult neurobehavioral test battery (NBTB)
Four behavioral assays were used to assess function in adult
zebrafish that had been exposed to RA or VPA during early
development. The test battery consisted of startle habituation,
novel environment exploration, social affiliation, and predator
escape/avoidance assays; all are described below. All fish
completed each assay, which were introduced in the same order
for each fish.
Startle habituation. Typical fish (and other animals) will exhibit
a diminished response to a repeated stimulus, ‘‘habituation’’.
Habituation curves, therefore, provide information about neuro-
plasticity and adaptation to an environmental stimulus. Here,
habituation learning and sensorimotor function were quantified by
exploiting a tap-elicited startle reflex (Eddins et al., 2009). These
methods have been described in detail elsewhere (see Eddins et al.,
2009), but briefly, eight fish were tested simultaneously in a
2  4 array of swim arenas (exposure group was counterbalanced
within each session of 8 fish). Below each arena was a centrally
located push solenoid that delivered a sudden physical tap to the
test environment when activated. A digital video camera was
centrally positioned above the arena display, 75-cm above the
water level. The video output from the camera was imported into a
computer running EthoVisionTM tracking software, which allowed
for each fish to be located 6 times/s. Motor startle responses were
assessed for 10 tap-trials with 1 min intervals between taps to
determine the initial startle response and subsequent habituation
with repeated stimulus presentation. The dependent measures
were distance traveled (cm) for the 5-s both before and after each
stimulus delivery.
Novel environment exploration. Anxiety-like behavior and
exploration of an unfamiliar environment were assessed using a
novel tank arrangement, which has also been previously described
(see Bencan and Levin, 2008 or Levin and Cerutti, 2008). When
placed in an unfamiliar tank, zebrafish, like many other prey fish,
avoid the more exposed upper regions of the tank and instead dive
to the bottom. As zebrafish acclimate to a tank devoid of aversive
stimuli, they gradually rise and begin to explore the tank. Across
each minute of a 5-min trial, the distance from the tank floor (a
measure of spatial location) and the total distance traveled were
quantified to assess tank exploration.
Social affiliation. To assess social behavior, a shoal location
reversal task following a prototype developed by Gerlai and
colleagues (Saverino and Gerlai, 2008; Fernandes et al., 2015) was
used to quantify shoaling behavior. Typically, zebrafish have
strong shoaling tendencies and will rapidly approach a group of
moving conspecifics (i.e. a ‘‘shoal’’) and engage in a characteristic
pattern of swimming (i.e. ‘‘shoaling’’). The testing apparatus here
25
consisted of a 46-cm clear tank positioned beneath a video camera
and between two monitors. Each monitor was programmed to
show 1 min video clips of a group of 10–15 live shoaling zebrafish,
filmed earlier in the lab.
Individual fish were isolated in tanks surrounded by opaque
dividers; after 1-h of isolation, they were transferred to the testing
tank. Each trial began with a 1-min control period, after which time
one monitor displayed a 1-min video of shoaling fish. Subsequent-
ly, the shoal location was reversed: the opposite monitor began to
display the shoal video while the first monitor transitioned to a
stationary control image, matched for color and illumination to the
video image. This reversal process took place a total of 3 times in
the 5-min trial, for a total of two left presentations and two right
presentations (starting location was counterbalanced between
subjects). EthoVisionTM software was used to calculate tank
location (i.e. distance from the active shoal) and total distance
traveled.
Predator escape and avoidance. To analyze fear-like behavior and
escape in zebrafish, the image of an expanding blue dot was used to
simulate the approach of a predator. The dot, which increased in
diameter from 1.3-cm to 30.5-cm in 5-s, was displayed on a
computer monitor located on one side of a rectangular 1.5-L tank. A
single fish occupied the test tank and following a 1-min tank
acclimation, the video alternated between a 1-min ‘‘predator on’’
condition (in which the stimulus was presented 12 times) and a
1-min ‘‘predator off’’ condition (in which the screen was blank).
Each 5-min trial was comprised of four alternations between
conditions (‘‘on’’–‘‘off’’–‘‘on’’–‘‘off’’) and an initial 1-min acclima-
tion/baseline period (see Luca and Gerlai, 2012). Typically,
zebrafish will rapidly flee the presentation of the stimulus and
restrict swimming to the tank end farthest from the ‘‘predator’’
stimulus; however, when the stimulus is off, control fish will
typically explore the tank space and occasionally approach the
tank wall associated with the predator stimulus. Utilizing the on/
off paradigm allows for the examination of both predator escape
(when the stimulus is on) and predator avoidance (when the
stimulus is off). Total distance traveled and tank location measures
(i.e. distance from the ‘‘predator’’ wall) were used to quantify
behavior under this arrangement.
2.4. Statistical analysis
Dependent measures corresponding to each assay are described
above. When appropriate, log transformations were performed on
raw data if variability increased proportionally with magnitude or
if distributions were positively or negatively skewed. All statistical
analyses were performed using Superanova/Statview (SAS; Cary,
NC, USA) and all graphs were made using SigmaPlot (SYSTAT
Software Inc., Richmond, CA). The Type 1 error rate (a) was set at
0.05. A repeated-measures analysis of variance (RMANOVA) was
performed for each dependent variable of interest. Dose served as
between-subject factors; tap number, time block within session, or
stimulus condition served as within subject factors. When
appropriate, e.g. where the adjustment to the degrees of freedom
was less than 0.8 Greenhouse–Geiser adjustments to degrees of
freedom were used to account for lack of sphericity in the dataset.
Dunnett’s comparisons (two-tailed) were used to test for
differences between the dose groups and control, and are reported
below.
3. Results
3.1. Survival/malformation
A number of eggs are typically removed or excluded from a
spawning batch during the first 72-hpf as some eggs are not
26 J.M. Bailey et al. / NeuroToxicology 52 (2016) 23–33
fertilized (and misidentified as fertilized) or development is
arrested very early following fertilization. Here, survival among
control larvae was 60% from harvested eggs to adulthood, which is
typical survival for the breeding pairs used here and comparable to
rates achieved by our own group and elsewhere (e.g. Bourrachot
et al., 2014). There was no loss due to death or malformation after
10-dpf (100% of 10-dpf fish survived to adulthood), which is also
consistent with previous studies. Moreover, animals from the
exposure conditions did not exhibit malformations or death at a
higher rate than controls (p’s > 0.05); as noted above, daily
(0–6 dpf) monitoring of embryos and subsequent tracking of
survival through adulthood were undertaken. Survival plots for RA
(Fig. 1A) and VPA (Fig. 1B) exposures represent the proportion of
fish remaining after dead/unfertilized and malformed embryos
were excluded for each of the first 6-dpf (while larvae were being
examined carefully under a dissecting microscope).
3.2. Larval activity
RA exposure at the 0.2 nM dose caused a significant (p < 0.05)
increase in swimming activity relative to control (Fig. 2A). VPA had
a differential effect (p < 0.0005) during the light and dark phases.
None of the doses caused significant changes relative to controls
during the light phase, but during the dark phase the 50 mM VPA
dose caused a significant (p < 0.01) reduction in activity relative to
control (Fig. 2B). Importantly, these fish were not immobile; rather,
they swam at a near constant rate over the course of the 50 min
trial regardless of lighting condition, in contrast with all other
groups that swam substantially less during the light phases than
during the dark phases.
The follow-up study with added intermediate doses showed a
significant (p < 0.0005) main effect of treatment on activity
(Fig. 2C). Dunnett’s tests comparisons of treatment groups to
controls showed that this study replicated the effect that the
0.2 nM RA exposure caused significant (p < 0.05) overall hyper-
activity (+17.4% relative to control). The higher and lower doses in
this finer-grained dose–response analysis did not produce signifi-
cant hyperactivity. In the follow-up study, the 30 mM VPA dose
caused significant (p < 0.01) overall hypoactivity ( 47.7% relative
to control). Since there was a significant (p < 0.0005) interaction of
exposure  light–dark condition in the follow-up study, analyses
of the simple main exposure effects in each lighting condition were
conducted. This showed that the 15 mM VPA exposure caused
significant (p < 0.05) hyperactivity (+37.3% relative to control)
during the light condition, while the 30 mM VPA exposure caused a
[(Fig._1)TD$IG]
A B
Proportion Remaining
0
0.02
0.2
2.0
RA (nM)
Fig. 1. The proportion of zebrafish larvae remaining after unfertilized eggs, embryos with arrested development and dead or malformed zebrafish were removed from the RA
(0–5 dpf) treated group (Panel A) and from the VPA (0–5 dpf) treated group (Panel B).
significant (p < 0.01) hypoactivity ( 62.0% relative to control)
during the dark condition.
3.3. Startle and habituation
Neither RA nor VPA exposure resulted in a significantly
different startle response to the tap stimulus. There was significant
(p < 0.0005) habituation over the course of the session in both drug
exposure groups (Fig. 3A and B).
3.4. Novel tank exploration
None of the RA doses significantly affected the swimming speed
in the novel tank nor did any RA dose significantly change the
bottom dwelling. Main effects of time (i.e. trial minute) were seen
on both swimming speed and swimming location measures
(p < 0.0005) with the swimming speed and the distance from
the bottom increasing as the session progressed, indicating that all
fish generally displayed the typical dive response followed by
increased exploration (Fig. 4A and B).
None of the VPA doses caused significant differences from
control with either swimming speed or distance from the bottom
measures. There were significant main effects of time with both
swimming speed (p < 0.0005) and distance from the bottom
(p < 0.01) with increases in both measures over the course of the
session (Fig. 4C and D). There was a significant dose by time
interaction on swimming speed (p < 0.05).
3.5. Social affiliation/shoaling behavior
RA exposure at the 2 nM dose significantly (p < 0.05) decreased
approach to the video of shoaling conspecifics compared to control
fish, indicating diminished social behavior (Fig. 5A–B). The 5 mM
VPA dose also caused a significant (p < 0.05) decrease in approach to
the shoal (Fig. 5C and D). In both studies there were significant main
effects (p < 0.05) of time within session with the fish showing
greater approach in the early part of the session than later. There
were no significant RA or VPA effects detected on swimming speed.
3.6. Predator escape and avoidance
Neither RA nor VPA exposure resulted in differential fleeing
from the predator stimulus as measured by distance from the
predator stimulus. As expected, a main effect of stimulus condition
(‘‘predator stimulus on’’, ‘‘predator stimulus off’’ and habituation)
1.0
0.8
0.6
0
0.5
5
0.4
50
0 dpf 1 dpf 2 dpf 3 dpf 4 dpf 5 dpf 6 dpf
VPA (uM)
[(Fig._2)TD$IG] J.M. Bailey et al. / NeuroToxicology 52 (2016) 23–33 27

Fig. 2. Larval activity data are plotted as a function of time, across 10-min condition blocks, for each group of RA treated fish (Panel A) and VPA treated fish (Panel B). Panel C
displays the mean activity during the light and dark portions of the test in the follow-up study with additional doses of RA and VPA. The lines below the asterisks across the
light and dark bars indicate significant comparisons of average activity across lighting conditions of the exposed groups to control to convey the Dunnett’s comparisons within
the significant main effect of exposure. Asterisks directly over individual light and dark bars without the lines indicate significant comparisons of that particular lighting
condition back to control performance for that lighting condition to convey the Dunnett’s comparisons within the significant interaction of exposure  lighting condition.
Error bars represent SEM.
[(Fig._3)TD$IG]
28 J.M. Bailey et al. / NeuroToxicology 52 (2016) 23–33

Sensorimotor Habituation
50
A

Following Stimulus Delievery (cm)

40

Distance Traveled 30

20

Tap 1-2
Tap 3-4
10
Tap 5-6
Tap 7-8
Tap 9-10

0
0 0.02 0.2 2
Renoic Acid (nM)
50
B
Following Stimulus Delivery (cm)

40
Distance Traveled

30

20

Tap 1-2
Tap 3-4
10
Tap 5-6
Tap 7-8
Tap 9-10

0
0 0.5 5.0 50.0

VPA (uM)

Fig. 3. Distance traveled following the tap stimulus presentation is plotted for each RA (Panel A) and VPA (Panel B) dose condition (grouped bars). Each bar represents the
mean of two tap presentations. Error bars represent SEM.

appeared for fish treated with both RA and VPA (p < 0.001) (Fig. 6A
and B).
4. Discussion
Here we administered a range of doses of two chemicals that
affect retinoid signaling, either directly or indirectly, to developing
zebrafish. The doses selected were anchored by the existing
literature in zebrafish (Wang et al., 2014; Aluru et al., 2013; Farooq
et al., 2012) to be within a sub-threshold range for producing
increased lethality and overt dysmorphogenesis. For instance,
concentration-dependent increases in malformations from 4 to
8 nM retinoic acid (developmental exposure) (doses below 4 nM
did not induce malformation) were reported in zebrafish by Wang
et al. (2014) and Farooq et al. (2012) demonstrated VPA-induced
cardiac edema (but not skeletal malformation) at 30 mM in
zebrafish. The lower doses used in the present study are of particular
interest to us as most are below those that elicit structural
malformation and might prove relevant for understanding some
subtle or unexplained behavioral dysfunction in humans following
low-dose developmental exposures. We employed a design that
captures not only short-term or acute behavioral effects but that also
permits the evaluation of long-term effects on behavioral function.
These were important aspects of the design as our goal was to
determine if doses below those which cause morphological changes
(morphological abnormality is often associated with retinoid
disruption at high doses) might produce lasting behavioral effects
in zebrafish. We were interested in capturing aberrations in both
simple and complex function in organisms that were otherwise
morphologically normal. We anticipated that the exposures used
here might be of particular clinical relevance as maternal sources of
Vitamin A or VPA exposure could affect humans at very low doses,
which we speculate might explain later life behavioral consequences
relevant to early developmental disabilities.
The data reported here show that early life retinoid disruption
(directly via RA or indirectly via VPA) has consequences on several
behavioral domains including social affiliation, novel tank
exploration and swimming activity during distinct conditions of
[(Fig._4)TD$IG] J.M. Bailey et al. / NeuroToxicology 52 (2016) 23–33 29

Fig. 4. Mean distance traveled for each min of the trial is plotted by RA dose (Panel A) and VPA dose (Panel C) condition (grouped bars). Mean distance from the tank floor (a
measure of tank location) is plotted by RA dose (Panel B) and VPA dose (Panel D) condition (grouped bars). Error bars represent SEM.

illumination (larval data). The effects here are reported at doses
that did not cause characteristic, overt physical malformations
associated with higher dose retinoid disruption. While both RA and
VPA were disruptive to larval swimming, their effects later in life
manifested primarily as alterations in shoaling behavior (or, social
affiliation). Excess VPA exposure was associated with hyperactivity
during the novel tank exploration assay, whereas Dunnett’s
comparisons approached but failed to reach statistical significance
for RA exposures on this assay, which trended toward hypoactivity.
Both compounds, however, diminished social affiliation on the
shoaling assay. Moreover, the disruption of RA and VPA on social
responding was pronounced and not likely due to motoric or visual
impairment because all zebrafish fled the predator image similarly
on the escape/avoidance assay. This effect suggests that these
exposures might be affecting sensitivity to reinforcement, such
that RA and VPA might diminish the reinforcing efficacy of
appetitive stimuli in the environment. From this vantage point,
associating with conspecifics is less reinforcing or rewarding for
the RA/VPA-exposed organism. This interpretation might explain
the absence of effects on the predator escape assay in particular,
which takes advantage of aversive stimulus properties and to some
extent even the startle habituation and novel tank exploration
procedures, both of which are devoid of appetitive or rewarding
stimuli. This interpretation is consistent with the conceptual
framework that the reinforcing properties and the aversive
properties of stimuli are thought to rely on divergent neural
[(Fig._5)TD$IG]
30 J.M. Bailey et al. / NeuroToxicology 52 (2016) 23–33

Shoaling (First Minute Inset)

20 12

10

Distance from Shoal (cm)

15
4

2
*
0
B
0 6e-3 0.06 0.6

10

5 Min 1
Min 2
Min 3
Min 4
A Min 5
0
0 0.02 0.2 2

RA (nM)

20 12

10

6
Distance from Shoal (cm)

15 4

2
*
0 D
0 0.5 5.0 50.0

10

5 Min 1
Min 2
Min 3
Min 4
C Min 5
0
0 0.5 5.0 50.0

VPA (uM)
Fig. 5. Mean distance from the shoal (cm) for each min of the trial is plotted by RA dose (Panel A) and VPA dose (Panel C) condition (grouped bars), inset graphs plot the mean
distance from the shoal (cm) for first 1 min of the session, which highlights the initial difference among groups for RA (Panel B) and VPA (Panel D) conditions. Error bars
represent SEM. An ‘‘*’’ denotes p < 0.05, difference from control (Dunnett’s post hoc).

pathways (Samad et al., 1997). One alternate explanation for Moreover, while unlikely, the very small (million-fold dilution)
the absence of effects emerging under aversive control might be quantity of DMSO that is required to facilitate RA might have
that performance on the predator escape assay, for instance, had manifested as behavioral alterations. Regardless of differences
reached some ceiling, or maximal possible response, thereby between controls from VPA and RA, each treatment group was only
preventing the emergence of subtle deficits. compared to its own control group.
Regarding the performance differences between the VPA and Additionally, the larval disruption reported here, as captured by
RA control animals, we have hypothesized two factors that might Daniovision, predicted adult impairment. VPA was associated with
explain some of this variability – although some natural variability more motoric alterations on the adult tests then RA and VPA
is consistently detected among zebrafish across behavioral assays. caused aberrant motoric behavior across illumination conditions
Here, specifically, a portion of this variability might be explained on the larval assay, which primarily captures motor function (as
by temporal nature of these experiments as the VPA cohort well as visual capacity), albeit the effect on larval activity was most
preceded the RA cohort by approximately two weeks and pronounced at the highest dose of VPA. RA exposure had more
corresponded to a different breeding night than the RA cohort. subtle motor effects during larvae screening and fewer motoric
[(Fig._6)TD$IG] J.M. Bailey et al. / NeuroToxicology 52 (2016) 23–33
Fig. 6. Mean distance to predator stimulus is plotted by RA dose (Panel A) and VPA dose (Panel B) condition (grouped bars). Error bars represent SEM.
changes during adulthood. In sum, the larval activity assay
captured an inverted U-shaped dose effect curve with RA excess
with moderate but not higher or lower doses causing significant
larval hyperactivity, an effect that was replicated in a follow-up
study. VPA caused a narrowing in the changes in activity in
response to the light–dark changes in the larval motility test. Of
course, it is also worth noting that the larval screen is not a clear
analog to any one of the adult tasks in particular, but rather is
meant to serve as a general predictor of function. The fully
expressed character of the behavioral dysfunction necessitates
assessment of the complex behavioral function in adults. In
previous studies we have found that this long-term assessment of
complex behavior can be more sensitive in detecting behavioral
impairment than the short term simple behavioral change in larvae
(Eddins et al., 2010; Levin et al., 2014; Yen et al., 2011).
Although the teratogenicities of RA and VPA have been posited
to converge on cellular signals important to embryonic develop-
ment such as the Hox genes (Kawanishi et al., 2003; Faiella et al.,
2000; Menegola et al., 2006), the doses here were not associated
with morphological abnormalities. This suggests that Hox genes
may not have been affected in this study; therefore, the behavioral
effects reported here might not have resulted via this convergent
mechanism of action. It may thus be unsurprising that VPA and RA
produced somewhat differential effects on behavior – after all, the
alterations to gene expression produced by VPA (via HDAC
inhibition) and RA (via its nuclear receptor) are numerous, and
it is possible that genes other than those on which the two
compounds converge are producing the neurobehavioral terato-
genicity reported in this study.
In many ways, the data presented here address the apparent
gap in the literature (e.g. Adams, 2010) pertaining to the social and
emotional consequences of retinoid disruption. As early as the
1950s, animal studies have associated specific malformations (e.g.
spina bifida, hydrocephalus) with hypervitaminosis A (e.g. Cohlan,
1953, 1954) and have noted that the developmental window of
exposure determines gross malformation (Kalter and Warkany,
1961). Subsequently, lower dose studies demonstrated that excess
vitamin A can cause behavioral impairment to emerge even in the
absence of overt structural malformation (Vorhees, 1974; Vorhees
et al., 1978, 1979). Subtle indices of malfunction manifested in rats
given relatively low doses of vitamin A early in development. These
effects include abnormalities in operant behavior (Hutchings et al.,
1973; Hutchings and Gaston, 1974), maze performance (Butcher
et al., 1972; Vorhees et al., 1979), and Pavlovian conditioning
(Vorhees et al., 1979). Moreover, vitamin A excess in rat models
31
between days 15 and 16 of gestation showed persistent learning
impairments (Hutchings et al., 1973) and mice exposed to RA in the
neonatal period showed behavioral abnormalities explained by
adverse consequences in the frontal cortex and hippocampus (Luo
et al., 2004). Even transient disruption of neurodevelopmental
processes by vitamin A excess during mid-gestation can have
lasting functional effects (Vacca and Hutchings, 1977).
It is relatively unknown, however, how other domains of
behavior, particularly emotional and psychological endpoints, are
affected by early life disruption to retinoid signaling. For instance,
behavioral data from brainstem nuclei projection patterns suggest
that chemicals like RA could contribute to the etiology of autism
spectrum disorder (ASD) (Rodier et al., 1996). It is also believed
that valproic acid given during mid-gestation can affect dopami-
nergic and serotonergic systems (Chen et al., 2007), and lower
doses of valproic acid have been linked to increased rates of autism
(Ornoy, 2009). Some connections have been hypothesized to exist
between excess vitamin A and suicidal behavior, for example, in
patients using isotretinoin (e.g. Marqueling and Zane, 2007).
Moreover, it has been demonstrated that regions of the brain that
are high in retinoic acid receptors are also areas known to be
involved with stress and depression (Bremmer and McCaffery,
2008) and that retinoic acid is required for proper regulation of
dopaminergic signaling (Bremmer and McCaffery, 2008; Tafti and
Ghyselinck, 2007). In many ways, the data presented here
compliment this existing literature, and highlight some of the
understudied effects (e.g. the reinforcing efficacy of social stimuli)
of retinoid disruption.
Finally, it is worth emphasized the utility of the zebrafish model
for behavioral studies. Zebrafish can be quite useful as a
complementary model system, which is an intermediate between
cell based in vitro models and classic rodent models; this is
particularly evident within neurobehavioral toxicology. Because
behavior is the primary means by which an organism interacts
with its environment (Weiss and Cory-Slechta, 1994), it is
necessary that behavior remains a subject matter of interest as
only it reflects the summed and integrated capacity of an organism
to handle the environment within which it exists (Weiss, 1978). As
demonstrated here, zebrafish can provide relatively complex
behavioral data. This comes at a lower cost compared to murine
species and the species offers high-throughput mechanistic
(cellular and molecular) studies to complement behavioral data.
Within the current study, the zebrafish model was useful in
demonstrating the persisting neurobehavioral toxicity caused by
disruptions of RAR signaling during early development.
32 J.M. Bailey et al. / NeuroToxicology 52 (2016) 23–33
Conflict of interest
The authors declare that there are no conflicts of interest.
Transparency document
The Transparency document associated with this article can be
found in the online version.
Acknowledgements
Research was supported from the Duke University Superfund
Research Center ES 010356. JMB was supported by the Duke-RJR
Leon Golberg Toxicology Fellowship.
References
Adams, J., 1993. Structure–activity and dose–response relationships in the
neural and behavioral teratogenesis of retinoids. Neurotoxicol. Teratol. 15,
193–202.
Adams, J., 2010. The neurobehavioral teratology of retinoids: a 50-year history.
Birth Defects Res. 88, 895–905.
Ahmad, F., Noldus, L., Tegelenbosch, R., Richardson, M., 2012. Zebrafish embryos
and larvae in behavioural assays. Behaviour 149, 1241–1281.
Aluru, N., Deak, K.L., Jenny, M.J., Hahn, M.E., 2013. Developmental exposure to
valproic acid alters the expression of microRNAs involved in
neurodevelopment in zebrafish. Neurotoxicol. Teratol. 40, 46–58.
Bencan, Z., Levin, E.D., 2008. The role of a7 and a7 2 nicotinic receptors in the
nicotine-induced anxiolytic effect in zebrafish. Physiol. Behav. 95, 408–412.
Boudadi, E., Stower, H., Halsall, J.A., Rutledge, C.E., Leeb, M., Wutz, A., O’Neill,
L.P., Nightingale, K.P., Turner, B.M., 2013. The histone deacetylase inhibitor
sodium valproate causes limited transcriptional change in mouse embryonic
stem cells but selectively overrides Polycomb-mediated Hoxb silencing.
Epigenetics Chromatin 6, 1–16.
Bourrachot, S., Brion, F., Pereira, S., Floriani, M., Camlleri, V., Cavalié, I., Palluel,
O., Guilermin, A., 2014. Effects of depleted uranium on the reproductive
success and F1 generation survival of zebrafish (Danio rerio). Aquat. Toxicol.
154, 1–11.
Bremmer, J.D., McCaffery, P., 2008. The neurobiology of retinoic acid in affective
disorders. Prog. Neuro-Psychoph. 32, 315–331.
Butcher, R.E., Brunner, R.L., Roth, T., Kimmel, C.A., 1972. A learning impairment
associated with maternal hypervitaminosis-A in rats. Life Sci. I 11, 141–145.
Chen, P.S., Wang, C.C., Bortner, C.D., Peng, G.S., Wu, X., Pang, H., et al., 2007.
Valproic acid and other histone deacetylase inhibitors induce microglial
apoptosis and attenuate lipopolysaccharide-induced dopaminergic
neurotoxicity. Neuroscience 149, 203–212.
Coberly, S., Lammer, E., Alashari, M., 1996. Retinoic acid embryopathy: case
report and review of literature. Pediatr. Pathol. Lab Med. 6 (5), 823–836.
Cohlan, S.Q., 1953. Excessive intake of vitamin A as a cause of congenital
anomalies in the rat. Science 117, 535–536.
Cohlan, S.Q., 1954. Congenital anomalies in the rat produced by excessive intake
of vitamin A during pregnancy. Pediatrics 13, 556–567.
Duerbeck, N.B., Dowling, D.D., 2012. Vitamin A: too much of a good thing?
Obstet. Gynecol. Surv. 67 (2), 122–128.
Easton, A.H., Goulson, D., 2013. The neonicotinoid insecticide imidacloprid
repels pollinating flies and beetles at field-realistic concentrations. PLOS
ONE 8, 1–4.
Eddins, D., Cerutti, D., Williams, P., Linney, E., Levin, E., 2010. Zebrafish
provide a sensitive model of persisting neurobehavioral effects of
developmental chlorpyrifos exposure: Comparison with nicotine and
pilocarpine effects and relationship to dopamine deficits.
Neurotoxicol. Teratol. 32, 99–108.
Eddins, D., Petro, A., Williams, P., Cerutti, D.T., Levin, E.D., 2009. Nicotine effects
on learning in zebrafish: the role of dopaminergic systems.
Psychopharmacology (Berl) 202 (1–3), 103–109.
Faiella, A., Wernig, M., Consalez, G.G., Hostick, U., Hofmann, C., Hustert, E.,
Boncinelli, E., Balling, R., Nadeau, J.H., 2000. A mouse model for valproate
teratogenicity: parental effects, homeotic transformations, and altered HOX
expression. Hum. Mol. Genet. 9 (2), 227–236.
Farooq, M., El-Faham, A., Ahmad, M., Sami, A.A., Wadaan, M.A., 2012.
Teratogenic profile of valproic acid and newly synthesized derivatives in
zebrafish embryos. Int. J. Agric. Biol. 14 (6), 894–900.
Fernandes, Y., Rampersad, M., Jia, J., Gerlai, R., 2015. The effect of the
number and size of animated conspecific images on shoaling responses of
zebrafish. Pharmacol. Biochem. Behav., http://dx.doi.org/10.1016/
j.pbb.2015.01.011.
Gerlai, R., Lahav, M., Guo, S., Rosenthal, A., 2000. Drinks like a fish: zebrafish
(Danio rerio) as a behavior genetic model to study alcohol effects.
Pharmacol. Biochem. Behav. 67, 773–782.
Hathcock, J.N., Hattan, D.G., Jenkins, M.Y., McDonald, J.T., Sundaresan, P.R.,
Wilkening, V.L., 1990. Evaluation of vitamin A toxicity. Am. J. Clin. Nutr. 52,
183–202.
Hayman, R.M., Dalziel, S.R., 2012. Acute vitamin A toxicity: a report of three
paediatric cases. J. Paediatr. Child Health 48 (3), E98–E100.
Hutchings, D.E., Gaston, J., 1974. The effects of vitamin A excess administrated
during the mid-fetal period on learning and development in rat offspring.
Dev. Psychobiol. 7, 225–233.
Hutchings, D.E., Gibbon, J., Kaufman, M.A., 1973. Maternal vitamin A excess
during the early fetal period: Effects on learning and development in the
offspring. Dev. Psychobiol. 6 (5), 445–457.
Kalter, H., Warkany, J., 1961. Experimental production of congenital
malformations in strains of inbred mice by maternal treatment with
hypervitaminosis A. Am. J. Pathol. 38, 1–21.
Kawanishi, C.Y., Hartig, P., Bobseine, K.L., Schmid, J., Cardon, M., Massenburg, G.,
Chernoff, N., 2003. Axial skeletal and Hox expression domain alterations
induced by retinoic acid, valproic acid, and bromoxynil during murine
development. J. Biochem. Mol. Toxicol. 17 (6), 346–356.
Kimura-Kuroda, J., Komuta, Y., Kuroda, Y., Hayashi, M., Kawano, H., 2012.
Nicotine-like effects of the neonicotinoid insecticides acetamiprid
and imidacloprid on cerebellar neurons from neonatal rats. PLoS ONE 7,
1–11.
Koch, S., Göpfert-Geyer, I., Jäger-Roman, E., et al., 1983. Anti-epileptic agents
during pregnancy. A prospective study on the course of pregnancy,
malformations and child development. Dtsch. Med. Wochenschr. 108 (7),
250–257.
Lammer, E.J., Chen, D.T., Hoar, R.M., et al., 1985. Retinoic acid embryopathy.
N. Engl. J. Med. 313, 837–841.
Levin, E., Cerutti, D., 2008. Behavioral neuroscience of zebrafish. In: Buccafusco,
J. (Ed.), Methods of Behavioral Analysis in Neuroscience. CRC Press,
Augusta.
Levin, E., Cauly, M., Johnson, J., Cooper, E.M., Stapleton, H., Ferguson, P.L., Seidler,
F.J., Slotkin, T.A., 2014. Prenatal dexamethasone augments the
neurobehavioral teratology of chlorpyrifos: significance for maternal stress
and preterm labor. Neurotoxicol. Teratol. 41, 35–42.
Li, Q., Foote, M., Chen, J., 2014. Effects of histone deacetylase inhibitor valproic
acid on skeletal myocyte development. Sci. Rep. 4 (7207), 1–4.
Luca, R.M., Gerlai, R., 2012. In search of optimal fear inducing stimuli:
Differential behavioral responses to computer animated images in zebrafish.
Behav. Brain Res. 226 (1), 66–76.
Luo, T., Wagner, E., Crandall, J.E., Drager, U.C., 2004. A retinoic-acid
critical period in the early postnatal mouse brain. Biol. Psychiatry. 56,
971–980.
Marqueling, A.L., Zane, L.T., 2007. Depression and suicidal behavior in acne
patients treated with isotretinoin: a systematic review. Semin. Cutan. Med.
Surg. 26, 210–220.
Menegola, E., Di Renzo, F., Broccia, M.L., Giavini, E., 2006. Inhibition of histone
deacetylase as a new mechanism of teratogenesis. Birth Defects Res. C
Embryo Today 78 (4), 345–353.
Ornoy, A., 2009. Valproic acid in pregnancy: how much are we endangering the
embryo and fetus? Reprod. Toxicol. 28, 1–10.
Rhinn, M., Dollé, P., 2012. Retinoic acid signalling during development.
Development 139, 843–858.
Rodier, P.M., Ingram, J.L., Tisdale, B., Nelson, S., Romano, J., 1996. Embryological
origin for autism: developmental anomalies of the cranial nerve motor
nuclei. J. Comp. Neurol. 370, 247–261.
Samad, T.A., Krezel, W., Chambon, P., Borrelli, E., 1997. Regulation of
dopaminergic pathways by retionids: activation of the D2 receptor promoter
by members of the retinoic acid receptor-retinoid X receptor family.
Proc. Natl. Acad. Sci. U. S. A. 94, 14349–14354.
Samarut, E., Fraher, D., Laudet, V., Gibert, Y., 2015. ZebRA: An overview of
retinoic acid signaling during zebrafish development. Biochim. Biophys. Acta
1849 (2), 73–83.
Saverino, C., Gerlai, R., 2008. The social zebrafish: behavioral responses to
conspecific, heterospecific, and computer animated fish. Behav. Brain Res.
191, 77–87.
Shaw, G.M., Velie, E.M., Schaffer, D., Lammer, E.J., 1997. Periconceptional intake
of vitamin A among women and risk of neural tube defect-affected
pregnancies. Teratology 55, 132–133.
Shenefelt, R.E., 1972. Gross congenital malformations, animal model: treatment
of various species with a large dose of vitamin A at known stages in
pregnancy. Am. J. Pathol. 66, 589–592.
Steele, C.E., Trasler, D.G., New, D.A., 1983. An in vivo/in vitro evaluation of the
teratogenic action of excess vitamin A. Teratology 28, 209–214.
Tafti, M., Ghyselinck, N.B., 2007. Functional implication of the vitamin A
signaling pathway in the brain. Arch. Neurol. 64, 1706–1711.
Vacca, L., Hutchings, D.E., 1977. Effect of maternal vitamin A excess on S-100
in neonatal rat cerebellum: a preliminary study. Dev. Psychobiol. 10,
171–176.
Vorhees, C.V., 1974. Some behavioral effects of maternal hypervitaminosis A in
rats. Teratology 10, 269–274.
Vorhees, C.V., Brunner, R.L., Butcher, R.E., 1979. Psychotropic drugs as behavioral
teratogens. Science 205, 1220–1225.
Vorhees, C.V., Brunner, R.L., McDaniel, C.R., Butcher, R.E., 1978. The relationship
of gestational age to vitamin A induced postnatal dysfunction. Teratology 17,
271–276.
 J.M. Bailey et al. / NeuroToxicology 52 (2016) 23–33 33

Wang, Y., Chen, J., Du, C., Huang, C., Dong, Q., 2014. Characterization of retonic
acid-induced neurobehavioral effects in developing zebrafish.
Environ. Toxicol. Chem. 33, 431–437.
Weiss, B., Cory-Slechta, D.A., 1994. Assessment of behavioral toxicity. In: Hayes,
A.W. (Ed.), Principles and Methods of Toxicology. 3rd ed. Raven Press, New
York, pp. 1091–1155.
Weiss, B., 1978. The whole animal as an assay system. In: Toribara, T.Y.,
Coleman, J.R., Dahneke, B.E., Feldman, J. (Eds.), Environmental Pollutants.
Plenum, New York, pp. 53–66.
Yen, J., Donerly, S., Levin, E.D., Linney, E.A., 2011. Differential
acetylcholinesterase inhibition of chlorpyrifos, diazinon, and parathion in
larval zebrafish. Neurotoxicol. Teratol. 33, 735–741.