- Research article
- Open Access
Identification of a neuronal population in the telencephalon essential for fear conditioning in zebrafish
© Kawakami et al. 2018
Received: 11 August 2017
Accepted: 7 March 2018
Published: 25 April 2018
Fear conditioning is a form of learning essential for animal survival and used as a behavioral paradigm to study the mechanisms of learning and memory. In mammals, the amygdala plays a crucial role in fear conditioning. In teleost, the medial zone of the dorsal telencephalon (Dm) has been postulated to be a homolog of the mammalian amygdala by anatomical and ablation studies, showing a role in conditioned avoidance response. However, the neuronal populations required for a conditioned avoidance response via the Dm have not been functionally or genetically defined.
We aimed to identify the neuronal population essential for fear conditioning through a genetic approach in zebrafish. First, we performed large-scale gene trap and enhancer trap screens, and created transgenic fish lines that expressed Gal4FF, an engineered version of the Gal4 transcription activator, in specific regions in the brain. We then crossed these Gal4FF-expressing fish with the effector line carrying the botulinum neurotoxin gene downstream of the Gal4 binding sequence UAS, and analyzed the double transgenic fish for active avoidance fear conditioning. We identified 16 transgenic lines with Gal4FF expression in various brain areas showing reduced performance in avoidance responses. Two of them had Gal4 expression in populations of neurons located in subregions of the Dm, which we named 120A-Dm neurons. Inhibition of the 120A-Dm neurons also caused reduced performance in Pavlovian fear conditioning. The 120A-Dm neurons were mostly glutamatergic and had projections to other brain regions, including the hypothalamus and ventral telencephalon.
Herein, we identified a subpopulation of neurons in the zebrafish Dm essential for fear conditioning. We propose that these are functional equivalents of neurons in the mammalian pallial amygdala, mediating the conditioned stimulus–unconditioned stimulus association. Thus, the study establishes a basis for understanding the evolutionary conservation and diversification of functional neural circuits mediating fear conditioning in vertebrates.
Fear conditioning is a type of learning through which animals learn to predict an aversive event from a correlated environmental cue. In mammals, the amygdala plays essential roles in this type of learning [1, 2]. The mammalian amygdala is a complex and anatomically heterogeneous structure, consisting of approximately 20 subnuclei that are derivatives of the pallial and subpallial portions of the telencephalon. The pallial amygdala consists of cortical nuclei and basolateral nuclei (BLA), containing predominantly glutamatergic neurons, whereas the subpallial amygdala consists of medial and central nuclei (CeA), containing predominantly GABAergic neurons [3–5]. The roles of these nuclei in fear conditioning have been studied extensively by producing nuclei-specific lesions. The BLA has been shown to serve as the sensory interface essential for the association of a conditioned (CS) and an unconditioned stimulus (US). The CeA then receives inputs from BLA through intra-amygdaloid circuitry and serves as the primary output structure, with projections to other brain regions and controlling fear responses [1, 2, 6, 7].
Fear conditioning is an evolutionarily conserved behavior, and both active avoidance and Pavlovian fear conditioning have been described in teleost fish [8–11]. In teleost, it has been hypothesized that the forebrain is formed by a mechanism called eversion, while the mammalian forebrain is formed by evagination. Additionally, the teleost telencephalon is composed of area dorsalis and area ventralis, which are homologous to the pallium and subpallium and are rich in glutamatergic and GABAergic neurons, respectively [12–15]. From neuroanatomical and hodological studies, it has been proposed that the medial zone of the dorsal telencephalon (Dm) is a homolog of the mammalian amygdala [16, 17]. A functional study on the telencephalic region important for fear conditioning was performed by ablation experiments , wherein goldfish were trained in active avoidance fear conditioning and, after acquisition of the conditioned avoidance response, the medial pallium (MP), including the Dm area, was ablated by surgery. The MP-lesioned fish exhibited a deficit in performing the avoidance response, indicating that the MP area was essential for retention of the conditioned avoidance response. Although the functional study supported the hypothesis that the teleost Dm is a homolog of the mammalian amygdala, the lesions created were quite large, and specific neuronal populations or circuits essential for fear conditioning have yet to be explored.
In zebrafish, Lau et al.  analyzed c-fos expression patterns in the brain when fish performed a light/dark choice (light-avoidance) behavior, and found that c-fos expression was detected in cells in Dm and other brain regions, including the hypothalamus and ventral telencephalon. von Trotha et al.  also analyzed c-fos expression in the zebrafish brain during administration of amphetamine, a drug of abuse, and an amphetamine-induced place preference behavior, and detected c-fos expression in the Dm area. Thus, these studies suggest the involvement of neurons in Dm in emotional and motivational behaviors. However, the cells found in these studies are not genetically labeled, and are therefore not manipulatable, and the requirements for behaviors are unknown.
Herein, we aim to identify the neuronal population essential for fear conditioning through a genetic approach in zebrafish. In previous studies [21, 22], we developed transposon-mediated gene trap and enhancer trap methods, and generated transgenic fish lines that expressed Gal4FF, an engineered Gal4 transcription activator, in specific organs, tissues and cells, including specific neuronal populations. Further, we demonstrated that, by taking advantage of the Gal4-UAS binary system, the activity of such specific neurons can be inhibited by targeted expression of the tetanus neurotoxin gene [21, 23]. In this study, we applied this powerful approach to explore the adult brain function. First, we performed large-scale gene trap and enhancer trap screens and identified transgenic fish lines that expressed Gal4FF in various different regions in the adult brain. Second, we selected lines expressing Gal4FF predominantly in the forebrain, crossed them with the UAS-botulinum neurotoxin fish that contained a modified botulinum toxin (BoTx) gene downstream of UAS , and analyzed behaviors of the double transgenic fish by using fear conditioning paradigms. Finally, we found transgenic fish lines expressing Gal4FF in a subpopulation of neurons in Dm that showed reduced performance in fear conditioning when crossed with the UAS:BoTx fish. Thus, the present study genetically identified the neuronal population in zebrafish essential for fear conditioning, which may be a functional equivalent of the mammalian amygdala.
Identification of transgenic fish with Gal4FF expression in the adult brain
Identification of Gal4FF transgenic fish with deficits in two-way active avoidance fear conditioning
Movie S1. Active avoidance fear conditioning of wild type fish. The movie shows an example of the analysis. On day 1, wild type fish were placed in a white acrylic tank, and US (electric shock) was given 10 s after CS (green LED) was on. On day 5, the fish successfully escaped to another compartment after CS was on. (MOV 3374 kb)
The emx3 enhancer trap lines showed deficits in two-way active avoidance fear conditioning when crossed with the UAS-neurotoxin line
Transposon integration sites in transgenic lines with Gal4FF expression in Dm
Trapped gene/ nearest gene
fgf17 / zgc:123194-001
To examine whether SAGFF120A;UAS:zBoTxBLC:GFP fish had a deficit in acquisition of active avoidance conditioning, but not in possible consolidation processes during the night, we developed another experimental procedure, in which five sessions (20 trials for active avoidance conditioning per session) were conducted within 1 day. Additionally, with this procedure, wild type fish could perform the avoidance response in more than 60% of the trials after the second session, while SAGFF120A;UAS:zBoTxBLC:GFP fish exhibited a much reduced performance (Fig. 6e–g), indicating that double transgenic fish had a deficit prominently in the acquisition process.
Other behavioral phenotypes in SAGFF120A;UAS:zBoTxBLC:GFP fish
To exclude the possibility that SAGFF120A;UAS:zBoTxBLC:GFP fish might have a deficit in the motor system, we analyzed the locomotor activity during free swimming of wild type and double transgenic fish. We detected comparable levels of locomotor activity in both wild type and double transgenic fish (Fig. 7d).
The mammalian amygdala mediates both Pavlovian and active avoidance fear conditioning [1, 2]. Pavlovian fear conditioning has also been described in fish [10, 11]. We tested whether SAGFF120A;UAS:zBoTxBLC:GFP fish also show a deficit in Pavlovian fear conditioning. In this procedure, we placed fish in a small rectangular tank and gave US (electric shock) 9 s after CS (light) was on. After repeating US coupling with CS five times, CS was administered and the turn activity was measured. We found that, while wild type fish showed an approximately three-fold increase in the turn activity after training, SAGFF120A fish exhibited significantly reduced turn activities (Fig. 7e and Additional file 5: Movie S2), indicating that the Gal4FF-expressing cells in SAGFF120A were essential for both active avoidance and Pavlovian fear conditioning.
Movie S2. Pavlovian fear conditioning of wild type fish. The movie shows an example of the analysis. Before conditioning: Wild type fish were placed in a white acrylic box and only CS (green LED) was given for 10 s. The turning activity during CS was measured. During conditioning: US (electric shock) was given for 1 s, 9 s after CS was on. After conditioning: CS was given for 10 s, and the turning activity was measured. (MOV 2802 kb)
In mammals, it has been reported that lesions in the basolateral amygdala lead to a deficit in an innate unconditioned response (freezing) to a natural dangerous stimulus (e.g., a ball of cat hairs for rats) . It has been known that zebrafish display robust innate fear response to the alarm substance included in the skin extract, with fish exhibiting erratic movement followed by freezing [29, 30]. To test whether SAGFF120A;UAS:zBoTxBLC:GFP fish exhibit the alarm response, we analyzed locomotor activities of wild type (n = 9) and SAGFF120A;UAS:zBoTxBLC:GFP (n = 8) fish upon administration of the skin extract (Fig. 7f–i). Both wild type and double transgenic fish responded to the skin extract, and exhibited erratic movement (1.8- to 2.0-fold increase of the average speed) followed by freezing (Additional file 6: Movie S3). However, we detected significant differences in the average speed and freezing duration in the phase of post-erratic movement between wild type and SAGFF120A;UAS:zBoTxBLC:GFP fish; namely, SAGFF120A;UAS:zBoTxBLC:GFP fish showed an increased average speed and decreased freezing duration, suggesting that Gal4FF-expressing cells may play a role in modulation of the freezing behavior (Fig. 7h, i).
Movie S3. Alarm responses of wild type fish and SAGFF120A;UAS:zBoTxBLCGFP fish. Behaviors of wild type and SAGFF120A;UAS:zBoTxBLCGFP fish were videotaped upon addition of skin extract. (MOV 8664 kb)
Characterization of the neuronal population in Dm (120A-Dm neurons) essential for fear conditioning
To examine if GFP-positive cells are indeed neurons, we analyzed brain sections from SAGFF120A;UAS:GFP fish by immunohistochemistry using anti-GFP and anti-NeuN (neuronal marker) antibodies. Overall, 99% of GFP-positive cells were NeuN-positive (2394/2423), revealing that most of the GFP-positive cells were neurons (Fig. 8bd–d). We also found that 16% (1986/12282) of the NeuN-positive cells in the Dm area were GFP-positive, indicating that only a subpopulation of neurons in Dm were labeled in the SAGFF120A line. The GFP-positive (Gal4FF-positive) cells in the Dm area in the SAGFF120A line are hereafter referred to as 120A-Dm neurons.
We then performed in situ hybridization using the vglut1/2.1/2.2 and gad67 probes. In the zebrafish telencephalon, the pallium (dorsal telencephalon) and subpallium (ventral telencephalon) were rich in glutamatergic and GABAergic neurons, respectively (Fig. 8e–g), in agreement with previous reports [14, 15]. We analyzed brain sections from SAGFF120A;UAS:GFP fish using the anti-GFP antibody and vglut1/2.1/2.2 probes, and found 94% (352/374) of the GFP-positive cells to be glutamatergic (Fig. 8h–j).
Characterization of projections of the 120A-Dm neurons
Further, we prepared a cleared brain sample from SAGFF120A;UAS:GFP fish by the Scale method , and analyzed it with light-sheet microscopy. This analysis also visualized the projections from the Dm area to the ventral telencephalon and hypothalamus. In the hypothalamic area, the projections proceeded laterally and terminated in the lateral hypothalamic nucleus (LH), the anterior tuberal nucleus (ATN), and dorsal zone of periventricular hypothalamus (Hd) (Fig. 9i and Additional file 7: Movie S4).
Movie S4. 3D image of the brain from SAGFF120A;UAS:GFP fish. A GFP fluorescence image of a transparent brain from SAGFF120A;UAS:GFP fish analyzed by light-sheet microscopy is shown. (MOV 6882 kb)
Functional and neurochemical similarities between the 120A-Dm neurons and the pallial amygdala nuclei
It has been postulated that the medial zone of the dorsal telencephalon (Dm) in fish is a homolog of the mammalian amygdala based on neuroanatomical studies [12, 16, 17] and ablation experiments in goldfish . However, the neuronal population and circuitry had not yet been identified. In the present study, we performed a genetic approach using zebrafish and, for the first time, identified a subpopulation of neurons located in the Dm area, which we named 120A-Dm neurons, essential for acquisition of both active avoidance and Pavlovian fear conditioning.
The mammalian amygdala consists of pallial (cortical) and subpallial (striatal) portions, and is further subdivided into multiple nuclei. The BLA, which are included in the pallial portion, contain predominantly glutamatergic neurons and are essential for the CS–US association ; namely, it was shown that the BLA lesions caused deficits in both Pavlovian and active avoidance fear conditioning [1, 5, 33]. The 120A-Dm neurons identified in the teleost Dm were also mostly glutamatergic and essential for both Pavlovian and active avoidance fear conditioning. From these functional and neurochemical similarities, we suggest that the 120A-Dm neurons are the functional equivalent of the pallial amygdala, and presumably neurons in BLA. It is not known whether the entire population of the 120A-Dm neurons or only part of them are essential for fear conditioning. Further subdivision of the 120A-Dm neurons will be required to answer this question.
In mammals, BLA lesions cause a deficit in an innate unconditioned response to a natural dangerous stimulus (for instance, cat hairs for rats) . Additionally, there was a contradictory report describing that inactivation of BLA impaired the learned, but not innate, fear response in rats . Herein, we tested whether the 120A-Dm neurons are involved in the innate fear response in zebrafish by analyzing reactions to a skin extract [29, 30]. Similarly to wild type fish, SAGFF120A;UAS:zBoTxBLC:GFP fish could respond to the skin extract and perform erratic movement and freezing. However, during the post-erratic movement phase, fish showed reduced freezing behaviors in comparison to wild type fish, suggesting that 120A-Dm neurons may play a role in modulation of the freezing behavior. Consistent with this, it was reported that cells in the Dm are activated upon administration of a skin extract . Further, it has been shown that the alarm response- or alarm substance-induced fear conditioning can be modulated by social buffering  or administration of the endocannabinoid receptor CB1 agonist . Transgenic fish that expressed Gal4FF in 120A-Dm neurons should allow us to investigate a neuronal basis of these behaviors as well as other motivational and emotional behaviors in zebrafish, such as light/dark choice or drug-seeking behaviors [19, 20], that are thought to be mediated by the amygdala-like functions of Dm.
In addition to the Dm, the present study highlighted other forebrain regions possibly important for fear conditioning. For instance, we found a relatively large number of lines that showed reduced avoidance responses and that commonly had Gal4FF expression in the ventral nucleus of the ventral telencephalon (Vv) or the preoptic area. Vv has been postulated to be a homolog of the septal nuclei of mammals [13, 37], which also play a crucial role in fear learning . The preoptic area has been postulated to be a homolog of the mammalian paraventricular nucleus of the hypothalamus (PVN) , containing the magnocellular neurosecretory system that mediates fear responses . hspGFFDMC56B fish had rather specific Gal4FF expression in the preoptic area and should be used for further studies to explore their role in fear conditioning.
emx3-expressing neurons are essential for fear conditioning
In the SAGFF70A and SAGFF120A lines, Gal4FF was expressed in a pattern similar to that of the emx3 gene. The zebrafish emx3 gene is expressed in the dorsal telencephalon at the embryonic stage and in the dorsomedial region in the adult brain [25–27]. Knockdown of the emx3 function by morpholino impairs expression of dorsal telencephalic marker genes . However, the function of the emx3-expressing cells had not been analyzed. The present study revealed the role of the emx3-expressing neurons in fear conditioning. It should be noted that, in our approach, the botulinum neurotoxin was continuously expressed throughout development. Although we could not detect gross morphological changes in the GFP-positive neurons in SAGFF120A;UAS:GFP;UAS:zBoTxBLC:GFP fish, a conditional system that prohibits neuronal activities only in the adulthood is required to examine the possibility that the toxin expression during developmental stages may have caused the observed behavioral deficits. Efforts are currently in progress along this line.
The mouse genome has two paralogs, Emx1 and Emx2 . Emx1 is expressed in the dorsal telencephalon in the developing brain, and Emx1-expressing cells give rise to excitatory neurons in the pallium, including pallial portions of the amygdala . Emx2 is expressed earlier and more broadly, and plays a major role in the formation of the medial limbic cortex . The Emx1-expressing cells in the developing telencephalon in chicken and Xenopus also contribute to cells in amygdalar nuclei [45, 46]. Thus, roles of cells expressing the emx family genes in the developing brain may be conserved during vertebrate evolution. However, the function of Emx-expressing cells in the adult brain was not characterized. It should be interesting to investigate Emx-expressing neurons in the adult brain in other vertebrates to see whether those neuronal populations harbor essential roles in fear conditioning as well.
Projections of 120A-Dm neurons to the hypothalamus
We found that 120A-Dm neurons had major projections to the hypothalamic area. In mammals, the hypothalamus is important in fear responses, controlling heart rate and blood pressure [1, 2]. We assume that the Dm–hypothalamus connection should also play an important role in mediating fear responses in fish. In previous work using goldfish, efferent projections from Dm to the hypothalamic area, including the ATN and dorsal zone of periventricular Hd, were identified by anterograde labeling , and minor outputs from Dm to the LH were detected by retrograde labeling . The present study clearly visualized projections of 120A-Dm neurons that terminated in the hypothalamic area, including ATN, LH, and Hd, consistently with the results obtained in the tracer experiments in goldfish.
In mammals, the CeA is the major output center, has projections to the lateral hypothalamus, and predominantly contains GABAergic neurons [1, 2, 5]. In contrast, 120A-Dm neurons are mostly glutamatergic and thought to be a functional equivalent of the pallial amygdala. Further studies on the roles of excitatory projections from Dm to the hypothalamus should provide new insights into the conservation and diversification of limbico-hypothalamic connections during evolution.
Projections of 120A-Dm neurons to the other telencephalic regions
We found the projection of 120A-Dm neurons also terminated in the neuropil area of the EN and preoptic area, Vd, and Vs, suggesting that these are possible targets. The EN consists of a dorsal GABAergic part and a ventral glutamatergic part, which have been hypothesized to be homologous to the EN of non-primate mammals (internal segment of the globus pallidus of primates) and the bed nucleus of the stria medullaris, respectively . In goldfish and zebrafish, the ventral glutamatergic neurons have been shown to project to the Hb nuclei [11, 48], and it was recently shown that the Hb-median raphe circuit in zebrafish is essential for active avoidance conditioning . Thus, we hypothesize that the projection of 120A-Dm neurons to the EN may play a role in mediating active avoidance responses. Consistent with this hypothesis, hspGFF38B and hspGFF55B fish, which strongly expressed Gal4FF in the Hb and EN, respectively, also exhibited deficits in active avoidance conditioning when crossed with the UAS:neurotoxin line (Additional file 2: Figure S1). Further analyses using these transgenic fish should reveal the role of the Dm-EN-Hb circuit in active avoidance responses. Vd and Vs are structures located in the subpallium and rich in GABAergic neurons, and have been hypothesized as homologs of the mammalian striatum and CeA, respectively . Thus, these connections may correspond to connections of BLA to the striatum, and the intra-amygdaloid connection of BLA to CeA, which have been described in mammals [1, 2, 49].
The genetic approach reveals functional neuronal circuits mediating adult behavioral phenotypes
Herein, we succeeded in performing a genetic approach to study a learning behavior in zebrafish. We think this success mainly relies on three factors. Firstly, our trap lines expressed Gal4FF specifically and strongly in the adult brain. We have shown that Gal4FF is a strong but less-toxic transcription activator in zebrafish [21, 50] and, since then, performed gene and enhancer trap genetic screens to label specific cell types by Gal4FF expression at the larval stages . The present study demonstrated that the gene and enhancer trap approaches are applicable to generate specific Gal4FF expression patterns at the adult stage as well. Secondly, the UAS:zBoTxBLC:GFP line expressed the neurotoxin reliably and reproducibly in combination with various Gal4FF lines. It was reported that some UAS effector lines, especially those containing 14xUAS, suffered from silencing effects . In contrast, we have been using 5xUAS for UAS-effector lines , and have not experienced such severe silencing effects. Further, when we created UAS:zBoTxBLC:GFP fish, we generated more than 30 different insertions, selected transgenic fish that showed the strongest expression by crossing them with several different Gal4FF driver lines, and established the best line with a single transposon insertion. The UAS:zBoTXBLC:GFP line thus established worked effectively in the larval stages  as well as in the adult stage (this study). Thirdly, our protocol developed for active avoidance fear conditioning has worked very efficiently and reproducibly, enabling the identification of transgenic lines with reduced learning activities out of many candidate lines. In summary, the present study demonstrated that the genetic approach combined with a behavioral paradigm is powerful to dissect functional neuronal circuits in the adult zebrafish brain, and should be applicable to the study of other brain circuits and behaviors.
Fear conditioning is commonly observed in vertebrate species. In teleost, it has been postulated that the medial zone of the dorsal telencephalon (Dm) is a homolog of the mammalian amygdala, and essential for retention of the conditioned avoidance responses. However, Dm is a broad area and functional neuronal populations had not yet been identified. Herein, we identified a subpopulation of neurons in Dm essential for fear conditioning through a genetic approach in zebrafish. These neurons are mostly glutamatergic and have projections to other brain regions, including the hypothalamic area and ventral telencephalon. We propose that these should be functional equivalents of neurons in the mammalian pallial amygdala, mediating a CS–US association. Thus, we established a basis for understanding the evolutionary conservation and diversification of functional neural circuits mediating fear conditioning in vertebrates.
Transgenic fish that expressed Gal4FF were generated by the gene trap and enhancer trap methods . Transposon integration sites in transgenic fish lines were analyzed by Southern blot hybridization and inverse PCR as described previously . UAS:GFP fish were used to visualize Gal4FF expression . The T2SUASzBoTXBLCGFP construct containing 5xUAS, TATA sequence, the codon-optimized botulinum toxin B light chain gene  fused to the EGFP gene, and SV40 polyA between cis-sequences of Tol2 was created and injected to fertilized eggs to generate the UAS:zBoTXBLC:GFP transgenic fish.
Analysis of Gal4FF expression in the adult brain
The GFP expression patterns in the adult brain were first observed under a fluorescence microscope (MZ 16FA, Leica Microsystems). The heads were then fixed in 4% paraformaldehyde and dissected to take the brains out of the skulls as described previously . The isolated brains were observed under a fluorescence microscope (MZ 16FA, Leica Microsystems). The fixed brain samples were embedded in 1% agarose in 0.1 M phosphate buffered saline (PBS; pH 7.4) and 100 μm-thick serial coronal sections were made by using a vibratome. The slices were collected in 24-well plates and mounted on slide glasses (Matsunami) using PermaFluor Aqueous Mounting Medium (Thermoscientific). Sections were observed under an upright epifluorescence microscope (Axio Imager Z1, Zeiss).
Preparation of fish for behavioral analyses
Fish aged from 5 months to 1.5 years were used for behavioral studies. Prior to behavioral assays, fish were moved to a behavioral assay room and kept in isolation in 2-L tanks for 2 days.
Two-way active avoidance fear conditioning (5-day procedure)
A white opaque acrylic tank (length 41 cm × width 17 cm × height 12 cm) with transparent walls at both ends, a trapezoidal wedge (10 cm at the top and 20 cm at the bottom × width 17 cm × height 5 cm) in the center of the tank, green LEDs (3.3 V DC, 2 A), and a pair of platinum mesh electrodes (12 V AC) were used. Behaviors were monitored and analyzed by programs created by using LabView 8.6 (National Instruments). Habituation was performed for 15 min per day for 2 days in the shuttle box. Conditioning involved (1) 2 min in the shuttle box; (2) initiating CS (green LED) and US 10 s later (12 V AC electric shock); and (3) turning off of CS and US 5 s after initial US; (4) if fish escaped while CS was on, US was not given; and (5) if the fish moved to another compartment while US was on, both CS and US were turned off. The process was repeated 10 times with an inter-trial interval of 25 ± 5 s per day on 5 consecutive days. The avoidance index was calculated as the number of successful escape responses per 10 trials. The two-way active avoidance conditioning was also performed as blind experiments, in which the fish identities were not known to the experimenter (Fig. 6).
Two-way active avoidance fear conditioning (1-day procedure)
The same setup as the 5-day procedure was used. Habituation was performed for 1 h per day for 2 days in the shuttle box. Conditioning involved (1) 5 min in the shuttle box; (2) initiating CS and US 10 s later (9 V AC electric shock); (3) turning off of both CS and US 5 s after US; (4) if fish escaped while CS was on, US was not given; and (5) if the fish moved to another compartment while US was on, both CS and US were turned off. The process was repeated 20 times with intervals of 25 ± 5 s per session, and five sessions were performed with an inter-trial interval of 3 min. The avoidance index was calculated as the number of successful escape responses per 20 trials.
The light response
The response to a light stimulus, which was described previously for larval zebrafish , was measured by using adult fish. A white opaque acrylic box (length 12 cm × width 17 cm × height 12 cm) with a transparent wall on one-side and equipped with a green LED light (3.3 V, 2A DC) was used. The water level was 5 cm in depth. Fish behavior was monitored and analyzed by programs created using LabView 8.6 (National Instruments). Fish were kept in the apparatus for 10 min and the green LED light was then turned on for 10 s. Fish locomotion was recorded at 27 fps, and the locomotor activities (speed) of the fish 100 ms before and after light-on were analyzed. Movie analysis was carried out with ImageJ 1.48v (US National Institute of Health) .
Fish were habituated for 5 min in an opaque tank (length 33 cm × width 19 cm × height 15 cm). Behaviors were recorded for 10 min and analyzed by using a program created with LabView 8.6 (National Instruments). Locomotor activities were calculated as distances travelled in 10 min.
Pavlovian fear conditioning
A white opaque acrylic box (length 12 cm × width 17 cm × height 12 cm) with a transparent wall on one-side equipped with a green LED light (3.3 V, 2A DC) and a pair of platinum mesh electrodes (9 V AC) was used. Behaviors were monitored and analyzed by programs created with LabView 8.6 (National Instruments). Habituation was performed by (1) placing fish for 10 min in the tank, (2) then initiating CS for 10 s, five times, at 50-s intervals, (3) followed by 10 min of free-swimming. Conditioning involved initiating CS and, 9 s later, initiating US for 1 s, followed by turning off of both CS and US. The process was repeated 5 times at 50-s intervals and the entire conditioning process was repeated once. The test involved (1) 10 min of free-swimming after training and (2) delivery of CS five times at 50-s intervals. The behavior was recorded at 27 fps and the movies were analyzed with ImageJ . Turning angles were determined by measuring the change of the head-tail axis of the fish. Turning with angles greater than 90° within six frames (0.22 s) was defined as a conditioned response. The number of conditioned responses during the 10 s before and after CS was counted, and fold changes in turning frequencies were calculated with or without conditioning.
The alarm response
Skin extracts were prepared fresh and kept on ice on the day of use . Adult fish (> 3 months old) were anesthetized in Tricaine (0.025%) and quickly sacrificed by decapitation. Excess water was removed from the skin using a paper towel, and 15 shallow cuts were made on each side of the trunk, avoiding contamination of the blood. Cuts were washed with distilled water, and 10 mL of skin extract were collected from each fish. To test alarm response, fish were habituated for 10 min in a 2-L tank (length 25 cm × width 6.5 cm × height 16 cm) equipped with delivery tubes at both ends. Then, 2 mL of the skin extract was applied to the tank. Behavior was recorded for 3 min before and after addition of the skin extract. The locomotor activity of wild type and the double transgenic fish was divided into three phases, namely baseline (B, before addition of the skin extract), erratic movement (from addition of the skin extract to the time when the average speed dropped to the baseline level; 38 s for wild type and 24 s for the double transgenic fish), post-erratic movement, and analyzed. Freezing was defined as locomotor activities with a speed of less than 5 mm/s. The movies were analyzed using ImageJ .
For active avoidance performance (5-day procedure), analysis of variance (ANOVA) and Dunnet’s multiple comparison tests were performed to test the statistical significance between test samples and control. For active avoidance performance (1-day procedure), ANOVA and Tukey’s multiple comparison tests were performed to test the statistical significance between test samples and control. For locomotive activity, Kruskal–Wallis test was performed. In Pavlovian fear conditioning, two-way ANOVA and Tukey’s multiple comparison post-hoc tests were performed. No data points were excluded in these analyses. All the statistical tests were performed by using PRISM 6 (GraphPad software).
Immunohistochemistry and in situ hybridization
Coronal/sagittal sections of 100 μm in thickness were used for immunohistochemistry. The samples were treated for 1 h at room temperature in blocking buffer, 0.5% Tween-20 or Triton X-100, 3% bovine serum albumin (Sigma) in PBS, and then incubated overnight with primary antibodies diluted in blocking buffer at 4 °C. The samples were then washed with 0.5% Tween-20 or Triton X-100 in PBS and incubated with secondary antibodies. Rabbit anti-GFP polyclonal (1:500 dilution, A6455; Invitrogen, RRID:AB_221570, LOT number = 1,650,113), mouse anti-NeuN monoclonal (1:200 dilution, MAB377; Millipore, RRID:AB_2298772, LOT number = 2,428,671), and mouse monoclonal anti-MAP2 antibody [AP-20] (1:300 dilution, ab11268; Abcam, RRID:AB_297886) were used for the primary antibodies. AlexaFluor 488 goat anti-rabbit IgG (1:400 dilution, A11008; Invitrogen, RRID:AB_143165) and AlexaFluor 633 goat anti-mouse IgG (1:1000 dilution, A21050; Invitrogen, RRID:AB_141431) were used for the secondary antibodies. For in situ hybridization, digoxigenin-labeled probes were synthesized by using the emx3 and the vglut1/2.1/2.2 cDNAs as templates. In situ hybridization was performed on coronal slices of fixed brain using a protocol described previously  with modifications. Prehybridization and hybridization were performed at 65 °C for 2 h in Hyb(+) solution or at 65 °C overnight in Hyb(+) solution containing approximately 100 ng of digoxigenin-labeled probes, respectively. The samples were washed with 66% Hyb (−)/2X SSCT at 65 °C for 30 min, 33% Hyb(−)/2X SSCT at 65 °C for 30 min, 2X SSCT at 65 °C for 15 min, and with 0.2X SSCT at 65 °C for 30 min twice. Hyb(−): 50% formamide, 5X SSC (Gibco), 0.1% Tween-20 (Pierce). Hyb(+): Hyb(−) with 5 mg/mL RNA purified from torula yeast (Sigma) and 50 μL/mL heparin (Sigma). The samples were then incubated in blocking solution (2% blocking reagent in PBST) (Roche) at 4 °C overnight and then incubated in 1:5000 dilution of anti-digoxigenin-AP Fab-fragments (11093274910; Roche, RRID:AB_514497) at 4 °C overnight. For emx3, signals were detected using BM purple (Roche). The reaction was stopped by washing with PBST. The slices were mounted on slide glasses (Matsunami) using glycerol-gelatin mounting medium and observed under a microscope (Imager Z1). Images were taken with an AxioCam MRc5 (Zeiss) camera and analyzed with Axio Vision Ver4.1 imaging software (Zeiss). For vglut1/2.1/2.2, signals were detected with Fast Red Tablets (Roche). The slices were mounted on slide glasses using PermaFluor Aqueous Mounting Medium (Thermoscientific) and observed with a laser scanning confocal microscope (FV-1000-D; Olympus) or a Zeiss confocal microscope with Yokogawa CSU-W1 laser scanning unit (Yokogawa). Images were processed with ImageJ .
Preparation of a cleared brain and light-sheet microscopy
The Scale method  was applied in the preparation of a cleared brain from SAGFF120A;UAS:GFP fish. The head was fixed with 4% paraformaldehyde/PBS (Wako) at 4 °C for overnight, and then dissected. The brain sample was washed with PBS, incubated in 20% sucrose/PBS at 4 °C for 2 days, embedded in OCT compound (Sakura), and frozen with liquid nitrogen. The sample was then thawed and washed with PBS, transferred into the ScaleA2 solution, and kept at 4 °C for more than 3 weeks. The ScaleA2 solution was changed every 2 days. The sample was observed with a light sheet fluorescence microscope (Zeiss Light sheet Z.1) with a 5× NA 0.16 lens. Fluorescence was measured with excitation (488 nm) and emission (SP 550 (Ch1) and LP585 (Ch2)) filters, and the GFP signal was obtained by subtracting Ch2 from Ch1. For image analysis, ZEN (Zeiss) and IMARIS 7.0 (Bitplane) were used.
We thank N. Kishimoto and K. Sawamoto for instructions on brain dissection, H. Okamoto for advice on Pavlovian conditioning, and members of the Kawakami lab for fish maintenance and assistance in screening.
This work was partly supported by ERC Starting Grant 335561 (EY), National BioResource Project from Japan Agency for Medical Research and Development (AMED) (KK), and JSPS KAKENHI Grant Numbers JP15H02370, JP16H01651 and JP18H04988 (K.K.).
Availability of data and materials
All data generated during this study are included in this published article and its supplementary information files, and all materials generated during this study are available upon request.
PL and KK designed the experiments and wrote the manuscript. PL, HT, DA, YK, and EY performed experiments and analyzed the data. MLS generated the UAS:zBoTxBLC:GFP transgenic line. MIt, MIw, and HW performed anatomical analyses. AM assisted in the development of behavioral assay systems. All authors read and approved the final manuscript.
This experiment was approved by the Institutional Animal Care and Use Committee (IACUC, approval identification number 27–2), and complied with the Guide for the Care and Use of Laboratory Animals of the IACUC.
The authors declare that they have no competing interests.
Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.
Open AccessThis article is distributed under the terms of the Creative Commons Attribution 4.0 International License (http://creativecommons.org/licenses/by/4.0/), which permits unrestricted use, distribution, and reproduction in any medium, provided you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons license, and indicate if changes were made. The Creative Commons Public Domain Dedication waiver (http://creativecommons.org/publicdomain/zero/1.0/) applies to the data made available in this article, unless otherwise stated.
- LeDoux JE. Emotion circuits in the brain. Annu Rev Neurosci. 2000;23:155–84.View ArticlePubMedGoogle Scholar
- Maren S. Neurobiology of Pavlovian fear conditioning. Annu Rev Neurosci. 2001;24:897–931.View ArticlePubMedGoogle Scholar
- Sah P, Faber ES, Lopez De Armentia M, Power J. The amygdaloid complex: anatomy and physiology. Physiol Rev. 2003;83(3):803–34.View ArticlePubMedGoogle Scholar
- Martinez-Garcia F, Novejarque A, Lanuza E. The evolution of the amygdala in vertebrates. In: Kaas JH, editor. Evolutionary Neuroscience. Cambridge, MA: Academic Press; 2009. p. 313–92.Google Scholar
- Marek R, Strobel C, Bredy TW, Sah P. The amygdala and medial prefrontal cortex: partners in the fear circuit. J Physiol. 2013;591(10):2381–91.View ArticlePubMedPubMed CentralGoogle Scholar
- Kapp BS, Frysinger RC, Gallagher M, Haselton JR. Amygdala central nucleus lesions: effect on heart rate conditioning in the rabbit. Physiol Behav. 1979;23(6):1109–17.View ArticlePubMedGoogle Scholar
- LeDoux JE, Cicchetti P, Xagoraris A, Romanski LM. The lateral amygdaloid nucleus: sensory interface of the amygdala in fear conditioning. J Neurosci. 1990;10(4):1062–9.View ArticlePubMedGoogle Scholar
- Horner JL, Longo N, Bitterman ME. A shuttle box for fish and a control circuit of general applicability. Am J Psychol. 1961;74:114–20.View ArticlePubMedGoogle Scholar
- Pradel G, Schachner M, Schmidt R. Inhibition of memory consolidation by antibodies against cell adhesion molecules after active avoidance conditioning in zebrafish. J Neurobiol. 1999;39(2):197–206.View ArticlePubMedGoogle Scholar
- Eisenberg M, Kobilo T, Berman DE, Dudai Y. Stability of retrieved memory: inverse correlation with trace dominance. Science. 2003;301(5636):1102–4.View ArticlePubMedGoogle Scholar
- Amo R, Fredes F, Kinoshita M, Aoki R, Aizawa H, Agetsuma M, Aoki T, Shiraki T, Kakinuma H, Matsuda M, et al. The habenulo-raphe serotonergic circuit encodes an aversive expectation value essential for adaptive active avoidance of danger. Neuron. 2014;84(5):1034–48.View ArticlePubMedGoogle Scholar
- Nieuwenhuys R, Meek J. The telencephalon of actinopterygian fishes. In: Jones EG, Peters A, editors. Comparative Structure and Evolution of the Cerebral Cortex. New York: Springer; 1990. p. 31–73.Google Scholar
- Wullimann MF, Mueller T. Teleostean and mammalian forebrains contrasted: evidence from genes to behavior. J Comp Neurol. 2004;475(2):143–62.View ArticlePubMedGoogle Scholar
- Mueller T, Guo S. The distribution of GAD67-mRNA in the adult zebrafish (teleost) forebrain reveals a prosomeric pattern and suggests previously unidentified homologies to tetrapods. J Comp Neurol. 2009;516(6):553–68.View ArticlePubMedPubMed CentralGoogle Scholar
- Aoki T, Kinoshita M, Aoki R, Agetsuma M, Aizawa H, Yamazaki M, Takahoko M, Amo R, Arata A, Higashijima S, et al. Imaging of neural ensemble for the retrieval of a learned behavioral program. Neuron. 2013;78(5):881–94.View ArticlePubMedGoogle Scholar
- Braford MR Jr. Comparative aspects of forebrain organization in the ray-finned fishes: touchstones or not? Brain Behav Evol. 1995;46(4–5):259–74.View ArticlePubMedGoogle Scholar
- Northcutt RG. Connections of the lateral and medial divisions of the goldfish telencephalic pallium. J Comp Neurol. 2006;494(6):903–43.View ArticlePubMedGoogle Scholar
- Portavella M, Torres B, Salas C. Avoidance response in goldfish: emotional and temporal involvement of medial and lateral telencephalic pallium. J Neurosci. 2004;24(9):2335–42.View ArticlePubMedGoogle Scholar
- Lau BY, Mathur P, Gould GG, Guo S. Identification of a brain center whose activity discriminates a choice behavior in zebrafish. Proc Natl Acad Sci U S A. 2011;108(6):2581–6.View ArticlePubMedPubMed CentralGoogle Scholar
- von Trotha JW, Vernier P, Bally-Cuif L. Emotions and motivated behavior converge on an amygdala-like structure in the zebrafish. Eur J Neurosci. 2014;40(9):3302–15.View ArticlePubMedPubMed CentralGoogle Scholar
- Asakawa K, Suster ML, Mizusawa K, Nagayoshi S, Kotani T, Urasaki A, Kishimoto Y, Hibi M, Kawakami K. Genetic dissection of neural circuits by Tol2 transposon-mediated Gal4 gene and enhancer trapping in zebrafish. Proc Natl Acad Sci U S A. 2008;105(4):1255–60.View ArticlePubMedPubMed CentralGoogle Scholar
- Kawakami K, Abe G, Asada T, Asakawa K, Fukuda R, Ito A, Lal P, Mouri N, Muto A, Suster ML, et al. zTrap: zebrafish gene trap and enhancer trap database. BMC Dev Biol. 2010;10:105.View ArticlePubMedPubMed CentralGoogle Scholar
- Koide T, Miyasaka N, Morimoto K, Asakawa K, Urasaki A, Kawakami K, Yoshihara Y. Olfactory neural circuitry for attraction to amino acids revealed by transposon-mediated gene trap approach in zebrafish. Proc Natl Acad Sci U S A. 2009;106(24):9884–9.View ArticlePubMedPubMed CentralGoogle Scholar
- Sternberg JR, Severi KE, Fidelin K, Gomez J, Ihara H, Alcheikh Y, Hubbard JM, Kawakami K, Suster M, Wyart C. Optimization of a neurotoxin to investigate the contribution of excitatory interneurons to speed modulation in vivo. Curr Biol. 2016;26(17):2319–28.View ArticlePubMedGoogle Scholar
- Morita T, Nitta H, Kiyama Y, Mori H, Mishina M. Differential expression of two zebrafish emx homeoprotein mRNAs in the developing brain. Neurosci Lett. 1995;198(2):131–4.View ArticlePubMedGoogle Scholar
- Adolf B, Chapouton P, Lam CS, Topp S, Tannhauser B, Strahle U, Gotz M, Bally-Cuif L. Conserved and acquired features of adult neurogenesis in the zebrafish telencephalon. Dev Biol. 2006;295(1):278–93.View ArticlePubMedGoogle Scholar
- Ganz J, Kroehne V, Freudenreich D, Machate A, Geffarth M, Braasch I, Kaslin J, Brand M. Subdivisions of the adult zebrafish pallium based on molecular marker analysis. F1000Res. 2014;3:308.PubMedGoogle Scholar
- Vazdarjanova A, Cahill L, McGaugh JL. Disrupting basolateral amygdala function impairs unconditioned freezing and avoidance in rats. Eur J Neurosci. 2001;14(4):709–18.View ArticlePubMedGoogle Scholar
- Jesuthasan SJ, Mathuru AS. The alarm response in zebrafish: innate fear in a vertebrate genetic model. J Neurogenet. 2008;22(3):211–28.View ArticlePubMedGoogle Scholar
- Speedie N, Gerlai R. Alarm substance induced behavioral responses in zebrafish (Danio rerio). Behav Brain Res. 2008;188(1):168–77.View ArticlePubMedGoogle Scholar
- Kroehne V, Freudenreich D, Hans S, Kaslin J, Brand M. Regeneration of the adult zebrafish brain from neurogenic radial glia-type progenitors. Development. 2011;138(22):4831–41.View ArticlePubMedGoogle Scholar
- Hama H, Kurokawa H, Kawano H, Ando R, Shimogori T, Noda H, Fukami K, Sakaue-Sawano A, Miyawaki A. Scale: a chemical approach for fluorescence imaging and reconstruction of transparent mouse brain. Nat Neurosci. 2011;14(11):1481–8.View ArticlePubMedGoogle Scholar
- Amorapanth P, LeDoux JE, Nader K. Different lateral amygdala outputs mediate reactions and actions elicited by a fear-arousing stimulus. Nat Neurosci. 2000;3(1):74–9.View ArticlePubMedGoogle Scholar
- Ribeiro AM, Barbosa FF, Munguba H, Costa MS, Cavalcante JS, Silva RH. Basolateral amygdala inactivation impairs learned (but not innate) fear response in rats. Neurobiol Learn Mem. 2011;95(4):433–40.View ArticlePubMedGoogle Scholar
- Ruhl T, Zeymer M, von der Emde G. Cannabinoid modulation of zebrafish fear learning and its functional analysis investigated by c-Fos expression. Pharmacol Biochem Behav. 2017;153:18–31.View ArticlePubMedGoogle Scholar
- Faustino AI, Tacao-Monteiro A, Oliveira RF. Mechanisms of social buffering of fear in zebrafish. Sci Rep. 2017;7:44329.View ArticlePubMedPubMed CentralGoogle Scholar
- O'Connell LA, Hofmann HA. The vertebrate mesolimbic reward system and social behavior network: a comparative synthesis. J Comp Neurol. 2011;519(18):3599–639.View ArticlePubMedGoogle Scholar
- Calandreau L, Jaffard R, Desmedt A. Dissociated roles for the lateral and medial septum in elemental and contextual fear conditioning. Learn Mem. 2007;14(6):422–9.View ArticlePubMedPubMed CentralGoogle Scholar
- Herget U, Ryu S. Coexpression analysis of nine neuropeptides in the neurosecretory preoptic area of larval zebrafish. Front Neuroanat. 2015;9:2.View ArticlePubMedPubMed CentralGoogle Scholar
- Viviani D, Charlet A, van den Burg E, Robinet C, Hurni N, Abatis M, Magara F, Stoop R. Oxytocin selectively gates fear responses through distinct outputs from the central amygdala. Science. 2011;333(6038):104–7.View ArticlePubMedGoogle Scholar
- Viktorin G, Chiuchitu C, Rissler M, Varga ZM, Westerfield M. Emx3 is required for the differentiation of dorsal telencephalic neurons. Dev Dyn. 2009;238(8):1984–98.View ArticlePubMedPubMed CentralGoogle Scholar
- Simeone A, Gulisano M, Acampora D, Stornaiuolo A, Rambaldi M, Boncinelli E. Two vertebrate homeobox genes related to the Drosophila empty spiracles gene are expressed in the embryonic cerebral cortex. EMBO J. 1992;11(7):2541–50.PubMedPubMed CentralGoogle Scholar
- Gorski JA, Talley T, Qiu M, Puelles L, Rubenstein JL, Jones KR. Cortical excitatory neurons and glia, but not GABAergic neurons, are produced in the Emx1-expressing lineage. J Neurosci. 2002;22(15):6309–14.View ArticlePubMedGoogle Scholar
- Yoshida M, Suda Y, Matsuo I, Miyamoto N, Takeda N, Kuratani S, Aizawa S. Emx1 and Emx2 functions in development of dorsal telencephalon. Development. 1997;124(1):101–11.PubMedGoogle Scholar
- Puelles L, Kuwana E, Puelles E, Bulfone A, Shimamura K, Keleher J, Smiga S, Rubenstein JL. Pallial and subpallial derivatives in the embryonic chick and mouse telencephalon, traced by the expression of the genes Dlx-2, Emx-1, Nkx-2.1, Pax-6, and Tbr-1. J Comp Neurol. 2000;424(3):409–38.View ArticlePubMedGoogle Scholar
- Brox A, Puelles L, Ferreiro B, Medina L. Expression of the genes Emx1, Tbr1, and Eomes (Tbr2) in the telencephalon of Xenopus laevis confirms the existence of a ventral pallial division in all tetrapods. J Comp Neurol. 2004;474(4):562–77.View ArticlePubMedGoogle Scholar
- Rink E, Wullimann MF. Some forebrain connections of the gustatory system in the goldfish Carassius auratus visualized by separate DiI application to the hypothalamic inferior lobe and the torus lateralis. J Comp Neurol. 1998;394(2):152–70.View ArticlePubMedGoogle Scholar
- Villani L, Zironi I, Guarnieri T. Telencephalo-habenulo-interpeduncular connections in the goldfish: a DiI study. Brain Behav Evol. 1996;48(4):205–12.View ArticlePubMedGoogle Scholar
- Wall NR, De La Parra M, Callaway EM, Kreitzer AC. Differential innervation of direct- and indirect-pathway striatal projection neurons. Neuron. 2013;79(2):347–60.View ArticlePubMedPubMed CentralGoogle Scholar
- Asakawa K, Kawakami K. The Tol2-mediated Gal4-UAS method for gene and enhancer trapping in zebrafish. Methods. 2009;49(3):275–81.View ArticlePubMedPubMed CentralGoogle Scholar
- Kawakami K, Asakawa K, Hibi M, Itoh M, Muto A, Wada H. Gal4 driver transgenic zebrafish: powerful tools to study developmental biology, organogenesis, and neuroscience. Adv Genet. 2016;95:65–87.PubMedGoogle Scholar
- Goll MG, Anderson R, Stainier DY, Spradling AC, Halpern ME. Transcriptional silencing and reactivation in transgenic zebrafish. Genetics. 2009;182(3):747–55.View ArticlePubMedPubMed CentralGoogle Scholar
- Urasaki A, Kawakami K. Analysis of genes and genome by the Tol2-mediated gene and enhancer trap methods. Methods Mol Biol. 2009;546:85–102.View ArticlePubMedGoogle Scholar
- Kurazono H, Mochida S, Binz T, Eisel U, Quanz M, Grebenstein O, Wernars K, Poulain B, Tauc L, Niemann H. Minimal essential domains specifying toxicity of the light chains of tetanus toxin and botulinum neurotoxin type A. J Biol Chem. 1992;267(21):14721–9.PubMedGoogle Scholar
- Kishimoto N, Alfaro-Cervello C, Shimizu K, Asakawa K, Urasaki A, Nonaka S, Kawakami K, Garcia-Verdugo JM, Sawamoto K. Migration of neuronal precursors from the telencephalic ventricular zone into the olfactory bulb in adult zebrafish. J Comp Neurol. 2011;519(17):3549–65.View ArticlePubMedGoogle Scholar
- Burgess HA, Granato M. Modulation of locomotor activity in larval zebrafish during light adaptation. J Exp Biol. 2007;210(Pt 14):2526–39.View ArticlePubMedGoogle Scholar
- Image J Web Link. https://imagej.nih.gov/ij/. Accessed 20 Mar 2018.
- Nagayoshi S, Hayashi E, Abe G, Osato N, Asakawa K, Urasaki A, Horikawa K, Ikeo K, Takeda H, Kawakami K. Insertional mutagenesis by the Tol2 transposon-mediated enhancer trap approach generated mutations in two developmental genes: tcf7 and synembryn-like. Development. 2008;135(1):159–69.View ArticlePubMedGoogle Scholar
- Wullimann MF, Rupp B, Reichert H. Neuroanatomy of the Zebrafish Brain. Basel: Birkhäuser Verlag; 1996.View ArticleGoogle Scholar