Background
ASD is a genetically and clinically heterogeneous group of neurodevelopmental disorders representing various subtypes of altered social communication, unusually restricted interests, or repetitive behavior [
1]. Next-generation sequencing approaches have identified additional nonsense, frameshift, and insertion/deletion mutations in ASD or intellectual disability cases [
2‐
4].
In humans,
DYRK1A is located on chromosome 21q22.13 in the “Down Syndrome Critical Region (DSCR)” at 21q22.1–q22.3 [
5]. This gene has been proposed as a major contributor to the pathogenesis of Down syndrome, Alzheimer’s disease, and Huntington’s disease [
6‐
8]. However, truncation of
DYRK1A due to balanced chromosome translocations was previously reported in two unrelated individuals with overlapping phenotypes of developmental delay and microcephaly [
9]. Subsequently, mutations in
DYRK1A are also associated with primary microcephaly, intellectual disability, and ASD [
10‐
13]. In this report, we describe a newly affected individual with a heterozygous 21 kb intragenic deletion which involves the last five exons of
DYRK1A; the individual exhibits ASD in addition to learning difficulties and microcephaly.
Since these distinct cognitive phenotypes could arise from either increase or decrease of gene dosage, overexpression and KO techniques of
DYRK1A were applied to animal models in order to elucidate the underlying mechanism. Intellectual disability coupled with microcephaly was recapitulated in a
Dyrk1a overexpressing murine model which mimicked Down syndrome patients who possess an extra copy of chromosome 21 [
14,
15].
Dyrk1a null mutants exhibit generalized growth delay, including an overall reduction in the size of the developing brain as well as embryonic lethality during mid-gestation [
16‐
18]. Heterozygous mutants show decreased neonatal viability and reduced brain size from birth to adulthood. Neurobehavioral analysis revealed that heterozygous mutants in adulthood are deficient in motor function and learning [
18‐
20]; however, none of these murine model studies present sufficient evidence to directly link
Dyrk1a dysfunction with autism in the context of social interaction of an ASD animal model.
To understand the molecular mechanisms underlying microcephaly and ASD, we established an in vivo KO model using zebrafish. The zebrafish (
Danio rerio) is a tractable vertebrate model in biological research, especially in the fields of neuroscience [
21,
22]. Recent scientific reports show conservation of brain structures between zebrafish and humans, such as the amygdala, hippocampus, habenula, and hypothalamus [
23]. Moreover,
Danio rerio displays broad complex behaviors in aspects of learning, cognition, aggression, anxiety, and social interaction [
22]. The zebrafish and human genomes are well conserved with more than 80% of human disease genes represented in the zebrafish model [
24]. Thus, the zebrafish is a useful tool in elucidating the function of novel genes involved in head formation or neurogenesis [
25,
26] and, more recently, for validating the function of human candidate genes involved in microcephaly, intellectual disability, and ASD [
27‐
31].
We employed targeted KO of the zebrafish
DYRK1A orthologue and found that
dyrk1aa KO zebrafish exhibits microcephaly and impaired social behavior which is a key representative feature of ASD. Also, we report on the development of two approaches in assessing behavioral phenotypes of the zebrafish ASD model. Since social behavioral analysis in the context of ASD has not been reported on any other
DYRK1A animal model, we undertook the analysis of social and group behavioral interactions in the
dyrk1aa KO zebrafish. Several social interaction tests have already been addressed which assess the social behavior of zebrafish [
32,
33]; however, we have improved upon these social interaction assays by newly developing the shoaling bowl assay in which a flat-round bowl provides a convenient means for assessing group behavior in zebrafish autism models.
Methods
Clinical report
The proband was noted as being small for gestational age according to regular ultrasound scans. The affected female of northern European ancestry, now age 11 and half years, was born at 37-week gestation by emergency Cesarean section due to a drop in heart rate. Her birth weight was 1.9 kg. Due to breathing problems around the time of birth, the subject required suction at birth and did not cry. Afterward, she was administered oxygen and housed in the Special Care Baby Unit. The subject had a computed tomography (CT) brain scan at 1 year 3 months which showed mild cerebral atrophy involving mainly the frontal lobes. At age 3 years 1 month, she had a magnetic resonance imaging (MRI) scan and microcephaly was noted. Her head circumference has always been at − 5 standard deviations being below the 0.4th percentile. Her MRI showed increased X-ray CLC spaces which is a reflection of a moderate degree of cerebral volume loss, more so in the white matter than in gray. There were also some abnormal subcortical high signals in both temporal lobes; however, no overlaying abnormality was present. Also, there was evidence of thinning in the corpus callosum as well as a degree of volume loss in the medulla oblongata compared to previous scans. In addition, a mild dilatation of the lateral ventricles probably represented white matter loss. There was a small high-signal area seen in the white matter of the right parietal lobe representing gliosis.
Developmentally, the subject could sit alone at 8 months, roll from front to back by 1 year, commando crawl at 14 months, crawl properly around 16 ½ months, pull to stand at 15 months, walk around furniture at 1 year 7 months, and walk alone at 2 years despite having immature gait. She did not require the need of walking aids, but her legs, hips, knees, and ankles have always been very stiff. The subject’s parents approached their doctor when she was 3 years 6 months old, and she was diagnosed, at 6 years of age, with scoliosis of the back, differing leg lengths, inflexibility, and possible cerebral palsy. She has not had any serious head injuries but started having seizures at about 13 months. At the age of 2 ½, she was diagnosed with epilepsy and continues to have four or five serious seizures a year; each lasts over an hour. At 9 years 6 months, she had increasing difficulty in straightening her knees completely and walking, requiring frequent use of a wheelchair. At her latest clinical assessment, she displayed an increase in tone in her upper extremities and continues to be ambulatory. Clinically, her spinal deformity has not worsened as confirmed by full spine X-rays. She has a limb length discrepancy—shorter on the left than the right—with some pelvic obliquity. She also has bilateral valgus ankle joints and recurrent chest infections.
The stiffness in her arms and wrists makes dressing and undressing difficult. She has a combination of diagnoses including global development delay, ASD, learning difficulties, and illiteracy. Also, her level of speech at 9 years 8 months of age was that of a 3- to 4-year-old and required attendance to a special school. She continues to present high levels of challenging behaviors associated with distress and anxiety, continued problems in sociability, and little interest in her peers preferring to play on her own. She does not like crowds and might kick and shout at people if they invade her personal space. While she can be clingy to her parents, she makes very little eye contact and is fixated on particular items such as footballs and goggles; the latter of which she has at least 17 pairs that she wears at home but never at the swimming pool. She flaps her hands when excited, and she continues to have problems with attention and decreased concentration.
Microarray analysis
Array comparative genomic hybridization (CGH) was carried out using a BlueGnome 8x60k International Standard Cytogenomic Array (ISCA) design oligonucleotide microarray. Test DNA was referenced against same-sex control DNA, and data was analyzed in BlueFuse Multi v2.2. This platform should detect the majority of copy number imbalances > 15 kb in 500 disease gene/telomeric regions and > 180 kb in the genomic backbone and may detect smaller imbalances in some instances. The derivative log ratio (DLR) quality score given for this hybridization is 0.21. Probes are mapped to GRCh37/hg19.
Generation of dyrk1aa KO zebrafish
We identified the zebrafish
dyrk1aa gene and its exon/intron boundaries by searching the Ensembl database (GRCz10 Ensembl gene ID: ENSDARG00000063570; transcript ID: ENSDART00000100073). The
dyrk1aa (7 bp deletion) KO fish was generated using TALEN, as previously reported [
34]. A TALEN pair targeting exon 5 of
dyrk1aa (left target site: 5′-tgg gtc gcc atc aag atc at-3′; right target site: 5′- gcc ttc ctg aat cag gct ca-3′) was designed and assembled by ToolGen Inc. (
http://toolgen.com/). In vitro
-transcribed RNA of the TALEN pair (100 ng each) was microinjected into 1~2 cell stage of fertilized zebrafish eggs, which were then grown to 4-month-old adulthood. A stable mutant line,
dyrk1aakrb1, was identified and genotyped by direct PCR and sequencing performed using two sets of nested primers: the outer primer pair 5′-cca gca aca aga agg aga gg-3′ (forward) and 5′-agc cct gat ctt tcc agg tt-3′ (reverse) and the inner primer pair 5′-tta caa cga cgg cta tga cg-3′ (forward) and 5′-ttc atc tcg gtg tcg tgc t-3′ (reverse). The PCR amplification conditions were as follows: for primary PCR, 35 cycles of 95 °C, 20 s; 59 °C, 40 s; 72 °C, 1 min; and for secondary PCR, 25 cycles of 95 °C 20 s; 55 °C, 40 s; 72 °C, 30 s. The progeny were propagated through a series of out-crossings with wild type (WT) fish; these animals were eventually in-crossed to obtain homozygous KOs. The KO zebrafish line is deposited in the KCTC (
http://biorp.kribb.re.kr/) with deposit number, BP1294898.
Brain histology and expression analysis
To ascertain brain histology, 7-month-old male WT and KO fish were fixed in 4% paraformaldehyde (PFA) solution overnight, then compared for body length. Among fish of the same size and age, brains were isolated and imaged and sizes were measured using ImageJ software. After dehydration in ethanol and clearing in xylene, brains were infiltrated with paraffin, embedded, and serial-sectioned. The sections (10-um thick) were stained with hematoxylin-eosin. The total area and ventricle area of the brain in the sections were measured using ImageJ and the ratio (ventricle area/total area ×100) was calculated. In situ hybridization was performed as previously described [
35] using the following digoxigenin (DIG RNA labeling kit, Roche)-labeled antisense probes:
sox2,
neurog1,
ccnd1,
c-fos,
crh,
oxt,
th1,
vglut2.2, and
gad1b. For
c-fos analysis, 7-month-old male WT and KO zebrafish were fixed in 4% PFA solution immediately after social interaction test. For
crh analysis, 7-month-old male WT and KO fish were fixed after social isolation. For
oxt,
th1,
vglut2.2, and
gad1b analysis, 7-month-old male WT and KO fish from their home tank were fixed. To detect cell death, 3-week-old zebrafish larvae were fixed in 4% PFA solution for 4 h at room temperature. Fixed larvae were embedded in agar-sucrose solution (1.5% agar, 5% sucrose). The agar blocks containing the larvae were sunk in 30% sucrose solution and processed for transverse cryostat serial-sectioning. The sections (25-um thick) were immuno-stained with an antibody against activated caspase-3 (BD Biosciences), which marks apoptotic cell death.
Behavioral tests for early larval zebrafish
Dark flash test
Dark flash test was performed as previously reported [
36]. Free swimming 6 dpf larvae were placed in a 24-well plate (SPL life Sciences— each well contains a single larva— then inserted into the DanioVision Observation Chamber (Noldus). To induce freezing/startle response, dark flash pulses illuminated the plate for 30 s followed by lights off for 30 s (flash-off dark condition). This scheme was repeated five times. Locomotive response to visual stimuli was measured by video-tracking analysis using EthoVision XT7 software (Noldus). For analysis of locomotor activity, raw data was converted into total distance moved (cm) by each larva per 10 s time-bins. After behavioral assay, each zebrafish larva was genotyped using genomic PCR.
Sleep and waking activity
Sleep and waking activity was measured as previously described [
37].
dyrk1aa KO embryos and control WT embryos were raised in a light- and temperature-controlled incubator. Five-day-old larvae were placed in a 24-well plate in the observation chamber of Danio Vision tracking system for acclimation under controlled lighting conditions (12 h–12 h light-dark cycles). Starting from 5 dpf, locomotion of each larva during day and night phases were tracked and analyzed by EthoVision XT7 software over a course of 2 days. Locomotor activity was analyzed by converting raw data into the velocity (cm/s) of each larva per 30-min time-bins.
Social and group behavior tests for adult zebrafish
Novel tank assay
Novel tank assay was performed as previously described [
38]. Each 7-month-old male WT or KO zebrafish was placed in a transparent tank with dimensions measuring 24 × 15 × 15 cm. We replicated the novel tank assay with eight WT and eight KO fish. The back side of the tank was covered with a white sheet to aid data recording. We used a three-compartment novel tank with top, bottom, and middle virtual zones. All behavior tests were recorded for a period of 10 min from the lateral viewpoint of the tank using a video camera (Sony, HDR-CX190). Fish were returned to their home tanks immediately after completion of the test. The raw data was analyzed using EthoVision XT7 software.
Social interaction assay
The social interaction test was modified and improved upon from a previous study [
32]. The tank was divided into two sections by placing a metal mesh or an acrylic plate separator at the first quarter of the tank. To conduct the social interaction test, the first section of the tank was designated as the social cue. The second section was used as the space to investigate the behavior of tester fish. In every experiment, we used different 7-month-old male fish for both the social cue and tester to maintain similar conditions. We replicated this experiment with 30 WT and 30 KO tester fish, in total. The second section was divided further into four equal sub-zones; the zone nearest to the social cue was designated zone “I”, the second nearest zone “II”, the third zone “III”, and the last zone “IV”. The hollow-rectangular pattern of the metal mesh separator (0.3 × 0.3 cm) created a gray shadow, while the acrylic plate was transparent. All behavioral tests were performed between 13:00 and 17:00 h using water from a tank adjusted to the holding room temperature. All experimental fish were raised in a social environment. One day prior to each test, fish were transferred to a different tank in an isolated environment. All behavioral tests were recorded from the lateral viewpoint of the tank, for a period of 15 min using a video camera.
Shoaling bowl assay
Fish form groups in a behavior called shoaling [
39‐
41]. To test whether
dyrk1aa KO zebrafish show altered shoaling behavior, a group of 7-month-old fish (
n = 3–7 fish per group) was placed together and monitored by video tracking. We introduced a unique and simple device to test and quantify shoaling behavior. First, we examined several types of bowls (with different shapes, sizes, depths, and colors) and selected a round, flat bottom, white bowl for further experiments (upper half diameter, 33 cm; bottom diameter, 24 cm; height, 11 cm; and water depth, 3.2 cm). All tests for group behavior were recorded for a period of 15 min using a video camera at a fixed height with a top view of the bowl. The recorded videos were analyzed using 31 screenshots made every 10 s for 10–15 min measuring the distances between individual fish in the group using the ImageJ program.
Statistical analysis
In all experiments, comparisons between WT and KO fish were done using a two-tailed, Student’s t test. Data are expressed as mean ± standard error of the mean (SEM). In all tests, p < 0.05 was considered to be significant. * indicates p < 0.05, ** indicates p < 0.01, and ***p < 0.001.
Discussion
Although some functional roles of DYRK1A have been implied in mouse studies [
19,
20], so far there have been no reported behavioral studies of adult knockout animals with respect to autism. In this study, we generated a KO zebrafish line for
dyrk1aa after the discovery of an intragenic microdeletion of
DYRK1A in an individual with microcephaly and autism. We demonstrated through social behavioral tests that
dyrk1aa KO zebrafish exhibit social impairments reproductive of human ASD phenotypes.
The
DYRK1A gene is well conserved in vertebrates, including fish, rodents, and humans. Haploinsufficiency of
DYRK1A in humans results in microcephaly and ASD [
12], while knockout of
Dyrk1a in mice leads to premature death during early development [
18]. In the
dyrk1aa KO zebrafish, we found similar microcephaly and ASD-like phenotypes, yet the fish were viable without embryonic lethality. This discrepancy may be explained in part by reason of the two orthologous
DYRK1A genes in zebrafish,
dyrk1aa (NM_001080689) and
dyrk1ab (NM_001347831), caused by whole genome duplication of zebrafish [
59]. Thus, we can speculate that
dyrk1ab may compensate the early lethal phenotype and allow the survival of
dyrk1aa KO zebrafish into adulthood. We can confirm this possibility by generating a double KO line of both genes in further studies.
Previous murine model studies have been unable to link altered brain structure of Dyrk1a dysfunction with social behavior as a direct physiological model of ASD. The structural defect in our dyrk1aa zebrafish mutant is reminiscent not only of the Dyrk1a mouse, but also of other zebrafish models of autism candidate genes. They display significant structural abnormalities including microcephaly and cell death in anterior structures. Historically, linking these altered physiological states to behavioral deficits has been hampered by two major limitations. First is the paucity of bona fide genetic models for autism in zebrafish. Secondly, reported tracking programs to investigate adult fish behavior in 3D is subject to extensive variability, in large part due to the speed at which multiple fish are moving in three dimensions.
To overcome these limitations, we introduced two social behavioral tests: the social interaction and shoaling assay. In the social interaction assay, we optimized the (a) number of fish, (b) time window of monitoring, and (c) composition of separator material. Zebrafish are active animals and have a wide range of locomotion moving from side to side or from top to bottom in their tank. We found that a group of three fish, rather than 1–2 fish, was ideal for the social cue to facilitate recognition, provide better cueing effect, and elicit stronger interaction of tester fish. Previous work has shown that the ability to view and recognize others is an important factor of social cueing [
60‐
62]. We confirmed these observations by demonstrating that a transparent acrylic plate separator provided better recognition of social cues to tester fish than a metal mesh. Utilizing this assay, we showed that
dyrk1aa KO zebrafish have impaired social interaction as seen by frequent movements towards the far zones. Taken together, this newly optimized social interaction assay provides a useful means of investigating social interaction of zebrafish models in neurobehavioral disorders.
Next, we developed a novel shoaling assay, called the “shoaling bowl assay”. Shoaling behavior is considered an adaptive and effective natural anti-predatory response, which has been utilized in behavioral analyses in vertebrates [
32,
39]. This behavior mimics the tendency of zebrafish to live together and is a robust tool for measuring social behavior of group animals. We showed that the “mini shoal”, formed at the edge of the round bowl, is a preferred location for zebrafish to move together as a group along the narrow space of the shoal. We tested shoaling behavior in different group sizes (3–7 fish). A minimum group of three fish was used for the shoaling assay given that the fish maintained a constant distance between individuals regardless of the size of the group. With a minimum number of animals and a two-dimensional (2D) approach, analysis of social cohesion in a flat round bowl avoids the complexity of group behavior in a three-dimensional (3D) tank which is the current standard [
63,
64]. To our knowledge, the altered social cohesion of
dyrk1aa KO is the first experimental demonstration that shoaling behavior of animals can be regulated by a single gene. The interrelationship between anxiety and social cohesion in animal group behavior will be an interesting topic in further studies since collective animal behavior (huddling, flocking, or shoaling) is a defensive strategy employed by many species in response to predatory threat. Our findings open up a new avenue for the study of this evolutionarily important behavior at the molecular and neural circuit levels.
To understand the molecular mechanisms involved in behavioral alterations of KO fish, we analyzed the expression of various neural markers. Among them, we found significant changes in the expression of
c-fos and
crh mRNAs in specific brain regions of
dyrk1aa KO fish. Neuronal activity of KO fish, as indicated by
c-fos expression, was lower than that of WT fish in the ventral hypothalamic region during social interaction tests, which suggests that the KO fish brain is less activated by social cues. In addition,
crh expression level in the PO area of the hypothalamus of KO fish in the acute social isolation test was found to be lower than that of WT, demonstrating low responsiveness to stress in the context of social isolation. In mammals, the hypothalamic region is a known source of stress hormone secretion, such as CRH, and has been shown to be largely involved in social interaction behaviors [
65]. Thus, we can conclude that the reduction in the size of the
dyrk1aa KO fish brain exerts structural changes in the neural circuitry involved in executing of proper behavioral responses to external stress signals which is a vital decision-making aspect of social interactions. In future studies, we plan to examine in further detail the neural circuitry directly involved in
dyrk1aa function and autism.
In this paper, we have optimized a widely used social interaction test and newly developed the shoaling bowl assay as a convenient method to study group behavior. Furthermore, we showed that these tests can be effectively applied to the study of disease model animals in zebrafish. Together, these data demonstrate that dyrk1aa KO zebrafish not only recapitulate the neuroanatomical defects of humans with DYRK1A mutations but also exhibit similar hallmarks of impairments in social behavior.