Transcriptomic Analysis in the Striatum Reveals an Involvement of Nurr1 in Social Behavior of Prenatally Valproic Acid-Exposed Mice

Background The striatum of the basal ganglia is the major subcortical component of the mammalian forebrain. Magnetic resonance imaging (MRI) studies have implicated surface deformation of the striatum in the brains of patients with autism spectrum disorders (ASD) and their correlation with behavioral phenotypes. Methods Using RNA sequencing (RNA-Seq), we analyzed transcriptome alteration in striatal tissues from 10-week-old prenatally valproic acid (VPA)-exposed BALB/c male mice. To investigate the relationship of Nurr1 with synaptic and social decit, we activated the expression level of nuclear receptor related 1 protein (Nurr1) by i.p. injection of amodiaquine (AQ) for 2 weeks. Furthermore, we employed lentiviral system to inhibit the Nurr1 expression and provide evidence for the role of Nurr1 in social behavior. Transcriptomic analysis showed higher levels of genes related to synaptic function of neurons in striatal tissues from the prenatally VPA-exposed mice. Among those genes, Nurr1 expression was signicantly upregulated. The treatment with AQ, which has been known to be a ligand for Nurr1, to saline-exposed mice mimicked social decits and synaptic abnormality observed in prenatally VPA-exposed mice. Moreover, viral inhibition of Nurr1 markedly improved social decits in prenatally VPA-exposed mice. This study did not identify the mechanism of how Nurr1 activation regulates striatum-related circuit. Identication of the mechanism will provide explanation for behavioral impairments in prenatally VPA-exposed mice.

Introduction adapters using the TruSeq RNA Library prep Kit (Illumina, CA, USA). The suitable fragments automatically puri ed by BluePippin 2% agarose gel cassette (Sage Science, MA, USA) were selected as templates for PCR ampli cation. The nal library sizes and qualities were evaluated electrophoretically with an Agilent High Sensitivity DNA kit (Agilent Technologies, CA, USA) and the fragment was found to be between 350-450 bp. Subsequently, the library was sequenced using an Illumina HiSeq2500 sequencer (Illumina, CA, USA). Statistics for each gene in each of the differential expression analysis, including FDR corrected pvalues are found in Table S1.
Bioinformatic analysis The gene ontology analyses of differentially expressed genes were performed using DAVID software (version 6.8). Differentially expressed genes were also analyzed for phenotypes using mouse genome informatics mammalian phenotype analysis in Enrichr (http://amp.pharm.mssm.edu/Enrichr/). p-value 0.05 was used as cutoff.
AQ treatment For an in vitro study, primary striatal neuron cultures were treated with 100 nM of AQ (Sigma-Aldrich, MO, USA) for 24 h. AQ was diluted in 0.9% saline and prepared before treatment.
For in vivo study, 6-week-old prenatally SAL or VPA-exposed mice were intraperitoneally injected with AQ (20 mg/kg), twice per day at 12 h intervals, for 2 weeks. The AQ dose (20 mg/kg) used in this study was referred from previous report regarding the activating effect of AQ on Nurr1 in rodents [29]. AQ was diluted in 0.9% saline and prepared before administration. Mice underwent behavioral testing 1 week after the nal injection.
Stereotaxic injection of lentivirus Nurr1 shRNA lentiviruses and its control shRNA lentiviruses were purchased from Sirion Biotech (Martinsried, Germany). Mice received bilateral stereotaxic injections of virus (1.5 μl per side) into the striatum (coordinates: AP +0.3, ML ±1.9, DV -3.25 mm) at rates of 0.15 μl/min at each site (Kopf instruments, CA, USA). The needle was left in place for an additional 5 min and then was withdrawn gently.
Golgi staining After mice were transcardially perfused using heparin (100 U/ml) PBS solution, brains were dissected. Golgi staining was performed using the FD Rapid GolgiStain Kit (FD Neurotechnologies, MD, USA) according to manufacturer's instructions to label neurons. The brains were sectioned in the coronal plane at 100 μm thickness on a cryostat. Images of neurons in the DMS or DLS were acquired using a LSM 510 confocal microscope (Zeiss, Oberkochen, Germany) with a Plan-Neo uar 100x/1.30 N.A. oil immersion objective and the bright-eld setting. To assess spine density and phenotype, 8--10 cells of each slice were randomly selected. 2-3 dendrites per neuron were analyzed. Stacks of 512 x 512 pixel 3-D images with an interval of 1 μm were then taken for each cell to include all visible dendritic branches in the Zen software. After 3D neuronal reconstruction, the secondary and tertiary dendrite spines were measured, wherein the distance to the soma varied from 20-80 µm.
Quantitative reverse transcription polymerase chain reaction (qPCR) Total RNA was extracted from the whole striatum using the Qiazol reagent (Qiagen, Hilden, Germany). RNA was converted to cDNA using AccuPower RocketScript RT PreMix (Bioneer, Daejeon, Korea). qPCR was performed using a CFX96 (Bio-Rad, CA, USA). Results are presented as △△Ct-values normalized to the 18S rRNA. Primers were designed using NCBI primer blast software (http://www.ncbi.nlm.nih.gov/tools/primer-blast/). The speci city of the primer pairs was tested by PCR, and the PCR product were examined by agarose gel electrophoresis.
Western blot The striatum of mice was homogenized in ice-cold RIPA buffer (Elpis Biotech, Daejeon, Korea) with freshly added protease inhibitors (Roche, IN, USA), and phosphatase inhibitors (1 mM PMSF, 1 mM Na3VO4, 5 mM NaF). Protein was quanti ed using a bicinchoninic acid assay kit (Thermo Fisher Scienti c, IL, USA). 20-50 μg of proteins were resolved on a 10% SDS-PAGE gel or tris-tricine gel and transferred to nitrocellulose or polyvinylidene uoride membrane, followed by blocking with 5% skim milk.
Behavioral assays -Self-righting test was held on postnatal day 5-9 (P5-9) as described in previous literature [30]. Each mouse was placed in the supine position and gently held with all four limbs extended outwards at which time it was released. Time taken to right was recorded by the latency for all four paws touching the surface. A maximum score of 30 sec was recorded when the mouse failed to right in that period.
-Maternal scent preference test was conducted on P14 as described in previous research [31]. Each pup was moved from home cage to a fresh transparent polycarbonate cage (20 × 30 × 15 cm). The left third of the test cage was lled to a depth of 3 cm with litter from the mother's cage, the center third contained clean litter, and the right third contained litter from the cage of a stranger dam. The placement of the test litters (mother and stranger) was alternated across subjects to control for any side preferences. Three 1 min trials, with inter-trial intervals of 10 sec, were administered for each pup. For the rst trial, pups were placed in the center of the fresh litter facing the back wall of the test cage. For the second trial, pups were placed in the center of the fresh litter facing the section containing its mother's cage litter. For the third trial, pups faced the section containing the litter of the stranger dam. Time spent in each section of the cage was recorded and averaged across the 3 trials. The pup was considered to be inside a section when all four paws were touching the litter within the speci ed region.
-Social interaction was assayed using the 3 chamber test. The apparatus was constructed of a Plexiglas box (60×45×22 cm) partitioned into 3 chambers with retractable doorways. Openings between the compartments allowed the animals access all three chambers. In the rst phase, a mouse was placed in the center chamber and was allowed to freely explore with an age-matched male (familiar) for 10 min. In the second phase, the test mouse was gently guided to the center chamber, and the entrances were blocked. An age-matched stranger mouse then was placed in apposite chamber from the familiar mouse, and then the test mouse was similarly allowed to explore with the familiar and stranger mouse for an additional 10 min. The apparatus was cleaned with 70% ethanol between trials.
-Self grooming was performed at 9-10 weeks of age, and each mouse was placed individually into a clean transparent polycarbonate cage (20 × 30 × 15 cm) with a video camera placed 15 cm away from the cage. The duration of the test was 10 min after 10 min habituation. The time spent grooming was measured.
-Rotarod test was conducted with mice (9-10 weeks) following previously published research [17]. The test consisted of three trials per day over the course of 3 days. Rotarods were accelerated from 4-40 rpm in 300 s. Each trial ended when a mouse fell off, made one complete backward revolution while hanging on, or reached 300 sec.
-Open eld test was conducted at 9-10 weeks of age. A square plastic box (100 cm × 100 cm × 40 cm) was used for this general locomotor activity test. The mice were put into the arena and its movements monitored with a video camera for 30 min. Tracking of mouse behavior was done using EthoVision XT (Noldus) tracking system. The open eld was thoroughly cleaned with 70% alcohol between test animals.
Immuno uorescence and image analysis Mice were anesthetized using Zoletil/Xylazine and perfused using heparin (100 U/ml) phosphate buffered saline (PBS) solution. The brains from 9 to 10-week-old prenatally VPA-or saline-exposed mice were then removed and post-xed in 4% paraformaldehyde (PFA) at 4 °C for 24 h before they were transferred to 30% sucrose-PBS 0.1 M, pH 7.3 solution at 4 °C. Afterwards, the brains were sectioned into 30 μm-thick coronal sections using a cryostat (Thermo Fisher Scienti c, IL, USA), and three slices per mouse were used in all IF analyses (n=3-4 mice/staining). The brain slices were incubated in 10 mM sodium citrate buffer (pH 6.0) for 10 min at 95°C for antigen retrieval, and blocked in PBS containing 2% BSA or 10% serum and 0.3% Triton X-100 for 1 h at room temperature (RT). Sections were then incubated in blocking buffer containing a primary antibodies diluted in blocking buffer at 4°C for overnight. Next, sections were incubated with the secondary antibodiesin PBS for 2-3 h at RT protected from light. Finally, sections were stained with Topro3 (diluted 1:1000; Thermo Fisher Scienti c, IL, USA) or DAPI (1:1000, D1306, Thermo Fisher) in PBS. After nal rinsing, sections were mounted and cover-slipped using mounting medium (#345789, Merck, Darmstadt, Germany). The images were acquired on an LSM510 confocal microscope (Zeiss, Oberkochen, Germany) using a Plan-Neo uar 40x/0.90 N.A. with a water immersion objective or on a Nikon A1 confocal microscope (Nikon, Melville, NY, USA) with a Plan uor 20 × lens (0.75 numerical aperture). For quanti cation, 2-3 striatal regions were randomly selected for confocal imaging, wherein the intensity of each region was analyzed. The primary antibodies used were Nurr1 (#PA5-13416, Thermo Fisher Scienti c, IL, USA), NeuN (#MAB377, Millipore, CA, USA), Iba-1 (#NB100-1028, Novusbio, CO, USA). Secondary antibodies used were goat anti-rabbit Alexa 555, goat anti-mouse Alexa 488, goat anti-rabbit Alexa 488, and donkey anti-goat Alexa 555 (Thermo Fisher Scienti c, IL, USA).
Statistical analysis The data are expressed as means ± SEM values and were analyzed with the SPSS 23 software (IBM, Chicago, IL, USA) using the Kruskal-Wallis test, one-way-ANOVA with LSD post-hoc analysis, two-way-ANOVA with LSD post-hoc analysis, or repeated measures (RM)-ANOVA with Bonferroni post-hoc analysis. The results were considered to be statistically signi cant if p < 0.05. n means a number of mice analyzed unless stated otherwise.

Results
Abnormalities in the density and morphology of dendritic spines in the striatum of prenatally VPAexposed 10-week-old mice Morphological abnormality of the striatum has been one of the most consistent abnormalities reported in ASD [32,33]. Spine number and morphology can act as a marker of synaptic plasticity. To analyze the effects of VPA on dendritic structure in the striatum, Golgi staining was performed with the brain tissues of 10-week-old mice. Notably, VPA mice had a signi cantly reduced spine density compared to SAL mice in DMS (p = 0.036), not in DLS (Fig. 1a). Fig. 1b shows representative images taken from Golgi-stained neurons in the DMS and DLS. Next we investigated the morphology of the spines. Dendritic spines are classi ed by their length and the width of the spine head as lopodia, thin, stubby, mushroom and branch type [34]. The lopodia, and thin types were categorized as immature spines, and the stubby, mushroom, and branch types were categorized as mature spines. The mature spine density was signi cantly decreased (mature, p = 0.033) (Fig. 1c). In contrast, the immature spine density in DMS of VPA mice was not different as compared to that of SAL mice (Fig. 1d).
The principal neuron type in the striatum is the medium spiny neurons expressing dopamine receptors. We further analyzed differential expression of D1 or D2 to examine whether the prenatal VPA exposure is involved in dysregulation of dopamine receptor expression. Expression level of D2, which could indicate the population of D2 medium spiny neurons, was increased in the striatal tissues of prenatally VPA-exposed mice (p = 0.038) (Fig. 1e). D1 expression level was decreased (p = 0.032) (Fig.  1f).
Main inputs to the striatum are glutamatergic signal from cerebral cortex and thalamus and dopaminergic signal from substantia nigra / ventral tegmental area (SN/VTA). Vesicular glutamate transporter 1 (VGLUT1) and VGLUT2 are localized to the excitatory nerve terminals projecting from the cortical and the thalamic area, respectively, in the striatum [35]. Expression of VGLUT1, which represents corticostriatal terminals in the striatum, was found to be decreased in the striatum of VPA mice (p = 0.017) (Fig. 1g). Since the terminal marker for the input from cortex to striatum (VGLUT1) was found to be decreased, we performed the dendritic spine counting with the prefrontal cortex sections. Prefrontal cortex has been well known to be implicated in a social recognition. It was observed that VPA mice exhibited lower total and mature spine density (total, p = 0.016, mature, p = 0.00001) in the prefrontal cortex compared to SAL mice (Fig. S1). Expression of VGLUT2, which represents thalamostriatal terminals in the striatum, was found to be decreased in the striatum of VPA mice (p = 0.042) (Fig. 1h). Expression of DAT, which is the marker for nigrostriatal dopaminergic nerve terminal, was found to be decreased in the striatum of VPA mice (p = 0.021) (Fig. 1i). The dysfunction of glutamatergic and dopaminergic terminals in the striatum may cause reduction in spine density in DMS of VPA mice.
Striatal transcriptome analysis of prenatally VPA-exposed 10-week-old mice To identify transcriptome pro les in the striatum of prenatally VPA or saline-exposed mice, RNA-Seq with the striatal mRNA from 10-week-old VPA mice was performed. Speci cally, 348 genes were upregulated, and 258 genes were downregulated in the striatum of the VPA mice compared to that of SAL mice (Fig.  2a). Top 10 differentially expressed genes are listed in Fig. 2b.
Next, we analyzed gene ontology (GO) of genes that had p-value less than 0.05 to identify molecular and physiological signature. Interestingly, upregulated genes were related to "ion channel activity" (potassium channel and cation channel) and "chemical synaptic transmission" which are important to excitability of neurons, and "axogenesis". Analyzing GO term by cellular component showed that upregulated genes are expressed mostly neuron speci c part like "dendrite membrane", "axon", "dendrite", and "node of Ranvier". Upregulated genes for enrichment of mouse phenotype terms were also tested. Two ASD related phenotypes ("abnormal anxiety related response" and "excessive scratching") were enriched. In the additional GO term analysis with downregulated genes, "protein targeting to endoplasmic reticulum or membrane", "biosynthesis", "mRNA catabolic process" by biological process, "binding of platelet-derived growth factor, actin, tropomyosin, and calcium ion", and "neuropeptide hormone activity" by molecular function, and "cytosolic ribosome" by cellular component were related. Downregulated genes showed relevance with bone development-related ("exostosis", "abnormal ischium morphology", "abnormal pubis morphology", and "absent common crus") mouse phenotype (Fig. 2c, d).
Since upregulated gene sets displayed more relevance with neuronal function and ASD symptoms than downregulated gene sets, we speci cally focused on upregulated genes. In order to verify those genes, top 10 upregulated genes were analyzed using qPCR with striatal mRNA of 10-week-old mice. Nurr1 mRNA (p = 0.007) and Cbln1 mRNA (p = 0.014) expression levels were found to be increased by qPCR (Fig. 2e, f). Hspa1a mRNA level was found to be increased by RNA-Seq, but showed only tendency to increase by qPCR (Fig. 2g).
To demonstrate the changes in protein expression, we performed a western blot analysis on the two genes which were con rmed to be increased with RNA-Seq as well as qPCR. We identi ed that only the Nurr1 protein level was signi cantly increased in the striatum of 10-week-old VPA mice (p = 0.047) (Fig.  2h-i).
Effects of AQ, an Nurr1 agonist, on molecular changes in primary striatal neuron cultures Next, we tested whether Nurr1 activation induce the molecular changes in the primary striatal neuron cultures, which was shown in the striatum of VPA mice. To nd an optimal concentration that doesn't induce cytotoxicity, the cell viabilities were evaluated after treatment of AQ at various concentrations (10 nM-10 μM) for 24 h by performing an MTT assay. Treatment of concentration over 500 nM AQ was signi cantly cytotoxic to primary striatal neurons (VPA; p = 0.003) (Fig. S2a, b). Based on these results, primary striatal neuron cultures from SAL mice or VPA mice were treated with 100 nM of AQ for following experiments. D2 expression level was increased by 100 nM of AQ in primary striatal neuron cultures of SAL mice, not in cultures of VPA mice (p = 0.043) (Fig. S2c, d).

Administration of AQ induces ASD-like behaviors in 10-week-old mice
In order to characterize the effect of Nurr1 activation on behavior, SAL mice were injected with 20 mg/kg of AQ for 14 days and subjected to a battery of behavior tests. The body weight and the brain weight of the AQ-injected SAL mice was similar to that of the vehicle-injected SAL mice (Fig. 3a-c). To examine the role of AQ in ASD-like behaviors, social behavioral abnormality and repetitive behavior were investigated by three chamber and self-grooming tests, respectively. The AQ-injected SAL mice displayed no difference in time spent in familiar zone and stranger zone (SAL Veh; p = 0.00005, SAL AQ; p = 0.317) (Fig. 3d). The AQ-injected SAL mice showed no signi cant alterations in repetitive grooming behavior (Fig. 3e). The rotarod test was performed to evaluate motor coordination and accelerated motor learning. The performances of the AQ-injected SAL mice on trials 1, 2, and 3 were signi cantly worse than those of the vehicle-injected SAL mice (T1; p = 0.04, T2; p = 0.014, T3; p = 0.028) (Fig. S3a). During open eld test, the AQ-injected SAL mice showed comparable motor function with vehicle group (Fig. S3b). Collectively, these results indicate that Nurr1 affects sociability, rather than repetitive behavior.
The synaptic and molecular changes in the striatum of AQ-injected 10-week-old mice To elucidate the underlying functional mechanisms responsible for the decreased social activity in AQinjected SAL mice, the dendritic spine density in DMS was measured. The density of mature spines was decreased in DMS of AQ-injected SAL mice (p = 0.007) (Fig. 4c). The density of total spines and immature spines weren't signi cantly different (Fig. 4a, d). Fig. 4b displays representative images taken from Golgi-stained neurons in the DMS.
To determine the molecular change related to medium spiny neurons, expression of GAD67, D2, and D1 were examined in the SAL and VPA mice injected with vehicle or AQ. GAD67 expression level was signi cantly increased by AQ injection in SAL mice (p = 0.021) (Fig. 4e). D2 and D1 expression level and the relative expression level of D2 compared to D1 (p = 0.095) was not signi cantly different (Fig. 4f-h).
Lentiviral Nurr1 knockdown in the striatum rescues autism-like social de cits in prenatally VPA-exposed 10-week-old mice To examine the therapeutic potential of Nurr1 knockdown, we stereotaxically injected lentiviruses expressing shRNA targeting Nurr1 into striatum and assessed the resulting behavioral consequences.
Given that 84.51 ± 2.33% were neurons and 8.02 ± 1.08% were microglia (p = 1.43 x 10 -14 ) among Nurr1 expressing cells (Fig. S4), we designed the vector expressing shRNA under Syn promoter. Astrocytes were found to reside only in or around the striosomes, and they do not colocalize with Nurr1 positive cells in striatum (data not shown). The knockdown of Nurr1 was con rmed in viral-infected striatum tissue from VPA mice (p = 4.75 x 10 -8 ) (Fig. 5a-c). Effect of shRNA was insigni cant since the basal intensity was so low in SAL mice. In 3-chamber test, VPA sh-Nurr1 mice exhibited the signi cantly increased social interaction time, compared to VPA sh-sc mice (Fig. 5d, SAL sh-sc, p = 0.020; SAL sh-Nurr1, p = 0.014; VPA sh-sc, p = 0.205; VPA sh-Nurr1; p = 0.039). Taken together, these results indicated that Nurr1 knockdown could rescue the social de cits in VPA mice.
To investigate the effects of Nurr1 knockdown on dendritic structure in DMS, we analyzed the spine density, nding that total (p = 0.007) and mature (p = 0.00003) spine densities were signi cantly increased in DMS of VPA sh-Nurr1 mice. In contrast, the immature spine density in DMS of VPA sh-Nurr1 mice was not different as compared to that of VPA sh-sc mice (Fig. 6a-d). Additionally, we performed a western blot analysis, wherein D1 (p = 0.04) and DAT (p = 0.009) expression levels were found to be signi cantly increased in the striatum of VPA sh-Nurr1 mice as compared to that of VPA sh-sc mice (Fig.   6e, f). D2 expression level was not signi cantly different (Fig. 6g). However, the relative expression level of D2 compared to D1 was signi cantly increased (p = 0.023) (Fig. 6h).

Nurr1 expression is also increased in the striatum of PatDp +/mice, a genetic animal model of ASD
To probe the importance of Nurr1 in pathophysiology of ASD, we also assessed the Nurr1 expression in a commonly used ASD genetic mouse model, PatDp +/mice that carries a 6.3 Mb paternal duplication homologous to the human 15q11-q13 locus. Chromosomal abnormalities in this region are known to cause ASD, Prader-Willi syndrome, and Angelman syndrome in humans [36]. Interestingly, Nurr1 expression level was also increased in the striatal tissues from PatDp +/mice (p = 0.008) (Fig. S5).

Discussion
MRI study with high-functioning ASD subjects aged between 6 years and 25 years revealed that the volume of caudate increased with development [37]. The growth rate of striatal structures for individuals with ASD increased compared to control subjects in a longitudinal MRI research. This effect was speci c to caudate nucleus and correlated with the insistence on same cluster of repetitive behavior at the preschool age [38]. Striatal functional connectivity is also aberrant in ASD patients. A resting-state positron emission tomography study detected weaker correlations in glucose consumption between the frontal cortical regions and the striatum in young adults with ASD [39]. However, a resting-state functional MRI implicated the functional connectivity between the striatum and the associative and limbic cortex increased in children with ASD [40]. These suggest the connectivity between the striatum and other brain regions is differed in ASD patients and is related to autistic behavioral phenotype. Changes in the dorsal striatum input has been reported to be involved in promoting sociability de cits and repetitive behaviors [14]. It has also been reported that glutamatergic innervation from the neocortex and the thalamus is known to modulate dendritic morphology in medium spiny neurons [41]. Moreover, the cortical and thalamic glutamatergic input also potentiates the output of neurotransmission of the striatal GABAergic neurons [42]. In fact, human functional MRI experiments have shown that the striatum becomes active in relation to other's reward situations and during social learning [11]. Corticostriatal synapses in strisomal neurons were reduced in the striatum of prenatally VPA-exposed mice [43]. Mice with autism-linked mutation in the DAT, a presynaptic transporter for DA reuptake from synaptic cleft, which is located at the membranes of dopaminergic nerve terminals, exhibited repetitive behaviors and de cits in social interaction [44]. Based on these reports, we checked the synaptic alteration in the striatum and input signals to the striatum. Dendritic spine density was decreased in prenatally VPAexposed striatum and this effect was speci c to mature type of spines. Furthermore, glutamatergic and dopaminergic inputs to the striatum were decreased in prenatally VPA-exposed mice (Fig. 1). Although several reports used VGLUT1 and VGLUT2 as presynaptic markers for the corticostriatal and thalamostriatal circuits and DAT as a presynaptic marker for dopaminergic inputs to the striatum, our data could not be su cient to interpret the exact signi cance of alterations in synaptic innervation into the striatum in the contribution to behavioral phenotypes of VPA mice. Further research seems to be needed regarding the contribution of the alterations in striatal circuit in VPA-induced AS phenotypes in near future. In our behavioral studies, VPA mice showed a developmental delay (Fig. S6a, g, h), an impairment of social interaction (Fig. S6b-c), a tendency of increased grooming (Fig. S6d), a signi cantly increased repetitive motor routine learning rate (Fig. S6e), and a signi cantly decreased motor function (Fig. S6h). Taken together, our ndings indicate that altered synaptic plasticity and altered glutamatergic and dopaminergic inputs in the striatum could modulate the striatal neurotransmission output. This could be relevant regarding the behavioral phenotypes in prenatally VPA-exposed mice.
In this study, our RNA-seq analyses identi ed the upregulation of synaptic and neuronal function related genes and the downregulation of genes related to binding and process of protein, and bone development.
Upon validating the top 10 differentially upregulated genes using qPCR, only two genes were upregulated. This was because we made the list of differentially upregulated genes following the order of q-value, not fold change. In our results, prenatally VPA-exposed mice show increased Nurr1 mRNA and protein levels in the striatum (Fig. 2). Nurr1 (Nr4a2), Nur77 (Nr4a1, NGFI-B), and Nor-1 (Nr4a3) are orphan nuclear receptors and conform Nur subfamily. Nurr1 is expressed exclusively in brain tissue, unlike Nur77 and Nor-1. Several lines of evidence have indicated that Nurr1 is important in the development and differentiation of dopaminergic neurons, neurogenesis, and learning and memory. Nurr1 is rst expressed at E10.5 in the mouse [45], and interacts with Pitx-3 inducing the differentiation of dopaminergic precursor cells to tyrosine hydroxylase positive dopaminergic neurons [46].
Dysregulation of Nurr1 expression was reported in some neurodevelopmental diseases. Recently, it has been reported that Nurr1 is also involved in the pathogenesis of schizophrenia by regulating the expression of Cnr1 which codes for the cannabinoid receptor 1, which is involved in brain functions such as emotional responses, motivated behavior, cognitive processing and motor control [47]. In addition, comparing genome sequencing data and exome sequencing data from ASD families, a frameshift variant in Nurr1 was discovered [22].  [49]. Thus, our ndings support the hypothesis that Nurr1 dysregulation has a role in neuropsychiatric disorders.
A Nurr1 agonist, AQ, is an anti-malaria drug that stimulates the transcriptional function of Nurr1 through physical interaction with its ligand binding domain [29].
Recently several studies reported that the pharmacological stimulation of Nurr1 using AQ improved cognitive function via enhancement of hippocampal neurogenesis [50,51]. Interestingly, AQ-injected SAL mice displayed dysfunctions in social interaction, whereas the injection did not affect repetitive behavior or motor function signi cantly (Fig. 3).
In humans, activity in the striatal circuits has been correlated with social de cits relevant to autism [13,52]. In our results, AQ-injected SAL mice showed decreased dendritic spine density in DMS and this effect was speci c to mature type of spines. In addition, the administration of AQ increased the expression of GAD 67, GABAergic neuron marker (Fig. 4). Furthermore, it has been reported that striatal D2 overexpression leads to a de cit in inhibitory transmission and dopamine sensitivity [53], suggesting that the molecular changes in AQ-injected SAL mice striatum could contribute to prefrontal cortex GABAergic system hypofunction and to reduced social novelty.
To further illustrate the roles of Nurr1 in social behavior, we performed intrastriatal injections of Nurr1-shRNA expressing lentiviruses. As depicted in Fig. 5, VPA mice injected with sh-Nurr1 lentiviruses showed reduced expression of Nurr1 in striatal neurons and rescued social interaction. Next, we con rmed that VPA mice injected with sh-Nurr1 lentiviruses exhibited increased total and mature spine density in DMS.
In addition, the expression levels of DAT and D1 were increased in the striatum of VPA mice injected with sh-Nurr1 lentiviruses. We also found that the relative expression level of D2 compared to D1 was signi cantly decreased. As well known, medium spiny neurons are composed of D1 and D2 expressing GABAergic neurons. D1s (exciatatory) are predominantly expressed on GABAergic medium spiny neurons in the dorsal striatum as part of the "direct pathway" to globus pallidus interna (GPi) and substantia nigra pars reticularis (SNpr) whereas D2s (inhibitory) are predominantly expressed on medium spiny neurons that primarily project to globus pallidus externa (GPe). A recent study with the post-mortem brains of ASD reported signi cant increases in D2 mRNA within medium spiny neurons in both the caudate and putamen, in correlation with our results with lentivirus injection shown in Fig. 6. These results indicate alterations in the indirect pathway of the basal ganglia, with possible implications for the E/I balance in the direct/indirect feedback pathways through thalamic and motor cortical areas [54]. Our study is the rst study that displays Nurr1 involved in the pathogenesis of ASD, indicating that the upregulation of Nurr1 expression in the striatum from VPA mice is the cause of the altered dendritic spine density of mature forms and may contribute to the autistic behavior. These ndings imply that the regulation of Nurr1 could be a therapeutic strategy for ASD.

Limitations
In this study, the relevance of the increase in Nurr1 expression with synaptic and behavioral de cits in prenatally VPA-exposed mice was investigated. To understand how Nurr1 regulates behaviors and synaptic plasticity in striatum, future studies on the changes in striatum-related brain circuits are needed.

Conclusions
Transcriptional analysis and its validation showed increased expression of Nurr1 in the striatum of prenatally VPA-exposed mice. Here we describe a novel role of Nurr1 in synaptic and behavioral phenotypes of ASD. Nurr1 expression was also increased in another ASD mouse model, PatDp +/mice ( Fig. S4). Taken together, our ndings implicate Nurr1 as a potential therapeutic target of ASD.  level (SAL, n = 11; VPA, n = 11). Data are presented as mean ± SEM. *p < 0.05 compared to SAL mice, unpaired t-test. Figure 2 signi cantly upregulated and 258 signi cantly downregulated genes (Log2 fold change). (n = 3 / group) b List of top 10 upregulated and downregulated differentially expressed genes (based on the fold change) from the RNA-Seq analysis. c-d GO and mouse genome informatics mammalian phenotype analysis of the upregulated genes and downregulated genes. e-g mRNA expression of Nurr1 (SAL, n = 9;

Figure 6
Lentiviral Nurr1 knockdown in the striatum rescues the abnormalities in dendritic spine density and inputs in the striatum of prenatally VPA-exposed 10-week-old mice. a Quanti cation of total dendritic spine density. b Representative images of Golgi-Cox stained neurons in DMS from VPA sh-sc and VPA sh-Nurr1