저작자표시 - 비영리 - 변경금지 2.0 대한민국 이용자는아래의조건을따르는경우에한하여자유롭게 이저작물을복제, 배포, 전송, 전시, 공연및방송할수있습니다. 다음과같은조건을따라야합니다 : 저작자표시. 귀하는원저작자를표시하여야합니다. 비영리. 귀하는이저작물을영리목적으로이용할

저작자표시 - 비영리 - 변경금지 2.0 대한민국 이용자는아래의조건을따르는경우에한하여자유롭게 이저작물을복제, 배포, 전송, 전시, 공연및방송할수있습니다. 다음과같은조건을따라야합니다 : 저작자표시. 귀하는원저작자를표시하여야합니다. 비영리. 귀하는이저작물을영리목적으로이용할수없습니다. 변경금지. 귀하는이저작물을개작, 변형또는가공할수없습니다. 귀하는, 이저작물의재이용이나배포의경우, 이저작물에적용된이용허락조건을명확하게나타내어야합니다. 저작권자로부터별도의허가를받으면이러한조건들은적용되지않습니다. 저작권법에따른이용자의권리는위의내용에의하여영향을받지않습니다. 이것은이용허락규약 (Legal Code) 을이해하기쉽게요약한것입니다. Disclaimer

Brain activity and connectivity in rat model of autism induced by prenatal exposure to valproate Hojin Cho Department of Medical Science The Graduate School, Yonsei University

Brain activity and connectivity in rat model of autism induced by prenatal exposure to valproate Directed by Professor Chul Hoon Kim The Doctoral Dissertation submitted to the Department of Medical Science, the Graduate School of Yonsei University in partial fulfillment of the requirements for the degree of Doctor of Philosophy Hojin Cho December 2016

This certifies that the Doctoral Dissertation of Hojin Cho is approved. -------------------------------------------------------------- Thesis Supervisor: Chul Hoon Kim -------------------------------------------------------------- Thesis Committee Member#1: Dong Koo Kim -------------------------------------------------------------- Thesis Committee Member#2: Hae-Jeong Park -------------------------------------------------------------- Thesis Committee Member#3: Hosung Jung -------------------------------------------------------------- Thesis Committee Member#4: Eosu Kim The Graduate School Yonsei University December 2016

ACKNOWLEDGEMENTS 박사학위과정을돌이켜보면많은분들께서여러가지로많은도움을주셨기때문에이과정을잘마칠수있었다는생각이듭니다. 그동안미처전하지못한감사하다는마음을지면을통하여표현하기에는부족하지만, 도움을주셨던많은분들께이렇게짧게나마감사하다는말씀을드리고싶습니다. 학위기간동안글로표현하기어려울정도로많은가르침을주신김철훈교수님께감사드립니다. 언제나제가보지못하던것을볼수있게해주시고새로운생각을할수있도록많은도움을주신김동구교수님께감사드립니다. 학생때부터지금까지많은시간동안여러가지로이끌어주신박해정교수님께도감사드립니다. 논문을진행하는동안여러가지로조언해주시고도움을주신정호성교수님과김어수교수님께감사드립니다. 약리학교실에있는동안에많은가르침을주셨던김경환교수님, 안영수교수님, 이민구교수님, 박경수교수님, 김주영교수님, 김형범교수님께이자리를빌려감사의마음을드립니다. 전공의때부터지금까지많은도움을주시고이자리에있을수있도록이끌어주신이종두교수님, 윤미진교수님, 강원준교수님께감사드립니다. 약리학교실에있는동안에부족한저에게많은도움을주신선 후배님들과동료들에게감사드립니다. 처음실험실에들어왔을때부터많은도움을주었던헌영이형, 진세형, 같이실험실에들어와서많은힘이되었던홍인이형, 은석이, 형순이, 정환이, 그리고같은실험실에서많은시간을함께한실험실동료들, 특히

정호, 제호, 그리고조아련박사님께감사한마음을전하고싶습니다. 처음실험이라는세계를접할때많은도움을주신최민아박사님, 그리고바쁜와중에도많은시간을힘든내색없이도와주신오맹근선생님, 무리하면서까지많은도움을주었던혜원이에게도감사하다는말을전하고싶습니다. 그리고짧지않은시간동안함께하면서언제나든든한힘이되어준태윤이형, 준상이, 윤대, 호준이에게도평소에미처말하지못했지만항상감사하고있었다는마음을전하고싶습니다. 끝으로지금까지늘그랬듯이항상곁에서지켜주고흔들리지 않도록도와준가족에게감사한마음을전하고싶습니다. 다시한번모든분들께진심으로감사드립니다. 2016 년 12 월

TABLE OF CONTENTS ABSTRACT 1 I. INTRODUCTION 3 II. MATERIALS AND METHODS 7 1. Animals 7 2. Pup body weight 7 3. Three-chamber social approach test 7 4. 18 F-FDG PET acquisition 8 5. Image processing 9 6. Voxel-based statistical analyses 9 7. Node definition 9 8. Connectivity estimated by correlation 12 9. Connectivity estimated by sparse inverse covariance estimation 12 10. Statistical analysis 12 III. RESULTS 13 1. Pup body weight gain 13 2. Three-chamber social approach test 16 3. Injected 18 F-FDG activity 18 4. Voxel-based statistical analyses 20 5. Connectivity estimated by correlation 23 6. Connectivity estimated by sparse inverse covariance estimation 25 7. Node degree 28

8. Connectivity strength 31 IV. DISCUSSION 35 V. CONCLUSION 41 REFERENCES 42 ABSTRACT (IN KOREAN) 50 PUBLICATION LIST 52

LIST OF FIGURES Figure 1. Prenatal exposure to VPA result in reduced body weight in rat pups 15 Figure 2. Sociability or social preference for novelty was impaired by VPA treatment 17 Figure 3. Injected 18 F-FDG activity did not differ between all four subgroups 19 Figure 4. Statistical parametric mapping revealed altered brain activity by VPA treatment 21 Figure 5. Correlation analysis did not show any connectivity changes by VPA treatment 24 Figure 6. SICE identified connectivity changes by VPA treatment 26 Figure 7. Aberrant number of connections was found in VPA-treated female rats 30 Figure 8. VPA treatment results in significant changes in connectivity strength 33

LIST OF TABLES Table 1. Volume of interest for connectivity estimation 10 Table 2. Brain regions with significant metabolic differences in VPA-treated rats compared to control rats 22 Table 3. Number of connections in whole brain, within each hemisphere, and between hemispheres 27 Table 4. Changes in node degree in VPA-treated female rats 29 Table 5. Changes in connectivity strength by VPA treatment 32

ABSTRACT Brain activity and connectivity in rat model of autism induced by prenatal exposure to valproate Hojin Cho Department of Medical Science The Graduate School, Yonsei University (Directed by Professor Chul Hoon Kim) Autism spectrum disorder (ASD) is a severe lifelong neurodevelopmental disorder characterized by early-onset impairments in social communication and social interaction, and restrictive and repetitive behavior, interests, or activities. Currently no biological markers are available to make a reliable diagnosis of ASD. I used valproic acid (VPA) rat model to explore the possible alterations in metabolic brain activity and connectivity. Female Sprague Dawley rats were given a single intraperitoneal injection of sodium valproate or normal saline on embryonic day 12.5. To evaluate autistic-like behaviors in pups, social interaction was examined during postnatal weeks 4-6 using three-chamber social approach test. Resting-state 18 F-fluorodeoxyglucose (FDG) positron emission tomography (PET) scans were acquired at postnatal weeks 6 or 62. To assess the changes in metabolic brain activity by VPA treatment, Statistical Parametric Mapping (SPM) was performed. I also examined if change in metabolic brain connectivity was 1

accompanied with alterations of regional metabolic activity. Metabolic connectivity was modeled using both conventional correlation analysis, and sparse inverse covariance estimation (SICE), a partial correlation analysis. VPAtreated rats exhibited impairments in social behaviors, and this difference was more pronounced in male than female rats. Preference for social novelty was impaired in VPA-treated male rats, while sociability was diminished in VPAtreated female rats. I found that metabolic activity and connectivity was significantly changed by VPA treatment. Changes in metabolic connectivity was revealed by SICE while conventional correlation analysis did not show any difference. VPA-treated male rats had significantly decreased metabolic activity in the olfactory bulb, and had decreased metabolic connectivity between the left insular cortex and left amygdala, which constitute the salience network. There were no brain regions with decreased metabolic activity in VPA-treated female rats. In contrast, VPA-treated female rats had reduced metabolic connectivity between the thalamus and midbrain, and between the right medial prefrontal cortex and left caudoputamen. Such alterations in metabolic activity and connectivity may represent neurobiological substrates of autistic-like behavior, particularly in males, and may serve as a pathognomonic sign in VPA rat models of ASD. As such this study supports the idea that non-invasive brain imaging may serve as an imaging endophenotype that could aid diagnosis of ASD, classification of severity, and possibly reveal insights to neurobiological underpinnings in autistic-like behavior. Key words : autism spectrum disorder, rat, valproic acid, metabolic connectivity, fluorodeoxyglucose, positron emission tomography, sparse inverse covariance estimation 2

Brain activity and connectivity in rat model of autism induced by prenatal exposure to valproate Hojin Cho Department of Medical Science The Graduate School, Yonsei University (Directed by Professor Chul Hoon Kim) I. INTRODUCTION Autism spectrum disorder (ASD) is a severe lifelong neurodevelopmental disorder characterized by early-onset impairments in social communication and social interaction, and restrictive and repetitive behavior, interests, or activities. 1 ASD is one of the most common disorders among children. According to estimates released by the Centers for Disease Control and Prevention, ASD affects one in 68 children. ASD is five times more prevalent in boys than in girls. 2 Early diagnosis is essential for early intervention, which provides the opportunity for optimal development and greatly improves prognosis. 3,4 However, ASD comprises a highly heterogeneous set of disorders, which poses a considerable challenge to make an accurate diagnosis, especially in preverbal children. 5 The variable manifestations of ASD stem from the level of functioning and comorbidity, also renders the diagnosis difficult. 6 As there are no reliable biological markers for ASD, the diagnosis is made based on behavioral signs and 3

symptoms. There is no single known cause for ASD. Both the genetic and environmental factors are thought to play roles in the development of ASD. 7 Valproic acid (VPA) is one of the most commonly prescribed teratogenic drugs in women of child bearing potential. It is used for the treatment of epilepsy and other neuropsychological disorders. VPA has been shown to increase in brain concentration of gamma-aminobutyric acid (GABA), the major inhibitory neurotransmitter. 8 The mechanism of action of VPA has also been attributed to the direct inhibition of voltage-gated sodium channel and negative modulation of glutamatergic signaling. It was also demonstrated that VPA acts as a nonselective histone deacetylase (HDAC) inhibitor, and this contributes to the teratogenic effects of VPA. 9 Prenatal exposure to VPA is known to be associated with increased risk of ASD in human. 10,11 Fetal valproate exposure in rodent showed brain abnormalities similar to those found in patients with ASD: reduced in neuron number in the oculomotor, trigeminal, abducens, and hypoglossus nerves; and small cerebellum with decreased number of Purkinje cells. Rats prenatally exposed to VPA displayed reduced social interaction, increased repetitive/stereotyped behaviors, and early signs of neurodevelopmental delay. 12 The VPA rat model has been shown to have construct validity, face validity, and predictive validity. 13 However, it remains unclear how VPA exposure leads to neurobiological alterations responsible for the development of autistic-like behaviors. A growing body of evidence supports the notion that ASD is associated with altered brain connectivity. 14 Decreases in white matter integrity and impaired long-range connectivity are well known findings in humans with ASD. 15,16 However, the connectivity changes associated with such findings and their relation to the behavioral phenotypes of ASD are still uncertain. 17-19 Furthermore, findings from neuroimaging studies are largely inconsistent and even contradictory across studies. 18,20 This may be due to the heterogeneous nature of 4

subtypes and origins of the disorders studied. Thus, a well-controlled study investigating changes in neural circuitry in a homogeneous type of ASD would reveal a more concrete relation between autistic-like behaviors and neurobiological alterations. In addition to heterogeneity, a strong male predominance has been observed in ASD, which affects boys five times more frequently than girls. 2 This difference in prevalence between boys and girls is thought to arise from less pronounced repetitive and stereotyped behaviors in girls and dissimilar levels of functioning between boys and girls, or male biased diagnostic criteria, and social and cultural expectations. 21,22 Sex-specific differences are thought to potentiate risk in males and/or attenuate risk in females induced by genetic and environmental factors. 2 However, little attention has been paid to how ASD in males differs from that in females, particularly with how differences in neurobiology may underlie differences in behavior. Although VPA is known to affect rats in a sex-specific way, 23 no clear neurobiological mechanism underlying this difference has been identified. In the present study, I explored alterations in the brain connectivity of VPAtreated and control rats using resting-state 18 F-fluorodeoxyglucose (FDG) positron emission tomography (PET). Functional connectivity refers the temporal correlation of the activities among distinct brain regions. It is generally inferred from blood oxygenation level-dependent (BOLD) functional magnetic resonance imaging (fmri) or electro- or magnetoencephalogram (EEG/MEG). 24 In particular, BOLD fmri has received much attention after the discovery of spatially organized low-frequency spontaneous fluctuations in the absence of explicit task. 25,26 Recent investigations regarding brain circuitry have begun to utilize resting-state functional connectivity mapping with fmri in animals. 27 However, unlike studies involving humans, animal studies using resting-state fmri inevitably require sedation of animals, which limits exploration of restingstate functional connectivity. 28 In contrast, 18 F-FDG uptake reflects activity in the 5

awake state since animals are sedated after the completion of 18 F-FDG distribution during the resting state. Furthermore, regional 18 F-FDG uptake can be considered a more direct measure of neuronal activity BOLD signals in fmri, as neurons exhibit preferential uptake of glucose in an activity-dependent manner. 29 In contrast, the BOLD signal is an indirect measure of neural activity. It depends on the change in concentration of deoxyhemoglobin in the microvasculature, and is affected by cerebral blood flow (CBF), cerebral blood volume (CBV), and cerebral blood oxygen consumption (CMRO 2) in relation to neural activity. 30,31 Although 18 F-FDG PET images do not provide time series data for the conventional evaluation of functional connectivity, introduction of crosssectional analysis of 18 F-FDG PET imaging allows researchers to evaluate differences in both regional activity and connectivity between groups. As 18 F- FDG PET reveals neuronal activity at steady state in hourly terms, PET-based metabolic connectivity can be estimated by measuring regional variations of metabolic activity in the brain. 32,33 First I performed resting-state 18 F-FDG PET in VPA-treated rats and applied Statistical Parametric Mapping (SPM) to explore if there were any brain regions with altered metabolic activity in VPA-treated rats. I also examined if change in metabolic connectivity was accompanied with alterations of regional metabolic activity. I adopted a recently proposed analysis approach for estimating inverse covariance, called sparse inverse covariance estimation (SICE), a group-level partial correlation analysis of metabolic activity, which enables a more comprehensive understanding of brain connectivity, as compared to conventional intensity-based analysis. 34,35 I tested whether the metabolic connectivity estimated by both conventional correlation analysis and SICE can identify the altered brain connectivity in VPA rat model of ASD in relation to autistic-like behaviors. 6

II. MATERIALS AND METHODS 1. Animals Animal protocols were approved by the Yonsei University College of Medicine Institutional Animal Care and Use Committee (IACUC). Pregnant Sprague-Dawley (SD) female rats (Orient Bio Inc., Gyeonggi-do, South Korea) were randomly assigned to receive a single subcutaneous injection of sodium valproate (Sigma) dissolved in saline (400 mg/kg), or saline alone on embryonic day 12.5 (E12.5), as previously described. 36 Dams were housed individually and were allowed to raise their own litters. The rats were divided into two groups (adolescence and adulthood) based on the age at imaging. Litters in adolescent group were culled to four male and four female pups on postnatal day 2. Litters in adult group were not culled. Pups were weaned at postnatal day 21 (P21), and were housed with 2-3 rats in same-sex and same-treatment cages. Standard plastic laboratory cages were used with bedding and ad libitum access to food and water, handled twice a week. Animals were kept in a 12-hr light-dark schedule with lights on at 8:00 am, in rooms under controlled humidity and temperature. 2. Pup body weight Body weight gain of the pup was measured weekly for 5 wk for adolescent group (postnatal week (PNW) 1-5), and for 31 wk for adult group (PNW 2-32). I used a mixed effect model to analyze the effect of VPA treatment on pup body weight over time. The model was built using R (version 3.2.2) and the lmer() function in the lme4 package (version 1.1-10). 3. Three-chamber social approach test Three-chamber social approach test was performed as described previously with minor modifications. 37 The test consisted of three phases. In the first phase (habituation), a test rat was placed in the center of the empty three-chamber apparatus with two small wire cages in the left or right chamber 7

to habituate for 5 min. In the second phase (sociability test), an age- and sexmatched rat that had never been exposed to the test rat, was placed in one of the two wire cages. The empty wire cage was presented as a novel object. Then the two entrances were opened to allow the test rat in the center to freely explore each of the three chambers for 10 min. In the third phase (preference for social novelty test), the test rat was gently guided to the center chamber, and another age- and sex-matched novel rat was placed in the empty wire cage. Then the two entrances were opened to allow the test rat in the center to freely explore for another 10 min. All tests were performed between 9 AM and 6 PM. Time spent in each chamber was measured by the video tracking software (EthoVision XT 11.5, Noldus Information Technology, Wageningen, The Netherlands). In adolescent group, the test was performed on PNW4, and adult group test was performed on PNW6. 18 4. F-FDG PET acquisition PET scans were performed on PNW6 in adolescent group, and on PNW62 in adult group. All rats were deprived of food for 12-18 hr before the scanning, but had access to water at all times. The rats were injected intraperitoneally with 37 MBq of 18 F-FDG under light isoflurane anesthesia. Then the rats were moved back to the plastic cages, and woke up immediately from anesthesia. The cage was placed in a quiet room with dim light for 50 min. PET scans were acquired on Siemens Inveon scanner (Siemens, Knoxville, TN, USA). Static emission scan was started at 60 min after injection under isoflurane anesthesia. Emission scan was acquired for 20 min, followed by 20 min of transmission scan. All PET scans were performed between 9 AM and 6 PM. The images were reconstructed using the ordered subsets expectation maximization (OSEM) algorithm with attenuation, scatter, and random correction. The voxel size was 0.776 0.776 0.796 mm. 8

5. Image processing Spatial processing was performed using Statistical Parametric Mapping (SPM8, Wellcome Department of Cognitive Neurology, London, UK). All reconstructed images were spatially normalized to the 18 F-FDG rat brain template (PMOD 3.6, PMOD Technologies Ltd, Zürich, Switzerland). As the core symptoms of ASD and comorbidities tend to persist throughout the lifetime, 38,39 I assumed that if there are distinct features in the brain which represent autistic-like behavior, the features do not vary with age as long as the symptoms persist. Therefore, I did not consider the effect of aging in further imaging analysis. 6. Voxel-based statistical analyses Voxel-based statistical analyses were performed with SPM. Spatially normalized images were smoothed with a 1.6 1.6 1.6 mm 3 full-width-halfmaximum (FWHM) Gaussian kernel. Proportional scaling was used for global normalization. For statistical inferences, an P-value < 0.05 (family-wise error (FWE) corrected) with an extent threshold of 50 continuous voxels was considered significant. 7. Node definition The images were segmented using predefined volume of interests (VOIs) in PMOD software. I chose 58 anatomical VOIs that served as nodes throughout the brain (Table 1). 18 F-FDG uptake in the VOIs was normalized to global brain activity. 9

Table 1. Volume of interest for connectivity estimation Name Motor cortex left Somatosensory cortex left Auditory cortex left Visual cortex left Cingulate cortex left Frontal association cortex left Medial prefrontal cortex left Orbitofrontal cortex left Insular cortex left Retrosplenial cortex left Parietal association cortex left Olfactory bulb left Anterodorsal hippocampus left Posteroventral hippocampus left Entorhinal cortex left Amygdala left Caudoputamen left Nucleus accumbens left Septum left Thalamus left Hypothalamus left Superior colliculus left Inferior colliculus left Midbrain left Periaqueductal gray left Ventral tegmental area left Pons left Medulla left 10 Abbreviation MO-L SS-L AUD-L VIS-L CG-L FA-L MPF-L OF-L INS-L RSP-L PA-L OB-L HA-L HP-L ENT-L AMG-L CP-L ACB-L S-L TH-L HY-L SC-L IC-L MB-L PAG-L VTA-L P-L M-L

Cerebellum left Motor cortex right Somatosensory cortex right Auditory cortex right Visual cortex right Cingulate cortex right Frontal association cortex right Medial prefrontal cortex right Orbitofrontal cortex right Insular cortex right Retrosplenial cortex right Parietal association cortex right Olfactory bulb right Anterodorsal hippocampus right Posteroventral hippocampus right Entorhinal cortex right Amygdala right Caudoputamen right Nucleus accumbens right Septum right Thalamus right Hypothalamus right Superior colliculus right Inferior colliculus right Midbrain right Periaqueductal gray right Ventral tegmental area right Pons right Medulla right Cerebellum right CB-L MO-R SS-R AUD-R VIS-R CG-R FA-R MPF-R OF-R INS-R RSP-R PA-R OB-R HA-R HP-R ENT-R AMG-R CP-R ACB-R S-R TH-R HY-R SC-R IC-R MB-R PAG-R VTA-R P-R M-R CB-R 11

8. Connectivity estimated by correlation For each pair of VOIs, Pearson s correlation coefficient was calculated in an inter-subject manner. Then, correlation matrix was generated for each group, VPA-treated and control rats in each sex. Correlation matrices for each group were transformed to Z scores using Fisher transformation. Permutation test was performed to test statistical differences between groups, followed by Fisher transformation. The subject labels for each group were permuted 10,000 times. Significant differences were determined by applying a false-discovery rate (FDR) threshold of 0.05. 9. Connectivity estimated by sparse inverse covariance estimation The SICE was used to estimate metabolic connectivity structure, and stability approach to regularization selection (StARS) was applied to determine the optimal network. 34,35 I adopted a constrained optimization algorithm to estimate the strength of connections in the graph obtained by SICE. 40 To test statistical differences between VPA-treated and control rats, I obtained a null distribution by permuting subject labels and the optimal network was reestimated for each permuted groups. The entire subject labels were permuted 10,000 times. A FDR threshold of 0.05 was used as significance cutoff. I compared the sparsity and connectivity strength in the network between groups. 10. Statistical analysis Statistical analyses were performed using R (version 3.2.2). Unless otherwise specified, data were analyzed with two-tailed unpaired Student s t- test or Mann-Whitney U test as appropriate. I also used one-way analysis of variance (ANOVA) with injected 18 F-FDG activity as a between-group factor. The null hypothesis was rejected for alpha greater than 5%. Details on statistical method applied in imaging analyses are described in each section. 12

III. RESULTS 1. Pup body weight gain Rats treated with VPA had lower body weight both in adolescent and adult groups. I found a statistically significant effect of VPA treatment both in adolescent (t-value = -6.71, P < 0.001) and adult (t-value = -2.97, P = 0.004) groups. Post hoc analysis using two-tailed unpaired Student s t-test or Mann- Whitney U test showed lower body weight in VPA-treated rats on PNWs from 1 to 5 in adolescent group in both sexes (n=8 except for VPA-treated male rats (n=7)) (Figure 1A). In adult group, lower body weight was seen in VPA-treated male rats (n=15) compared to control male rats (n=9) on PNWs 12, 21, and 28. VPA-treated female rats (n=11) had lower body weight compared to control female rats (n=11) on PNWs 11, 12, 19, 20, 22-24, 27, 31, and 32 (Figure 1B). 13

Figure 1. Prenatal exposure to VPA result in reduced body weight in rat pups. (A) Weight gain in adolescent group. (B) Weight gain in adult group. There was a significant effect of VPA treatment on body weight. VPA-treated male (n=7, adolescent VPA-treated male; n=15, adult VPA-treated male) and female rats (n=8, adolescent VPA-treated female; n=11, adult VPA-treated female) were lighter than controls (n=8, adolescent control male; n=9, adult control male; n=8, 14

adolescent control female; n=11, adult control female) both in adolescent and adult groups. Female rats were lighter than male rats. Data shown as mean and standard error of the mean (SEM). In each time point, data were analyzed with two-tailed unpaired Student s t-test or Mann-Whitney U test as appropriate. VPA, valproic acid; *P < 0.05, **P < 0.01, ***P < 0.001. 15

2. Three-chamber social approach test VPA treatment found to have no significant effect in sociability test. Both in adolescent and adult group, all rats preferred to explore the first stranger (S1) over an empty wire cage presented as a novel object (O), irrespective of VPA treatment or sex (Figures 2A, B). I found that VPA-treated male rats displayed reduced social preference for novelty. In adolescent group, there was a trend that VPA-treated male rats spent less time with the second stranger (S2) compared to controls, however VPA-treated male rats still spent significantly more time with the S2 than with the S1, as controls do (Figure 2C). In adult group, control male rats preferred to spend more time with the S2 over S1. In contrast, the preference to the S2 over S1 was absent in VPA-treated male rats (Figure 2D). VPA-treated female rats in adolescent group showed preference to the S1 over S2, while control female rats did not show preference to the S1 (Figure 2C). As all adult female rats had preference to the S2 over S1, there was no significant effect of VPA treatment (Figure 2D). In brief, although VPAtreated adolescent male rats maintained preference for the S2 over S1, VPAtreated male rats showed autistic-like behaviors in general. 16

Figure 2. Sociability or social preference for novelty was impaired by VPA treatment. (A) Sociability test in adolescent group (n=8, adolescent control male; n=7, adolescent VPA-treated female; n=8, adolescent control female; n=8, adolescent VPA-treated female). (B) Sociability test in adult group (n=9, adult control male; n=15, adult VPA-treated male; n=11, adult control female; n=11, adult VPA-treated female). (C) Social preference for novelty test in adolescent group. (D) Social preference for novelty test in adult group. VPA treatment had no effect on sociability test. VPA-treated adult male rats lost preference to the S2 over S1. VPA-treated female rats spent more time with the S1 over S2. Data represent mean ± SEM. Data were analyzed with two-tailed unpaired Student s t-test or Mann-Whitney U test as appropriate. VPA, valproic acid; O, object; S1, stranger 1; S2, stranger 2; SEM, standard error of the mean. *P < 0.05, **P < 0.01, ***P < 0.001. 17

3. Injected 18 F-FDG activity Injected 18 F-FDG activity in control males (n=17, 40.52 ± 0.40 MBq), VPA-treated males (n=22, 40.12 ± 0.34 MBq), control females (n=19, 40.48 ± 0.35 MBq), and VPA-treated females (n=19, 40.48 ± 0.31 MBq) were not differed across groups. All values are mean ± standard error of the mean (SEM). 18

Figure 3. Injected 18 F-FDG activity did not differ between all four subgroups. Injected dose in control males (n=17), VPA-treated males (n=22), control females (n=19), and VPA-treated females (n=19) were not different across groups. Data represent mean ± SEM. Data were analyzed with one-way ANOVA. VPA, valproic acid; SEM, standard error of the mean; ANOVA, analysis of variance. 19

4. Voxel-based statistical analyses VPA-treated male rats (n=22) had reduced metabolism in the olfactory bulb compared to control male rats (n=17) at P < 0.05 corrected for familywise error (FWE) (Figure 4A). The metabolism in the bilateral somatosensory cortices, thalami, and cerebellum was increased in VPA-treated male rats than control male rats (Figure 4B). In VPA-treated female rats (n=19), no region was decreased in metabolism compared to control female rats (n=19). Increased metabolism was observed in the left caudoputamen and left medulla in VPA-treated female rats compared to control female rats (Figures 4C, D). Peak coordinates in Paxinos space for each cluster are shown in Table 2. 20

Figure 4. Statistical parametric mapping revealed altered brain activity by VPA treatment. (A) Metabolism in the olfactory bulb was decreased in VPAtreated male rats (n=22) compared to control male rats (n=17). (B) Metabolism in the bilateral somatosensory cortices, caudoputamen, and cerebellum was increased in VPA-treated male rats. (C-D) VPA-treated female rats (n=19) had increased metabolism in the left caudoputamen and left medulla compared to control female rats (n=19). Color bar indicates t-values. Voxels survived at family-wise error (FWE) corrected P < 0.05 were shown. VPA, valproic acid. 21

Table 2. Brain regions with significant metabolic differences in VPAtreated rats compared to control rats Region Paxinos coordinates t-value k x y z VPA-treated male Decreased metabolism * Olfactory bulb, bilateral 1.1 5.2-5.8 4.7 962 Increased metabolism * Somatosensory cortex, left -6.7-2.0-3.0 5.2 597 Somatosensory cortex, right and caudoputamen, right 6.7-1.4-3.4 5.1 1160 Cerebellum, bilateral -0.3-14.0-4.6 4.7 108 Caudoputamen, left -4.3-1.0-4.6 4.5 187 VPA-treated female Increased metabolism * Medulla, left -1.1-13.0-8.2 4.5 198 Caudoputamen, left -3.3-0.4-4.6 4.4 64 * Survived at family-wise error (FWE) corrected P < 0.05. VPA, valproic acid. 22

5. Connectivity estimated by correlation Correlation matrices for each group in both sexes were generated: control males (n=17), VPA-treated males (n=22), control females (n=19), and VPAtreated females (n=19) (Figures 5A-D). Permutation test revealed that no connection was significantly different to show the effect of VPA treatment in both sexes. 23

Figure 5. Correlation analysis did not show any connectivity changes by VPA treatment. (A) Control males (n=17). (B) VPA-treated males (n=22). (C) Control females (n=19). (D) VPA-treated females (n=19). VPA, valproic acid. 24

6. Connectivity estimated by sparse inverse covariance estimation I estimated the most robust and stable connections using SICE for each group, VPA-treated and control rats in both sexes (Figures 5A-D). There was no significant difference in total number of connections in the brain between VPAtreated and control rats in both sexes. The number of connections within each hemisphere and between hemispheres were also not differed between VPAtreated and control rats in both sexes (Table 3). 25

Figure 6. SICE identified connectivity changes by VPA treatment. (A) Control males (n=17). (B) VPA-treated males (n=22). (C) Control females (n=19). (D) VPA-treated females (n=19). SICE, sparse inverse covariance estimation; VPA, valproic acid. 26

Table 3. Number of connections in whole brain, within each hemisphere, and between hemispheres Number of connections P Control VPA-treated Male Whole brain 60 51 0.515 Within left cerebral hemisphere 8 10 0.284 Within right cerebral hemisphere 7 10 0.242 Between cerebral hemispheres 18 24 0.174 Female Whole brain 84 69 0.310 Within left cerebral hemisphere 10 8 0.401 Within right cerebral hemisphere 9 8 0.438 Between cerebral hemispheres 33 24 0.151 Data were analyzed using permutation test. VPA, valproic acid. 27

7. Node degree VPA-treated male rats and control male rats were not differed in node degree. In VPA-treated female rats, the number of connections involving the left thalamus was significantly increased than in control female rats (Table 4 and Figure 7). No node with decreased number of connections was found in VPA-treated female rats compared to control female rats. 28

Table 4. Changes in node degree in VPA-treated female rats Node Number of connections P Control female VPA-treated female Thalamus, left 0 3 <0.001 Data were analyzed using permutation test. VPA, valproic acid. 29

Figure 7. Aberrant number of connections was found in VPA-treated female rats. (A) Control females (n=19). (B) VPA-treated females (n=19). Compared to control females, VPA-treated female rats had an increased number of connections in the left thalamus. Data were analyzed using permutation test. MB, midbrain; TH, thalamus; VPA, valproic acid; L, left; R, right. 30

8. Connectivity strength In VPA-treated male rats, the connectivity strength between the left insular cortex and left amygdala was significantly reduced than in control male rats, which is important in salience processing. 41 The connectivity strength between the left pons and left medulla, and between the right nucleus accumbens and right hypothalamus were increased in VPA-treated male rats compared to control male rats. VPA-treated female rats had decreased connectivity strength between the thalamus and midbrain, bilaterally, and between the right medial prefrontal cortex and left caudoputamen compared to control female rats. The connectivity strength between the right medial prefrontal cortex and right caudoputamen, and between the right somatosensory cortex and right hypothalamus were increased in VPAtreated female rats (Table 5 and Figure 8) 31

Table 5. Changes in connectivity strength by VPA treatment Comparisons Connectivity strength P Connections VPAtreated Control VPA-treated males < Control males Insular cortex, left Amygdala, left 0 1.96 0.001 VPA-treated males > Control males Pons, left Medulla, left 0-2.61 <0.001 Nucleus accumbens, right Hypothalamus, right 0-2.55 <0.001 VPA-treated females < Control females Thalamus, right Midbrain, right -1.61 0 <0.001 Thalamus, left Midbrain, left -0.93 0 0.001 Medial prefrontal cortex, right Caudoputamen, left 0 3.33 0.001 VPA-treated females > Control females Medial prefrontal cortex, right Caudoputamen, right 0-4.52 <0.001 Somatosensory cortex, right Hypothalamus, right 2.08 0 0.002 VPA, valproic acid. 32

Figure 8. VPA treatment results in significant changes in connectivity strength. (A) In VPA-treated male rats (n=22), the connection between the left insular cortex and left amygdala was impaired compared to control males (n=17). (B) The 33

connections between the left pons and left medulla, and between the right nucleus accumbens and right hypothalamus were decreased in control male rats than in VPA-treated male rats. (C) VPA-treated female rats (n=19) had decreased connections between the thalamus and midbrain, bilaterally, and between the right medial prefrontal cortex and left caudoputamen compared to control female rats (n=19). (D) The connections between the right medial prefrontal cortex and right caudoputamen, and between the right somatosensory cortex and right hypothalamus were decreased in control female rats than in VPA-treated female rats. Data were analyzed using permutation test, followed by false discovery rate corrections (corrected P 0.05). ACB, nucleus accumbens; AMG, amygdala; CP, caudoputamen; HY, hypothalamus; INS, insular cortex; M, medulla; MB, midbrain; MPF, medial prefrontal cortex; P, pons; SS, somatosensory cortex; TH, thalamus; VPA, valproic acid; L, left; R, right. 34

IV. DISCUSSION ASD is heterogeneous in nature, and common pathognomonic signs remain elusive. Many neuroimaging efforts aimed at identifying commonalities in patients with ASD are also hampered by the vast diversity of ASD cases. Longitudinal follow-up studies have shown that abnormally increased brain volumes and accelerated rates of brain growth during early childhood were observed in only a small minority of children with ASD. 42,43 Also, there was no significant difference between ASD and typical development in terms of the growth curve of total corpus callosum volume. 44 Moreover, according to a recently conducted study with a large sample size, corpus callosum size was not decreased in ASD. 18 Among the few studies that have investigated resting-state brain glucose metabolism in humans with ASD, there has been no consensus across 18 F-FDG PET investigations. Indeed, some studies have observed no differences between patients with ASD and controls, 45 while others have reported fewer positive correlations between frontal and parietal regions, 46 or widespread hypermetabolism in individuals with ASD. 47 These studies are mostly based on regional metabolic differences, with the exception of one study that explored resting-state metabolic connectivity in humans with ASD. 46 In that study, correlation analysis was used to estimate the metabolic connectivity among frontal cortices, parietal cortices, and subcortical structures, and patients with ASD showed impaired interactions between frontal/parietal regions and the neostriatum and thalamus. I suspected that these inconsistent results may stem from the heterogeneous nature of ASD cases. Thus, I focused on a homogeneous type of ASD induced by prenatal exposure to VPA. I hypothesized that alterations in functional connectivity would provide insight into the neurobiological basis of ASD, particularly that caused by VPA exposure. I also explored sex-specific changes in behavior and metabolic networks following homogeneous VPA treatment. I demonstrated that VPA-treated male rats displayed reduced social behavior, had decreased metabolic 35

activity in the olfactory bulb, and had impaired metabolic connectivity between left insular cortex and left amygdala compared to control male rats. Prenatal VPA exposure causes changes in neurodevelopmental trajectory. In vitro VPA exposure of chick neural crest cells resulted in loss of N-cadherin expression, leading to neural tube defect. 48 The neural tube normally closes during 4th weeks in the human. 49 In the rat, E12.5 corresponds to this period, in which the neurogenesis occurs. VPA exposure in this vulnerable period results in abnormal neural development. 50 As VPA act as a nonselective HDAC inhibitor, it can induce increase in synapse numbers and a robust facilitation of excitatory synapse maturation. 51 VPA also enhance neural proliferation and promotes neurite growth, which in turn contributes to abnormal neural development. 52,53 VPA treatment caused reduced body weight gain in the pups. Both the VPAtreated male and female rats had less body weight than controls. It is not clear whether the body weight gain is associated with autistic-like behaviors. Few studies have explored the body weight gain in human with ASD, 54 and results from VPAtreated rat are mixed. 12,55 Further study is needed to clarify the relationship between VPA treatment and body weight gain. VPA-treated male rats in adult group lost the preference to the novel rat over familiar rat, while VPA-treated male rats in adolescent group maintained the preference, although the degree of the preference was less than that of controls. It is not clear whether the male rats in adolescent group will lose the preference if the three-chamber social approach test is performed more later. In contrast, VPA-treated female rats in adult group had the preference to the novel rat over familiar rat, as control rats do, while the preference was absent in VPA-treated female rats in adolescent group. The loss of preference for social novelty has not been considered as robust a social phenotype of ASD as the loss of sociability. Nevertheless, the loss of preference for social novelty may be one indicator of functionally significant impairment in social interaction, a core deficit in patients with ASD. 1 In this regard, loss of preference for social novelty could be one possible aspect of this functionally 36

significant impairment. It is also of note that there are occasions when time spent in each chamber is not significant but sniffing time is significant. 56 The behavioral discrepancy between VPA-treated adolescent males and adult males is needed to be validated, as there are multiple measures to test autistic-like behaviors in two symptom domains of ASD. 57 VPA-treated male rats had significantly decreased metabolism in the olfactory bulb. Diminished function of olfactory system has implicated both in rat and human with ASD. In VPA-treated male rats, nest-seeking response was delayed. 12 Children with ASD did not showed the preference to pleasant odorants over unpleasant odorants. In contrast, typically developing children had the preference to pleasant odorants over unpleasant odorants. 58 Increasing evidence supports that social chemosignaling is an important mediator of social interaction both in human and rodent, as olfaction fundamentally influences social behaviors. 59,60 Given that the sniff response is similar between humans and rodents, 61 decreased metabolism in the olfactory bulb may serve as a potential biomarker for ASD. However, whether this finding is specific to ASD or common across various neurodevelopmental disorders remains to be elucidated. Metabolism in the bilateral somatosensory cortices, caudoputamen, and cerebellum was increased in VPA-treated male rats. Longitudinal studies have shown that an increase in growth rate of striatum, which was specific to caudate nucleus, was correlated with repetitive behavior in humans with ASD. 62,63 Right caudate volume also implicated in repetitive behaviors in ASD. 64 Increased metabolism in the caudoputamen in VPA-treated male rats is in line with these observations. Increasing evidence supports the notion that the cerebellar region not only mediates motor function but also influences social processing. 65 In humans with ASD, the cerebellum was shown to have defects early in life as core ASD deficits appear, especially in the vermis, which were correlated with social impairments. 66 However, whether increase in metabolism in the cerebellum is 37

correlated with cerebellar defects in relation to ASD remains unclear. The role of somatosensory cortex in ASD is also needed to be explored. There was no significant effect of VPA in node degree in male rats. In VPAtreated rats, the connectivity strength between the left insular cortex and left amygdala was significantly reduced than in control males, which suggests that VPAtreated male rats had impaired salience processing as compared to controls. Salience network processes multiple stimuli perceived by the brain and determines which stimulus is stands out than others, and allows to show relevant behaviors. The insular cortex plays a central role in salience processing. The insular cortex is an information processing hub, and the anterior insular cortex interprets external sensory cue with internal bodily states. The insular cortex has extensive structural connections with the amygdala, orbitofrontal cortex, olfactory cortex, anterior cingulate cortex, and superior temporal sulcus. In particular, the activity of the insular cortex, often together with amygdala activity, reflects one's interoception, as well as the bodily perceptions of external stimuli. 41 Previous studies revealed that altered salience network connectivity or reduced salience network integrity were associated with ASD core symptoms. 67,68 Reduced salience processing may serve as a pathognomonic sign in VPA rat model of ASD. The significance of increased connections between the left pons and left medulla, and between the right nucleus accumbens and right hypothalamus, which were shown in VPA-treated males, is needed to be clarified. VPA-treated female rats had increased metabolism in the left caudoputamen and left medulla, had increased metabolic connectivity between the right medial prefrontal cortex and right caudoputamen, and between the right somatosensory cortex and right hypothalamus, but had decreased metabolic connectivity between the thalamus and midbrain, bilaterally, and between the right medial prefrontal cortex and left caudoputamen. Whether these regions play a significant role in the development of autistic-like behavior is not clear. 38

VPA-treated female rats showed less prominent behavioral derangement and associated changes in brain connectivity, and these findings are in line with those of a previous study, which indicated that VPA affects rats in a sex-specific manner. 23 Whereas women are less affected by ASD than men, women show a broad spectrum of clinical presentation. While women with low-functioning ASD exhibit more severe cognitive impairment, women with high-functioning ASD may be underdiagnosed or misdiagnosed. 69 Such behavioral variability makes group comparison following VPA treatment difficult. The changes in brain activity and connectivity correlated with autistic-like behavior in the present study were lateralized rather than symmetric. Previous studies have also recognized abnormal brain lateralization in patients with ASD. Atypical cerebral asymmetry and the absence of the left hemispheric dominance for language has been reported as a feature in patients with ASD, although this alteration is not specific to ASD. 70 In this study, I could not find shared features between changes in brain activity and connectivity. However, a number of studies have already shown that changes in brain activity and connectivity often accompany one another in patients with ASD. 15 Subtle changes were possibly obscured by strict statistical criteria. My results indicated that multiple brain regions were involved in the development of autistic-like behavior in VPA-exposed rats. The olfactory bulb processes social chemosignaling. 59,60 The insular cortex and amygdala constitute the salience network and are involved in social cognition. 71,72 The cerebellum has also been implicated in social and emotional processing, language, and cognition. 65 Taken together, these findings suggest that inappropriate responses to social stimuli from deficient salience processing and abnormal social processing in the cerebellum contribute to autistic-like behavior. Although VPA exposure represents only one possible risk factor for ASD, the VPA rat model may shed light on brain alterations that occur during the development of ASD. 39

In a methodological aspect, brain connectivity estimated by SICE revealed more differences among groups compared to the estimation using conventional correlation analysis, as group comparison based on Pearson s correlation analysis failed to show any significant differences. A fundamental issue is that the Pearson s correlation method cannot factor out latent effects of a third and/or fourth node that may modulate connectivity between the two nodes. This makes interpretation of the correlative activities unclear, whether they are from the intrinsic structural connection or from polysynaptic induction, common modulatory effects, or common feed-forward projections via the thalamus. 73 By its definition, inverse covariance measures connection between two brain regions while controlling for the effects from other brain regions. 74 Thus, inverse covariance measures are more efficient in revealing direct associations between brain areas. If a covariance between two brain regions is considerably large and is shared by both groups, comparison using conventional correlation analysis may not capture the group differences. However, as SICE models brain connectivity at the group level, it is not possible to estimate brain connectivity at the individual level, limiting its use in aiding the diagnosis of ASD. Another limitation is that, as the number of subjects was not sufficient to model brain connectivity in each age group, it remains to be validated whether the robust brain connectivity observed in VPA-treated rats changes with age. Another limitation is that nodes were empirically defined based on anatomy. A data-driven method called independent component analysis (ICA), which does not require prior knowledge, may provide more appropriate chance to explore connectivity of the entire brain, especially in terms of substructures. 75 40

V. CONCLUSION VPA-treated rats exhibited impairments in social behaviors, and this difference was more pronounced in male than female rats. Preference for social novelty was impaired in VPA-treated male rats, while sociability was diminished in VPA-treated female rats. I found that metabolic activity and connectivity was significantly changed by VPA treatment. Changes in metabolic connectivity was revealed by SICE while conventional correlation analysis did not show any difference. VPA-treated male rats had significantly decreased metabolic activity in the olfactory bulb, and had decreased metabolic connectivity between the left insular cortex and left amygdala, which constitute the salience network. There were no brain regions with decreased metabolic activity in VPA-treated female rats. In contrast, VPA-treated female rats had reduced connectivity between the thalamus and midbrain, and between the right medial prefrontal cortex and left caudoputamen. Such alterations in metabolic activity and connectivity may represent neurobiological substrates of autistic-like behavior, particularly in males, and may serve as a pathognomonic sign in VPA rat models of ASD. As such this study supports the idea that non-invasive brain imaging may serve as an imaging endophenotype that could aid diagnosis of ASD, classification of severity, and possibly reveal insights to neurobiological underpinnings in autisticlike behavior. 41

REFERENCES 1. American Psychiatric Association., American Psychiatric Association. DSM-5 Task Force. Diagnostic and statistical manual of mental disorders : DSM-5. 5th ed. Washington, D.C.: American Psychiatric Association; 2013. 2. Developmental Disabilities Monitoring Network Surveillance Year Principal I, Centers for Disease C, Prevention. Prevalence of autism spectrum disorder among children aged 8 years - autism and developmental disabilities monitoring network, 11 sites, United States, 2010. MMWR Surveill Summ 2014;63:1-21. 3. Dawson G, Jones EJ, Merkle K, Venema K, Lowy R, Faja S, et al. Early behavioral intervention is associated with normalized brain activity in young children with autism. J Am Acad Child Adolesc Psychiatry 2012;51:1150-9. 4. Estes A, Munson J, Rogers SJ, Greenson J, Winter J, Dawson G. Long- Term Outcomes of Early Intervention in 6-Year-Old Children With Autism Spectrum Disorder. J Am Acad Child Adolesc Psychiatry 2015;54:580-7. 5. Shattuck PT, Durkin M, Maenner M, Newschaffer C, Mandell DS, Wiggins L, et al. Timing of identification among children with an autism spectrum disorder: findings from a population-based surveillance study. J Am Acad Child Adolesc Psychiatry 2009;48:474-83. 6. Simonoff E, Pickles A, Charman T, Chandler S, Loucas T, Baird G. Psychiatric disorders in children with autism spectrum disorders: prevalence, comorbidity, and associated factors in a population-derived sample. J Am Acad Child Adolesc Psychiatry 2008;47:921-9. 7. Lai MC, Lombardo MV, Baron-Cohen S. Autism. Lancet 2014;383:896-910. 8. Dufour-Rainfray D, Vourc'h P, Tourlet S, Guilloteau D, Chalon S, Andres 42