The Oxytocin Receptor Gene (OXTR) Is Associated with Autism Spectrum Disorder: a Meta-Analysis

Home , Oxytocin (medication), Oxytocin receptor

ORIGINAL ARTICLE The oxytocin receptor gene (OXTR) is associated with autism spectrum disorder: a meta-analysis

D LoParo and ID Waldman

The oxytocin receptor gene (OXTR) has been studied as a risk factor for autism spectrum disorder (ASD) owing to converging evidence from multiple levels of analysis that oxytocin (OXT) has an important role in the regulation of affiliative behavior and social bonding in both nonhuman mammals and humans. Inconsistency in the effect sizes of the OXTR variants included in association studies render it unclear whether OXTR is truly associated with ASD, and, if so, which OXTR single-nucleotide polymorphisms (SNPs) are associated. Thus, a meta-analytic review of extant studies is needed to determine whether OXTR shows association with ASD, and to elucidate which specific SNPs have a significant effect on ASD. The current meta-analysis of 16 OXTR SNPs included 3941 individuals with ASD from 11 independent samples, although analyses of each individual SNP included a subset of this total. We found significant associations between ASD and the SNPs rs7632287, rs237887, rs2268491 and rs2254298. OXTR was also significantly associated with ASD in a gene-based test. The current meta-analysis is the largest and most comprehensive investigation of the association of OXTR with ASD and the findings suggest directions for future studies of the etiology of ASD.

Molecular Psychiatry (2015) 20, 640–646; doi:10.1038/mp.2014.77; published online 5 August 2014

INTRODUCTION sizes.8 Nevertheless, several genes and genomic regions have Autism spectrum disorder (ASD) refers to a group of neuro- shown relatively consistent association with ASD. developmental disorders characterized by developmental delays, Geneticists have used several methodologies in an attempt to impaired functioning in social and communicative skills and uncover the sources of heritability in ASD. Genome-wide linkage the presence of restricted, repetitive behavior.1 Researchers have studies have identified dozens of genomic regions likely to harbor – 2,3 autism risk genes.9 Several genome-wide association studies estimated the heritability of ASD at 55 80%, indicating that – fl (GWAS) of ASD also have been conducted,10 12 although repli- genetic in uences are responsible for most of its etiology. 8,13 Neuropeptide genes have been investigated in relation to ASD cated results have yet to emerge. An area of recent growth is owing to evidence that oxytocin (OXT) and arginine vasopressin the exploration of copy-number variants in individuals with have an important role in the regulation of affiliative behavior and ASD. Recent studies have shown that large, rare copy-number social bonding in both nonhuman mammals and humans4,5 and variants occur more frequently in ASD individuals than in controls14 and that 5–11% of ASD individuals have at least one are differentially expressed in the blood of autistic individuals in 14 comparison with non-autistic individuals.6,7 The oxytocin receptor rare or de novo copy-number variant, most of which were unique to each individual.14,15 Beyond copy-number variants, gene (OXTR) is the neuropeptide gene most frequently studied for 16–18 association with ASD. This gene contains many single-nucleotide recent exome sequencing studies found that de novo point polymorphisms (SNPs), and each association study of OXTR and mutations (single-base substitutions) were more frequent in individuals with ASD, particularly in genes involved in cell structure ASD has analyzed a different set of these variants, leading to and protein binding in the brain.8 Although these genome-wide, considerable heterogeneity in the literature. Further, inconsistency ‘hypothesis-free’ approaches have begun to yield results, many in the effect sizes of the OXTR variants renders it unclear whether researchers also study specific genes based on their hypothesized OXTR is truly associated with ASD, and, if so, which OXTR SNPs are role in the etiology of ASD. associated. Thus, we conducted a meta-analytic review of the Genes in several biological pathways have been examined current literature to determine whether OXTR and its constituent for association with ASD, with a particular focus on pathways SNPs are associated with ASD. involved in neuronal and cortical functioning. For example, studies have found that ASD is significantly associated with the genes for 9,19 Genetics of ASD hepatocyte growth factor and its receptor, MET, the gene that 20 fi codes for the protein Reelin, and genes in serotonergic, Although there has been considerable research devoted to nding 21 specific genes that underlie ASD, most of its heritability remains GABAergic and glutamatergic pathways. Researchers have also unexplained.8 ASD is likely a highly polygenic condition, with no investigated biological factors thought to be particularly involved one gene expected to explain a substantial proportion of its in social and communicative functioning, such as OXT. variance.8 Further, variants associated with ASD are often rare or even de novo mutations (spontaneous mutations that neither Oxytocin and social behavior parent possessed or transmitted), making the reliable identifica- OXT is a peptide constituted by oligopepetide chains of nine tion of associated genes difficult, especially without large sample amino acids.22 The gene that codes for the OXT receptor, OXTR,

Psychology Department, Emory University, Atlanta, GA, USA. Correspondence: D LoParo, Psychology Department, Emory University, 36 Eagle Row, Atlanta, GA 30322, USA. E-mail: [email protected] Received 12 December 2013; revised 8 May 2014; accepted 17 June 2014; published online 5 August 2014 OXTR is associated with autism spectrum disorder D LoParo and ID Waldman 641 spans 19.2 kilobases on chromosome 3 and is expressed in association studies of OXTR SNPs with ASD,23,45–53 comprising 13 humans throughout the body in reproductive structures22 as well independent samples, 11 of which used family-based designs. as in neural regions, such as the frontal cortex, amygdala, Family-based designs examine the over-transmission of a putative hypothalamus and olfactory nucleus.5,23 high-risk allele at an SNP from heterozygous parents to their Nonhuman mammal research has established a role for OXT in affected offspring, under the assumption that if an allele is social behavior. OXT receptors are expressed in neural regions transmitted more often than expected by chance (i.e., 50%) then it related to social behavior, pair-bonding, social memory and social is associated with the disorder. Each of these studies tested a aggression in nonhuman mammals.24 In particular, OXT neurons different set of OXTR SNPs, and no SNP has been found to be in the hypothalamus project to the nucleus accumbens and significantly associated with ASD in every study in which it has posterior pituitary in prairie voles.25 Researchers have found that been examined, rendering a qualitative synthesis of this literature OXT seems to promote approach behavior by inhibiting difficult. Also, no studies to date have performed a gene-based instinctual avoidance of closeness and defensive behavior.5,26 test to evaluate the association between OXTR and ASD. Gene- Disrupting oxytocin receptor functioning also seems to disrupt based tests combine significance levels across multiple SNPs and social behavior. For example, OXTR knockout mice have impaired the linkage disequilibrium among them, thus reducing type I error social memory and recognition,27 impaired mother-offspring and increasing statistical power.54 Owing to the considerable interaction28 and increased aggression.28 This evidence suggests heterogeneity across studies in the SNPs tested and their evidence that OXT has a causal role in the development and maintenance of for association, a meta-analysis is needed to determine whether social functioning in rodent species. These findings have OXTR as a whole and specific OXTR SNPs are associated with ASD. prompted researchers to investigate the role that OXT and OXTR have in human social behavior. The present meta-analysis OXT in humans has been shown to impact the processing of The primary aims of the current meta-analysis are to evaluate the social stimuli. For example, intranasal OXT administration has been main effects of OXTR and variants therein on ASD, and to test for shown to increase calmness and decrease anxiety in response to heterogeneity in the effects of these variants across studies. a social stressor,29 increase ability to infer affective states from 30 Unfortunately, too few studies currently exist to reliably explore viewing the eye regions of emotional faces and increase social variables that may explain between-study heterogeneity. None- referencing to eye regions.31 Further, OXT in humans increases fi 32 theless, characterizing the magnitude of heterogeneity will facili- af liative behavior and trust in social situations. It is worth tate the future examination of study-level moderators as samples noting that commonly used methods for measuring peripheral increase. We also use the extant data to perform a gene-based test oxytocin can be unreliable, and it is not known how much OXT of association between OXTR and ASD. reaches the brain after intranasal administration,33 thus these results should be interpreted with caution. Nonetheless, findings in the human literature suggest that OXT has a complex but METHODS fi facilitative effect on social behavior, paralleling ndings in the Literature search and selection strategy animal literature. These results have led researchers to investigate A systematic literature search was conducted to collect studies relevant for whether genetic variation in the human OXTR gene is associated the current meta-analysis. First, we searched three online databases with social functioning. 22 (PubMed, PsycInfo and Google Scholar) for articles published in English on OXTR contains several dozen SNPs that have been geno- associations between variants in OXTR and ASD. Search terms used to typed in association studies of various traits and behaviors identify studies were a combination of terms identifying the gene (‘OXTR’, in humans. Researchers have found that various OXTR SNPs ‘OXTR gene’ and ‘oxytocin receptor gene’) and terms identifying the are associated with generous behavior,34 prosociality ratings,35 phenotype of interest (‘autism’, ‘ASD’ and ‘Asperger’‘pervasive develop- empathy and social communication,36 amygdala and hypo- mental disorder’). thalamus functioning,36 reduced physiological reactivity to After identifying a set of publications returned from each search, we stress,37 increased benefits from social support38 and greater collated the results and eliminated duplicates. We then reviewed the titles, 39 abstracts, keywords and texts of all studies to exclude studies that were parenting sensitivity, among other social and biological traits. clearly irrelevant. We searched the literature review and reference sections Taken together, these results suggest associations between OXTR of the publications to find additional publications that were missed in the and various aspects of human social behavior, although the original search. Finally, we applied criteria to the set of studies to mechanisms underlying OXTR’sinfluence and the contributions determine their inclusion or exclusion in the present meta-analysis. of particular OXTR SNPs is unclear. Inclusion and exclusion criteria Oxytocin in ASD The following criteria were applied to select studies for inclusion in the Recent research on the relation between OXT and ASD has meta-analyses: (1) the study must have examined the association between produced promising, if preliminary, results. Young children at least one variant in OXTR and autism disorder, Asperger’s syndrome, fi with ASD have reduced OXT plasma levels6,40 and increased pervasive developmental disorder not otherwise speci ed or ASD broadly unprocessed OXT peptides,40 suggesting that low levels of OXT using either a case-control or family-based association design; (2) the study must have reported data from at least one independent sample; (3) the are associated with ASD. Intranasal or intravenous administrations study must have included ASD diagnoses made by either psychiatric of OXT in individuals with ASD have been shown to increase assessment or using standard diagnostic instruments. Diagnostic instru- 41 performance on an emotion recognition task, recognition of ments used in studies included in the meta-analysis included the Autism social cooperation, trust, and preference42 and the time gazing at Diagnostic Interview-Revised,55 the Autism Diagnosis Observation Sche- eyes while viewing pictures of faces.42 Further, one study found dule-Generic,56 the Child Autism Rating Scale57 and the Developmental, that an OXTR SNP (rs237887) explains 10% of the variance in Dimensional, and Diagnostic Interview.58 The Autism Diagnostic Interview- face-recognition memory in families with an autistic child.43 These Revised is a semi-structured interview for caregivers of children and adults results suggest that OXT may both have a role in the etiology of focused on the three ASD symptom clusters. The Autism Diagnosis Observation Schedule-Generic is a semi-structured observational measure ASD and also be a treatment target. that assesses communication, social interaction and play in children. The Given the aforementioned results, as well as findings of a 44 Child Autism Rating Scale is a behavior rating scale that can be completed genetic linkage peak directly over the OXTR gene, researchers by clinicians, teachers or parents. The Developmental, Dimensional, and have investigated whether variation in OXTR is associated with Diagnostic Interview is a computerized procedure administered by trained ASD. At the time of this study, there have been 10 genetic interviewers to parents that measures autistic features dimensionally.

© 2015 Macmillan Publishers Limited Molecular Psychiatry (2015), 640 – 646 OXTR is associated with autism spectrum disorder D LoParo and ID Waldman 642 The following criteria were applied to exclude studies from the meta- of significant findings in each set of studies by summing the power of each analyses: (1) the study did not use either case-control or family-based study and then to compare that value to the observed number of association designs; and (2) the sample did not contain individuals with significant findings with a χ2-test.64 ASD diagnoses (e.g., the study measured autism-related traits in exclusively non-disordered populations). We identified a total of 10 eligible studies RESULTS with 13 independent samples.23,44–52 One study was excluded because its sample contained solely non-disordered individuals,52 leaving a total of 9 Meta-analyses of OXTR and ASD studies with 12 independent samples that were eligible for data extraction. After applying inclusion and exclusion criteria (see Materials and Methods), data from 8 studies with 11 independent – – Data extraction samples23,43 44,46 50 were included in the meta-analysis for OXTR. The following data were extracted from each article when possible: first Across these samples, 57 different SNPs were analyzed for author, year of publication, study design, country, predominant ethnicity of association with OXTR, 16 of which were tested in 4 or more participants, mean age, proportion of males, risk and non-risk allele count samples and thus included in our meta-analysis (see Table 1). or allele frequencies in cases and controls (for case-control studies) or Table 2 shows the pooled odds ratios, their 95% confidence transmitted and non-transmitted risk allele counts from heterozygous intervals and P-values and heterogeneity tests and indices parents to probands (for family-based studies), effect sizes and diagnostic (Q-statistics, P-values and I2) for the pooled estimates for criteria used. The data were double-extracted independently and each OXTR SNP. Of the 16 SNPs meta-analyzed, significant the results were cross-checked for discrepancies. Several discrepancies in pooled odds ratios were found for 4 SNPs: rs7632287 (‘A’ allele the study-level data were found and corrected. No discrepancies were is risk-inducing), rs237887 (‘A’ allele is risk-inducing), rs2268491 found in the SNP-level data. Given the large number of markers tested in ‘ ’ ‘ ’ many of the studies, several publications omitted the requisite information ( T allele is risk-inducing) and rs2254298 ( A allele is risk- for inclusion of a particular SNP in a meta-analysis. In these cases, authors inducing). were contacted via email and asked to provide the missing data. After this For rs7632287, the OR was significant and large: OR = 1.44 process, we were able to include 8 studies with 11 independent (95% CI: 1.23–1.68, P = 0.000005) (see Figure 1a) and there was no – – samples23,44 45,47 51 in the meta-analysis. One study contained two significant between-study heterogeneity (Q-statistic: χ2 = 0.92, independent samples48 and another contained three independent 2 fi 23 P = 0.819, I = 0%). For rs237887, the OR was signi cant but small: samples, all of which were included and treated as independent. OR = 0.88 (95% CI: 0.79–0.98, P = 0.024) (see Figure 1b) and there was no significant between-study heterogeneity (Q-statistic: Meta-analytic methods χ2 = 2.07, P = 0.839, I2 = 0%). For rs2268491, the OR was significant OXTR SNPs were selected for the meta-analysis if they were tested for and moderate: OR = 1.19 (95% CI: 1.05–1.36, P = 0.0075) (see association with ASD in at least four independent samples. Separate meta- Figure 1c) and there was no significant between-study hetero- analyses for each SNP then were performed using the meta-analysis geneity (Q-statistic: χ2 = 4.13, P = 0.531, I2 = 0%). For rs2254298, the 59 software program, Metasoft (http://genetics.cs.ucla.edu/meta). Metasoft OR was significant under the alternative random-effect model: fi allows users to input effect sizes and standard errors, conducts xed- and OR = 1.15 (95% CI: 0.93–1.43, P = 0.0038) (see Figure 1d), but non- fi random-effects meta-analyses and outputs pooled xed-effects and significant under the traditional random-effects model (P = 0.16). random-effects effect sizes and corresponding tests of significance and 2 There also was substantial heterogeneity in the estimated effect heterogeneity (i.e., the Q-statistic and I index). Metasoft also outputs χ2 2 P-values from Han & Eskin’s58 alternative random-effects model that is sizes (Q-statistic: = 11.2, P = 0.048, I = 55%). optimized to detect associations under heterogeneity by assuming We next conducted gene-based tests of association between no heterogeneity under the null hypothesis. This model has greater OXTR and ASD separately in Asian- and European-background power than traditional random-effects models and even fixed-effects samples to account for different LD patterns by ethnicity. The models when heterogeneity is substantial.59 We chose to report both gene-based test for the Asian samples was significant (traditional traditional and alternative random-effect P-values for all analyses. Fixed- random effects P = 0.019, alternative random effects P = 0.016), effects and random-effect P-values were identical when no heterogeneity whereas the gene-based test in the European-background was present. samples was highly significant (traditional random effects P = 0.000067, alternative random effects P = 0.000045). Meta- Gene-based tests of association analytic combination of the gene-based test results from the Gene-based tests of association between ASD and OXTR as a whole were two ethnicities was also highly significant (traditional random conducted separately by ethnicity (i.e., Asian- and European-background effects P = 0.0000077, alternative random effects P = 0.0000053), samples) to account for differing linkage disequilibrium (LD) patterns by suggesting that OXTR is associated with ASD. ethnicity using both traditional and alternative random-effect model P-values. Gene-based tests were conducted using KGG (http://statgenpro. fi psychiatry.hku.hk/limx/kgg/),60 a software package that combines SNP- Tests of publication bias and excess signi cance based P-values with estimates of linkage disequilibrium among SNPs We tested the sets of studies of the four SNPs with signi- within the gene to provide a gene-based P-value. Gene-based tests were ficant pooled ORs for excess significance64 and publication bias.63 conducted using the hybrid set-based test method,61 a combination of a There was no evidence of excess significance for rs7632287 2 scaled χ -test and an extended Simes test that integrates SNP association (P = 0.620), rs237887 (P = 0.444), rs2268491 (P = 0.927) or P-values and LD among the SNPs within a gene to obtain an overall P-value rs2254298 (P = 0.386). There was no evidence of publication bias for the entire gene. The Asian- and European-background samples were 22 for rs7632287 (P = 0.419), rs237887 (P = 0.649) or rs2268491 tested separately using the HapMap linkage disequilibrium information fi appropriate for each ethnicity. The results were then combined across (P = 0.995), although rs2254298 showed evidence of signi cant 62 but slight publication bias (P = 0.077). The significance of this ethnicity using MetaP, a program that performs meta-analysis by 46 combining P-values from independent samples. We reported Stouffer’sz test is driven by the study by Jacob et al. which had an OR of P-values, which take sample size into account. 0.43, quite distant from the other estimates (see Figure 1). Remo- ving this study rendered the test of publication bias nonsignificant Tests of publication bias and excess significance (P = 0.437), increased the estimated effect size (OR = 1.26) (95% CI: 63 1.10–1.43, P = 0.0006), and rendered the heterogeneity between We performed Egger’s test for publication bias and Ioannidis’s test for 2 2 fi 64 fi studies nonsignificant (Q-statistic: χ = 4.17, P = 0.3837, I = 4%). excess signi cance for the SNPs with signi cant pooled odds ratios. fi Specifically, Egger’s test for publication bias tests the significance of the Removing this study also slightly increased the signi cance of the intercept in a linear regression of each study’s standardized odd ratio on the gene-based test in the European-background sample inverse of the s.e. of each study’s odds ratio with an alpha level of α = 0.10.63 (P = 0.000047) and the meta-analytically combined sample Ioannidis’s test for excess significance is to calculate the expected number (P = 0.0000047).

Molecular Psychiatry (2015), 640 – 646 © 2015 Macmillan Publishers Limited OXTR is associated with autism spectrum disorder D LoParo and ID Waldman 643 DISCUSSION Yes Yes Yes Yes Yes We performed a meta-analysis of the association of ASD with 16 SNPs in OXTR using data from 3941 individuals with ASD from 11 independent samples (although analyses of each individual SNP only included a subset of this total). We found associations between ASD and rs7632287 (‘A’ allele is risk-inducing), rs237887 (‘A’ allele is risk-inducing), rs2268491 (‘T’ allele is risk-inducing) and rs2254298 (‘A’ allele is risk-inducing), as well as OXTR as a whole.

Interpretation of OXTR ﬁndings The association of the ‘A’ allele of rs7632287 with ASD was found cant SNPs Usable

ﬁ to be particularly strong, consistent with converging research that this marker may be involved in human social behavior.50,52 Given — — — rs2268495 — its location in the promoter region of OXTR, rs7632287 may be involved in transcription factor binding, although it has not been experimentally demonstrated that it alters gene expression.50 ‘ ’

Scale; 3Di, Developmental, Dimensional, and Diagnostic The A allele was associated with lower pair-bonding and social skills scores across three normative samples,52 suggesting that rs7632287 has a role in normal and autism-related social behavior. Although the pooled OR for rs7632287 was relatively large (1.44), recent GWAS of ASD did not have sufﬁcient power to detect effects of that size,16–18,65 which may explain why SNPs in OXTR have yet to emerge as signiﬁcantly associated with ASD in GWAS. Given the present results, as GWAS increase in sample size SNPs in

CARS OXTR should begin to emerge as significantly associated with ASD. The association of the ‘A’ allele of rs237887 with ASD also has some corroborating evidence in the literature. One study found that although this allele was not associated with ASD diagnosis, it was associated with communication and daily living skills in ASD individuals.46 Another study found that the ‘A’ allele of rs237887 was associated with lower altruism in two separate behavior tasks,34 suggesting that rs237887 may underlie various aspects of human social behavior. This allele was also associated with reduced face- recognition memory in families with an autistic child.43 Two SNPs, rs2268491 is and rs2254298,49 are in very high LD (r240.95) and were both significantly associated with ASD in the meta-analysis, suggesting that signals from these two SNPs may reflect a common association with ASD. Some corroborating CaucasianJapanese 80% 88% 23 Psychiatrist 25 Psychiatrist rs237887, rs2264891, rs2254298, evidence exist that the ‘T’ allele of rs2268491 and the ‘A’ allele of and autism spectrum disorder rs2254298 are involved in social behavior. Specifically, rs2268491 has been found to be associated with aspects of empathy OXTR (although the ‘T’ allele was associated with higher empathy),66 and the ‘A’ allele of rs2254298 has been found to be associated with attachment security (with effect direction dependent on sample ethnicity),67 less sensitive parenting and lower plasma 68 controls controls Sample size Ethnicity Percent male Mean age Diagnostic criteria Signi OXT levels, suggesting that these SNPs may have a role in both normal and autism-related social behavior. There also was evidence of slight publication bias for rs2254298, as demonstrated fi

SNPs by signi cant positive skew around the estimated effect size. The significance of this test was driven by the study by Jacob et al.,46 which had an effect size that was quite discrepant from the other estimates (see Figure 1). Removing this study rendered the test of publication bias nonsignificant, increased the estimated effect size (OR increased from 1.15 to 1.26) and rendered the heterogeneity between studies minimal and nonsignificant (I2 dropped from 55 to 4%). Notably, the removed study included exclusively Caucasian individuals, whereas the other studies of rs2254298 with positive effect sizes included exclusively Asian individuals, highlighting the 20052007 Family2007 Family2008 Family2009 Family Case-control20102010 4 Family 2 Case-control 18 52010 195 families2010 3 57 Family families 1332010 174 families Family cases, 11 155 151 Family families 112010 280 Chinese cases,2011 Han 440 Family 215 families Caucasian Israeli Family 22 89% 20 Caucasian 18 100 families 79% 177 families Japanese 18 199 87% 7 84% families 25 60 Psychiatrist families 6 Caucasian 85% 1238importance families 11 Caucasian NR Psychiatrist, Caucasian ADI-R, ADOS-G ADI-R, Psychiatrist, 95% ADI-R, ADOS-G ADOS-G 18 79% Caucasian rs2254298 Caucasian 86% rs2268494, Psychiatrist rs1042778of 12 examining 10 82% rs2254298, NR rs53576 Psychiatrist, ADI-R, 6 ADOS-G Psychiatrist, ADI-R, ADOS-G rs2268493 Psychiatrist, 11 rs2270465 ADI-R, ADOS-G, NR ethnic rs11720238, No rs7632287, rs4564970 Psychiatrist, 3Di ADI-R, ADOS-G Yes heterogeneity Yes Yes rs2268493,in rs1042778, Yes rs7632287 effects Yes Yes in future association studies. The gene-based test for OXTR suggested that the gene is 52

51 highly associated with ASD overall (P = 0.0000053) as well as in 23 23 48 23 Characteristics of studies investigating associations between 49 both Asian- (P = 0.016) and European-background individuals 46 50 50 47

45 (P = 0.000045). Gene-based tests maximize the amount of genetic variation analyzed simultaneously and can provide substantially Liu B Campbell Wu Author Year Design Number of Liu A Wermter Jacob Lerer Yrigollen Chakrabarti Tansey A Tansey B Tansey C Interview, NR, not reported; SNPs, single-nucleotide polymorphisms. Table 1. Abbreviations: ADI-R, Autism Diagnostic Interview-Revised; ADOS-G, Autism Diagnostic Observation Schedule-Generic; CARS, Child Autism Rating increased power to detect association, particularly in moderate

Table 2. Meta-analytic results of SNPs in OXTR and autism spectrum disorder

SNP Position Minor Sample Family Cases/ Random CI (95%) Random Han-Eskin Q χ2 Q I2 allele N N controls effects effects P-value P-value N OR P-value

rs7632287 8766446 A 4 2769 0 1.44 1.23–1.68 0.000005* 0.000005* 0.927 0.8189 0% rs1042778 8769545 T 6 2984 280/440 0.97 0.87–1.09 0.6495 0.6495 3.607 0.6072 0% rs237885 8770543 G 8 3135 454/595 0.96 0.85–1.08 0.4583 0.4583 3.831 0.7990 0% rs11706648 8771547 C 4 2769 0 1.02 0.89–1.18 0.7552 0.7552 0.724 0.8675 0% rs237887 8772042 G 6 2984 280/440 0.88 0.79–0.98 0.0239* 0.0239* 2.073 0.8389 0% rs2268490 8772085 T 4 2769 0 1.13 0.93–1.34 0.2000 0.2000 2.396 0.4944 0% rs237888 8772095 C 4 2769 0 1.17 0.92–1.50 0.2051 0.2051 2.974 0.3957 0% rs4686301 8773586 T 4 536 0 1.15 0.92–1.43 0.2157 0.2157 1.541 0.6728 0% rs2268491 8775398 T 6 2984 280/440 1.19 1.05–1.36 0.0075* 0.0075* 4.133 0.5305 0% rs2268493 8775840 C 4 2699 280/440 0.98 0.71–1.33 0.8776 0.4958 5.944 0.1144 50% rs2254298 8777228 A 6 2900 280/440 1.15 0.93–1.43 0.1646 0.0038* 11.199 0.0476 55% rs53576 8779371 A 5 2800 280/440 0.91 0.76–1.09 0.3162 0.0699 9.037 0.0601 56% rs237894 8781531 C 4 536 0 1.03 0.84–1.27 0.7759 0.7759 2.068 0.5584 0% rs237895 8782423 T 4 536 0 1.21 0.98–1.48 0.0724 0.0724 2.070 0.5580 0% rs2268495 8782535 A 6 751 280/440 0.97 0.78–1.21 0.7921 0.2048 8.828 0.1161 43% rs4684302 8784222 T 4 2769 0 0.87 0.64–1.23 0.4687 0.0760 5.325 0.1495 44% Abbreviations: CI, conﬁdence interval; OR, odds ratio; SNPs, single-nucleotide polymorphisms. Note: Asterisks indicate signiﬁcance.

Figure 1. Depicts forest plots of odds ratios from the analyses of SNPs found to have significant effects on ASD. (a) Forest plot of odds ratios from the fixed-effects model for rs7632287. (b) Forest plot of odds ratios from fixed-effects model for rs237887. (c) Forest plot of odds ratios from the fixed-effects model for rs2268491. (d) Forest plot of odds ratios from the random-effects model for rs2254298.

sample sizes.53,54 The strength of this association suggests that Limitations although the effects of individual SNPs within OXTR may differ depending on sample size, ethnicity or other study-level factors, Several limitations should be taken into account when interpret- OXTR as a whole inﬂuences ASD. ing the results of this meta-analysis. Given that the SNP-based

Molecular Psychiatry (2015), 640 – 646 © 2015 Macmillan Publishers Limited OXTR is associated with autism spectrum disorder D LoParo and ID Waldman 645 meta-analyses comprised only 4–8 studies, a conservative inter- data necessary for inclusion in meta-analyses, comprehensive pretation of these results is that they reflect the current state of SNP selection will facilitate more systematic evaluations of the the literature and can guide future research, rather than association of OXTR and ASD. representing definitive estimates of the effect sizes attributable to each SNP. Further, for SNPs for which heterogeneity was detected, we had too few studies to reliably test moderators of CONCLUSIONS that heterogeneity, such as ethnicity, sex or intelligence quotient. The current meta-analysis found significant main effects of A second set of limitations arises from differences in the SNPs rs7632287, rs237887, rs2268491 and rs2254298 in OXTR on ASD. included in each individual study. Although many of the studies In a gene-based test, OXTR was highly associated with ASD. that examined associations between ASD and OXTR included Significant heterogeneity was found in the estimated effect sizes multiallelic or haplotype-based analyses, inconsistencies in the across studies for several SNPs in OXTR, suggesting that potential SNPs tested and results reported across studies rendered it moderator variables, such as ethnicity, may cause the observed impossible to meta-analytically evaluate any of these associations. differences in effect sizes across studies. Thus, although the Gene-based analyses encompassing multiple SNPs more compre- current results suggest that OXTR has a role in ASD, further hensively characterize variation across the genes and may thus investigation of this putative association is warranted. Future better capture risk for ASD. researchers should attempt to include as many previously studied variants as possible and comprehensively ‘tag’ OXTR to maximize Future directions the comparability across studies and make available data Several important issues remain for future research on the OXTR’s necessary for inclusion in meta-analyses. The current meta- analysis is the largest and most comprehensive investigation of role in ASD. First, little is known regarding the functionality of fi specific OXTR variants. Understanding which OXTR variants are the association of OXTR with ASD, and the ndings should help functional and how they affect OXTR’s expression in the brain is guide further investigations of the etiology of ASD. crucial for understanding OXTR’s effects on social behavior and autism-related traits. CONFLICT OF INTEREST fl Second, there is evidence that OXTR in uences social behavior in fl both normative and clinical samples. Future studies should examine The authors declare no con ict of interest. whether OXTR’s effects differ for normal-range social behavior versus clinical extremes such as ASD. Such research should include REFERENCES continuous measures of social behavior and compare effect sizes 1 Volkmar FR, Lord C, Bailey A, Schultz RT, Klin A Autism and pervasive develop- of OXTR between non-disordered and ASD samples. mental disorders. J Child Psychol Psychiatry 2004; 45: 135–170. Third, studies included in the current meta-analyses did not 2 Freitag CM, Staal W, Klauck SM, Duketis E, Waltes R. Genetics of autistic disorders: consistently include analyses of associations of OXTR with components review and clinical implications. Eur Child Adolesc Psychiatry 2010; 19:169–178. of ASD, or even differential associations with specificASDdiagnoses 3 Lichtenstein P, Carlström E, Råstam M, Gillberg C, Anckarsäter H. The genetics of (i.e., autistic disorder, Asperger syndrome, or pervasive developmental autism spectrum disorders and related neuropsychiatric disorders in childhood. disorder not otherwise specified). The studies that did include analyses Am J Psychiatry 2010; 167: 1357–1363. fi of components of ASD found that associations with OXTR tended to 4 Lim MM, Young LJ. Neuropeptidergic regulation of af liative behavior and social 50 – replicate across these components.46,49 Namely, although OXTR was bonding in animals. Horm Behav 2006; : 506 517. 5 Heinrichs M, Domes G. Neuropeptides and social behaviour: effects of oxytocin investigated owing to its hypothesized relation to social and 170 – fi and vasopressin in humans. Prog Brain Res 2008; :337 350. communicative de cits, studies have found OXTR to be associated 6 Modahl C, Green L, Fein D, Morris M, Waterhouse L, Feinstein C et al. Plasma 46,49 with stereotyped behaviors as well. Multivariate behavior genetic oxytocin levels in autistic children. Bio Psychiatry 1998; 43:270–277. studies have compared the degree of genetic overlap between the 7 Boso M, Emanuele E, Politi P, Pace A, Arra M, Ucelli di Nemi S et al. Reduced three symptom clusters of ASD and have found that each has a unique plasma apelin levels in patients with autistic spectrum disorder. Arch Med Res genetic component in addition to the genetic influences that they 2007; 38:70–74. share in common.69,70 Thus it is unclear whether OXTR is primarily 8 Yates D. Neurogenetics: unravelling the genetics of autism. Nat Rev Neurosci 2012; 13 related to the social and communicative aspects of ASD, or to ASD : 359. 9 Yang M, Gill M. A review of gene linkage, association and expression studies in more generally. Researchers should rigorously test associations of 25 fi autism and an assessment of convergent evidence. Int J Dev Neurosci 2007; : OXTR with speci c components of ASD as this might provide stronger 69–85. and more replicable results if OXTR’seffectsaremorespecific. 10 Lauritsen MB, Als TD, Dahl HA, Flint TJ, Wang AG, Vang M et al. A genome-wide Relatedly, we were unable to evaluate potential sources of hetero- search for alleles and haplotypes associated with autism and related pervasive geneity such as ethnicity, sex, age or intelligence quotient, given that developmental disorders on the Faroe Islands. Mol Psychiatry 2006; 11:37–46. these moderators were not examined in any of the included studies. 11 Ma D, Salyakina D, Jaworski JM, Konidari I, Whitehead PL, Andersen AN et al. A Researchers should more consistently test whether these moderators genome-wide association study of autism reveals a common novel risk locus have a significant effect on the relation between OXTR and ASD. at 5p14.1. Ann Hum Genet 2009; 73:263–273. Fourth, our results demonstrate that gene-based tests can 12 Hussman JP, Chung RH, Griswold AJ, Jaworski JM, Salyakina D, Ma D et al. A noise- reduction GWAS analysis implicates altered regulation of neurite outgrowth and evaluate association across a gene while taking into account guidance in autism. Mol Autism 2011; 2:1. population LD structure and thus facilitate interpretation of 13 Schaaf CP, Zoghbi HY. Solving the autism puzzle a few pieces at a time. Neuron separate SNP-based tests. As genome-wide data become more 2011; 70:806–808. widely available, researchers should consider using gene-based 14 Sebat J, Lakshmi B, Malhotra D, Troge J. Strong association of de novo copy and pathway-based approaches to develop a more comprehen- number mutations with autism. Science 2007; 316:445–449. sive framework of genetic risk for ASD. 15 Levy D, Ronemus M, Yamrom B, Lee YH, Leotta A, Kendall J et al. Rare de novo and Finally, much of the data from the current set of studies was transmitted copy-number variation in autistic spectrum disorders. Neuron 2011; 70:886–897. unusable in the meta-analysis owing to heterogeneity in sets of ’ SNPs analyzed. Determining a standard set of SNPs for future 16 O Roak BJ, Vives L, Girirajan S, Karakoc E, Krumm N, Coe BP et al. Sporadic autism exomes reveal a highly interconnected protein network of de novo mutations. research is essential for maximizing comparability across studies. Nature 2012; 485:246–250. Such a set should include previously studied SNPs as well as a set 17 Sanders SJ, Murtha MT, Gupta AR, Murdoch JD, Raubeson MJ, Willsey AJ et al. De of tagging SNPs that capture well the genetic variation across novo mutations revealed by whole-exome sequencing are strongly associated OXTR via high LD. In addition to researchers making available with autism. Nature 2012; 485:237–241.

© 2015 Macmillan Publishers Limited Molecular Psychiatry (2015), 640 – 646 OXTR is associated with autism spectrum disorder D LoParo and ID Waldman 646 18 Neale BM, Kou Y, Liu L, Ma'ayan A, Samocha KE, Sabo A et al. Patterns and rates of 45 Wu S, Jia M, Ruan Y, Liu J, Guo Y, Shuang M, Gong X et al. Positive association of exonic de novo mutations in autism spectrum disorders. Nature 2012; 485: the oxytocin receptor gene (OXTR) with autism in the Chinese Han population. 242–245. Biol Psychiatry 2005; 58:74–77. 19 Campbell DB, D'Oronzio R, Garbett K, Ebert PJ, Mirnics K, Levitt P et al. Disruption 46 Jacob S, Brune C, Carter C, Leventhal B. Association of the oxytocin receptor gene of cerebral cortex MET signaling in autism spectrum disorder. Ann Neurol 2007; OXTR in Caucasian children and adolescents with autism. Neurosci 2007; 417:6–9. 62: 243–250. 47 Lerer E, Levi S, Salomon S, Darvasi A, Yirmiya N, Ebstein RP. Association between 20 Skaar DA, Shao Y, Haines JL, Stenger JE, Jaworski J, Martin ER et al. Analysis of the the oxytocin receptor (OXTR) gene and autism: relationship to Vineland Adaptive RELN gene as a genetic risk factor for autism. Mol Psychiatry 2005; 10:563–571. Behavior Scales and cognition. Mol Psychiatry 2007; 13:980–988. 21 Zimmerman AW, Connors SL, Pardo-Villamizar CA. Neuroimmunology and 48 Yrigollen C, Han S, Kochetkova A, Babitz T. Genes controlling affiliative behavior as neurotransmitters in autism. In: Tuchman R, Rabin I (eds), Autism: A Neurological candidate genes for autism. Biol Psychiatry 2008; 63: 911–916. Disorder of Early Brain Development. Mac Keith Press: London, UK, 2006, 49 Wermter AK, Kamp-Becker I, Hesse P, Schulte-Körne G, Strauch K, Remschmidt H. pp 141–159. Evidence for the involvement of genetic variation in the oxytocin receptor gene 22 The International HapMap Consortium. A second generation human haplotype (OXTR) in the etiology of autistic disorders on high-functioning level. Am J Med map of over 3.1 million SNPs. Nature 2007; 449:851–861. Genet B Neuropsychiatr Genet 2010; 153B:629–639. 23 Tansey KE, Brookes KJ, Hill MJ, Cochrane LE, Gill M, Skuse Det al. Oxytocin receptor 50 Liu X, Kawamura Y, Shimada T, Otowa T, Koishi S, Sugiyama T et al. Association of (OXTR) does not play a major role in the aetiology of autism: genetic and the oxytocin receptor (OXTR) gene polymorphisms with autism spectrum disorder 474 – molecular studies. Neurosci Lett 2010; :163 167. (ASD) in the Japanese population. J Hum Genet 2010; 55:137–141. 24 Landgraf R, Neumann I. Vasopressin and oxytocin release within the brain: a 51 Campbell DB, Datta D, Jones ST, Batey Lee E, Sutcliffe JS, Hammock EA et al. dynamic concept of multiple and variable modes of neuropeptide communica- Association of oxytocin receptor (OXTR) gene variants with multiple phenotype 25 – tion. Front Neuroendocrinol 2004; :150 176. domains of autism spectrum disorder. J Neurodev Disord 2011; 3:101–112. 25 Ross H, Young L. Oxytocin and the neural mechanisms regulating social cognition 52 Chakrabarti B, Dudbridge F, Kent L, Wheelwright S, Hill-Cawthorne G, Allison C fi 30 – and af liative behavior. Front Neuroendocrinol 2009; : 534 547. et al. Genes related to sex steroids, neural growth, and social-emotional behavior 26 McGraw L, Young L. The prairie vole: an emerging model organism for under- are associated with autistic traits, empathy, and Asperger Syndrome. Autism Res 33 – standing the social brain. Trends Neurosci 2010; :1 14. 2009; 2: 157–177. 27 Lee HJ, Caldwell HK, Macbeth AH, Tolu SG, Young WSY. A conditional knockout 53 Walum H, Lichtenstein P, Neiderhiser JM, Reiss D, Ganiban JM, Spotts EL et al. 149 – mouse line of the oxytocin receptor. Endocrinol 2008; : 3256 3263. Variation in the oxytocin receptor gene is associated with pair-bonding and social 28 Nishimori K, Takayanagi Y, Yoshida M, Kasahara Y, Young LJ, Kawamata M. New behavior. Biol Psychiatry 2012; 71:419–426. aspects of oxytocin receptor function revealed by knockout mice: sociosexual 54 Luo L, Peng G, Zhu Y, Dong H, Amos CI, Xiong M. Genome-wide gene and behavior and control of energy balance. Prog Brain Res 2008; 180:79–90. pathway analysis. Eur J Hum Genet 2010; 18: 1045–1053. 29 Heinrichs M, Baumgartner T, Kirschbaum C, Ehlert U. Social support and oxytocin 55 Lord C, Rutter M, Couteur A. Autism Diagnostic Interview-Revised: a revised ver- interact to suppress cortisol and subjective responses to psychosocial stress. Biol sion of a diagnostic interview for caregivers of individuals with possible pervasive Psychiatry 2003; 54: 1389–1398. developmental disorders. J Autism Neurodev Disord 1994; 24: 659–685. 30 Domes G, Heinrichs M, Michel A, Berger C, Herpertz SC. Oxytocin improves 56 Lord C, Risi S, Lambrecht L, Cook EH Jr, Leventhal BL, DiLavore PC et al. The autism ‘mind-reading’ in humans. Biol Psychiatry 2007; 61: 731–733. diagnostic observation schedule-generic: a standard measure of social and 31 Guastella AJ, Mitchell PB, Dadds MR. Oxytocin increases gaze to the eye region of communication deficits associated with the spectrum of autism. J Autism Neu- human faces. Biol Psychiatry 2008; 63:3–5. rodev Disord 2000; 30:205–223. 32 Kosfeld M, Heinrichs M, Zak PJ, Fischbacher U, Fehr E. Oxytocin increases trust 57 Schopler E, Reichler RJ, DeVillis RF, Daly K. Toward objective classification of in humans. Nature 2005; 435:673–676. childhood autism: Childhood Autism Rating Scale (CARS). J Autism Neurodev 33 McCullugh ME, Churchland PS, Mendez AJ. Problems with measuring peripheral Disord 1980; 10:91–103. oxytocin: can the data on oxytocin ad human behavior be trusted?. Neurosci 58 Skuse D, Warrington R, Bishop D, Chowdhury U, Lau J, Mandy Wet al. The Biobehav Rev 2013; 37: 1485–1492. Developmental, Dimensional and Diagnostic Interview (3di): A novel computer- 34 Israel S et al. The oxytocin receptor contributes to prosocial fund allocations in the ized assessment for autism spectrum disorders. J Am Acad Child Adolesc Psychiatry dictator game and the social value orientations task. PLoS One 2009; 4:e5535. 45 – 35 Kogan A, Saslow LR, Impett EA, Oveis C, Keltner D, Rodrigues-Saturn S. Thin-slicing 2004; : 548 558. study of the oxytocin receptor (OXTR) gene and the evaluation and expression of 59 Han B, Eskin E. Random-effects model aimed at discovering associations in meta- 88 – the prosocial disposition. Proc Natl Acad Sci USA 2011; 108: 19189–19192. analysis of genome-wide association studies. Am J Hum Genet 2011; : 586 598. 36 Tost H, Kolachana B, Hakimi S, Lemaitre H, Verchinski BA, Mattay VS et al. 60 Li MX, Sham PC, Cherny SS, Son YQ. A knowledge-based weighting framework to 5 A common allele in the oxytocin receptor gene (OXTR) impacts prosocial boost the power of genome-wide association studies. PLoS One 2010; : e14480. temperament and human hypothalamic-limbic structure and function. Proc Natl 61 Li MX, Gui HS, Kwan JS, Sham PC. GATES: a rapid and powerful gene-based 88 Acad Sci USA 2010; 107: 13936–13941. association test using extended Simes procedure. Am J Hum Genet 2011; : – 37 Rodrigues SM, Saslow LR, Garcia N, John OP, Keltner D. Oxytocin receptor genetic 283 293. variation relates to empathy and stress reactivity in humans. Proc Natl Acad Sci 62 Ge D 2009 MetaP [Software]. Available from http://compute1.lsrc.duke.edu/soft USA 2009; 106: 21437–21441. wares/MetaP/metap.php. 38 Chen FS, Johnson SC. An oxytocin receptor gene variant predicts attachment 63 Egger M, Smith GD, Schneider M, Minder C. Bias in meta-analysis detected by a 315 – anxiety in females and autism-spectrum traits in males. Soc Psychol Personal Sci simple, graphical test. Brit Med J 1997; : 629 634. fi 2011; 3:93–99. 64 Ioannidis JPA, Trikalinos TA. An exploratory test for an excess of signi cant 39 Bakermans-Kranenburg MJ, van Ijzendoorn MH. Oxytocin receptor (OXTR) and findings. Clin Trials 2007; 4:245–253. serotonin transporter (5-HTT) genes associated with observed parenting. Soc 65 Anney R, Klei L, Pinto D, Regan R, Conroy J, Magalhaes TR et al. A genome-wide Cogn Affect Neurosci 2008; 3:128–134. scan for common alleles affecting risk for autism. Hum Mol Genet 2010; 19: 40 Green L, Fein D, Modahl C, Feinstein C, Waterhouse L, Morris M. Oxytocin and 4072–4082. autistic disorder: alterations in peptide forms. Biol Psychiatry 2001; 50:609–613. 66 Wu N, Li Z, Su Y. The association between oxytocin receptor gene polymorphism 41 Guastella A, Einfeld SL, Gray KM, Rinehart NJ, Tonge BJ, Lambert TJ et al. Intranasal (OXTR) and trait empathy. J Affect Disord 2012; 138:468–472. oxytocin improves emotion recognition for youth with autism spectrum dis- 67 Chen FS, Barth ME, Johnson SL, Gotlib IH, Johnson SC. Oxytocin receptor (OXTR) orders. Biol Psychiatry 2010; 6:692–694. polymorphisms and attachment in human infants. Front Psychol 2011; 2: 200. 42 Andari E, Duhamel JR, Zalla T, Herbrecht E, Leboyer M, Sirigu A. Promoting social 68 Feldman R, Zagoory-Sharon O, Weisman O, Schneiderman I, Gordon I, Maoz R behavior with oxytocin in high-functioning autism spectrum disorders. Proc Natl et al. Sensitive parenting is associated with plasma oxytocin and polymorphisms Acad Sci USA 2010; 107: 4389–4394. in the OXTR and CD38 genes. Biol Psychiatry 2012; 72: 175–18. 43 Skuse DH, Lori A, Cubells JF, Lee I, Conneely KN, Puura K et al. Common poly- 69 Ronald A, Happé F, Bolton P, Butcher LM, Price TS, Wheelwright S et al. Genetic morphism in the oxytocin receptor gene (OXTR) is associated with human social heterogeneity between the three components of the autism spectrum: a recognition skills. Proc Natl Acad Sci USA 2013; 111: 1987–1992. twin study. J Am Acad Child Psy 2006; 45:691–699. 44 Ylisaukko-oja T, Alarcón M, Cantor RM, Auranen M, Vanhala R, Kempas E et al. 70 Mazefsky CA, Williams DL, Minshew NJ. Variability in adaptive behavior in autism: Search for autism loci by combined analysis of autism genetic resource exchange Evidence for the importance of family history. J Abnorm Child Psych 2008; 36: and finnish families. Ann Neurol 2006; 59: 145–155. 591–599.