The role of oxytocin and oxytocin receptor gene variants in childhood-onset aggression


  • A. I. Malik,

    1. Institute of Medical Science, University of Toronto
    2. Child, Youth and Family Program, Centre for Addiction and Mental Health
    Search for more papers by this author
  • C. C. Zai,

    1. Department of Psychiatry, University of Toronto
    2. Neurogenetics Section, Neuroscience Research Department, Centre for Addiction and Mental Health, Toronto, Canada
    Search for more papers by this author
  • Z. Abu,

    1. Child, Youth and Family Program, Centre for Addiction and Mental Health
    Search for more papers by this author
  • B. Nowrouzi,

    1. Child, Youth and Family Program, Centre for Addiction and Mental Health
    Search for more papers by this author
  • J. H. Beitchman

    Corresponding author
    1. Institute of Medical Science, University of Toronto
    2. Child, Youth and Family Program, Centre for Addiction and Mental Health
    3. Department of Psychiatry, University of Toronto
      J. H. Beitchman, Centre for Addiction and Mental Health, 250 College Street, Room 125, Toronto, ON M5T 1R8, Canada. E-mail:
    Search for more papers by this author

J. H. Beitchman, Centre for Addiction and Mental Health, 250 College Street, Room 125, Toronto, ON M5T 1R8, Canada. E-mail:


Aggressive antisocial behaviours are the most common reasons why adolescents are referred to mental health clinics. Antisocial behaviours are costly in social and financial terms. The aetiology of aggressive behaviours is unknown but growing evidence suggests it is heritable, and certain genetic variants have been implicated as contributing factors. The purpose of this study was to determine whether genes regulating the hormone oxytocin (OXT) were associated with aggressive antisocial behaviour. The case-control study sample consisted of 160 cases of children displaying extreme, persistent and pervasive aggressive behaviour. This case sample was compared with 160 adult controls. We used polymerase chain reaction (PCR) to determine the genotype for three oxytocin gene (OXT) single nucleotide polymorphisms (SNPs): rs3761248, rs4813625 and rs877172; and five oxytocin receptor gene (OXTR) SNPs: rs6770632, rs11476, rs1042778, rs237902 and rs53576. Genotypic analyses were performed using stata, while differences in haplotypic and allelic frequencies were analysed using Unphased. We also performed within-case analyses (n = 236 aggressive cases) examining genotypic and allelic associations with callous-unemotional (CU) scores (as measured by the psychopathic screening device). OXTR SNPs rs6770632 and rs1042778 may be associated with extreme, persistent and pervasive aggressive behaviours in females and males, respectively. These and haplotype results suggest gender-specific effects of SNPs. No significant differences were detected with respect to CU behaviours. These results may help to elucidate the biochemical pathways associated with aggressive behaviours, which may aid in the development of novel medications.

Antisocial behaviours are the most common reasons why adolescents are referred to mental health clinics (Kazdin et al. 2006). Antisocial behaviours are costly in human and financial terms (Beitchman et al. 1992; Foster et al. 2005, 2007); public expenditures for children with conduct disorder (CD) have been estimated to be $70 000 more than expenditures for non-aggressive children in the USA, per child over a 7-year period. We are particularly interested in the category of aggressive antisocial behaviour.

The aetiology of aggressive behaviours remains elusive although environmental factors such as family instability, family conflict and parental abuse have been implicated (e.g. Aguilar et al. 2000). However, not all children who are maltreated develop aggressive behaviours (Widom 1997). There is growing evidence that genetic variation contributes to the variability in the development of antisocial aggressive behaviour (e.g. Tuvblad et al. 2011).

With regard to the genetics of aggressive behaviour, different studies have used varying phenotypes of aggressive behaviours, reporting discrepant results, e.g. discrepant results for serotonin and aggression (Twitchell et al. 2001 vs. Beitchman et al. 2006). In order to address this issue, we have identified an extreme, persistent and pervasive phenotype of aggressive behaviour, which we hypothesize will be stable, replicable and susceptible to genetic vulnerabilities. As such, potential genetic markers responsible for aggressive behaviour should be over-represented in children exhibiting this phenotype.

While the aggressive behaviours defined this way may be heritable, these behaviours are known to be heterogeneous and, thus, still represent a heterogeneous category of behaviours. One suggested method of defining important subgroups of aggressive antisocial youth focuses on the presence or absence of callous-unemotional (CU) traits (Frick & White 2008). CU traits represent a homogenous category of aggressive behaviours and constitute the broader clinical construct of psychopathy, marked by insincere charm, and lack of guilt and empathy (Frick et al. 1994, 2000).

Both aggressive and CU behaviours have been shown to be highly heritable, with heritability estimates of just below 50% (Larsson et al. 2006; Rhee & Waldman 2002) and about 80% when combined (Viding et al. 2005, 2008). Many studies suggest that aggressive behaviours may be associated with genetic polymorphisms that result in disrupted neurotransmitter systems. For example, Caspi et al. (2002) studied a functional polymorphism in the monamine oxidase A (MAO-A) gene and found that maltreated males with the genotype conferring high expression levels of MAO-A were less likely to develop antisocial problems. Other findings support associations between aggressive antisocial behaviours and genes such as serotonergic genes (e.g. Beitchman et al. 2006) and dopaminergic genes (e.g. Zai et al. 2012).

Oxytocin and aggressive behaviours

In humans, the oxytocin (OXT) nasal spray has been shown to amplify prosocial behaviours such as trust (Kosfeld et al. 2005; Mikolajczak et al. 2010). Dysregulation of OXT as a consequence of genetic variability may possibly disrupt prosocial behaviours and predispose an individual to antisocial behaviours.

There have been very few studies linking OXT to aggressive behaviours in humans. Fetissov et al. (2006) have found that levels of oxytocin-reactive auto-antibodies were increased in aggressive male subjects compared with controls, suggesting that hypo-oxytocinergic function may account for aggressive behaviour. Accordingly, Lee et al. (2009a) have found that oxytocin levels in cerebrospinal fluid (CSF) were inversely correlated with a life history of aggressive behaviours. Indeed, single nucleotide polymorphisms (SNPs) within the OXTR gene have already been identified that are significantly associated with CU behaviour in highly aggressive children (Beitchman et al. 2012).

We will examine the OXT and OXTR genes in association with (1) extreme, persistent and pervasive aggressive behaviour and (2) CU traits in highly aggressive children.



A total of 236 children were recruited throughout the duration of the study (162 males, 74 females). Inclusion criteria were as follows: age 6–16 (mean age ± standard deviation (SD), 11.45 ± 3.04 years); a minimum 2-year history of aggressive behaviours; at or above the 90th percentile on subscales of aggressive behaviours in both the Achenbach Child Behaviour Checklist (CBCL; Achenbach and Rescorla 2001; mean T score ± SD, 74.47 ± 6.38) and Teacher Report Form (TRF; mean T score ± SD, 73.52 ± 7.90; 34); an intelligence quotient of at least 70 (mean IQ score ± SD, 98.26 ± 14.64) as measured by a two-subtest short form of the Wechsler intelligence scale for Children – 3rd edition (Wechsler 1991). Children with chronic medical illnesses and psychiatric disorders, such as schizophrenia, mania, autism and Tourette's syndrome, were excluded. All participants' parents or guardians were asked to fill out a psychopathic screening device (PSD; Frick et al. 2000, 2001) and their scores on the callous-unemotional subscale were used for analysis (mean score ± SD, 1.06 ± 0.52). Ethnicity was assessed using parental self-reported data; the sample was 82% Caucasian, 8% African-Canadian and 10% other and mixed ethnicities.

Reviews of the participants' health records and administration of the Diagnostic Interview Schedule for Children (DISC) checklist to the participant's primary caretaker were used to obtain information about diagnoses for disruptive behaviour disorders according to criteria by the Diagnostic and Statistical Manual for Mental Disorders-fourth edition (DSM-IV). Of the children for whom complete diagnostic information was available, 83% had been diagnosed with a disruptive behaviour disorder [conduct disorder (CD) or oppositional defiant disorder (ODD)] and, of those, 74% were comorbid for attention deficit hyperactivity disorder (ADHD). On the basis of the Children's Depression Inventory, 16% of the subjects obtained clinically elevated depression scores (i.e. total T score >65).

DNA for child cases

Genomic DNA was extracted from saliva, buccal cells or blood preserved in ethylenediaminetetraacetic acid (EDTA) for each case. Saliva samples were collected using Oragene 500 DNA kits (DNA Genotek, Kanata, Canada) and extracted as per manufacturer's instructions. DNA was extracted from blood samples using a non-enzymatic, high-salt procedure by Lahiri and Nurnberger (1991), whereas the Qiagen DNA isolation kit (Qiagen Inc., CA, USA) was used for DNA extraction from cheek swabs.

Adult controls

Control DNA samples consisted of 160 healthy adults (mean age ± SD, 25.87 ± 8.99 years) controls with no reported history of childhood-onset aggressive behaviours. These adult controls were screened for the absence of past and current major psychiatric disorders using semi-structured interview for DSM-IV diagnoses (SCID-I; First et al. 1997; Maxwell 1992). Adult controls were matched by gender and ethnicity with the child cases for comparison purposes (160 pairs: 106 male pairs, 54 female pairs).


Oxytocin is a nonapeptide and the oxytocin-neurophysin I (or OXT) gene has been mapped to chromosome 20p13 and is separated from the vasopressin-neurophysin II (AVP) gene by about 12 kbp; it contains two introns and three exons and spans less than 1 kbp (Rao et al. 1992). The human OXTR is a 389-amino-acid polypeptide with seven transmembrane domains and belongs to the class I G protein-coupled receptor (GPCR) family (Gimpl and Fahrenholz 2001); the OXTR gene is located on chromosome 3p25, spans less than 20 kbp and contains three introns and four exons (Inoue et al. 1994).

All case and control DNA samples were genotyped for three oxytocin (OXT) SNPs with the following rs numbers and chromosome positions taken from Pubmed: rs3761248 (chromosome reference position = 2998393), rs4813625 (ChrPos = 2997720) and rs877172 (ChrPos = 2997890); and five SNPs in the oxytocin receptor (OXTR) gene: rs6770632 (ChrPos = 8768724), rs11476 (ChrPos = 8763198), rs1042778 (ChrPos = 8769545), rs237902 (ChrPos = 8784184) and rs53576 (ChrPos = 8779371). Technically, OXTR SNP rs11476 is actually found in the Caveolin 3 (CAV3) gene, located adjacent to the OXTR gene on chromosome 3p25. However, this SNP may be part of a regulatory sequence involved in the transcription of the OXTR gene.

All eight SNPs were genotyped using Life Technologies' TaqMan® assays as per manufacturer's (Life Technologies, Applied Biosystems Inc., Foster City, CA, USA) directions in a total volume of 10 µl and using 20 ng DNA template. Genotypes were subsequently determined using the Allelic Discrimination Program within the ABI7500 Prism Sequence Detection System (Life Technologies). Ten percent of the DNA samples were genotyped again for quality control purposes.

Statistical analyses

Values for Hardy–Weinberg equilibrium and linkage disequilibrium of all eight markers were generated using HaploView version 4.1 (Barrett et al. 2005). Genotypic differences in frequencies between adult controls and cases were assessed using stata v. 11.0 for Windows (StataCorp. 2009). For cells with expected counts less than 5, the two-tailed Fisher's exact P-value was reported; for all others, the P-value from the chi-squared test was reported.

Also, associations between genotypes and scores on measures such as CU traits were analysed using one-way analysis of variance (anova) in stata v. 11.0, and the Kruskall–Wallis non-parametric test was applied if the distribution of a measure failed the test of homogeneity of variance.

Allelic and haplotype analyses were performed using unphased v. 3.1 (Dudbridge 2003), using a two-window sliding-window strategy specifically for haplotypic analyses. Haplotypes with frequencies of less than 5% were excluded from the analyses.

Power calculations were carried out using quanto version 1.2.3 (Gauderman & Morrison 2006). Our matched case-control sample (160 pairs) had over 80% power to detect a genetic odds ratio (OR) of at least 1.70 (minor allele frequency = 0.20, two-sided alpha = 0.05, additive model). For within-case analyses, our sample of 236 aggressive child cases had over 80% power to detect an R2 of 0.033 (minor allele frequency = 0.20; two-sided alpha = 0.05; additive model; mean CU score ± SD: 1.07 ± 0.52).

This study was approved by the Research Ethics Board of the Centre for Addiction and Mental Health (CAMH) and all subjects and guardians provided written, informed consent before participating.


The three OXT and five OXTR SNPs were in Hardy–Weinberg equilibrium (P > 0.1116). Figure 1a,b shows the linkage disequilibrium values of three OXT SNPs and five OXTR SNPs, respectively. The Bonferroni-corrected alpha level becomes (0.05/8 SNPs) 0.00625. The following P-values are Bonferroni-corrected.

Figure 1.

Linkage disequilibrium values for SNPs. Figures indicate linkage disequilibrium D′ values for (a) 3 OXT SNPs and (b) 5 OXTR SNPs, using the entire case-control sample.

Case-control analyses

Allelic and genotypic analyses

The C allele for OXTR rs1042778 was over-represented in the male aggressive cases (P = 0.016) as compared with male adult controls, as was the CC genotype (P = 0.072).

Also, allele T for OXTR rs6770632 was over-represented in female aggressive cases as compared with female controls (P = 0.32). As well, genotype TT for rs6770632 was more prevalent in female cases, although these results did not reach significance. Please see Table 1 for a summary of results for genotypic and allelic analyses.

Table 1.  Genotypic and allelic frequencies of the three OXT SNPs and five OXTR SNPs for cases and controls
  1. Count of adult controls and count of aggressive cases in each genotypic or allelic group. χ2 values (df = 2) and P-values are provided under counts, except where a Fisher's exact test was performed. Bonferroni-corrected significant P-values are underlined. Asterisk (*) next to P-values indicates Fisher's exact test performed because of expected counts being less than 5 in one or more cells.

OXT rs4813625CC3327241998C13013291903942
  χ 2 (P)2.63 (1.000)1.69 (1.000)1.03 (1.000) P 1.0001.0001.000
OXT rs877172GG2614169105G1159274594133
  χ 2 (P)4.42 (0.88)2.61 (1.000)1.94 (1.000) P 0.4080.9201.000
OXT rs3761248AA9710364703333A2362451551638182
  χ 2 (P)1.000*1.000*1.000* P 1.0001.0001.000
OXTR rs11476AA23261218118A12612776845043
  χ 2 (P)0.38 (1.000)1.42 (1.000)0.96 (1.000) P 1.0001.0001.000
OXTR rs6770632GG969671582538G2432491711597290
  χ 2 (P)2.33 (1.000)0.968*0.208* P 1.0001.000 0.032
OXTR rs1042778AA233912251114A119149711014848
  χ 2 (P)6.24 (0.352)9.47 (0.072)1.41 (1.000) P 0.112 0.016 1.000
OXTR rs53576CC708848372231C2052251351477078
  χ 2 (P)4.42 (0.880)1.74 (1.000)1.000* P 0.5921.0001.000
OXTR rs237902AA676742432524A2001951261277468
  χ 2 (P)0.87 (1.000)1.000*1.000* P 1.0001.0001.000

Haplotype analyses

We used a sliding-window approach and tested two-marker haplotypes within the OXT and OXTR genes for all eight SNPs for the case-control analyses. The haplotype consisting of OXTR rs1042778 allele C and OXTR rs6770632 allele G was over-represented in male aggressive cases vs. controls (window P = 0.032, specific P = 0.0024, OR = 1.9). In addition, the haplotype consisting of OXTR rs1042778 allele C and OXTR rs53576 allele T was over-represented in male aggressive cases vs. controls (window P = 0.036, specific P = 0.016, OR = 2.5).

For a summary of all Bonferroni-corrected P-values, please refer to Figure 2a–c .

Figure 2.

Figures indicate significance levels (negative log of P-values;log(P)) for tests of association between genotypes, alleles and two-marker haplotypes. Dotted orange line indicates the threshold alpha value of 0.05; Bonferroni-corrected P-values are presented. Figures have been generated for (a) overall sample, (b) male subsample and (c) female subsample. Four-pointed stars indicate significant results for main allelic or genotypic effects. Colour scheme follows legend at the top right (that is, red star indicates significant allelic effects).

Within-case analyses

We were unable to find any significant genotypic, allelic and haplotypic associations with CU scores within the sample of 236 cases. These results are not shown.

Caucasian subsample

It should be noted that we performed analyses on the Caucasian subsample and found significant results at the same markers as the overall sample; this included the gender-specific differences in significance.


We have shown that OXTR polymorphisms are associated with extreme, persistent and pervasive childhood-onset aggressive behaviours. To our knowledge, this is the first study to show associations between OXTR gene variants and childhood-onset aggressive behaviours.

It is interesting to note that in the male subsample (i.e. male aggressive cases vs. adult controls), the allelic main effects for rs1042778 were still significant when compared to the Bonferroni-adjusted alpha value. Haplotypes in the male subsample also survived the Bonferroni correction, including a haplotype consisting of rs1042778 and rs6770632 and a haplotype consisting of rs1042778 and rs53576. In the female subsample, allelic effects for rs6770632 were significant even after being compared to the Bonferroni-adjusted alpha value.

Both OXTR SNPs rs1042778 and rs6770632, which were found to be associated with aggressive behaviour, are found in the 3′-untranslated region (UTR) of exon 4 in the OXTR gene. In the OXTR gene, exons 3 and 4 encode amino acids (Gimpl & Fahrenholz 2001). Exon 4 contains the sequence encoding the seventh transmembrane domain, the COOH terminus, and the entire 3′-non-coding region, including the polyadenylation signals, of the OXTR polypeptide.

Previous literature on OXTR SNPs

SNP OXTR rs1042778 has also been tested in autism-related studies; the C allele for rs1042778 has been found to be significantly associated with susceptibility to autistic spectrum disorders in children (Campbell et al. 2011; Lerer et al. 2008) although these findings did not survive correction for multiple analyses. We show that the C allele is more prevalent in aggressive children. Taken together, the findings suggest that the C allele for rs1042778 may be associated with disrupted social behaviour.

However, it should be noted that the C allele in rs1042778 (or the G allele on the reverse strand) has been associated with a higher number of prosocial responses in adult males and females (Israel et al. 2009). The reason for the difference in results may be due to the phenotypic differences in the two samples. Our sample of children (mean age = 11.45) was chosen specifically because of their clinically significant aggressive behaviour, whereas Israel et al.'s sample consisted of healthy adults (mean age = 26) in the university. The samples may therefore not be comparable. There may also be developmental and maturational effects that modulate the effects of these alleles although we are unaware of empirical evidence in support of this idea.

We found rs6770632 to have nominally significant genotypic effects and allelic effects (T allele more prevalent in cases) that are significant even after correction, but only in females. OXTR rs6770632 has previously been found to be non-significant in autism research (Lerer et al. 2008). The paper did not specify sex-specific results but the differences in results are most likely due to the differences in ages of the two samples and differences in the phenotype. For instance, Lerer et al.'s sample ranged from 2 to 33 years of age of individuals who were diagnosed with DSM-IV autistic disorder (N = 131) or pervasive developmental disorder-not otherwise specified (PDD-NOS; N = 21). Additionally, in contrast to our sample in which an IQ of less than 70 was an exclusion criteria, the average IQ in Lerer et al.'s sample was 67. Although the OXTR gene may affect both aggression and autism, perhaps different genetic variants are involved.

Implications of sex-specific differences

We found the results of this genetic association study to be gender-specific. The OXTR rs6770632 significant findings occurred in females only, whereas the association between OXTR rs1042778 and aggressive behaviours was significant only in males. Another interesting finding was that significant haplotypes in the male sample implicated the G allele in SNP rs6770632 as conferring risk for aggressive behaviours; however, in the female sample, the T allele conferred risk for aggressive behaviours. Previous studies have also found gender-dependent significant associations between OXTR SNPs and behavioural phenotypes (e.g. the role of OXTR SNPs in prosocial fund allocations; Israel et al. 2009).

These findings likely reflect sexual dimorphism. Expression of OXTR has been found to be relatively higher in females than in males (Carter 2007; Lee et al. 2009b; Zingg and Laporte 2003), possibly because of the differences in the effects of genetic variation in the OXTR gene. Also, the expression of OXT is regulated by steroid hormones, including oestrogen and thyroid hormones (Mohr & Schmitz 1991; Richard & Zingg 1990), which may contribute to lessened effects of certain SNPs in females as compared with males or vice versa.


There are a few limitations of this study, including the small sample size (160 cases and 160 controls for case-control analyses; 236 cases for within-cases analyses), which makes replication with a larger sample size necessary. Also, the analyses for CU traits were conducted in highly aggressive children because these children were hypothesized to exhibit a higher prevalence of CU traits. As this sample was selected for aggressive behaviour, there was limited variability in the distribution of CU traits which may have contributed to the failure to find significantly different genotypes in our within-cases analyses. However, to truly understand the genes involved in CU behaviour, future studies should also examine CU traits in clinical as well as non-clinical samples in which CU traits would be normally distributed.

Our use of adult controls may be considered a limitation. However, because it is unlikely that genetic sequence would change with age particularly at these SNPs, this should be an acceptable comparison group. Also, in the event that adult participants displayed aggressive behaviours as children but did not report it, case-control differences would have been weakened not amplified; hence, these findings are conservative.

Another limitation may be the smaller number of females compared with males in the sample. As an example, for OXTR rs1042778, the results were significant for the overall sample and the male sample but not the female sample (see Table 1 for P-values). This suggests that the results for the overall sample were being largely driven by the male sample. We have accordingly analysed the male and female subsamples separately. In doing so we reduced the sample size (i.e. we had smaller subsamples) and may have lost power. Hence, it is possible that some results that were non-significant may be significant in larger samples. Thus, these findings should be tested in a larger sample.

Speculations and implications for future investigations

Amygdalar volume has been implicated in anxiety and the perception of threats, or fear (Davis et al. 2010), which may be a trigger for aggressive behaviour. Interestingly, an OXTR SNP rs2254298 has been associated with limbic structure volumes, including bilateral amygdalar volume in healthy Asian adults but not in healthy Caucasian adults (Inoue et al. 2010; Tost et al. 2011). Unfortunately, the SNPs identified in our study have not been tested in relation to amygdalar volume in Caucasians. Consequently, it is not known if these SNPs contribute to aggressive behaviour through amygdalar volume or through processes other than the amygdala; alternatively, they may be in linkage disequilibrium with SNPs that are associated with amygdalar volume. These issues bear further research.

In addition, a hypo-oxytocinergic endophenotype may be involved in the development of antisocial behaviours. In support of this hypothesis, early adverse environments (i.e. maltreatment) may be associated with lower central oxytocin levels (Heim et al. 2009; Winslow et al. 2003), and aggressive behaviours are also associated with lower central and peripheral oxytocin levels (Fetissov et al. 2006; Lee et al. 2009a).

Perhaps maltreatment may result in the decrease of the biological activity of oxytocin, either through the regulation of oxytocin production or through the oxytocin receptor sensitivity; this hypo-oxytocinergic endophenotype may subsequently be associated with aggressive behaviours. It may therefore be deduced that OXTR SNPs rs1042778 and rs6770632 may contribute to a hypo-oxytocinergic endophenotype, either directly or through linkage disequilibrium with other SNPs.


Our research findings suggest a role of the OXTR gene in the development and/or in the expression of aggressive antisocial behaviours. No such role could be discerned for the SNPs of the OXT gene investigated in this report. These findings should inform future research directed at elucidating potential contributions of genetic factors in the aetiology of aggressive behaviour.

Ideally, genetic research may help unravel biochemical pathways associated with aggressive and antisocial behaviours and may ultimately aid in the development of novel medications.


We would like to acknowledge grants received from the Centre for Addiction and Mental Health (CAMH) Foundation, the Natural Sciences and Engineering Research Council of Canada (NSERC) and the University of Toronto.