Background
Autism is a highly varied neurodevelopmental condition characterised by deficits in social interaction and communication, alongside unusually repetitive behaviour and extremely narrow interests. Other characteristics of autism include a resistance to unexpected change and atypical sensory sensitivity (DSM-5, 2013).
One of the main characteristics of autism is difficulties in social cognition, and in particular cognitive empathy [
1]. Cognitive empathy is defined as the ability to identify the mental state of the other [
2]. One of the most well-validated and widely used measures of cognitive empathy is the ‘Reading the Mind in the Eyes’ Test (Eyes Test) [
3]. Individuals diagnosed with autism tend to score lower on this measure than controls [
3,
4], and other measures of social cognition show the same pattern [
1,
3,
5]. A recent study found that certain subgroups within autism score lower than others [
6] and a whole genome association study of performance on the Eyes Test suggests modest but significant heritability [
7].
Autism manifests considerable heterogeneity, varying in clinical presentation across a spectrum of behaviours, as well as in levels of intellectual impairment and degree of delay in language development [
6,
8]. In line with this, the genetic aetiology of autism is heterogeneous and hundreds of genes are hypothesised to be implicated [
9], with about 50% of the genetic effect attributed to common genetic variations [
10]. This phenotypic and genetic heterogeneity is also evident in studies of the brain [
11]. Studies of the anatomical and functional brain differences in autism versus typically developing individuals yield mixed results [
12,
13]. One way of gaining additional insight into the biological basis of autism is by combining data regarding genetic variation and brain imaging in a single analysis [
14,
15]. This is the aim of the current investigation.
Oxytocin
Here, we focused specifically on the oxytocin receptor (
OXTR) gene, as it has been previously linked with social cognition and behaviour in the typical population, as well as with autism [
16]. Oxytocin (OXT) is a nonapeptide with a long evolutionary history and a well-established role in animal and human social behaviour and cognition [
17]. OXT has a major role in the ‘social brain’, i.e. in brain regions that have been clearly associated with social cognition [
18,
19]. The ‘social brain’ includes areas such as the amygdala, insula, medial prefrontal cortex, superior temporal sulcus, anterior cingulate cortex, temporoparietal junction, and inferior parietal lobule [
20,
21]. The effects of OXT on the social brain are likely mediated by its receptor—the OXTR. Indeed, a recent study using in vivo arterial spin labelling to identify changes in cerebral blood flow following intranasal administration of OXT implicated many areas within the social brain [
22], suggesting a broad expression pattern of the
OXTR. An analysis of
OXTR expression patterns using RNAseq revealed that it is broadly expressed in subcortical and cortical regions [
23].
Association of OXTR with both social cognition and autism
Single nucleotide polymorphisms (SNPs) in the
OXTR have been associated with autism in different populations [
24‐
30], especially related to the social domain in autism [
31,
32], although null-findings are also reported [
33‐
35]. Epigenetic markers on the
OXTR have also been associated with autism [
36,
37], and oxytocin (OXT) administration has been shown to improve social symptoms in autism [
38‐
45] (but see also [
46]). Similar findings of association between
OXTR and social cognition have been reported for typical populations, including an association with performance on the Eyes Test [
47,
48] and empathy [
49,
50], as well as with prosocial behaviour [
51,
52], partner bonding [
53], parent-child relationship [
54], and others.
The involvement of the OXT system in social cognition is further supported by studies in which participants are given intranasal doses of OXT. In these studies, OXT administration is shown to increase cognitive empathy, including as measured by the Eyes Test [
43,
55,
56]. Taken together, these findings suggest two conclusions. First, OXT is associated with social cognition and social behaviour across the entire spectrum of social ability—both in typical population and in autism. Second, OXT and
OXTR are also associated with individual differences in social cognition and behaviour [
57]. For example, OXT administration had a stronger effect in improving empathic accuracy for those scoring higher on the autism spectrum quotient (AQ; [
58]), i.e. typical males with higher levels of autistic traits [
59]. Similarly, those who had the most impaired eye-contact also improved the most after receiving a dose of OXT [
44].
Oxytocin in the brain
Most studies of the role of OXT in the brain are conducted in typical populations and utilise intranasal administration of OXT. These studies typically find that OXT induces decreased activation in the amygdala during emotion processing, although this effect may differ in men and women [
18,
60,
61]. Few studies have examined the effects of OXT in autism, and these usually show that OXT administration is associated with the recovery of a typical pattern of activation in certain brain areas [
40,
62]. Moreover, a recent study extended these findings by showing that the effect of OXT administration on brain function (increased activity and connectivity between dorsal anterior cingulate cortex (ACC) and dorsomedial prefrontal cortex (dmPFC)) is dependent on
OXTR genotype [
63].
One other previous study combined imaging and genetics to study
OXTR in autism, and that study focused on the reward circuitry and especially the nucleus accumbens (NAcc) [
64]. Their findings show an
OXTR-dependent change in the connectivity of the reward circuitry in children with autism during resting state. Apart from this study, all other imaging genetics studies of
OXTR conducted in typical Caucasian or non-Caucasian populations [
50,
65‐
73] implicate the structure and function of the amygdala and the hypothalamus as most associated with
OXTR genotype, but also find associations with other parts of the social brain, such as the striatum and dmPFC. Moreover, a recent study reported a sex-specific association between the
OXTR SNP rs2254298 and connectivity in the default mode network (DMN) [
72]. Due to the broad distribution of the
OXTR and evidence for its potential to affect many brain areas [
22,
23,
60,
72], we chose an unconstrained whole-brain analysis approach. The aim of the current study was to better understand the complex interaction between oxytocin genotype, brain function, and autism, by integrating
OXTR genotype and brain imaging data in a sample of adolescents (aged 12.01–18.53 years) with and without an autism diagnosis. Here, we focused on social cognition, and specifically the ability to recognise emotions, which is a hallmark difficulty in autism [
3]. This is one of the first studies, to our knowledge, to take an imaging genetics approach to better understand the oxytocin-related aetiology of social cognition deficits in autism.
Results
Participants in the autism group were significantly less accurate in their mental state judgements on the Eyes Test than the control group (M = 24.03, SD = 6.99 and M = 27.76, SD = 2.29 respectively, t (61)= − 2.90, Cohen’s d = 0.72, p = .005). However, once genotype was considered as well (and sex and age were controlled for, as in the imaging analyses below), this effect disappeared (all p values > .083).
Genotype distributions
Autism and control groups did not differ on genotype distribution for any of the examined SNPs, AQ ratings, or RMET scores. The exception was for rs7632287, for which the major frequency (low risk) genotype appeared more frequently than expected in the autism group, and scored higher on the AQ, as compared to the low-frequency (high risk) genotype (see Table
3).
Table 3
Distributions by genotype
rs53576 | GG | 15 | 6 | 31 | 80.76 |
A | 23 | 26 | 23.67 | 81.1 |
| Χ2 = 0.059 | F = 3.11 | F = .005 |
rs2268491 | CC | 27 | 28 | 23.84 | 80.72 |
T | 12 | 5 | 31.12 | 82.62 |
| Χ2 = 0.12 | F = 2.73 | F = .15 |
rs2254298 | GG | 27 | 27 | 24.09 | 80.43 |
A | 11 | 5 | 31.13 | 85.42 |
| Χ2 = 0.186 | F = 2.38 | F = 1.07 |
rs7632287 | GG | 28 | 15 | 29.72 | 82.86 |
A | 11 | 18 | 19.38 | 79.13 |
| Χ2 = 5.156* | F = 7.89* | F = .758 |
rs2228485 | TT | 22 | 16 | 27.08 | 81.74 |
C | 17 | 16 | 24.27 | 80.21 |
| Χ2 = 0.290 | F = .535 | F = .725 |
Imaging genetic analysis
For each of the five SNPs that were analysed, a main effect of diagnosis, genotype, and the interaction between the two factors was examined (2 × 2). In none of the analyses was diagnosis a significant predictor of activation. We report the nominal p value for these analyses and interpret the results based on a more stringent Bonferroni-corrected significance criterion of p = .01, in order to control for the testing of the five SNPs.
Discussion
The current investigation aimed to understand the links between oxytocin receptor genotype, brain activity in response to an explicit cognitive empathy task, and autism. As autism encompasses a spectrum of manifestations, we expected to find subgrouping within those diagnosed with autism. We found that diagnostic status interacts with OXTR genotype to predict activation within the right supramarginal gyrus and the right inferior parietal lobule during a mental state judgment task. Two (rs2268491 and rs2254298, with rs53576 showing the same effect nominally) out of the five SNPs examined showed a similar effect of differential activation based on diagnostic status, and two of the SNPs (rs2254298 and rs53576) were associated with activation, all within the rSMG and rIPL. Interestingly, the interaction showed a crossover effect, whereby an allele showing overactivation in the control group shows underactivation in the autism group and vice versa. The effects not only localised to the rSMG, but largely to the same cluster within the rSMG (with some distinct effect for the interaction between rs53576 and diagnosis, which was only nominally significant). The degree of the overlap in effect between rs2254298 and rs2268491 is expected as these two SNPs are in high LD (R2 = .98, D’ = 1). Other SNP pairs show moderate LD (rs53576 with rs2268491; R2 = .021, D’ = .56 and with rs2254298; R2 = .019, D’ = .54). As such, these consistent findings suggest an overall effect for the OXTR gene within this possibly functional locus, and in particular implicate the rSMG. Moreover, activation within an anatomically defined rSMG predicted diagnostic status in analysis of two of the three analysed OXTR SNPs (rs2254298, rs2268491), and the interaction between activation and genotype was significant for all three SNPs.
The effects of
OXTR genotype localise to the rSMG, and this is in line with a study showing that
OXTR methylation is associated with activity in the supramarginal gyrus and the dorsal anterior cingulate cortex (ACC) [
37]. Interestingly, an fMRI study that investigated emotional egocentricity bias (EEB) found that overcoming such bias, i.e. being able to empathise with another even when the other’s feelings differ from your own, is related to hyperactivity of the rSMG. Moreover, disrupting the activity of the rSMG using transcranial magnetic stimulation (TMS) resulted in increased bias [
88]. In the above-mentioned study, EEB was manipulated using a touch paradigm in which participants rated the pleasantness of tactile stimulation for themselves and another participant while experiencing either congruent or incongruent stimulation. The difference between congruent and incongruent conditions for self and other was used as the outcome measure. Another study reported no behavioural differences in performance on the EEB task between adults with and without an autism diagnosis, nor did they find differences in resting state rSMG connectivity between groups [
89]. The findings of the current study suggest that the
OXTR can explain some of the within-group variability in self-other differentiation, which is not otherwise captured by comparing individuals with and without autism.
Other studies that examined
OXTR genotype within typical populations usually implicate areas of the social brain other than the rSMG, such as the amygdala or the ACC [
50,
67‐
69,
90]. One possible explanation for the results lies in the specific contrast used in the current analysis. We compared participants’ brain activity in response to similar images but using different prompts—sex judgments versus mental states judgments. It is possible that during the sex judgments, an automatic, implicit processing of mental states was performed. Therefore, the unexpected activation pattern in the current study could be a result of this different type of comparison. Importantly, an analysis of the full sample, from which the current sample was drawn, revealed differential activation within the inferior frontal gyrus, temporal pole, and retrosubicular area [
74]. It is the addition of the genotype information (within a subsample) which revealed a different effect. Therefore, a careful interpretation of the current findings in light of the aforementioned previous research is that differential activity in the rSMG, together with genotype, is a marker of self-other distinction, crucial for the ability to correctly interpret the other’s mental state (as opposed to over-relying on one’s own mental state). Recent research suggests that the rSMG is connected to other brain areas involved in empathy (anterior insula and anterior cingulate cortex) and is responsible for self-other differentiation in relation to empathic processing [
91]. It has been recently proposed that the observed deficit in empathy in autism could be due to a reduced ability to differentiate between self and other in the social domain [
92]. This interpretation of the findings is in line with research pointing to the role of oxytocin as modulating the salience of social stimuli [
57,
93,
94], and perhaps more accurately shifting one’s focus from self to other.
Several limitations beg a cautious interpretation of the current findings. It is important to note that an adolescent sample was used in this study. On the one hand, this constitutes a limitation, as developmental and pubertal stages have not been directly assessed. However, we dealt with this by controlling for age and sex, as well as creating a study-specific brain template based on the age and sex composition of the participants. On the other hand, the adolescent sample is a strength of this study as few studies have focused on this age group. In addition, while brain overgrowth in the early stages of development has been repeatedly associated with autism [
95], many of these brain differences tend to disappear as children grow older, and brain volume during adolescence is comparable to that of typically developing children [
96].Although, one study reported on accelarated cortical thinning during adolescence, as compared to typical adolescents [
99]. Later in life brain volume decreases faster in autism as compared to typical adults [
97,
98].Although another study reported on accelerated cortical thinning during adolescence, as compared to typical adolescents [
99]. As such, the current study highlights effects that persist into later stages of development, but findings should be interpreted with caution, and future research would benefit from investigating similar effects in other age groups, taking a developmental approach. Another limitation is that of sample size, although our sample size was modest as compared to other imaging genetics studies. We emphasise that, due to the exploratory nature of the current study, replication and extension studies are needed to substantiate the current findings.
Conclusions
This is one of the first studies, to our knowledge, to incorporate OXTR genotype and brain function data in order to better understand the biological underpinnings of social cognition and cognitive empathy in autism. The current study further supports the involvement of oxytocin in the aetiology of autism and simultaneously suggests a mechanism for this effect, through activation of the rSMG, an important part of the social brain, in response to a test of cognitive empathy. Future studies, utilising larger samples, are needed to substantiate this effect and can be further used to answer additional questions, for example, regarding the role of sex and circulating levels of oxytocin on these effects. Given the preliminary findings that implicate oxytocin as a therapeutic target, a greater understanding of the mechanism by which oxytocin is involved in autism from genetics to brain function, and how it contributes to variability within autism, can advance the development of precise therapeutic (both medical and non-medical) interventions.