A functional polymorphism of the OXTR gene is associated with autistic traits in Caucasian and Asian populations
Abstract
There is increasing evidence for associations between polymorphisms of the oxytocin receptor (OXTR) gene and autism spectrum disorder, but to date no study has established links with autistic traits in healthy subjects and potential cultural differences. The present research firstly investigated associations between three widely studied OXTR SNPs and autistic and empathic traits (rs53576 (G/A); rs2254298 (G/A); rs2268498 (T/C)) in two independent studies on male and female Caucasian (n = 537) and Chinese students (n = 280). Autistic and empathic traits were measured in all subjects in the two independent groups using the Autism -Spectrum Quotient (AQ) and the Interpersonal Reactivity Index (IRI) respectively, together with their sub-scales. For both sites, genotyping of the OXTR SNPs was conducted on buccal swab samples using a Cobas Z 480 Light Cycler following automated DNA extraction. Associations at the genotype level with autism trait scores were found in Caucasian subjects for rs2268498 only, with TT carriers having the lowest AQ scores compared with those carrying at least one C-allele. This finding was independently replicated in the Chinese sample although a smaller proportion carried the C-allele compared with the Caucasian sample. Some minor associations were found between empathy trait scores and the three SNPs but were not consistent between the samples. These findings show for the first time that the rs2268498 SNP localized in the promoter flanking region of the OXTR gene is associated with autistic traits in different ethnic/cultural groups. This provides further support for the role of the OXTR gene in relation to autism.
1 INTRODUCTION
The neuropeptide oxytocin (OXT) plays a key role in social cognition and motivation and its intranasal administration can enhance a range of different aspects of social cognition in both healthy (e.g., Fischer-Shofty, Shamay-Tsoory, Harari, & Levkovitz, 2010; Marsh, Henry, Pine, & Blair, 2010; Schulze et al., 2011) and clinical populations such as individuals with autism spectrum disorder (ASD) (e.g., Anagnostou et al., 2012; Guastella & Hickie, 2016; Guastella et al., 2010; Yatawara, Einfeld, Hickie, Davenport, & Guastella, 2015). The OXT receptor (OXTR) gene is located on chromosome 3p25 and includes three intronic and four exonic regions (see Ebstein, Knafo, Mankuta, Chew, & San Lai, 2012). Polymorphisms in the OXTR gene have been associated with a range of personality and social behaviors as well as ASD, although there is variability between studies and cultural differences and gender may be influential (Aspé-Sánchez, Moreno, Rivera, Rossi, & Ewer, 2016; Bakermans-Kranenburg & van Ijzendoorn, 2014; Kranz et al., 2016; Kumsta & Heinrichs, 2013; LoParo & Waldman, 2015; Yamasue, 2013). The roles of the different single nucleotide polymorphisms (SNPs) for OXTR function are also largely unknown.
Two of the most widely studied OXTR SNPs are rs53576 (dbsnp: G to A) and rs2254298 (dbsnp: G to A; both SNPs are located on intron 3), although a recent meta-analysis has questioned their role in social cognition (Bakermans-Kranenburg & van Ijzendoorn, 2014). Also of interest for the present work, another meta-analysis could not find evidence for a role of these SNPs in autism (Warrier, Chee, Smith, Chakrabarti, & Baron-Cohen, 2015). Aside from power issues in some of the earlier mentioned studies, it needs to be mentioned that the allelic distribution of rs53576 and its association with social cognition also differs between Caucasian and Asian populations [where it is known that prevalences of genotypes differ (Kim et al., 2010)]. On the other hand, both rs53576 and rs2254298 have been associated with ASD (LoParo & Waldman, 2015; Wu et al., 2005) and recent additional evidence has reported that the A-allele of rs53576 is associated with lower empathy in Israeli subjects (Uzefovsky et al., 2015). However, in a large cohort of Chinese subjects rs53576 was only found to be associated with some aspects of cognitive empathy, when additionally including its interaction with trait interdependence (Luo et al., 2015). Furthermore, to date no specific roles for either rs53576 or rs2254298 have been identified in terms of OXTR function.
Recent studies have increasingly demonstrated that rs2268498 (dbsnp: T to C) located in the promoter flanking region of the OXTR gene may play an important role in social cognition, with TT carriers showing better face emotion recognition (Melchers, Montag, Markett, & Reuter, 2013), interpersonal perception (Melchers, Montag, Felten, & Reuter, 2015) and empathic concern (a subscale of the Interpersonal Reactivity Index (IRI)) (Christ, Carlo, & Stoltenberg, 2016). Moreover, TT carriers have higher activity in the inferior occipital lobe while processing negative facial expressions (O‘Connell et al., 2012) suggesting they may process more information from emotional faces. By contrast, the C-allele is associated with higher empathic accuracy in a pain perception task (Laursen et al., 2014) and C-allele carriers rate accidentally committed harm as more blameworthy compared to TT carriers (Walter et al., 2012). Importantly, on a functional expression level the T-allele of rs2268498 is associated with lower OXTR mRNA expression (Reuter et al., 2016).
Autistic symptoms can be assessed across both healthy and clinical populations using dimensional measures such as the Autism-Spectrum Quotient (AQ) (Ruzich et al., 2015). To date only one study has investigated associations between OXTR SNPs and AQ scores and failed to find any significant associations with 17 different SNPs or 75 imputed SNPs (Verhallen et al., 2017). However, associations with rs53576, rs2254298 and rs2268498 were not specifically examined, although rs53576 was included in their group of imputed SNPs. It is therefore still unclear as to whether these three SNPs are associated with autistic traits.
The objective of the current studies was to implement a dimensional trait approach to establish potential associations between these three specific OXTR SNPs and autistic and empathic traits as well as possible cultural influences. Both self-report measures investigating autistic and empathic traits were applied, because the literature suggests that these traits are inversely associated (Wheelwright et al., 2006) and autistic patients score lower on empathic traits than healthy controls (Baron-Cohen & Wheelwright, 2004). To this end we have firstly systematically investigated the influence of the SNPs on individual differences in scores on two widely used autistic and empathy trait questionnaires in a large Caucasian (German) population sample, and then secondly carried out an independent replication study in an Asian sample (Chinese). Although the literature is quite heterogenous with respect to rs53576 and rs2254298, there is more homogeneity with respect to rs2268498. We therefore hypothesized that the TT genotype of rs2268498 would be associated with lowest autism and highest empathic trait scores. Our hypothesis was derived from the fact that four (above cited) studies associated the TT genotype with prosocial skills/traits compared to only two studies favoring the C-allele.
2 MATERIAL AND METHODS
2.1 Participants
A total of 537 mostly healthy, mostly Caucasian, (5 non-Caucasian: 1 African, 1 Arabic, 1 Ethiopian, 1 Tamil, and 1 Tataric) participants (154 males and 383 females; mean ± SD — age: 23.28 ± 5.12 years) completed self-report questionnaires for autistic and empathy traits and provided buccal swabs to investigate three SNPs on the OXTR gene. All participants were recruited from the Ulm Gene Brain Behavior Project (UGBBP). Excluding the non-Caucasian participants did not change the overall results and thus the whole sample was included in the analyses. The mean BDI-II (Beck's Depression Inventory — II) score of the Caucasian sample was 8.76 (SD = 7.84). The study was approved by the local ethics committee of Ulm University, Ulm, Germany.
The second independent sample of participants was recruited from the University of Electronic Science and Technology of China (UESTC) in Chengdu. A total of 280 mostly Han Chinese participants (189 males and 91 females; mean ± SD — age = 22.13 ± 2.33 years; 266 Han Chinese, 2 Hui, 3 Tujia, 2 Zhuang, 2 Yi, 2 Man, 1 Miao, 2 Mongol) completed the autism and empathy trait questionnaires and also provided buccal swabs to analyze the OXTR gene variants. Excluding the non-Han-Chinese participants did not change the main results. Thus, the whole sample was included in the analyses. All participants belong to the Chengdu Gene Brain Behavior Project (CGBBP). The mean BDI-II score of the Chinese sample was 6.39 (SD = 8.02). The study was approved by the local ethics committee of UESTC in Chengdu, China.
All study procedures were in accordance with the latest revision of the Declaration of Helsinki. All participants provided written informed consent.
2.2 Autism and empathy trait questionnaires
To investigate autistic traits in the Caucasian sample we administered a German version of the 50-item Autism-Spectrum Quotient (AQ) (Baron-Cohen, Hoekstra, Knickmeyer, & Wheelwright, 2006; Baron-Cohen, Wheelwright, Skinner, Martin, & Clubley, 2001). For both the total AQ score and its sub-scales higher scores indicate a stronger overall trait or sub-component. Internal consistency for the total AQ score was excellent (α = 0.77). For the Chinese sample a Mandarin version of the 50-item AQ was used (Liu, 2008) which also had an acceptable internal consistency (α = 0.68). When calculating the five original proposed AQ sub-scales some of their internal consistencies were unreliable (< 0.50) and so we used the three AQ sub-scales proposed by Austin ((poor) social skills, details/pattern, and (poor) communication/mindreading) based on 26 of the questions of the original AQ questionnaire. Moreover, we also built one small AQ score out of these 26 items (Austin, 2005). In short, individuals with high scores on the social skills dimension do not enjoy social situations. Individuals with high scores on the details/pattern dimension are more interested in numbers and patterns rather than social information and those with high scores on the communication/mindreading facet have problems with these specific skills. The internal consistencies for these subscales were generally acceptable (German: sum score of 26 items (small AQ score): α = 0.73, social skills α = 0.77, details/pattern α = 0.60, communication/mindreading α = 0.51; Chinese: small AQ score: α = 0.62, social skills α = 0.73, details/pattern α = 0.52, communication/mindreading α = 0.50). For consistency we calculated associations with AQ scores based both on the original 50 items and also the 26 items used the sub-scales. To assess empathy traits, we administered German and Mandarin versions of the Interpersonal Reactivity Index (IRI) (Davis, 1983; Siu & Shek, 2005). Internal consistencies were excellent for the German version (perspective taking α = 0.79, fantasy α = 0.84, empathic concern α = 0.85 and personal distress α = 0.77) but lower but still acceptable for the Chinese one (perspective taking α = 0.63, fantasy α = 0.61, empathic concern α = 0.63 and personal distress α = 0.56).
2.3 Genotyping
Genotyping of the OXTR SNPs was conducted on a Cobas Z 480 Light Cycler (Roche Diagnostics, Mannheim, Germany). A prior DNA extraction step was performed on a MagNA Pure 96 robot (Roche Diagnostics, Mannheim) and the simple probes were designed by TIBMolBiol (Berlin, Germany). The equipment and procedures for genetic analyses were the same at Ulm University (Ulm, Germany) and UESTC (Chengdu, China).
2.4 Statistical analyses
In an initial step, we tested for gender and age effects on autism and empathy trait scores in the two different samples using MANOVAs and correlations. We carried out MANOVAs for economic reasons here (instead of using single ANOVAs because the gender analysis was also not the focus of the present research). Next, we tested for genotype effects (on genotype and allele levels) with separate ANOVA models for each SNP (given that the SNPs have been largely investigated independently before). Finally these analyses were followed by including gender as a factor into the models to test for gene by gender interaction effects on the dependent variables. In order to control for multiple testing, we decided to apply Bonferroni corrections for three tests resulting in α = 0.017 (α = 0.05/3; for each SNP tested). Of note, we set up a directed hypothesis with respect to the OXTR rs2268498, therefore we need no further correction steps with respect to this SNP. One further note: We argue that if replication of a genetic finding on any of the scales can be observed despite different prevalences of the different genotypes in the investigated participants (Caucasian vs. Han-Chinese), the different cultural background, and differences in sociodemographic-variables, this makes a strong case in terms of generalizability of the observed effects (and thus multiple testing would not present a severe issue).
3 RESULTS
Since there were no robust/consistent findings across the two study samples in relation to IRI dimensions, these OXTR-IRI analyses are presented in the Supplementary File and the main focus is subsequently only on associations with the total AQ score (built from all the 50 items). Results regarding the small AQ score and the sub-scales according to Austin (2005) are also presented in the Supplementary File.
3.1 Gender differences and effects of age in the Caucasian and Chinese sample
Gender and age effects on the scores for the total AQ and AQ sub-scales as well as the IRI dimensions are shown in Supplementary Tables S1 and S2 in the Supplementary File.
3.2 Distribution of OXTR SNPS
Table 1 shows the distribution of all investigated SNPs in both the Caucasian and Chinese samples. All distributions were in Hardy Weinberg Equilibrium. For rs53576 the GG genotype is less common in the Chinese compared to the Caucasian sample whereas the opposite is true for the AA genotype (replicating previous findings — e.g., Kim et al., 2010; Tost et al., 2010). For rs2268498 we found that the CC genotype is also less common in the Chinese compared to the Caucasian sample. This is line with data available at pubmed.com (Reference SNP, 2016).
| SNP | Genotypes | Hardy-Weinberg | ||
|---|---|---|---|---|
| Germany | ||||
| rs53576 | GG = 229 | GA = 242 | AA = 65 | Chi2 = 0.01, p = 0.931 |
| rs2254298 | GG = 420 | GA = 111 | AA = 6 | Chi2 = 0.20, p = 0.657 |
| rs2268498 | TT = 149 | TC = 280 | CC = 108 | Chi2 = 1.29, p = 0.257 |
| China | ||||
| rs53576 | GG = 29 | GA = 108 | AA = 130 | Chi2 = 0.83, p = 0.361 |
| rs2254298 | GG = 134 | GA = 117 | AA = 29 | Chi2 = 0.21, p = 0.645 |
| rs2268498 | TT = 142 | TC = 106 | CC = 28 | Chi2 = 1.51, p = 0.220 |
- OXTR, oxytocin receptor; SNP, single nucleotide polymorphism.
- Different sample sizes are caused by unclear, and therefore excluded, results regarding the genotypes of some participants in the different SNPs.
No significant statistical interactions between the three individual SNPs were found which could pass correction for multiple testing on total AQ scores in either the Caucasian or the Chinese sample. Thus, we only report associations with individual SNPs. For the individual SNPs, when gender was included as an independent variable, no meaningful gender by genotype or gender by allele effects were observed which could meaningfully survive correction for multiple testing. However, it should be noted that some of the cell sizes were too small to test reliably for interaction effects in our samples.
The mean values and standard deviations as well as F- and p-values for the total AQ score separately for nation and genotypes are presented in Table 2 and graphically in Figure 1 (Caucasian sample) and Figure 2 (Chinese sample). For the interested reader we also present the genotype effects (separately for nation) on the sub-scales of the AQ in Supplementary Table S3 in the Supplementary File.
| Total AQ score | ||||
|---|---|---|---|---|
| Genotype | F, P | |||
| Germany | ||||
| rs53576 | GG: 15.78 ± 6.01 | GA: 16.86 ± 6.26 | AA: 17.00 ± 5.86 | F(2,533) = 2.18, p = 0.114 |
| rs2254298 | GG: 16.46 ± 6.26 | GA: 16.19 ± 5.42 | AA: 17.33 ± 8.80 | F(2,534) = 0.15, p = 0.858 |
| rs2268498 | TT: 15.28 ± 6.30 | TC: 16.93 ± 6.29 | CC: 16.64 ± 5.15 | F(2,534) = 3.69, p = 0.026 |
| China | ||||
| rs53576 | GG: 20.69 ± 5.42 | GA: 21.23 ± 5.85 | AA: 21.22 ± 5.98 | F(2,264) = 0.11, p = 0.899 |
| rs2254298 | GG: 21.70 ± 6.14 | GA: 21.09 ± 5.57 | AA: 19.14 ± 4.75 | F(2,277) = 2.37, p = 0.095 |
| rs2268498 | TT: 20.23 ± 5.45 | TC: 21.47 ± 6.28 | CC: 24.93 ± 4.25 | F(2,273) = 8.20, p < 0.001 |
- M ± SD
3.3 Associations between total AQ scores and OXTR rs53576
For the Caucasian sample on a genotype level OXTR rs53576 was not significantly associated with the total AQ score (see, Table 2). Grouping the alleles into A+ (AA/GA) versus A− (GG) led to significant associations with the total AQ score (F(1,534) = 4.33, p = 0.038; A+: 16.89 ± 6.16; A−: 15.78 ± 6.01). The A+ group showed significantly higher scores than A− carriers. The reverse grouping (G + vs. G −) did not result in a significant finding.
For the Chinese sample no significant associations were found between the OXTR rs53576 SNP and the total AQ score (neither on the genotype nor on the allele level).
3.4 Associations between total AQ scores and OXTR rs2254298
For the Caucasian sample no significant associations were observed between OXTR rs2254298 and the total AQ score; neither on the genotype nor on the allele level.
For the Chinese sample OXTR rs2254298 did not quite achieve a significant association with the total AQ score at the genotype level (see, Table 2). When dividing into the allelic groups G+ (GG/GA) and G−(AA) the effect on the total AQ score did achieve significance (F(1,278) = 4.04, p =.045), with lower scores in the small group of G− carriers (G−: 19.14 ± 4.75; G+: 21.41 ± 5.88). When dividing into the allelic groups A+ (AA/GA) and A− (GG) no significant effects were found on the total AQ score.
3.5 Associations between total AQ scores and OXTR rs2268498
In the Caucasian sample a significant association between OXTR rs2268498 and the total AQ score was observed at the genotype level. CC and TC carriers had similar scores, with both descriptively (and TC significantly) being higher to those of TT carriers (see, Table 2). This effect was most pronounced for the grouping C+ (CC/TC) versus C− (TT): (F(1,535) = 7.21, p = 0.007). C− carriers (15.28 ± 6.30) had the lowest total AQ scores compared to carriers of at least one C allele (16.85 ± 5.99). This effect would also hold for correction procedures due to multiple testing (p = 0.05/3 = 0.017).
For the Chinese sample OXTR rs2268498 SNP was also significantly associated with the total AQ score on genotype level with TT carriers again having the lowest and CC carriers the highest scores (see, Table 2). This effect would pass correction for multiple testing (p = 0.05/3 = 0.017). We also tested C+ (CC/TC) versus C− (TT) carriers and this resulted in significant effects on the total AQ score (F(1,274) = 8.00, p =.005; C+: 22.19 ± 6.07; C−: 20.23 ± 5.45). The group of C+ carriers showed higher scores. The division into T+ (TT/TC) and T− (CC) was also analyzed and revealed even stronger effects (F(1,274) = 13.42; p < 0.001; T+: 20.76 ± 5.84; T−: 24.93 ± 4.25).
4 DISCUSSION
The present study investigated associations between three commonly studied OXTR SNPs and autistic and empathic traits in two independent samples of healthy male and female students from different cultures (Caucasian n = 537 and Chinese n = 280). Our main finding was that both Caucasian and Chinese samples exhibited a highly significant and robust association between the rs2268498 SNP and the total AQ score at the genotype level and this was independent of gender. In both samples carriers of the TT genotype of rs2268498 exhibited the lowest scores on the AQ and carriers of the C-allele (C+ group) the highest, with the CC and CT genotypes being less frequent in Chinese than in Caucasian groups. Interestingly however, an analysis of sub-scale components of the AQ revealed that while the Caucasian subjects only showed an association with attentional aspects of the autistic trait (details/pattern) the Chinese ones only did so for social skills (see, Supplementary Table S3). For OXTR rs53576 while there was an association at the genotype level for Caucasian subjects with the communication/mindreading scale, with AA carriers having the highest scores, there was no such association in the Chinese sample. On the contrary for OXTR rs2254298 while there was an association with the small AQ score and the social skills sub-scale in the Chinese sample, with GG carriers having the highest scores, this association was absent in the Caucasian sample. Some minor associations were also found with components of empathy and OXTR SNPs but these were not consistent across the two samples (see, Supplementary Table S4). There were also no significant gender interaction effects or interactions between the three SNPs.
The robust association between the rs2268498 SNP and total AQ score in two independent samples from different cultures further underlines its potential importance in the context of social cognition. Although the allelic distribution of this SNP differs in the Caucasian and Chinese samples, in that the C-allele is less frequent in Chinese, carriers of this allele in both populations have higher total AQ scores. On the other hand, the TT genotype in both populations is generally associated with low total AQ scores and therefore potentially higher prosocial interpersonal abilities (although in contrast to Christ et al. (2016) we did not find any association with IRI scores see further discussion below). Importantly, rs2268498 is located in the promoter-flanking region of the OXTR and the prosocial T-allele of rs2268498 is associated with lower OXTR mRNA expression (Reuter et al., 2016). The association of the T-allele with lower OXTR density might reflect an enhanced sensitivity to oxytocin resulting in a compensatory decrease in OXTRs, although it remains to be demonstrated whether TT-carriers are indeed more sensitive to endogenous or exogenously administered OXT. To date however, no studies have investigated associations between this SNP and autism in clinical populations. Mounting evidence exists that the nasal administration of oxytocin indeed might have prosocial effects not only in healthy humans, but also in autistic patients with respect to the recognition of emotions in faces and related paradigms (e.g., Guastella et al., 2010; Marsh et al., 2010). Therefore, a combination of nasal application of oxytocin taking into consideration an individual's OXTR rs2268498 genotype will present an interesting future research question.
In an attempt to identify more precisely which specific aspects of autistic traits were associated with rs2268498 we used the three factor sub-scales suggested by Austin (2005) which had better internal consistencies than the original five sub-scales that have been proposed across the two samples. Interestingly this revealed a difference between the two study samples, with the Caucasian one only showing a significant association with the sub-scale incorporating attentional aspects (i.e., details/pattern) whereas for the Chinese one the correlation was with the social-skills sub-scale (see, Supplementary Table S3). While this may reflect a genuine cultural difference in symptom association with this OXTR SNP, it is also clear that the Chinese sample had much higher scores than the Caucasian one in the social skills domain. However, scores on the details/pattern domain were equivalent in the two samples and yet a significant association was only found in the Caucasian sample.
Despite previous findings of associations between the OXTR rs2254298 SNP and ASD (LoParo & Waldman, 2015; Wu et al., 2005), in the current study we found no evidence for a similar association with autistic traits in healthy Caucasian subjects. However, in the Chinese sample there was a significant association at the genotype level both for the small AQ and social skill sub-scale scores, with A-allele carriers showing low scores and G-allele carriers high ones (see, Supplementary Table S3). This contrasts with a study on Japanese subjects which failed to find a significant association between rs2254298 and total AQ or social skill scores that could survive correction, although this was based on the full 50-question version of the AQ (Kawamura et al., 2011) and we also failed to find a significant association with the total AQ score using the full 50 question version.
Findings in relation to the most widely studied OXTR SNP rs53576, which has also been associated with ASD (Wu et al., 2005), were again inconsistent across samples. In the Caucasian sample while no associations between rs53576 and total AQ scores reached significance at the genotype level, at the allele level G-carriers tended to have the lowest scores. On the other hand, there was a significant association at both genotype and allele levels in the German sample with scores on the communication/mindreading scale with again G-carriers having the lowest scores and A-carriers the highest ones (see, Supplementary Table S3). However, for the Chinese sample there were no significant associations between rs53576 and AQ scores. Since the allelic distribution of rs53576 and its association with social cognition exhibit differences between Caucasian and Asian populations (Kim et al., 2010) our findings may suggest cultural differences in this SNP in relation to autistic traits.
In contrast to a recent study on subjects from the US (Christ et al., 2016), we found no significant associations between rs2268498 and trait empathy (IRI scores) in our two population samples (see, Supplementary Table S4). In view of the fact that several previous studies have reported associations between the T-allele and better face emotion recognition skills (Melchers et al., 2013; for an overview to study the molecular basis of facial expressions see Montag & Panksepp, 2016) it is possible that it is more influential in the domain of social and emotional recognition skills as opposed to empathy. For rs53576 there was a significant association at the allelic level in the Caucasian group for the fantasy sub-scale with highest scores for AA-carriers and lower for G-carriers. No significant association between rs53576 and any of the IRI dimensions could be observed in the Chinese sample (see, Supplementary Table S4). Another recent study on Chinese subjects has reported that rs53576 associations with the cognitive empathy dimensions of the IRI (perspective taking and fantasy) interact with interdependence traits (Luo et al., 2015) and therefore the relationship between this SNP and empathy may be complex and dependent upon multiple factors. For rs2254298, no significant association on genotype level with the IRI could be observed in the Caucasian sample (other than on the allele level with the empathic concern dimension but only based on a group of six individuals — see, Supplementary File). For the Chinese sample there was a significant association at both genotype and allele levels but with the personal distress sub-scale, with AA-carriers having the lowest scores and GG-carriers the highest ones (see, Supplementary Table S4). Thus, overall there were no robust and consistent patterns of associations between these OXTR SNPs and IRI-based assessments of empathic traits.
While in the larger Caucasian sample the expected significant gender-differences in autism (males > females) (Ruzich et al., 2015) and empathy (females > males) (Eisenberg & Lennon, 1983) trait scores could be observed, in the Chinese sample this did not occur for autism and only for the IRI fantasy and empathic concern dimensions (see, Supplementary File: Supplementary Tables S1 and S2). This may have reflected the smaller sample size or the recruitment of students from a more science-based Chinese university, although other studies have reported both higher AQ scores and reduced gender differences in Asian compared to Causasian subjects (Freeth, Sheppard, Rmachandran, & Milne, 2013) and gender differences were also not found in Japanese subjects (Kawamura et al., 2011). Similarly, other studies on Chinese subjects have reported no overall differences between male and female students on empathy traits measured by total IRI scores, and only minor ones for a few of its sub-scales similar to our current findings (Siu & Shek, 2005; Wu, Li, & Su, 2012). Importantly though in the context of the current study no meaningful influences of gender were found on OXTR SNP associations with either autism or empathy trait scores. For the interested reader we present correlations between the AQ scales and the IRI dimensions in the Supplementary File (see, Supplementary Table S5 and S6).
In conclusion, the present study provides the first evidence that the OXTR SNP rs2268498 is associated with autism trait scores in independent samples of Caucasian and Chinese subjects. In both cases C-allele carriers have higher scores and interestingly this allele occurs less frequently in the Chinese population. Despite a number of previous studies reporting associations between rs53576 and rs2254298 neither of these OXTR SNPs show robust associations with either autistic or empathic traits across the two population samples. Since rs2268498 is in the promoter-flanking region of the OXTR and has been reported to influence OXTR mRNA expression, this SNP presents an important future focus for investigating links between the OXTR and social cognition and psychopathology across different cultures. Future studies will also need to test, if the effects of nasal administration differ according to carrying the TT or C variants of OXTR rs2268498. Based on a dimensional disorder model the present study examined associations in a healthy student sample. It remains to be determined whether the OXTR rs2268498 is associated with the level of autistic symptoms in clinical populations.
ACKNOWLEDGMENTS
CM is funded by a Heisenberg grant (DFG, MO2363/3-1) from the German Research Foundation (Deutsche Forschungsgemeinschaft). Moreover, the present work was funded by a research grant awarded to CM by the German Research Foundation (Deutsche Forschungsgemeinschaft, MO2363/2-1). CS is stipend of the German Academic Scholarship Foundation (Studienstiftung des deutschen Volkes). KMK is supported by the National Natural Science Foundation of China (NSFC) (grant 31530032).
CONFLICT OF INTEREST
The authors declare that no conflicts of interest exist.
AUTHORS CONTRIBUTIONS
CM and MM designed the German part of the study. CM, BB, and KMK designed the Chinese part of the study. CM, CS, and SJ recruited the participants in Germany. CS and RL recruited the participants in China. Genetic analysis in Germany were implemented and carried out by CM and SJ. Genetic analysis in China were implemented and carried out by CM, CS, and RL. CM drafted the first version of the manuscript and carried out the statistical analyses. All statistical analyses were independently checked by CS. KMK completed the second version of the draft, which then was checked and accepted by all authors. JX, WX, and AG made amendments when revising the draft. All authors agree to be accountable for the content of the work.
