Sample and genotyping
A total of 229 right-handed healthy Han Chinese (male = 114; mean age ± SD =21.58 ± 2.343) were included in the final analysis (see Figure S2 for exclusion flowchart). Distribution of the 4 OXTR (rs2254298, rs2268491, rs2268498, and rs53576) genotypes were in the Hardy-Weinberg equilibrium (HWE) and alleles of the SNPs were divided into two groups as in previous studies [37,43] to increase statistical power and avoid statistical inference errors (see Table S1).
Behavioral results
Two subjects were additionally excluded due to incomplete data for the implicit emotion recognition task, leading to 227 subjects (male = 112; mean age SD = 21.57 2.351 years) included for behavioral data analyses (see Supplementary for details of analytic approach). ANOVA results for the implicit emotion recognition task presented in Figure 2, suggest that subjects attentively processed the facial stimuli (gender discrimination) and that presentation of negative emotional faces resulted in a lower accuracy and longer RTs. No significant associations were found between explicit face emotion recognition accuracy or RTs or recognition memory for faces or arousal or intensity ratings and ASQ scores in tests conducted after the scan (see Supplementary).
Findings from MVPA analyses
The SVR-based MVPA analysis revealed activity patterns of the right midbrain (extending into limbic areas) comprising the ventral tegmental area (VTA), anterior cingulate cortex (ACC), and hypothalamus (MNIxyz = 3/-3/-18, k = 40), the left anterior orbitofrontal cortex (OFC) (MNIxyz = -39/57/-9, k = 11), left caudate (MNIxyz = -3/-6/0, k=20), right dorsal medial frontal cortex (dmPFC; MNIxyz = 3/42/42, k = 12), left postcentral gyrus (PoCG; MNIxyz = -45/-33/51, k=45), and the left superior frontal gyrus (SFG; MNIxyz = -21/63/0, k=6) could accurately predict individual autistic traits during processing of angry (versus neutral) expressions (Figure 3a). Additionally, during processing fear (versus neutral) expressions, individual autistic trait scores could be decoded from activity in the left MFG (MNIxyz = -36/18/42, k = 8) (Figure 3b). Individual autistic-trait scores could also be decoded from neural activity towards sad (versus neutral) expressions in the left MFG (MNIxyz = -36/24/33, k = 7), the right midbrain (MNIxyz = -3/-30/-9, k = 13), and left hypothalamus (MNIxyz = -3/-3/-15, k = 30) (Figure 3c).
Analysis of FC associations with ASQ and influence of OXTR genotype
The gPPI analyses using a seed-to-whole brain approach (seeds defined according to the MVPA analyses) revealed no significant FCs associated with trait autism during processing of specific emotional faces irrespective of OXTR genotype. However, there were significant interaction effects of ASQ and genotype for rs2268491 and rs2254298 on FC during processing of angry expressions.
Examination of rs2268491 genotype (T+/T -) × ASQ score interactions revealed effects for the angry > neutral contrast in coupling of left OFC (seed, x = -39, y = 57, z = -9) with bilateral MCC (MNIxyz = 0/-9/45, k =24, T=5.11, pFWE-peak = 0.005) (Table 1). Parameter estimates from the significant clusters extracted to visualize interaction effects found opposite effects of ASQ scores on FC strengths in the T+ and T- groups: a positive association between ASQ and the FC strength in the T- group and a negative association in the T+ group (T-: r = 0.318, p < 0.01; T+: r = -0.321, p < 0.01) (Figure 4). Additionally, ASQ and rs2268491 interactively impacted FC strengths between the right midbrain (seed, x = 3, y = -3, z = -18) and right supramarginal gyrus (SMG) (MNIxyz = 63/-39/30, k =14, T=4.92, pFWE-peak = 0.012) (Table 1). Parameter estimates were subsequently extracted and showed a positive association between ASQ and the FC strength in the T- group but negative association in the T+ group (T-: r = 0.263, p < 0.01; T+: r = -0.356, p < 0.01) (Figure 4).
Table 1 FC modulated by OXTR genotype during processing of angry > neutral face emotion.
Region
|
k
|
t-value
|
pFWE-0.05
|
x
|
y
|
z
|
rs2254298
|
seed: OFC (-39/57/-9)
|
|
|
MCC, R
|
14
|
5.10
|
0.006
|
15
|
-21
|
42
|
MCC, L/ MCC, R
|
23
|
5.05
|
0.007
|
0
|
-9
|
45
|
rs2268491
|
seed: OFC (-39/57/-9)
|
|
|
MCC, L/ MCC, R
|
24
|
5.11
|
0.005
|
0
|
-9
|
45
|
seed: Midbrain (3/-3/-18)
|
|
|
SMG, R
|
14
|
4.92
|
0.012
|
63
|
-39
|
30
|
Note: Statistical significance thresholding was applied with whole-brain peak level at pFWE < 0.05 and a minimum voxel size of k > 10. OFC = orbitofrontal cortex; MCC = middle cingulate cortex; SMG = supramarginal gyrus; L = left; R = right.
Significant interaction effects of genotype of rs2254298 (A+/A-) × ASQ scores were also observed during processing angry face emotion in FC between left OFC (seed, x = -39, y = 57, z = -9) and left MCC (extended to the right MCC) (MNIxyz = 0/-9/45, k =23, T=5.05, pFWE-peak = 0.007), and additional clusters located in the right MCC (MNIxyz = 15/-21/42, k =14, T=5.10, pFWE-peak = 0.006) (Table 1). Parameter estimates were extracted and further revealed opposite effects of ASQ on these two couplings in A+ and A- carriers. ASQ was positively associated with the left OFC - left MCC coupling in A- carriers while A+ carriers exhibited a negative correlation (A-: r = 0.329, p < 0.01; A+: r = -0.311, p < 0.01) (Figure 4). Likewise, for A- carriers there was a positive correlation between autistic trait scores and left OFC - right MCC coupling but a negative correlation was in A+ carriers (A-: r = 0.399, p < 0.01; A+: r = -0.247, p < 0.01) (Figure 4).
No significant interaction effects were observed for other seeds nor during processing fear or sad expressions. The analyses for SNPs rs53576 and rs2268498 did not yield any significant effects.