Previous studies have confirmed the sex difference of gray matter asymmetry in
typically developing controls and the abnormal gray matter asymmetry in autism
spectrum disorders. However, whether and how sex differences of gray matter
asymmetry exist in autism spectrum disorders remains studied. This paper analyzes
the above issues and explores correlations between gray matter asymmetry and
autistic symptoms. Data from 72 children (36 males and 36 females) with autism
spectrum disorders and 72 typically developing-controls (36 males and 36 females)
between 8 and 14 years were included and obtained from the autism brain imaging
data exchange repository (autism brain imaging data exchange I and autism brain
imaging data exchange II). The voxel-based morphometry approach was used to
assess gray matter asymmetry in T1-weighted brain data, and gray matter asymmetry
was quantified as asymmetry index. A 2
Autism spectrum disorders (ASD) are a set of pervasive neurodevelopmental conditions characterized by social impairments, communication difficulties, and restricted, repetitive behaviors [1]. It is one of the most common developmental disorders, affecting about 1 in 160 people worldwide [2]. Despite its prevalence, studies have not been able to determine its exact etiology [3]. The imbalanced male-to-female prevalence ratio is one of the most apparent characteristics of ASD [2], and the sex bias for males is even higher in neuroimaging studies [4, 5]. There are also symptom differences between males and females with ASD [6]; males have lower levels of social impairments and externalizing problems [7] but more significant restricted behaviors compared to females with ASD [8]. Thus, sex is increasingly being recognized as the source of neuropathophysiological heterogeneity in ASD.
One of the most striking features of the human brain is its structural or functional lateralization or asymmetry. In most typically developing-controls (TDs), the left hemisphere specializes in motor control and language, while the right hemisphere is responsible for visual-spatial attention [9]. And in TDs, studies have found that numerous brain regions showed asymmetry differences in structure (especially gray matter) or function between males and females [10, 11]. Alterations of lateralization or asymmetry have been found in various mental and neurocognitive disorders, including schizophrenia and dyslexia [12, 13]. And atypical brain asymmetry has long been hypothesized for ASD, a supposition that is supported by multiple neuroimaging studies [14, 15, 16]. Several structure-based studies reported hemispheric asymmetry abnormalities in ASD, but discoveries of potential structural alterations for ASD have been inconsistent. For example, Groen et al. [17] reported increased hippocampal volume in ASD. On the other hand, Eilam-Stock et al. [18] found significantly decreased hippocampal volume in subjects with ASD compared with TDs. Therefore, further structure-based researches on the hemispheric brain asymmetry of ASD are necessary.
Due to the relatively high prevalence of ASD in males, almost all existing studies on the brain structure of ASD have focused on samples that are predominantly or exclusively male. However, there is growing evidence that males and females with ASD have different brain structural alteration patterns. For example, [19] reported that females might need higher detrimental loads, including brain structural alterations, before developing clinically relevant levels of autistic characteristics. In addition, ASD-associated neural anatomy in females may differ from that in males; Ecker et al. [20] detected that temporal lobe cortical thickness was decreased and increased in females and males with ASD, respectively. Unfortunately, potential sex differences in brain structure have often been ignored in ASD. This may be one of the main reasons for the inconsistent results of brain structure research in ASD. Thus, the inclusion of females with ASD is critical to help improve our understanding of brain structural alterations in ASD.
Based on the findings of the sex difference of gray matter (GM) asymmetry in TDs and the abnormality of GM asymmetry in ASD, quantitative or qualitative sex differences of the GM asymmetry alterations in ASD are probably to be expected [11, 21]. However, if the effects on GM asymmetry in males and females with ASD are present in only one sex or the opposite form, previous research on GM asymmetry of ASD that did not consider the sex factor may be unreliable, as the above effects may be offset in these researches. This significantly reduced the possibility of detecting actual effects. Therefore, a sex-specific analysis is urgently needed to determine whether and how the sex differences of brain structural asymmetry might be altered in ASD to better understand the neuroanatomy of the situation.
Recently, Kurth et al. [22] established a fully automated voxel-based morphometry (VBM) method to specifically analyze brain GM asymmetry. A key advantage over the standard VBM workflow is establishing the voxel-wise hemispheric correspondence, which is ensured by using Diffeomorphic Anatomical Registration Lie (DARTEL) algebra to convert spatial normalization into a symmetric space. Several studies have used this approach to investigate the effects of sex, meditation, alcohol dependence, and schizophrenia on GM asymmetry [22, 23, 24].
Here, we used the advanced VBM method [25] to conduct a sex-specific study of GM asymmetry in ASD. GM asymmetry was quantified as asymmetry index (AI). We also explored correlations between AI values extracted from the clusters with significant differences between the four groups and autistic symptoms (social impairments, communication difficulties, and restricted, repetitive behaviors) measured by the revised autism diagnostic interview (ADI-R) scale.
The participants included 72 children (36 males and 36 females) with ASD and 72
TDs (36 males and 36 females) between 8 and 14. We used Gpower
(https://www.psychologie.hhu.de/arbeitsgruppen/allgemeine-psychologie-und-arbeitspsychologie/gpower.html)
software to calculate the sample size. The power and the required effect size in
this paper are 0.85 and 0.24, respectively. All participants were obtained from
the autism brain imaging data exchange (ABIDE) repository (ABIDE I and ABIDE II).
The dataset consists of 1060 individuals with ASD and 1166 TDs, aged 6–65,
collected from 36 international sites. Data were collected from five sites (KKI,
NYU, OHSU, YALE, GU) that included five or more females with ASD who met the
criteria. All subjects with ASD had a clinician’s Diagnostic and Statistical
Manual of Mental Disorders (DSM)-IV-TR or DSM-V diagnosis. The majority of
participants were assessed using Autism Diagnostic Observation Schedule (ADOS)
modules 3 or 4, ADI-R, or both. The ABIDE database also provided information on
medications and comorbidities. The female ASD group was formed first because most
sites included a few females with ASD. To avoid biases due to the different
sites, random subjects were selected to form age-, sex- and site-matched groups.
All four groups were formed with the following criteria: (1) age range: 8–14
years, (2) full-scale intelligence quotient (FIQ)
All included data were acquired from 3.0-Tesla scanners, but the scanner types and image acquisition parameters varied across sites. The scan parameters and acquisition protocols are provided at http://fcon_1000.projects.nitrc.org/indi/abide/.
The high-resolution 3D T1-weighted images were preprocessed using the VBM8 toolbox (http://dbm.neuro.uni-jena.de/vbm8) package Statistical Parametric Mapping (SPM8, http://www.fil.ion.ucl.ac.uk/spm) software. Before preprocessing, visual checks were performed for all 3D T1-weighted images. Then, all preprocessing steps were performed following the protocol by Kurth et al. [22]: (1) The 3D T1-weighted images were segmented into GM, white matter (WM), and cerebrospinal fluid; (2) the segmented images were flipped at the midline; (3) the original and flipped GM and WM versions of each participant were used to create a symmetric DARTEL anatomical template; (4) the original and flipped GM and WM tissue segments of all participants were warped into the created symmetric template to align the brain of each participant in a shared space; and (5) to limit further data analysis to the right hemisphere, we used the symmetric template space to create a right-hemispheric mask in MRIcron (http://people.cas.sc.edu/rorden/mricron/index.html). First, in the sagittal window of MRIcron, we selected one plane to the right in the midline line of GM. This plane was marked as a volume of interest using the drawing tool. Then, masking was repeated in the next plane until the right hemisphere was finished. Finally, the file was saved in NIfTI format.
The AI value of GM volume was calculated with an automated procedure. This
procedure used the ‘calculate’ script provided by [22]. First, type ‘calculate’
in MATLAB’s command window to run the script and select ‘Step 6’, then the script
asks for the original warped GM images and the hemispheric mask generated in the
above steps [22]. This step generates the masked AI images of each participant.
The formula in this script is [22]:
AI = ((i1–i2)/((i1 + i2)
All AI maps were spatially smoothed with an 8-mm Gaussian smoothing kernel. All results were limited in the right hemisphere, and a positive AI indicated rightward asymmetry (larger right-hemispheric GM volume in a given cluster). Conversely, a negative AI indicated leftward asymmetry (larger left-hemispheric GM volume in a given cluster) [22, 26].
A 2
The AI values were extracted using Data Preprocessing Assistant for rs-fMRI
(DPARSF) software from the clusters with significant differences in the 2
Descriptive statistics are presented in Table 1. The four groups did not
significantly differ in age (p
Males with ASD | Females with ASD | Males TDs | Females TDs | Diagnosis effect | Sex effect | Diagnosis × Sex | |||||||||
n | mean |
n | mean |
n | mean |
n | mean |
F-value | p-value | F (t)-value | p-value | F-value | p-value | ||
Age (year) | 36 | 10.88 |
36 | 10.80 |
36 | 10.82 |
36 | 10.83 |
0.002 | 0.965 | 0.008 | 0.930 | 0.021 | 0.885 | |
FIQ | 36 | 107.64 |
36 | 102.61 |
36 | 117.75 |
36 | 116.64 |
25.466** | 0.000 | 1.647 | 0.201 | 0.670 | 0.414 | |
ADI-R | |||||||||||||||
Social | 33 | 19.52 |
30 | 18.91 |
0.251 | 0.677 | |||||||||
Verbal | 33 | 15.36 |
30 | 14.52 |
–0.150 | 0.452 | |||||||||
RRB | 33 | 5.68 |
30 | 5.21 |
-0.523 | 0.462 | |||||||||
ADOS | |||||||||||||||
Total | 14 | 10.08 |
18 | 12.28 |
–1.665 | 0.107 | |||||||||
Social | 14 | 3.46 |
18 | 3.33 |
0.282 | 0.782 | |||||||||
Communication | 14 | 6.23 |
18 | 8.11 |
–2.073 | 0.067 | |||||||||
RRB | 14 | 2.31 |
18 | 2.39 |
–0.137 | 0.892 | |||||||||
Comorbidity | |||||||||||||||
ADHD | 26 | 10/26 | 26 | 9/26 | 0.083 | 0.773 | |||||||||
Medication | 31 | 12/31 | 31 | 9/31 | 0.421 | 0.648 | |||||||||
Notes: ASD, autism spectrum disorders; TDs, typically developing controls; FIQ,
full-scale intelligence quotient; SD, standard deviation; ADI-R, revised autism
diagnostic interview scale; ADOS, autism diagnostic observation schedule; RRB,
restricted repetitive behaviors; ADHD, attention deficit hyperactivity disorder,
**p |
All brain regions in results were reported using the Automated Anatomical Labeling template.
Specific brain regions showed significant main effects for diagnosis. The ASD
patients had more leftward asymmetry than TDs for the parahippocampal gyrus and
postcentral gyrus (GRF correction, voxel-level p
The map for brain regions showing a statistically significant
main effect of diagnosis in the 2
Relative to males, females showed significantly more rightward asymmetry in the
middle temporal gyrus, inferior frontal gyrus, angular gyrus, and postcentral
gyrus and minor rightward asymmetry in the superior frontal gyrus (GRF
correction, voxel-level p
The map for brain regions showing a statistically significant
main effect of sex in the 2
Significant diagnosis
The map for brain regions showing a statistically significant
main effect of diagnosis
Diagnosis | (I)sex | (J)sex | Mean Difference (I–J) | Std. Error | Sig |
ASD | Male | Female | –0.239* | 0.068 | 0.001 |
Female | Male | 0.239* | 0.068 | 0.001 | |
TD | Male | Female | 0.162* | 0.068 | 0.019 |
Female | Male | –0.162* | 0.068 | 0.019 | |
Sex | (I)diagnosis | (J) diagnosis | Mean Difference (I–J) | Std. Error | Sig |
Male | ASD | TD | –0.245* | 0.068 | 0.000 |
TD | ASD | 0.245* | 0.068 | 0.000 | |
Female | ASD | TD | 0.155* | 0.068 | 0.024 |
TD | ASD | –0.155* | 0.068 | 0.024 | |
Notes: Based on estimated marginal means. * The mean difference is significant
at the 0.05 level. |
Diagnosis | (I)sex | (J)sex | Mean Difference (I–J) | Std. Error | Sig |
ASD | Male | Female | –0.316* | 0.092 | 0.001 |
Female | Male | 0.316* | 0.092 | 0.001 | |
TD | Male | Female | 0.168 | 0.092 | 0.070 |
Female | Male | –0.168 | 0.092 | 0.070 | |
Sex | (I)diagnosis | (J) diagnosis | Mean Difference (I–J) | Std. Error | Sig |
Male | ASD | TD | –0.179 | 0.092 | 0.054 |
TD | ASD | 0.179 | 0.092 | 0.054 | |
Female | ASD | TD | 0.305* | 0.092 | 0.001 |
TD | ASD | –0.305* | 0.092 | 0.001 | |
Notes: Based on estimated marginal means. * The mean difference is significant
at the 0.05 level. |
We observed significant correlations between extracted AI values and clinical severity in the ASD group. Specifically, males with ASD showed a positive association between the AI value in the middle occipital gyrus and more significant verbal impairment measured by ADI-R (r = 0.387, p = 0.026; Fig. 4). In addition, the AI value in the parahippocampal gyrus was positively associated with more severe social impairment in females with ASD (r = 0.422, p = 0.020; Fig. 5 and Table 5).
The correlation between the asymmetry index value in the middle occipital gyrus and the verbal impairment score measured by the revised autism diagnostic interview scale in males with autism spectrum disorders (r = 0.387, p = 0.026).
The revised autism diagnostic interview scale measured the correlation between the asymmetry index value in the parahippocampal gyrus and social impairment score in females with autism spectrum disorders (r = 0.422, p = 0.020).
Brodmann area | MNI coordinates of peak | Cluster size (voxels) | F-value | ASD | TDs | |||||
x | y | z | Male | Female | Male | Female | ||||
The main effect of diagnosis | ||||||||||
Parahippocampal gyrus | 23 | –5 | –18 | 173 | 10.17 | –0.15 |
–0.06 |
0.08 |
0.12 | |
Postcentral gyrus | 11 | –41 | 74 | 244 | 15.11 | –0.11 |
–0.10 |
0.09 |
0.07 | |
The main effect of sex | ||||||||||
Middle temporal gyrus | 60 | 8 | –17 | 214 | 13.18 | –0.21 |
0.09 |
–0.24 |
0.07 | |
Inferior frontal gyrus | 36 | 23 | –6 | 550 | 13.47 | –0.07 |
0.11 |
–0.01 |
0.21 | |
Angular gyrus | 45 | –59 | 38 | 224 | 12.59 | –0.13 |
0.16 |
–0.03 |
0.12 | |
Superior frontal gyruss | 12 | 45 | 41 | 229 | 11.05 | 0.11 |
–0.01 |
0.21 |
–0.04 | |
Postcentral gyrus | 45 | –36 | 65 | 287 | 14.90 | –0.17 |
0.04 |
–0.06 |
0.07 | |
Interaction effect | ||||||||||
Angular gyrus | 30 | –69 | 47 | 311 | 12.53 | –0.17 |
0.07 |
0.08 |
–0.08 | |
Middle occipital gyrus | 39 | –80 | 12 | 395 | 10.64 | –0.01 |
0.30 |
0.16 |
–0.00 | |
Notes: ASD, autism spectrum disorders; TDs, typically developing controls. |
ADI-R | Middle Temporal Gyrus | Inferior Frontal Gyrus | Angular gyrus |
Superior Frontal Gyrus | Postcentral gyrus |
Parahippocampal gyrus | Postcentral gyrus |
Angular gyrus |
Middle occipital gyrus | ||||||||||
r-value | p-value | r-value | p-value | r-value | p-value | r-value | p-value | r-value | p-value | r-value | p-value | r-value | p-value | r-value | p-value | r-value | p-value | ||
Male | |||||||||||||||||||
Social | 0.277 | 0.118 | –0.106 | 0.556 | –0.123 | 0.497 | 0.033 | 0.853 | 0.264 | 0.137 | 0.117 | 0.516 | 0.054 | 0.767 | –0.230 | 0.198 | 0.219 | 0.220 | |
Verbal | 0.233 | 0.192 | –0.270 | 0.128 | –0.040 | 0.825 | –0.030 | 0.869 | 0.145 | 0.421 | 0.032 | 0.681 | 0.006 | 0.975 | –0.249 | 0.163 | 0.387* | 0.026 | |
RRB | –0.047 | 0.797 | –0.122 | 0.499 | 0.035 | 0.846 | –0.048 | 0.972 | 0.299 | 0.091 | –0.147 | 0.413 | –0.029 | 0.874 | –0.182 | 0.311 | 0.295 | 0.095 | |
Female | |||||||||||||||||||
Social | –0.293 | 0.116 | –0.022 | 0.907 | –0.223 | 0.237 | 0.337 | 0.069 | –0.330 | 0.075 | 0.422* | 0.020 | 0.130 | 0.494 | –0.217 | 0.248 | –0.100 | 0.600 | |
Verbal | –0.125 | 0.511 | 0.161 | 0.394 | –0.097 | 0.610 | 0.101 | 0.595 | –0.260 | 0.160 | 0.168 | 0.376 | 0.217 | 0.250 | –0.018 | 0.925 | –0.085 | 0.650 | |
RRB | 0.112 | 0.555 | –0.085 | 0.655 | 0.257 | 0.171 | 0.271 | 0.148 | –0.018 | 0.924 | 0.045 | 0.815 | 0.202 | 0.283 | 0.120 | 0.528 | 0.193 | 0.308 | |
Notes: AI, asymmetry index; ADI-R, revised autism diagnostic interview scale;
RRB, restricted repetitive behaviors; * p |
We mapped differences in GM asymmetry between ASD and TDs and between the male and female children with ASD. Previous studies on cortical asymmetry in ASD found variable differences or no differences [27]. Although Postema et al. [27] reported that total hemispheric results showed generalized less leftward asymmetry of cortical thickness in ASD, the direction of GM asymmetrical alterations in specific brain regions remains to be established. Using the advanced VBM approach, we found more leftward asymmetry of GM volumes in the parahippocampal gyrus and postcentral gyrus in ASD.
The parahippocampal gyrus is essential for ASD. Abnormalities in this brain region have been reported in previous studies on ASD. Mei et al. [28] reported decreased density in the parahippocampal gyrus in ASD. Yang et al. [29] found that the GM volume in the parahippocampal gyrus in adults with ASD was significantly increased relative to TDs. In addition, the AI value in the parahippocampal gyrus was positively associated with more severe social impairment in females with ASD (r = 0.422, p = 0.020). This result is consistent with previous studies that linked structural alterations in the parahippocampal gyrus to social deficits in ASD [30], indicating that structural alterations of the parahippocampal gyrus indeed serve a junctional role in ASD etiology.
Concerning the postcentral gyrus, volume abnormalities reported in this brain region in ASD have been inconsistent [31]. Mizuno et al. [31] reported decreased left postcentral gyrus volume in ASD , while Brieber et al. [32] described increased volume in this brain region in patients with ASD comparing with TDs. This may be due to the small sample sizes, the different proportions of females with ASD, and the different image processing techniques among different studies. We used an advanced VBM method specifically designed to assess brain structural asymmetry, had a relatively large sample size, and considered the sex factor. Therefore, our findings are more reliable and can provide more substantial evidence to clarify the inconsistent findings reported earlier. The postcentral gyrus participates in somatosensory functions. Abnormal sensory processing can lead to social impairments, communication difficulties, and restricted, repetitive behaviors, which are core characteristics of ASD [33]. Therefore, the abnormality of the postcentral gyrus may cause the core symptoms of ASD.
The sex difference of GM asymmetry was found in the angular gyrus in ASD. Males and females with ASD showed more minor and more rightward asymmetry in the angular gyrus, respectively, compared with their sex-matched TDs. A similar reduction in the angular gyrus was previously detected in [34] with the mixed-sex ASD sample (mainly males with ASD). Thus, less rightward asymmetry in the angular gyrus seems to be unique to males with ASD. Interestingly, this brain area is involved in the theory of mind, which is essential to help the human brain reason about others and effectively communicate and navigate in the social world [35]. There are sex differences for empathy in ASD [36], and the theory of mind is necessary for empathy. Therefore, sex differences of GM asymmetry in the angular gyrus may be why males are less empathetic than females with ASD.
The sex difference of GM asymmetry was also observed in the middle occipital gyrus in subjects with ASD. Like the angular gyrus, females with ASD showed more, and males with ASD showed less rightward asymmetry relative to their respective TDs in the middle occipital gyrus. This is the first work to find that males and females with ASD exhibited GM asymmetry alterations in the opposite direction relative to their sex-matched TDs. Moreover, males and females with ASD showed different correlativity between ASD symptom severity and AI value in several brain regions, including the middle occipital gyrus. This again suggests that the relationship between ASD and GM asymmetry partly varies by sex. The middle occipital gyrus mainly involves constructing visual and motorial perception, and aberrant perceptual processing might be associated with social dysfunction in ASD [37]. Considering the novelty of these results, they should be interpreted carefully, and they need to be replicated in future studies.
The results do not support the “Female Protective Effect” theory, which states that females need higher detrimental loads, including brain structural alterations, before developing clinically relevant levels of autistic characteristics [19]. Instead, we detected opposite sex-associated effects in ASD for both two brain regions. Similarly, [38] found that neither behavioral nor neural data supported the idea that females with ASD had more severe abnormalities. Thus, our research provides evidence that sex differences of GM asymmetry in ASD may be qualitative rather than quantitative. Based on these findings, we suggest that researchers should not conduct cross-sex neuroimaging studies for ASD because they may raise the risk of offsetting present results in only one sex or opposite directions in males and females.
There are limitations in the current research. First, it was based on multi-site datasets, which may introduce additional sources of variance associated with participant characteristics or scan parameters. Secondly, our sample size was still insufficient; further studies with larger sample sizes are needed to confirm sex-specific brain structure alterations in subjects with ASD. Finally, participants’ clinical information was also limited, with ADOS and ADI-R scores available for only participants.
Using an advanced VBM approach, we identified diagnosis
GM, gray matter; TDs, typically developing controls; ASD, autism spectrum disorders; ABIDE, autism brain imaging data exchange; AI, asymmetry index; ADI-R, revised autism diagnostic interview; VBM, voxel-based morphometry; DARTEL, Diffeomorphic Anatomical Registration Lie; DSM, Diagnostic and Statistical Manual of Mental Disorders; ADOS, Autism Diagnostic Observation Schedule; FIQ, full-scale intelligence quotient; SPM, Statistical Parametric Mapping; WM, white matter; ANCOVA, analysis of covariance; GRF, Gaussian random field; DPARSF, Data Preprocessing Assistant for rs-fMRI; ADHD, attention deficit hyperactivity disorder; GU, Georgetown University; KKI, Kennedy Krieger Institute; NYU, New York University; OHSU, Oregon Health and Science University; YALE, Yale Child Study Center.
CCL, MMN, PYF and HBX conceived and designed the experiments; CCL and MMN analyzed the data; CCL and HBX wrote the paper.
The written informed consent was obtained from the site where they conducted the tests. The research was conducted with the Ethics Committee of the University of Zhongnan Hospital of Wuhan University (Ethical approval number: 2019072).
We would like to express our sincere appreciation to the participants who generously and courageously participated in this research.
This work is supported by the National Key Research and Development Program of China No. 2017YFC0108803 and the National Natural Science Foundation of China (Grant No. 81771819) and the Improvement Project for Theranostic ability on Difficulty miscellaneous disease (Tumor).
The authors declare no conflict of interest.