Advertisement

Autism Is Associated With Interindividual Variations of Gray and White Matter Morphology

Open AccessPublished:September 05, 2022DOI:https://doi.org/10.1016/j.bpsc.2022.08.011

      Abstract

      Background

      Although many studies have explored atypicalities in gray matter (GM) and white matter (WM) morphology of autism, most of them relied on unimodal analyses that did not benefit from the likelihood that different imaging modalities may reflect common neurobiology. We aimed to establish brain patterns of modalities that differentiate between individuals with and without autism and explore associations between these brain patterns and clinical measures in the autism group.

      Methods

      We studied 183 individuals with autism and 157 nonautistic individuals (age range, 6–30 years) in a large, deeply phenotyped autism dataset (EU-AIMS LEAP [European Autism Interventions—A Multicentre Study for Developing New Medications Longitudinal European Autism Project]). Linked independent component analysis was used to link all participants’ GM volume and WM diffusion tensor images, and group comparisons of modality shared variances were examined. Subsequently, we performed univariate and multivariate brain-behavior correlation analyses to separately explore the relationships between brain patterns and clinical profiles.

      Results

      One multimodal pattern was significantly related to autism. This pattern was primarily associated with GM volume in bilateral insula and frontal, precentral and postcentral, cingulate, and caudate areas and co-occurred with altered WM features in the superior longitudinal fasciculus. The brain-behavior correlation analyses showed a significant multivariate association primarily between brain patterns that involved variation of WM and symptoms of restricted and repetitive behavior in the autism group.

      Conclusions

      Our findings demonstrate the assets of integrated analyses of GM and WM alterations to study the brain mechanisms that underpin autism and show that the complex clinical autism phenotype can be interpreted by brain covariation patterns that are spread across the brain involving both cortical and subcortical areas.

      Keywords

      Autism spectrum disorder (autism) is a heterogeneous condition characterized by difficulties with social and communicative behaviors, repetitive, rigid behaviors, and altered sensory processes (
      American Psychiatric Association
      Diagnostic and Statistical Manual of Mental Disorders.
      ). In search of the brain basis of autism, the condition has been associated with multiple morphological differences in gray matter (GM) and white matter (WM) (
      • Aoki Y.
      • Yoncheva Y.N.
      • Chen B.
      • Nath T.
      • Sharp D.
      • Lazar M.
      • et al.
      Association of white matter structure with autism spectrum disorder and attention-deficit/hyperactivity disorder.
      ,
      • van Rooij D.
      • Anagnostou E.
      • Arango C.
      • Auzias G.
      • Behrmann M.
      • Busatto G.F.
      • et al.
      Cortical and subcortical brain morphometry differences between patients with autism spectrum disorder and healthy individuals across the lifespan: Results from the ENIGMA ASD Working Group.
      ), as reported by magnetic resonance imaging (MRI) studies. However, previous studies have shown heterogeneous findings of the alterations in both cortical (e.g., cortical thickness, surface area, volume) and subcortical (e.g., volume) morphometry in multiple brain regions, making it difficult to define the neural correlates of autism (
      • van Rooij D.
      • Anagnostou E.
      • Arango C.
      • Auzias G.
      • Behrmann M.
      • Busatto G.F.
      • et al.
      Cortical and subcortical brain morphometry differences between patients with autism spectrum disorder and healthy individuals across the lifespan: Results from the ENIGMA ASD Working Group.
      ,
      • Bedford S.A.
      • Park M.T.M.
      • Devenyi G.A.
      • Tullo S.
      • Germann J.
      • Patel R.
      • et al.
      Large-scale analyses of the relationship between sex, age and intelligence quotient heterogeneity and cortical morphometry in autism spectrum disorder.
      ,
      • Zhang W.
      • Groen W.
      • Mennes M.
      • Greven C.
      • Buitelaar J.
      • Rommelse N.
      Revisiting subcortical brain volume correlates of autism in the ABIDE dataset: Effects of age and sex.
      ). Additionally, voxelwise GM volume analyses revealed divergent results, for instance, in temporal areas in autism (
      • DeRamus T.P.
      • Kana R.K.
      Anatomical likelihood estimation meta-analysis of grey and white matter anomalies in autism spectrum disorders.
      ,
      • Carlisi C.O.
      • Norman L.J.
      • Lukito S.S.
      • Radua J.
      • Mataix-Cols D.
      • Rubia K.
      Comparative multimodal meta-analysis of structural and functional brain abnormalities in autism spectrum disorder and obsessive-compulsive disorder.
      ,
      • Mei T.
      • Llera A.
      • Floris D.L.
      • Forde N.J.
      • Tillmann J.
      • Durston S.
      • et al.
      Gray matter covariations and core symptoms of autism: The EU-AIMS Longitudinal European Autism Project.
      ). Studies of WM microstructural associations in autism are similarly heterogeneous in their findings (
      • Aoki Y.
      • Yoncheva Y.N.
      • Chen B.
      • Nath T.
      • Sharp D.
      • Lazar M.
      • et al.
      Association of white matter structure with autism spectrum disorder and attention-deficit/hyperactivity disorder.
      ,
      • Naigles L.R.
      • Johnson R.
      • Mastergeorge A.
      • Ozonoff S.
      • Rogers S.J.
      • Amaral D.G.
      • et al.
      Neural correlates of language variability in preschool-aged boys with autism spectrum disorder.
      ,
      • Tung Y.H.
      • Lin H.Y.
      • Chen C.L.
      • Shang C.Y.
      • Yang L.Y.
      • Hsu Y.C.
      • et al.
      Whole brain white matter tract deviation and idiosyncrasy from normative development in autism and ADHD and unaffected siblings link with dimensions of psychopathology and cognition.
      ). One explanation for discrepant and heterogeneous findings is that the studies differ widely in data analytic strategy—i.e., these studies rely on unimodal analysis techniques that ignore the signal of interest probably present in more than one modality (
      • Groves A.R.
      • Beckmann C.F.
      • Smith S.M.
      • Woolrich M.W.
      Linked independent component analysis for multimodal data fusion.
      ). Additionally, when integrated together, these modalities might provide additional analytical sensitivity.
      This prompted research to move beyond unimodality and incorporate and connect data from different imaging modalities. For example, Cauda et al. (
      • Cauda F.
      • Costa T.
      • Palermo S.
      • D’Agata F.
      • Diano M.
      • Bianco F.
      • et al.
      Concordance of white matter and gray matter abnormalities in autism spectrum disorders: A voxel-based meta-analysis study.
      ) suggested that GM variation in autism is generally accompanied by WM variation; Ecker et al. (
      • Ecker C.
      • Andrews D.
      • Dell’Acqua F.
      • Daly E.
      • Murphy C.
      • Catani M.
      • et al.
      Relationship between cortical gyrification, white matter connectivity, and autism spectrum disorder.
      ) showed higher axial diffusivity (L1) in the WM fiber tracts originating and/or terminating in the GM clusters with increased local gyrification in adults with autism. Despite the progression away from unimodal approaches, in essence, these MRI studies that correlate GM and WM measures do so after separate unimodal statistical analyses. This likely has less sensitivity to assess the biological variance than fully integrating multimodal data analysis across participants.
      It is assumed that a relatively high level of co-occurring neurobiology underlying different aspects of brain morphology is due to the complicated nature of autism. Therefore, efficient modeling of this potential shared variance would increase the chances to produce a more complete picture of autism in a specific perspective (i.e., brain morphology in our study). Here, we aimed to use an integrative multimodal approach, linked independent component analysis (LICA), to simultaneously incorporate several imaging modalities allowing the investigation of intersubject variability across modalities in one analysis (
      • Groves A.R.
      • Smith S.M.
      • Fjell A.M.
      • Tamnes C.K.
      • Walhovd K.B.
      • Douaud G.
      • et al.
      Benefits of multi-modal fusion analysis on a large-scale dataset: Life-span patterns of inter-subject variability in cortical morphometry and white matter microstructure.
      ,
      • Llera A.
      • Wolfers T.
      • Mulders P.
      • Beckmann C.F.
      Inter-individual differences in human brain structure and morphology link to variation in demographics and behavior.
      ), which enables the isolation of artifacts and may increase the sensitivity to correlate the remaining signals with variables of interest (
      • Groves A.R.
      • Smith S.M.
      • Fjell A.M.
      • Tamnes C.K.
      • Walhovd K.B.
      • Douaud G.
      • et al.
      Benefits of multi-modal fusion analysis on a large-scale dataset: Life-span patterns of inter-subject variability in cortical morphometry and white matter microstructure.
      ). So far, studies that highlight the underlying shared variance between modalities using LICA in autism remain scarce. Previous studies revealed case-control differences between adults with autism and typically developing individuals in linked patterns of voxel-based morphometry (VBM) and diffusion tensor imaging (DTI) measures in several brain regions (
      • Itahashi T.
      • Yamada T.
      • Nakamura M.
      • Watanabe H.
      • Yamagata B.
      • Jimbo D.
      • et al.
      Linked alterations in gray and white matter morphology in adults with high-functioning autism spectrum disorder: A multimodal brain imaging study.
      ,
      • Mueller S.
      • Keeser D.
      • Samson A.C.
      • Kirsch V.
      • Blautzik J.
      • Grothe M.
      • et al.
      Convergent findings of altered functional and structural brain connectivity in individuals with high functioning autism: A multimodal MRI study.
      ). However, these studies focused exclusively on autistic adults without intellectual disability and comprised relatively small sample sizes (<100 individuals) (
      • Itahashi T.
      • Yamada T.
      • Nakamura M.
      • Watanabe H.
      • Yamagata B.
      • Jimbo D.
      • et al.
      Linked alterations in gray and white matter morphology in adults with high-functioning autism spectrum disorder: A multimodal brain imaging study.
      ,
      • Mueller S.
      • Keeser D.
      • Samson A.C.
      • Kirsch V.
      • Blautzik J.
      • Grothe M.
      • et al.
      Convergent findings of altered functional and structural brain connectivity in individuals with high functioning autism: A multimodal MRI study.
      ). Autism is a highly diverse condition; we therefore investigated brain patterns in a broader, more representative autism sample, which might help better characterize brain patterns of autism, which is one of the aims of the current study. We hypothesized that the model analysis would reveal the autism-related regional correspondence between GM and WM or modality-specific effect.
      In addition to identifying categorical group differences, dimensional analyses, i.e., analyses of continuous scores of autism symptoms, might capture more of the heterogeneity of autism compared with categorical diagnostic labels. Many studies have demonstrated the univariate connections between GM or WM patterns and the core symptoms of autism [e.g., (
      • Aoki Y.
      • Yoncheva Y.N.
      • Chen B.
      • Nath T.
      • Sharp D.
      • Lazar M.
      • et al.
      Association of white matter structure with autism spectrum disorder and attention-deficit/hyperactivity disorder.
      ,
      • van Rooij D.
      • Anagnostou E.
      • Arango C.
      • Auzias G.
      • Behrmann M.
      • Busatto G.F.
      • et al.
      Cortical and subcortical brain morphometry differences between patients with autism spectrum disorder and healthy individuals across the lifespan: Results from the ENIGMA ASD Working Group.
      )]. Nonetheless, the relationships between brain substrates and clinical phenotypes are potentially the outcome of integrative effects across multiple autism symptom domains and brain areas, and therefore the multidimensional associations between brain covariations and core symptoms of autism need to be clarified. We therefore performed univariate analyses to identify the one-to-one dimensional associations and additionally implemented a multivariate analysis using canonical correlation analysis (CCA) to learn the integrated associations (
      • Hotelling H.
      Relations between two sets of variates.
      ). Similarly, we furthermore expected that CCA would help to elucidate the potential correlation between brain and behavior.
      This study was designed to overcome the aforementioned limitations of previous work by applying LICA to the Longitudinal European Autism Project (LEAP) dataset (
      • Loth E.
      • Charman T.
      • Mason L.
      • Tillmann J.
      • Jones E.J.H.
      • Wooldridge C.
      • et al.
      The EU-AIMS Longitudinal European Autism Project (LEAP): Design and methodologies to identify and validate stratification biomarkers for autism spectrum disorders.
      ) to link the sources of variance of voxelwise GM volume and WM diffusion tensor measures. The LEAP dataset provides a deeply phenotyped and comprehensively biologically assessed multisite sample of individuals with and without autism that allows relating the results of LICA to clinical characteristics of the participants. More specifically, we applied 1) a univariate approach to identify categorical group differences of linking brain patterns and subsequently their one-to-one relationships to continuous measures of autism symptoms and 2) a multivariate method (i.e., CCA) to further quantify the association between two sets of brain patterns and autistic traits in the autism group.

      Methods and Materials

      Participants

      The participants were part of the EU-AIMS (European Autism Interventions—A Multicentre Study for Developing New Medications) and AIMS-2-TRIALS (Autism Innovative Medicine Studies-2-Trials) LEAP dataset, a large multicenter study aimed at identifying and validating biomarkers in autism (
      • Loth E.
      • Charman T.
      • Mason L.
      • Tillmann J.
      • Jones E.J.H.
      • Wooldridge C.
      • et al.
      The EU-AIMS Longitudinal European Autism Project (LEAP): Design and methodologies to identify and validate stratification biomarkers for autism spectrum disorders.
      ,
      • Charman T.
      • Loth E.
      • Tillmann J.
      • Crawley D.
      • Wooldridge C.
      • Goyard D.
      • et al.
      The EU-AIMS Longitudinal European Autism Project (LEAP): Clinical characterisation.
      ). Individuals with autism were included based on an existing clinical diagnosis according to DSM-IV, DSM-IV-TR, DSM-5, or ICD-10. Each participant underwent clinical, cognitive, and MRI assessment at one of 6 collaborative centers. Readers are referred to earlier EU-AIMS LEAP publications (
      • Loth E.
      • Charman T.
      • Mason L.
      • Tillmann J.
      • Jones E.J.H.
      • Wooldridge C.
      • et al.
      The EU-AIMS Longitudinal European Autism Project (LEAP): Design and methodologies to identify and validate stratification biomarkers for autism spectrum disorders.
      ,
      • Charman T.
      • Loth E.
      • Tillmann J.
      • Crawley D.
      • Wooldridge C.
      • Goyard D.
      • et al.
      The EU-AIMS Longitudinal European Autism Project (LEAP): Clinical characterisation.
      ) for further details on experimental design and clinical characterization. In the present study, diffusion-weighted imaging data at time point 1 were available from participants in only 3 centers. Therefore, participants were selected who had both T1-weighted and diffusion-weighted imaging data available from the following centers: Institute of Psychiatry, Psychology and Neuroscience, King’s College London, United Kingdom; Radboud University Medical Centre, Nijmegen, the Netherlands; and Central Institute of Mental Health, Mannheim, Germany (Supplement Section 1).
      The final sample comprised 344 participants between 6 and 30 years of age, including 185 autistic individuals (133 male and 52 female, IQ ≥40) and 159 nonautistic individuals (99 male and 60 female, IQ ≥50). The demographic and clinical information of the final sample is summarized in Table 1. For detailed exclusion criteria, see Supplement Section 2.
      Table 1Demographic and Clinical Information of Participants
      IQ and symptom profiles reported are imputed data (30).
      Autism Group, n = 185Control Group, n = 159t2p Value
      Demographic Information
      Age, Years
      Statistical differences were assessed by two-sample t test. Degrees of freedom of the two t tests were 342.
      17.30 (5.22)17.51 (5.19)0.369.712
      IQ
      Statistical differences were assessed by two-sample t test. Degrees of freedom of the two t tests were 342.
      ,
      IQ ranged from 40 to 148 in the autism group and from 50 to 142 in the control group.
      98.90 (20.44)102.6 (19.10)1.769.079
       IQ ≥ 75105.47 (15.30)107.32 (14.04)1.083.028
       IQ < 7566.28 (6.95)63.88 (8.55)−0.994.329
       IQ ≥ 75/IQ < 75
      The differences were examined by χ2 test.
      154 (83.2%)/31 (16.8%)142 (89.3%)/17 (10.7%)2.620.106
      Sex, Female/Male
      The differences were examined by χ2 test.
      52 (28.1%)/133 (71.9%)60 (37.7%)/99 (62.3%)3.610.057
      Clinical Profiles
      ADI-R
       Social interaction16.54 (6.95)
       Communication13.35 (5.57)
       RRB4.07 (2.58)
      ADOS-2 CSSs
       Total5.40 (2.75)
       Social affect6.06 (2.64)
       RRB4.70 (2.77)
      SRS-2 Raw Score
      In SRS-2, RBS-R, and SSP questionnaires, we used parent-rated report.
      70.80 (11.55)55.78 (11.89)
      RBS-R
      In SRS-2, RBS-R, and SSP questionnaires, we used parent-rated report.
      15.54 (13.54)5.30 (6.05)
      SSP
      In SRS-2, RBS-R, and SSP questionnaires, we used parent-rated report.
      142.16 (23.63)166.66 (17.76)
      Values are mean (SD) or n (%).
      ADI-R, Autism Diagnostic Interview–Revised; ADOS-2, Autism Diagnostic Observational Schedule, Second Edition; CSSs, calibrated severity scores; RBS-R, Repetitive Behavior Scale–Revised; RRB, restricted repetitive behavior; SRS-2, Social Responsiveness Scale, Second Edition; SSP, Short Sensory Profile.
      a IQ and symptom profiles reported are imputed data (
      • Llera A.
      • Brammer M.
      • Oakley B.
      • Tillmann J.
      • Zabihi M.
      • Amelink J.S.
      • et al.
      Evaluation of data imputation strategies in complex, deeply-phenotyped data sets: The case of the EU-AIMS Longitudinal European Autism Project.
      ).
      b Statistical differences were assessed by two-sample t test. Degrees of freedom of the two t tests were 342.
      c IQ ranged from 40 to 148 in the autism group and from 50 to 142 in the control group.
      d The differences were examined by χ2 test.
      e In SRS-2, RBS-R, and SSP questionnaires, we used parent-rated report.

      Clinical Measures

      The Autism Diagnostic Interview–Revised (ADI-R) (
      • Rutter M.
      • Le Couteur A.
      • Lord C.
      Autism Diagnostic Interview–Revised.
      ) and the Autism Diagnostic Observational Schedule, Second Edition (ADOS-2) (
      • Lord C.
      • Rutter M.
      • DiLavore P.C.
      • Risi S.
      • Gotham K.
      • Bishop S.
      Autism Diagnostic Observation Schedule, Second Edition (ADOS-2) Manual (Part I): Modules 1–4.
      ) were used to measure the past (ever and previous 4–5 years) and current core autism symptom severities in social interaction, communication, and restricted repetitive behavior (RRB) domains. Specifically, the calibrated severity scores for subscales and total of the ADOS-2 were calculated to use in the following analyses (
      • Gotham K.
      • Pickles A.
      • Lord C.
      Standardizing ADOS scores for a measure of severity in autism spectrum disorders.
      ,
      • Hus V.
      • Gotham K.
      • Lord C.
      Standardizing ADOS domain scores: Separating severity of social affect and restricted and repetitive behaviors.
      ). Additionally, we used several parent-reported scales to assess autism behaviors, including the Social Responsiveness Scale, Second Edition (
      • Constantino J.N.
      • Gruber C.P.
      Social Responsiveness Scale.
      ) capturing social-communication variations, the Repetitive Behavior Scale–Revised (
      • Bodfish J.W.
      • Symons F.J.
      • Parker D.E.
      • Lewis M.H.
      Varieties of repetitive behavior in autism: Comparisons to mental retardation.
      ) identifying repetitive and rigid behaviors, and the Short Sensory Profile (SSP) (
      • Tomchek S.D.
      • Dunn W.
      Sensory processing in children with and without autism: A comparative study using the short sensory profile.
      ) evaluating sensory processing variations. Concerning the potential effect of the co-occurrence of attention-deficit/hyperactivity disorder, anxiety, and depression, we separately included the scores from an attention-deficit/hyperactivity disorder rating scale based on DSM-5 (
      • DuPaul G.J.
      • Power T.J.
      • Anastopoulos A.D.
      • Reid R.
      ADHD Rating Scale–5 for Children and Adolescents Checklists, Norms, and Clinical Interpretation.
      ) and anxiety and depression scores from the Development and Well-Being Assessment (
      • Goodman R.
      • Ford T.
      • Richards H.
      • Gatward R.
      • Meltzer H.
      The Development and Well-Being Assessment: Description and initial validation of an integrated assessment of child and adolescent psychopathology.
      ) as additional covariates in the post hoc analyses. There was a substantial amount of missing clinical data, which could greatly reduce the power of our analysis. To tackle the missing clinical data and fully harness the large LEAP sample size, we used imputed clinical data (
      • Llera A.
      • Brammer M.
      • Oakley B.
      • Tillmann J.
      • Zabihi M.
      • Amelink J.S.
      • et al.
      Evaluation of data imputation strategies in complex, deeply-phenotyped data sets: The case of the EU-AIMS Longitudinal European Autism Project.
      ). The imputation procedure was developed by our colleagues who considered the potential nonrandomness of missing data, and therefore developed quantitative measures to assess the quality of the imputations, and finally imputed data adopting a nonparametric tree regression model embedded in an iterative round-robin iterative schedule (
      • Llera A.
      • Brammer M.
      • Oakley B.
      • Tillmann J.
      • Zabihi M.
      • Amelink J.S.
      • et al.
      Evaluation of data imputation strategies in complex, deeply-phenotyped data sets: The case of the EU-AIMS Longitudinal European Autism Project.
      ). The details of missingness of the current sample are included in Supplement Section 3.

      MRI Data Acquisition

      All participants were scanned on 3T MRI scanners. High-resolution structural T1-weighted images were acquired with full head coverage at 1.2-mm thickness with 1.1- × 1.1-mm2 in-plane resolution. Diffusion-weighted imaging scans were acquired using echo-planar imaging sequence at 2-mm thickness with 2.0- × 2.0-mm2 in-plane resolution. MRI data acquisition parameters are included in Supplement Section 4.

      Image Processing

      GM Volume Estimation

      Structural T1 images were preprocessed according to the CAT12 toolbox (https://neuro-jena.github.io/cat//) pipeline in SPM12 (https://www.fil.ion.ucl.ac.uk/spm/software/spm12/) to obtain VBM data, which is a spatially unbiased whole-brain approach extracting voxelwise GM volume estimations. T1-weighted images were automatically segmented into GM, WM, and cerebrospinal fluid and affine registered to the Montreal Neurological Institute (MNI) template. A high-dimensional, nonlinear diffeomorphic registration algorithm (DARTEL) (
      • Ashburner J.
      A fast diffeomorphic image registration algorithm.
      ) was used to generate a study-specific template from GM and WM tissue segments of all participants and then to normalize all segmented GM maps to MNI space with 2-mm isotropic resolution. All GM images were smoothed with a 4-mm full width at half maximum isotropic Gaussian kernel.

      Diffusion Parameters

      Diffusion-weighted images from all sites were preprocessed using the same pipeline. Denoising was performed using the Marchenko-Pastur principal component analysis method (
      • Veraart J.
      • Fieremans E.
      • Novikov D.S.
      Diffusion MRI noise mapping using random matrix theory.
      ). Subsequently, Gibbs-ringing artifacts were removed according to Kellner et al. (
      • Kellner E.
      • Dhital B.
      • Kiselev V.G.
      • Reisert M.
      Gibbs-ringing artifact removal based on local subvoxel-shifts.
      ). FSL eddy was employed to correct the eddy current–induced distortions and subject motion (
      • Andersson J.L.R.
      • Sotiropoulos S.N.
      An integrated approach to correction for off-resonance effects and subject movement in diffusion MR imaging.
      ). To improve the final quality of data and recover most of the motion artifacts, we used intra-volume slice motion correction (
      • Andersson J.L.R.
      • Graham M.S.
      • Drobnjak I.
      • Zhang H.
      • Filippini N.
      • Bastiani M.
      Towards a comprehensive framework for movement and distortion correction of diffusion MR images: Within volume movement.
      ). Quality control reports were then generated for each subject and each site (
      • Bastiani M.
      • Cottaar M.
      • Fitzgibbon S.P.
      • Suri S.
      • Alfaro-Almagro F.
      • Sotiropoulos S.N.
      • et al.
      Automated quality control for within and between studies diffusion MRI data using a non-parametric framework for movement and distortion correction.
      ).
      Individual voxelwise fractional anisotropy (FA), mean diffusivity (MD), mode of anisotropy (MO), L1, and radial diffusivity (RA) maps were derived using dtifit in FSL (
      • Smith S.M.
      • Jenkinson M.
      • Woolrich M.W.
      • Beckmann C.F.
      • Behrens T.E.
      • Johansen-Berg H.
      • et al.
      Advances in functional and structural MR image analysis and implementation as FSL.
      ). These 5 DTI features were selected on account of the different aspects of WM microstructure; for example, FA measures the degree of anisotropic movement of water molecules, and L1 represents the magnitude of the diffusion in the primary direction, which are related to myelin structure or myelination. FA images were processed using a tract-based spatial statistics pipeline including registration of all images to FMRIB58_FA standard space, skeletonization of the mean group WM, and projection of each individual’s data onto the skeleton (
      • Smith S.M.
      • Jenkinson M.
      • Johansen-Berg H.
      • Rueckert D.
      • Nichols T.E.
      • Mackay C.E.
      • et al.
      Tract-based spatial statistics: Voxelwise analysis of multi-subject diffusion data.
      ). The mean skeleton image was thresholded at FA 0.2. Other DTI measures (MD, MO, L1, RA) were projected onto the FA skeleton using the tbss_non_FA option. All DTI data had 1-mm isotropic resolution when entering the following data fusion model. A full quality control report and additional preprocessing details of the GM and WM images are included in Supplement Section 2.

      Modalities Fusing Analysis

      The shared inter-participant variations across 6 features (i.e., VBM, FA, MD, MO, L1, RA) were explored using LICA (
      • Groves A.R.
      • Beckmann C.F.
      • Smith S.M.
      • Woolrich M.W.
      Linked independent component analysis for multimodal data fusion.
      ). LICA is able to factorize the multiple input modalities simultaneously into modalitywise independent components (ICs) while importantly constraining all decompositions to be linked through a shared participant-loading matrix, which describes the amount of contribution of each participant to a specific IC. In addition to the participant-loading matrix, this method provides, per IC, a vector reflecting the contribution (weight) of each modality and a spatial map per modality showing the extent of the spatial variation. All mathematical algorithms of LICA are detailed in Groves et al. (
      • Groves A.R.
      • Beckmann C.F.
      • Smith S.M.
      • Woolrich M.W.
      Linked independent component analysis for multimodal data fusion.
      ). As the model order is recommended to be less than 25% of the sample size (
      • Groves A.R.
      • Beckmann C.F.
      • Smith S.M.
      • Woolrich M.W.
      Linked independent component analysis for multimodal data fusion.
      ,
      • Groves A.R.
      • Smith S.M.
      • Fjell A.M.
      • Tamnes C.K.
      • Walhovd K.B.
      • Douaud G.
      • et al.
      Benefits of multi-modal fusion analysis on a large-scale dataset: Life-span patterns of inter-subject variability in cortical morphometry and white matter microstructure.
      ), 80-dimensional factorization was chosen to perform LICA. A multimodal index (
      • Francx W.
      • Llera A.
      • Mennes M.
      • Zwiers M.P.
      • Faraone S.V.
      • Oosterlaan J.
      • et al.
      Integrated analysis of gray and white matter alterations in attention-deficit/hyperactivity disorder.
      ) (Supplement Section 5) was calculated to present the contribution uniformity of the modalities in each IC. This results in a scalar value where 0 would equate to 100% unimodal contribution and 1 would mean equal contributions from all modalities.

      Statistical Approach

      The participant loadings characterize the interindividual variations of the unimodal/multimodal effects, and in the current study, they were used for the analyses of group differences between autistic and nonautistic individuals and for associations with behavioral measures. Results reported in the main text were performed using imputed data to maximize the statistical power. All analyses were replicated using the original nonimputed data.

      Case-Control Difference

      A generalized linear model (GLM) was used to examine group differences of the brain’s inter-participant variations in LICA outputs while controlling for age, sex, IQ, and scanner site. Multiple comparison (number of tests = 80) correction was implemented using false discovery rate (FDR) (p < .05) (
      • Benjamini Y.
      • Hochberg Y.
      Controlling the false discovery rate: A practical and powerful approach to multiple testing.
      ). In addition to considering the effects of co-occurring conditions on case-control difference, we separately investigated the age-by-group, IQ-by-group, sex-by-group, and site-by-group interactions and medication use effects on brain pattern(s) with the case-control difference.

      Brain-Behavior Associations

      Similarly, we used a GLM to explore the univariate associations between each IC and subscales of ADI-R and ADOS-2, Social Responsiveness Scale, Second Edition, Repetitive Behavior Scale–Revised, and SSP in the autism group while controlling for age, sex, IQ, and scanner site. We corrected for multiple comparisons (number of tests = 80 × number of [sub]scale[s]) with FDR (p < .05). Subsequently, we used one CCA (
      • Hotelling H.
      Relations between two sets of variates.
      ) to better picture the overall association between all brain ICs and subscales of ADI-R and ADOS-2 and total scores of Social Responsiveness Scale, Second Edition, Repetitive Behavior Scale–Revised, and SSP in the autism group. CCA is a multivariate approach to simultaneously learn 2 sets of linear projections corresponding to the brain ICs and the behavioral profiles, which maximizes the correlation between 2 sets of variables at the participant level. In such maximized correlation, the evaluation of brain-behavior relationships is based on the respective contribution of each IC and each behavioral profile to the correlation, which can be measured by the loading of each variable (transformed from canonical coefficient) described previously (
      • Haufe S.
      • Meinecke F.
      • Gorgen K.
      • Dahne S.
      • Haynes J.D.
      • Blankertz B.
      • et al.
      On the interpretation of weight vectors of linear models in multivariate neuroimaging.
      ). Additionally, the canonical variates are calculated respectively for the brain and the behavioral sets according to the product of the canonical coefficients and the original sets. In this study, we referred to each pair of canonical variates as CCA mode. Before entering the CCA model, age, sex, IQ, and scanning site were controlled for both brain and behavior profile sets using the Huh-Jhun residualization method (
      • Huh M.
      • Jhun M.
      Random permutation testing in multiple linear regression.
      ,
      • Winkler A.M.
      • Renaud O.
      • Smith S.M.
      • Nichols T.E.
      Permutation inference for canonical correlation analysis.
      ). The statistical significance of CCA modes was assessed by a complete permutation inference algorithm proposed by Winkler et al. (
      • Winkler A.M.
      • Renaud O.
      • Smith S.M.
      • Nichols T.E.
      Permutation inference for canonical correlation analysis.
      ), where both brain and behavior data were permuted separately across all participants with 10,000 iterations. For multiple testing correction of each CCA mode, we used a stepwise cumulative maximum approach (p < .05) [see details in (
      • Winkler A.M.
      • Renaud O.
      • Smith S.M.
      • Nichols T.E.
      Permutation inference for canonical correlation analysis.
      )]. We further tested the reliability of the CCA findings and the stability of each loading of the significant CCA mode(s) using a leave-one-subject-out approach.

      Results

      Group Effect of Brain Components

      We obtained 80 ICs from the multimodal integration analysis in our study. The modality contributions (for 80 ICs) and multimodal index of each IC are in Supplement Section 5. We subsequently used the participant loadings of the 80 ICs to test for group differences and found one component (IC58) with a significant case-control difference (β = −0.192, t337 = −3.595, FDR-corrected p = .030) (Figure 1). The respective contributions of the modalities in IC58 are 26.0% from VBM, 18.4% from FA, 17.9% from MO, 13.8% from L1, 13.7% from RA, and 10.2% from MD, indicating that various MRI features share variance associated with autism. In Figure 1, we present the summarized images of each modality’s spatial map of IC58. The spatial patterns show autism-related smaller GM volume in the bilateral insula, inferior frontal gyrus, orbitofrontal cortex, precentral and postcentral gyrus, lateral occipital cortex, inferior temporal gyrus, angular gyrus, posterior division of cingulate gyrus, and precuneus cortex and larger GM volume in calcarine cortex, bilateral middle frontal gyrus, caudate, and anterior division of cingulate gyrus. Correspondingly, autism-related DTI features were found in bilateral superior longitudinal fasciculus (SLF), corticospinal tract (CST), and inferior fronto-occipital fasciculus (IFOF). In addition to these fasciculi, RA and MD in the cingulum and anterior thalamic radiation were also implicated. Taken together, SLF and their adjacent GM volumes (i.e., involved frontal, precentral, and postcentral areas [Supplement Section 6]) in autism indicate that variations of GM volumes and WM microstructure are linked in these brain locations, rather than dependent on modality or tissue.
      Figure thumbnail gr1
      Figure 1The multimodal component shows significant case-control difference. The relative contribution of each feature is displayed in parentheses. The voxel-based morphometry (VBM) spatial map is thresholded at 5<|z|<10. Clusters of diffusion tensor imaging features were filled and thresholded at 3<|z|<10, then smoothed using a 0.3-mm Gaussian kernel in FSL for visualization purposes. FA, fractional anisotropy; L, left; L1, axial diffusivity; MD, mean diffusivity; MO, mode of anisotropy; R, right; RA, radial diffusivity.
      Post hoc, to assess the respective influences of co-occurring conditions, diagnosis-by-age, diagnosis-by-sex, or diagnosis-by-IQ interactions, and medication use on the multimodal IC found significantly associated with group, we additionally included them as separate covariates in the GLM of IC58. The analysis showed that the group effect of IC58 was robust to the inclusion of these additional covariates in the model (p < .01). However, we found a significant moderate site-by-diagnostic group interaction effect on the current result (G22 = 6.860, p = .032). This was driven by having significant effects in 2 of the 3 sites with no significant differences in the third. Details are in Supplement Section 7.

      Relating Brain Patterns to Behavior Profiles

      We conducted the univariate (GLM) and multivariate (CCA) correlation analyses on brain and behavior data in the autism group only. No significant univariate brain-behavior relationship in the autism group was found (FDR-corrected p > .120). We did, however, find a significant multivariate association pattern of CCA (r = 0.823, corrected p = .006) (Figure 2). The proportion of total variance explained by this multivariate pattern was 20.8% for brain ICs and 14.2% for behaviors. In this multivariate associated pattern, multimodal IC7 (canonical loading: −0.334) and IC78 (canonical loading: 0.283) showed strong contributions to the correlation with autism core symptoms, and from a phenotypic perspective, this multivariate pattern demonstrated a strong association with the ADI-R RRB and ADOS-2 RRB subscales. WM microstructure mainly dominated in IC7 and IC78. IC7 mainly included right inferior longitudinal fasciculus, IFOF, and CST, and IC78 primarily involved bilateral anterior thalamic radiation and SLF. These two predominant ICs highlight the involvement of WM in autism symptoms. The loadings of each brain component of this CCA mode can be found in Supplement Section 8. The leave-one-subject-out analysis indicated that the significant CCA mode of CCA analysis was reliably estimated (Supplement Section 9). We additionally ran a CCA model excluding SSP to probe the effect that the imputed SSP scores (42%) specifically may be having and found the entire structure of the output did not differ greatly (Supplement Section 10).
      Figure thumbnail gr2
      Figure 2The multivariate association pattern (i.e., canonical correlation analysis [CCA] mode) was found to be significant between the 2 sets of brain components and all behavioral profiles. (A) The scatterplot of this correlation (between the CCA variates); x- and y-axes represent the pair of CCA variates. One dot per each participant is coded with gradient color relating to the restricted repetitive behavior (RRB) subscale of the Autism Diagnostic Interview-Revised (ADI-R). (B) The loadings of each behavioral (sub)scale in this CCA mode. (C) The modality contributions to the components displayed in panel (D). (D) The 2 multimodal components with a strong contribution to the correlation with autism core symptoms; the top 2 loading modalities in each component are shown. The canonical loading of each component is shown in parentheses. The modality spatial maps are thresholded at 3<|z|<10. The CCA was performed only in the autism group. ADOS-2, Autism Diagnostic Observational Schedule, Second Edition; comm., communication; FA, fractional anisotropy; IC, independent component; L1, axial diffusivity; MD, mean diffusivity; MO, mode of anisotropy; RA, radial diffusivity; RBS-R, Repetitive Behavior Scale-Revised; SA, social affect; SRS-2, Social Responsiveness Scale, Second Edition; SSP, Short Sensory Profile; VBM, voxel-based morphometry.
      The results using nonimputed data of group effect and univariate brain-behavior association were similar to the main results. The different CCA patterns using nonimputed data were reasonable owing to the large amount of missingness (Supplement Section 3).

      Discussion

      We examined autism-related interindividual variance of integrated GM-WM morphology in a large European sample of individuals with and without autism across a broad age and IQ range. Analyses showed a significant diagnostic group effect of the linked GM-WM pattern that supports our hypothesis of the link between GM and WM morphology alterations in autistic individuals. In particular, the GM volume variation in precentral and postcentral areas converged with the WM microstructural variation in the SLF. This spotlights the shared variances between GM and WM morphology in these brain areas in autism and suggests that the structural associations in autism are not only limited to localized regions, but also involve the WM pathways connecting these brain areas. In a next set of analyses, we found a significant integrative association between brain patterns and autism core behaviors using CCA in the autism group, where the identified brain multimodal patterns underline the important role of WM morphology.
      Notably, the autism-specific VBM pattern on this multimodal analysis replicates to a certain extent our previous unimodal GM volume covariation study in a larger overlapped sample of the EU-AIMS project (
      • Mei T.
      • Llera A.
      • Floris D.L.
      • Forde N.J.
      • Tillmann J.
      • Durston S.
      • et al.
      Gray matter covariations and core symptoms of autism: The EU-AIMS Longitudinal European Autism Project.
      ). The areas of bilateral insula, inferior frontal gyrus, orbitofrontal cortex, and caudate form a steady autism-related covariation pattern in previous and current studies. These areas were demonstrated previously to relate to repetitive behaviors and reward-based decision-making abilities in autism (
      • Carlisi C.O.
      • Norman L.
      • Murphy C.M.
      • Christakou A.
      • Chantiluke K.
      • Giampietro V.
      • et al.
      Shared and disorder-specific neurocomputational mechanisms of decision-making in autism spectrum disorder and obsessive-compulsive disorder.
      ,
      • Langen M.
      • Durston S.
      • Kas M.J.
      • van Engeland H.
      • Staal W.G.
      The neurobiology of repetitive behavior: … and men.
      ). The covariation of insula and frontal areas in our studies indicates the consistency and stability of the co-occurring GM morphological alterations in autism. Benefiting from multimodal/multivariate approaches suggested by previous studies (
      • Mulders P.
      • Llera A.
      • Tendolkar I.
      • van Eijndhoven P.
      • Beckmann C.
      Personality profiles are associated with functional brain networks related to cognition and emotion.
      ,
      • Smith S.M.
      • Nichols T.E.
      Statistical challenges in “big data” human neuroimaging.
      ), the application of the LICA approach modeling the shared variances across modalities extends identified autism-related GM associations to precentral, postcentral, occipital, and temporal areas and additionally links with significant WM findings of DTI measures.
      Our results indicated one covarying set of brain GM and WM areas associated with autism diagnosis. In this multimodal set, GM volume in cortical and subcortical regions and microstructure in WM tracts (mainly SLF, CST, and IFOF) were implicated, and these regions/tracts have previously been identified in unimodal analyses (
      • Tung Y.H.
      • Lin H.Y.
      • Chen C.L.
      • Shang C.Y.
      • Yang L.Y.
      • Hsu Y.C.
      • et al.
      Whole brain white matter tract deviation and idiosyncrasy from normative development in autism and ADHD and unaffected siblings link with dimensions of psychopathology and cognition.
      ,
      • Carlisi C.O.
      • Norman L.
      • Murphy C.M.
      • Christakou A.
      • Chantiluke K.
      • Giampietro V.
      • et al.
      Shared and disorder-specific neurocomputational mechanisms of decision-making in autism spectrum disorder and obsessive-compulsive disorder.
      ,
      • Jung M.
      • Tu Y.
      • Lang C.A.
      • Ortiz A.
      • Park J.
      • Jorgenson K.
      • et al.
      Decreased structural connectivity and resting-state brain activity in the lateral occipital cortex is associated with social communication deficits in boys with autism spectrum disorder.
      ,
      • Lukito S.
      • Norman L.
      • Carlisi C.
      • Radua J.
      • Hart H.
      • Simonoff E.
      • et al.
      Comparative meta-analyses of brain structural and functional abnormalities during cognitive control in attention-deficit/hyperactivity disorder and autism spectrum disorder.
      ,
      • Thompson A.
      • Shahidiani A.
      • Fritz A.
      • O’Muircheartaigh J.
      • Walker L.
      • D’Almeida V.
      • et al.
      Age-related differences in white matter diffusion measures in autism spectrum condition.
      ). This broad range of brain regions along with large WM bundles associated with autism is in accord with the notion that the neural correlates of autism are widespread in brain regions and connectivity patterns (
      • Ecker C.
      • Bookheimer S.Y.
      • Murphy D.G.
      Neuroimaging in autism spectrum disorder: Brain structure and function across the lifespan.
      ,
      • Philip R.C.
      • Dauvermann M.R.
      • Whalley H.C.
      • Baynham K.
      • Lawrie S.M.
      • Stanfield A.C.
      A systematic review and meta-analysis of the fMRI investigation of autism spectrum disorders.
      ,
      • Ilioska I.
      • Oldehinkel M.
      • Llera A.
      • Chopra S.
      • Looden T.
      • Chauvin R.
      • et al.
      Connectome-wide mega-analysis reveals robust patterns of atypical functional connectivity in autism.
      ). This also corresponds with another multimodal autism study reporting extensive autism-related brain areas (
      • Itahashi T.
      • Yamada T.
      • Nakamura M.
      • Watanabe H.
      • Yamagata B.
      • Jimbo D.
      • et al.
      Linked alterations in gray and white matter morphology in adults with high-functioning autism spectrum disorder: A multimodal brain imaging study.
      ). The areas of this IC have been linked previously to both social and nonsocial cognitive difficulties in individuals with autism, varying from visual, sensory, and motor processing to high-order cognitive abilities (
      • Tung Y.H.
      • Lin H.Y.
      • Chen C.L.
      • Shang C.Y.
      • Yang L.Y.
      • Hsu Y.C.
      • et al.
      Whole brain white matter tract deviation and idiosyncrasy from normative development in autism and ADHD and unaffected siblings link with dimensions of psychopathology and cognition.
      ,
      • Kana R.K.
      • Keller T.A.
      • Minshew N.J.
      • Just M.A.
      Inhibitory control in high-functioning autism: Decreased activation and underconnectivity in inhibition networks.
      ,
      • McGrath J.
      • Johnson K.
      • O’Hanlon E.
      • Garavan H.
      • Gallagher L.
      • Leemans A.
      White matter and visuospatial processing in autism: A constrained spherical deconvolution tractography study.
      ,
      • Nomi J.S.
      • Molnar-Szakacs I.
      • Uddin L.Q.
      Insular function in autism: Update and future directions in neuroimaging and interventions.
      ,
      • Sato W.
      • Kochiyama T.
      • Uono S.
      • Yoshimura S.
      • Kubota Y.
      • Sawada R.
      • et al.
      Reduced gray matter volume in the social brain network in adults with autism spectrum disorder.
      ). For example, precentral and postcentral gyrus, SLF, and CST are related to (sensory-)motor processing and have been implicated in autism (
      • Tung Y.H.
      • Lin H.Y.
      • Chen C.L.
      • Shang C.Y.
      • Yang L.Y.
      • Hsu Y.C.
      • et al.
      Whole brain white matter tract deviation and idiosyncrasy from normative development in autism and ADHD and unaffected siblings link with dimensions of psychopathology and cognition.
      ,
      • Thompson A.
      • Shahidiani A.
      • Fritz A.
      • O’Muircheartaigh J.
      • Walker L.
      • D’Almeida V.
      • et al.
      Age-related differences in white matter diffusion measures in autism spectrum condition.
      ,
      • Thompson A.
      • Murphy D.
      • Dell’Acqua F.
      • Ecker C.
      • McAlonan G.
      • Howells H.
      • et al.
      Impaired communication between the motor and somatosensory homunculus is associated with poor manual dexterity in autism spectrum disorder.
      ). These adjacent affected areas (grouped areas of precentral and postcentral areas, SLF, and CST; grouped areas of lateral occipital cortex and IFOF occipital section) in our findings logically are in line with the brain organization principles during development, which states that nearby areas tend to be more interconnected (
      • Beul S.F.
      • Grant S.
      • Hilgetag C.C.
      A predictive model of the cat cortical connectome based on cytoarchitecture and distance.
      ,
      • Oligschlager S.
      • Xu T.
      • Baczkowski B.M.
      • Falkiewicz M.
      • Falchier A.
      • Linn G.
      • et al.
      Gradients of connectivity distance in the cerebral cortex of the macaque monkey.
      ). In summary, the autism diagnosis–related covarying GM-WM pattern reflects that autism is a complex condition associated with neural morphology. However, we did not find any significant univariate relationship between behavioral phenotypes and brain patterns. This is probably a result of the diverse phenotypes in our sample (i.e., complex and heterogeneous nature of autism); therefore, the compound variances of the symptom profiles cannot be explained by single unimodal/multimodal brain patterns. Additionally, imaging studies suggested that individuals with autism develop alternative processing strategies (
      • Philip R.C.
      • Dauvermann M.R.
      • Whalley H.C.
      • Baynham K.
      • Lawrie S.M.
      • Stanfield A.C.
      A systematic review and meta-analysis of the fMRI investigation of autism spectrum disorders.
      ) that might mix or neutralize the manifestations of behavioral phenotypes in autism moderating detection of well-established brain-behavior relationships. Furthermore, nonsignificant univariate relationships but one remarkable multivariate brain-behavior relationship in the current study may relate to the relatively mild autism traits in the LEAP cohort; for example, the average total score of ADOS-2 calibrated severity scores is lower than the clinical cutoffs, which was reported in the larger LEAP sample compared with other cohorts (
      • Charman T.
      • Loth E.
      • Tillmann J.
      • Crawley D.
      • Wooldridge C.
      • Goyard D.
      • et al.
      The EU-AIMS Longitudinal European Autism Project (LEAP): Clinical characterisation.
      ).
      The significant multivariate brain-behavior relationship in the current study is one prominent WM-dominated multivariate relationship between all brain patterns and all behavioral profiles. The top two ranking ICs emphasize the importance of WM connection to the core traits of autism. Multivariate/multimodal analysis increases the difficulty in interpreting findings, as it is challenging to clarify the direction of each association. Nonetheless, coinciding with previous studies (
      • Aoki Y.
      • Yoncheva Y.N.
      • Chen B.
      • Nath T.
      • Sharp D.
      • Lazar M.
      • et al.
      Association of white matter structure with autism spectrum disorder and attention-deficit/hyperactivity disorder.
      ,
      • Tung Y.H.
      • Lin H.Y.
      • Chen C.L.
      • Shang C.Y.
      • Yang L.Y.
      • Hsu Y.C.
      • et al.
      Whole brain white matter tract deviation and idiosyncrasy from normative development in autism and ADHD and unaffected siblings link with dimensions of psychopathology and cognition.
      ,
      • Thompson A.
      • Shahidiani A.
      • Fritz A.
      • O’Muircheartaigh J.
      • Walker L.
      • D’Almeida V.
      • et al.
      Age-related differences in white matter diffusion measures in autism spectrum condition.
      ,
      • Prigge M.B.D.
      • Lange N.
      • Bigler E.D.
      • King J.B.
      • Dean 3rd, D.C.
      • Adluru N.
      • et al.
      A 16-year study of longitudinal volumetric brain development in males with autism.
      ,
      • Ohta H.
      • Aoki Y.Y.
      • Itahashi T.
      • Kanai C.
      • Fujino J.
      • Nakamura M.
      • et al.
      White matter alterations in autism spectrum disorder and attention-deficit/hyperactivity disorder in relation to sensory profile.
      ), there are associations of inferior longitudinal fasciculus, IFOF, CST, and SLF microstructural measures with core symptoms/traits in autism. In line with previous findings, our work also shows laterality effects with much of the contribution from IC7 being right lateralized in autism (
      • McGrath J.
      • Johnson K.
      • O’Hanlon E.
      • Garavan H.
      • Gallagher L.
      • Leemans A.
      White matter and visuospatial processing in autism: A constrained spherical deconvolution tractography study.
      ,
      • Boets B.
      • Van Eylen L.
      • Sitek K.
      • Moors P.
      • Noens I.
      • Steyaert J.
      • et al.
      Alterations in the inferior longitudinal fasciculus in autism and associations with visual processing: A diffusion-weighted MRI study.
      ). Significantly, GM volume contributed only a small amount, which indicates that WM morphology has a stronger connection to the autism behavioral phenotypes compared with GM in this multivariate correlation. In our previous GM work, a multivariate correlation pattern exhibited a strong association between RRB scores of ADI-R and ADOS-2 and GM covariations in autism, while here, when including WM microstructural measures, the brain patterns demonstrated a strong association with RRB domains of the ADI-R and ADOS-2. This multivariate brain-behavior association needs further investigation to determine the relationship between the development of WM microstructure and behaviors, to determine the generalizability beyond the current sample, and to explore how different behavioral scales capture behavioral phenotypes in autism, which might expand our knowledge of current brain-behavior association patterns.
      Our findings should be interpreted with regard to several limitations. First, to generalize our pattern of brain alterations associated with autism requires replication in other large-scale datasets. Second, the current multimodal dataset included fewer participants than our previous work (
      • Mei T.
      • Llera A.
      • Floris D.L.
      • Forde N.J.
      • Tillmann J.
      • Durston S.
      • et al.
      Gray matter covariations and core symptoms of autism: The EU-AIMS Longitudinal European Autism Project.
      ), which may have lowered statistical power when detecting the group effects and brain-behavior associations in autism group. Despite that, this is still the largest multimodal MRI study of autism to date and includes a diverse sample of autistic and nonautistic participants. Third, limited to the cross-sectional nature of the current study, our findings are deficient to address the developmental effects on these brain patterns and their relations to the behavior profiles, as the structures of the brain (especially WM) change remarkably over puberty and with aging.
      In this study, we demonstrated autism-related interindividual covariations of GM volume in frontal, precentral, postcentral, and occipital areas and microstructure in associated WM fasciculi. Together, these GM and WM alterations are part of the underlying neural substrates of the phenotypes in autism. Subsequently, we highlighted the potential role of WM in relation to the core symptoms of autism. Further studies may link our GM-WM morphometric findings with brain function acquired from cognitive assessments and/or functional MRI data.

      Acknowledgments and Disclosures

      This work was primarily supported by the EU-AIMS consortium (European Autism Interventions), which receives support from Innovative Medicines Initiative Joint Undertaking (Grant No.115300), the resources of which are composed of financial contributions from the European Union’s Seventh Framework Programme (Grant No. FP7/2007-2013), from the European Federation of Pharmaceutical Industries and Associations companies’ in-kind contributions, and by the AIMS-2-TRIALS consortium, which has received funding from the Innovative Medicines Initiative 2 Joint Undertaking (Grant No. 777394). This Joint Undertaking receives support from the European Union’s Horizon 2020 research and innovation programme; European Federation of Pharmaceutical Industries and Associations; and Autism Speaks, Autistica, and Simons Foundation Autism Research Initiative. This work reflects the authors’ views, and neither Innovative Medicines Initiative nor the European Union, European Federation of Pharmaceutical Industries and Associations, or any associated partners are responsible for any use that may be made of the information contained therein.
      This work was also supported by the China Scholarship Council (Grant No. 201806010408 [to TM]), European Union’s Horizon 2020 Research and Innovation Programme under the Marie Skłodowska-Curie grant agreement (Grant No. 101025785 [to DLF]), European Union Seventh Framework Programme (Grant Nos. 602805 [AGGRESSOTYPE], 603016 [MATRICS], and 278948 [TACTICS] [to JKB], European Community’s Horizon 2020 Programme H2020/2014–2020 (Grant Nos. 643051 [MiND] and 642996 [BRAINVIEW] [to JKB] and Grant No. 847818 [CANDY; to JKB and CFB], Netherlands Organization for Scientific Research VICI (Grant No. 2020/TTW/00836465 [to CFB]), Wellcome Trust Collaborative Award (Grant No. 215573/Z/19/Z [to CFB]), and Autism Research Trust (to SB-C).
      We thank all participants and their families for participating in this study. We gratefully acknowledge the contributions of all members of the EU-AIMS LEAP group.
      A previous version of this article was published as a preprint on bioRxiv: https://doi.org/10.1101/2022.02.16.480649.
      TC has received consultancy fees from Roche and Servier and received book royalties from Guilford Press and Sage. DGMM has been a consultant to and advisory board member for Roche and Servier. He is not an employee or a stock shareholder of either of these companies. CFB is director and a shareholder of SBGneuro Ltd. TB has served in an advisory or consultancy role for ADHS digital, InfectoPharm, Lundbeck, Medice, Neurim Pharmaceuticals, Oberberg GmbH, Roche, and Takeda. He has received conference support or speaker’s fees from Medice and Takeda. He has received royalties from Hogrefe, Kohlhammer, CIP-Medien, and Oxford University Press. JKB has been a consultant to, advisory board member of, and a speaker for Janssen Cilag BV, Eli Lilly and Company, Shire, Lundbeck, Roche, and Servier. He is not an employee or a stock shareholder of any of these companies. He has no other financial or material support, including expert testimony, patents, or royalties. The present work is unrelated to the above grants and relationships. All other authors report no biomedical financial interests or potential conflicts of interest.

      Supplementary Material

      References

        • American Psychiatric Association
        Diagnostic and Statistical Manual of Mental Disorders.
        5th ed. American Psychiatric Association, Arlington, VA2013
        • Aoki Y.
        • Yoncheva Y.N.
        • Chen B.
        • Nath T.
        • Sharp D.
        • Lazar M.
        • et al.
        Association of white matter structure with autism spectrum disorder and attention-deficit/hyperactivity disorder.
        JAMA Psychiatry. 2017; 74: 1120-1128
        • van Rooij D.
        • Anagnostou E.
        • Arango C.
        • Auzias G.
        • Behrmann M.
        • Busatto G.F.
        • et al.
        Cortical and subcortical brain morphometry differences between patients with autism spectrum disorder and healthy individuals across the lifespan: Results from the ENIGMA ASD Working Group.
        Am J Psychiatry. 2018; 175: 359-369
        • Bedford S.A.
        • Park M.T.M.
        • Devenyi G.A.
        • Tullo S.
        • Germann J.
        • Patel R.
        • et al.
        Large-scale analyses of the relationship between sex, age and intelligence quotient heterogeneity and cortical morphometry in autism spectrum disorder.
        Mol Psychiatry. 2020; 25: 614-628
        • Zhang W.
        • Groen W.
        • Mennes M.
        • Greven C.
        • Buitelaar J.
        • Rommelse N.
        Revisiting subcortical brain volume correlates of autism in the ABIDE dataset: Effects of age and sex.
        Psychol Med. 2018; 48: 654-668
        • DeRamus T.P.
        • Kana R.K.
        Anatomical likelihood estimation meta-analysis of grey and white matter anomalies in autism spectrum disorders.
        Neuroimage Clin. 2015; 7: 525-536
        • Carlisi C.O.
        • Norman L.J.
        • Lukito S.S.
        • Radua J.
        • Mataix-Cols D.
        • Rubia K.
        Comparative multimodal meta-analysis of structural and functional brain abnormalities in autism spectrum disorder and obsessive-compulsive disorder.
        Biol Psychiatry. 2017; 82: 83-102
        • Mei T.
        • Llera A.
        • Floris D.L.
        • Forde N.J.
        • Tillmann J.
        • Durston S.
        • et al.
        Gray matter covariations and core symptoms of autism: The EU-AIMS Longitudinal European Autism Project.
        Mol Autism. 2020; 11: 86
        • Naigles L.R.
        • Johnson R.
        • Mastergeorge A.
        • Ozonoff S.
        • Rogers S.J.
        • Amaral D.G.
        • et al.
        Neural correlates of language variability in preschool-aged boys with autism spectrum disorder.
        Autism Res. 2017; 10: 1107-1119
        • Tung Y.H.
        • Lin H.Y.
        • Chen C.L.
        • Shang C.Y.
        • Yang L.Y.
        • Hsu Y.C.
        • et al.
        Whole brain white matter tract deviation and idiosyncrasy from normative development in autism and ADHD and unaffected siblings link with dimensions of psychopathology and cognition.
        Am J Psychiatry. 2021; 178: 730-743
        • Groves A.R.
        • Beckmann C.F.
        • Smith S.M.
        • Woolrich M.W.
        Linked independent component analysis for multimodal data fusion.
        Neuroimage. 2011; 54: 2198-2217
        • Cauda F.
        • Costa T.
        • Palermo S.
        • D’Agata F.
        • Diano M.
        • Bianco F.
        • et al.
        Concordance of white matter and gray matter abnormalities in autism spectrum disorders: A voxel-based meta-analysis study.
        Hum Brain Mapp. 2014; 35: 2073-2098
        • Ecker C.
        • Andrews D.
        • Dell’Acqua F.
        • Daly E.
        • Murphy C.
        • Catani M.
        • et al.
        Relationship between cortical gyrification, white matter connectivity, and autism spectrum disorder.
        Cereb Cortex. 2016; 26: 3297-3309
        • Groves A.R.
        • Smith S.M.
        • Fjell A.M.
        • Tamnes C.K.
        • Walhovd K.B.
        • Douaud G.
        • et al.
        Benefits of multi-modal fusion analysis on a large-scale dataset: Life-span patterns of inter-subject variability in cortical morphometry and white matter microstructure.
        Neuroimage. 2012; 63: 365-380
        • Llera A.
        • Wolfers T.
        • Mulders P.
        • Beckmann C.F.
        Inter-individual differences in human brain structure and morphology link to variation in demographics and behavior.
        Elife. 2019; 8e44443
        • Itahashi T.
        • Yamada T.
        • Nakamura M.
        • Watanabe H.
        • Yamagata B.
        • Jimbo D.
        • et al.
        Linked alterations in gray and white matter morphology in adults with high-functioning autism spectrum disorder: A multimodal brain imaging study.
        Neuroimage Clin. 2015; 7: 155-169
        • Mueller S.
        • Keeser D.
        • Samson A.C.
        • Kirsch V.
        • Blautzik J.
        • Grothe M.
        • et al.
        Convergent findings of altered functional and structural brain connectivity in individuals with high functioning autism: A multimodal MRI study.
        PLoS One. 2013; 8e67329
        • Hotelling H.
        Relations between two sets of variates.
        Biometrika. 1936; 28: 312-377
        • Loth E.
        • Charman T.
        • Mason L.
        • Tillmann J.
        • Jones E.J.H.
        • Wooldridge C.
        • et al.
        The EU-AIMS Longitudinal European Autism Project (LEAP): Design and methodologies to identify and validate stratification biomarkers for autism spectrum disorders.
        Mol Autism. 2017; 8: 24
        • Charman T.
        • Loth E.
        • Tillmann J.
        • Crawley D.
        • Wooldridge C.
        • Goyard D.
        • et al.
        The EU-AIMS Longitudinal European Autism Project (LEAP): Clinical characterisation.
        Mol Autism. 2017; 8: 27
        • Rutter M.
        • Le Couteur A.
        • Lord C.
        Autism Diagnostic Interview–Revised.
        Western Psychological Services, Los Angeles, CA2003
        • Lord C.
        • Rutter M.
        • DiLavore P.C.
        • Risi S.
        • Gotham K.
        • Bishop S.
        Autism Diagnostic Observation Schedule, Second Edition (ADOS-2) Manual (Part I): Modules 1–4.
        Western Psychological Services, Torrance, CA2012
        • Gotham K.
        • Pickles A.
        • Lord C.
        Standardizing ADOS scores for a measure of severity in autism spectrum disorders.
        J Autism Dev Disord. 2009; 39: 693-705
        • Hus V.
        • Gotham K.
        • Lord C.
        Standardizing ADOS domain scores: Separating severity of social affect and restricted and repetitive behaviors.
        J Autism Dev Disord. 2014; 44: 2400-2412
        • Constantino J.N.
        • Gruber C.P.
        Social Responsiveness Scale.
        2nd ed. Western Psychological Services, Torrance, CA2012
        • Bodfish J.W.
        • Symons F.J.
        • Parker D.E.
        • Lewis M.H.
        Varieties of repetitive behavior in autism: Comparisons to mental retardation.
        J Autism Dev Disord. 2000; 30: 237-243
        • Tomchek S.D.
        • Dunn W.
        Sensory processing in children with and without autism: A comparative study using the short sensory profile.
        Am J Occup Ther. 2007; 61: 190-200
        • DuPaul G.J.
        • Power T.J.
        • Anastopoulos A.D.
        • Reid R.
        ADHD Rating Scale–5 for Children and Adolescents Checklists, Norms, and Clinical Interpretation.
        Guilford Publications, New York2016
        • Goodman R.
        • Ford T.
        • Richards H.
        • Gatward R.
        • Meltzer H.
        The Development and Well-Being Assessment: Description and initial validation of an integrated assessment of child and adolescent psychopathology.
        J Child Psychol Psychiatry. 2000; 41: 645-655
        • Llera A.
        • Brammer M.
        • Oakley B.
        • Tillmann J.
        • Zabihi M.
        • Amelink J.S.
        • et al.
        Evaluation of data imputation strategies in complex, deeply-phenotyped data sets: The case of the EU-AIMS Longitudinal European Autism Project.
        BMC Med Res Methodol. 2022; 22229
        • Ashburner J.
        A fast diffeomorphic image registration algorithm.
        Neuroimage. 2007; 38: 95-113
        • Veraart J.
        • Fieremans E.
        • Novikov D.S.
        Diffusion MRI noise mapping using random matrix theory.
        Magn Reson Med. 2016; 76: 1582-1593
        • Kellner E.
        • Dhital B.
        • Kiselev V.G.
        • Reisert M.
        Gibbs-ringing artifact removal based on local subvoxel-shifts.
        Magn Reson Med. 2016; 76: 1574-1581
        • Andersson J.L.R.
        • Sotiropoulos S.N.
        An integrated approach to correction for off-resonance effects and subject movement in diffusion MR imaging.
        Neuroimage. 2016; 125: 1063-1078
        • Andersson J.L.R.
        • Graham M.S.
        • Drobnjak I.
        • Zhang H.
        • Filippini N.
        • Bastiani M.
        Towards a comprehensive framework for movement and distortion correction of diffusion MR images: Within volume movement.
        Neuroimage. 2017; 152: 450-466
        • Bastiani M.
        • Cottaar M.
        • Fitzgibbon S.P.
        • Suri S.
        • Alfaro-Almagro F.
        • Sotiropoulos S.N.
        • et al.
        Automated quality control for within and between studies diffusion MRI data using a non-parametric framework for movement and distortion correction.
        Neuroimage. 2019; 184: 801-812
        • Smith S.M.
        • Jenkinson M.
        • Woolrich M.W.
        • Beckmann C.F.
        • Behrens T.E.
        • Johansen-Berg H.
        • et al.
        Advances in functional and structural MR image analysis and implementation as FSL.
        Neuroimage. 2004; 23: S208-S219
        • Smith S.M.
        • Jenkinson M.
        • Johansen-Berg H.
        • Rueckert D.
        • Nichols T.E.
        • Mackay C.E.
        • et al.
        Tract-based spatial statistics: Voxelwise analysis of multi-subject diffusion data.
        Neuroimage. 2006; 31: 1487-1505
        • Francx W.
        • Llera A.
        • Mennes M.
        • Zwiers M.P.
        • Faraone S.V.
        • Oosterlaan J.
        • et al.
        Integrated analysis of gray and white matter alterations in attention-deficit/hyperactivity disorder.
        Neuroimage Clin. 2016; 11: 357-367
        • Benjamini Y.
        • Hochberg Y.
        Controlling the false discovery rate: A practical and powerful approach to multiple testing.
        J R Stat Soc Series B Stat Methodol. 1995; 57: 289-300
        • Haufe S.
        • Meinecke F.
        • Gorgen K.
        • Dahne S.
        • Haynes J.D.
        • Blankertz B.
        • et al.
        On the interpretation of weight vectors of linear models in multivariate neuroimaging.
        Neuroimage. 2014; 87: 96-110
        • Huh M.
        • Jhun M.
        Random permutation testing in multiple linear regression.
        Communications in Statistics - Theory and Methods. 2001; 30: 2023-2032
        • Winkler A.M.
        • Renaud O.
        • Smith S.M.
        • Nichols T.E.
        Permutation inference for canonical correlation analysis.
        Neuroimage. 2020; 220117065
        • Carlisi C.O.
        • Norman L.
        • Murphy C.M.
        • Christakou A.
        • Chantiluke K.
        • Giampietro V.
        • et al.
        Shared and disorder-specific neurocomputational mechanisms of decision-making in autism spectrum disorder and obsessive-compulsive disorder.
        Cereb Cortex. 2017; 27: 5804-5816
        • Langen M.
        • Durston S.
        • Kas M.J.
        • van Engeland H.
        • Staal W.G.
        The neurobiology of repetitive behavior: … and men.
        Neurosci Biobehav Rev. 2011; 35: 356-3565
        • Mulders P.
        • Llera A.
        • Tendolkar I.
        • van Eijndhoven P.
        • Beckmann C.
        Personality profiles are associated with functional brain networks related to cognition and emotion.
        Sci Rep. 2018; 813874
        • Smith S.M.
        • Nichols T.E.
        Statistical challenges in “big data” human neuroimaging.
        Neuron. 2018; 97: 263-268
        • Jung M.
        • Tu Y.
        • Lang C.A.
        • Ortiz A.
        • Park J.
        • Jorgenson K.
        • et al.
        Decreased structural connectivity and resting-state brain activity in the lateral occipital cortex is associated with social communication deficits in boys with autism spectrum disorder.
        Neuroimage. 2019; 190: 205-212
        • Lukito S.
        • Norman L.
        • Carlisi C.
        • Radua J.
        • Hart H.
        • Simonoff E.
        • et al.
        Comparative meta-analyses of brain structural and functional abnormalities during cognitive control in attention-deficit/hyperactivity disorder and autism spectrum disorder.
        Psychol Med. 2020; 50: 894-919
        • Thompson A.
        • Shahidiani A.
        • Fritz A.
        • O’Muircheartaigh J.
        • Walker L.
        • D’Almeida V.
        • et al.
        Age-related differences in white matter diffusion measures in autism spectrum condition.
        Mol Autism. 2020; 11: 36
        • Ecker C.
        • Bookheimer S.Y.
        • Murphy D.G.
        Neuroimaging in autism spectrum disorder: Brain structure and function across the lifespan.
        Lancet Neurol. 2015; 14: 1121-1134
        • Philip R.C.
        • Dauvermann M.R.
        • Whalley H.C.
        • Baynham K.
        • Lawrie S.M.
        • Stanfield A.C.
        A systematic review and meta-analysis of the fMRI investigation of autism spectrum disorders.
        Neurosci Biobehav Rev. 2012; 36: 901-942
        • Ilioska I.
        • Oldehinkel M.
        • Llera A.
        • Chopra S.
        • Looden T.
        • Chauvin R.
        • et al.
        Connectome-wide mega-analysis reveals robust patterns of atypical functional connectivity in autism.
        medRxiv. 2022; https://doi.org/10.1101/2022.01.09.22268936
        • Kana R.K.
        • Keller T.A.
        • Minshew N.J.
        • Just M.A.
        Inhibitory control in high-functioning autism: Decreased activation and underconnectivity in inhibition networks.
        Biol Psychiatry. 2007; 62: 198-206
        • McGrath J.
        • Johnson K.
        • O’Hanlon E.
        • Garavan H.
        • Gallagher L.
        • Leemans A.
        White matter and visuospatial processing in autism: A constrained spherical deconvolution tractography study.
        Autism Res. 2013; 6: 307-319
        • Nomi J.S.
        • Molnar-Szakacs I.
        • Uddin L.Q.
        Insular function in autism: Update and future directions in neuroimaging and interventions.
        Prog Neuropsychopharmacol Biol Psychiatry. 2019; 89: 412-426
        • Sato W.
        • Kochiyama T.
        • Uono S.
        • Yoshimura S.
        • Kubota Y.
        • Sawada R.
        • et al.
        Reduced gray matter volume in the social brain network in adults with autism spectrum disorder.
        Front Hum Neurosci. 2017; 11: 395
        • Thompson A.
        • Murphy D.
        • Dell’Acqua F.
        • Ecker C.
        • McAlonan G.
        • Howells H.
        • et al.
        Impaired communication between the motor and somatosensory homunculus is associated with poor manual dexterity in autism spectrum disorder.
        Biol Psychiatry. 2017; 81: 211-219
        • Beul S.F.
        • Grant S.
        • Hilgetag C.C.
        A predictive model of the cat cortical connectome based on cytoarchitecture and distance.
        Brain Struct Funct. 2015; 220: 3167-3184
        • Oligschlager S.
        • Xu T.
        • Baczkowski B.M.
        • Falkiewicz M.
        • Falchier A.
        • Linn G.
        • et al.
        Gradients of connectivity distance in the cerebral cortex of the macaque monkey.
        Brain Struct Funct. 2019; 224: 925-935
        • Prigge M.B.D.
        • Lange N.
        • Bigler E.D.
        • King J.B.
        • Dean 3rd, D.C.
        • Adluru N.
        • et al.
        A 16-year study of longitudinal volumetric brain development in males with autism.
        Neuroimage. 2021; 236118067
        • Ohta H.
        • Aoki Y.Y.
        • Itahashi T.
        • Kanai C.
        • Fujino J.
        • Nakamura M.
        • et al.
        White matter alterations in autism spectrum disorder and attention-deficit/hyperactivity disorder in relation to sensory profile.
        Mol Autism. 2020; 11: 77
        • Boets B.
        • Van Eylen L.
        • Sitek K.
        • Moors P.
        • Noens I.
        • Steyaert J.
        • et al.
        Alterations in the inferior longitudinal fasciculus in autism and associations with visual processing: A diffusion-weighted MRI study.
        Mol Autism. 2018; 9: 10