Atypical lateralization of motor circuit functional connectivity in children with autism is associated with motor deficits

Participants comprised 44 right-handed children with ASC (36 males, 8 females) and 80 right-handed typically developing children (58 males, 22 females) between 8 and 12 years of age (ASC mean = 10.23, standard deviation (SD) = 1.51, range = 8.01–12.99; controls mean = 10.15, SD = 1.08, range = 8.07–12.76) (see Additional file 1; for participant characteristics of the final sample, see Table 1). Individuals with ASC were recruited through advertisements placed within community-wide service groups, schools and hospitals, as well as from outpatient clinics at the Kennedy Krieger Institute in Baltimore. Participants with ASC received a clinical diagnosis according to the criteria of the Diagnostic and Statistical Manual of Mental Disorders–IV [51], and diagnoses were confirmed using the Autism Diagnostic Interview–Revised (ADI-R; [52]) and Module 3 of the Autism Diagnostic Observation Schedule (ADOS-G; [53]). Participants were excluded if they met any of the following criteria: (1) a history of seizures and/or traumatic brain injury; (2) a full-scale IQ (FIQ) less than 80 (in cases where the FIQ was less than 80, participants were still included if their verbal comprehension index (VCI) or perceptual reasoning index (PRI) was higher than 80), as assessed by the Wechsler Intelligence Scale for Children- fourth edition (WISC-III/IV; [54, 55]); (3) a developmental language disorder; (4) reading disability; (5) visual impairment and (6) neurologic disorder such as epilepsy. The Diagnostic Interview for Children and Adolescents, fourth edition (DICA-IV; [56]) was used to determine the presence of additional psychiatric diagnoses. Children with ASC met the following additional DICA-IV diagnoses: ADHD (16), OCD (4), specific phobia (9), generalized anxiety disorder (GAD) (4) and oppositional defiant disorder (ODD) (8). Twenty-four children with ASC were actively taking psychoactive medications, including stimulants (13), selective serotonin reuptake inhibitors (SSRI) (6), clonidine (2), risperidone (2) and lithium (2). Stimulant medications were discontinued the day prior to testing; all other medications were taken as prescribed. There was no diagnosis or family history of ASC in the typically developing control group. We included only individuals in the current sample with less than 3 mm translational and 3° rotational movement over the course of the resting scan.

Handedness was assessed using the Edinburgh Handedness Inventory (EHI; [58]), a self-completed questionnaire for determining hand preference. The sample comprised only right-handed individuals with EHI scores >40.

All participants performed a mock scan the day before the actual scan. All individuals underwent scanning on one of two 3-T Philips scanners (2D-SENSE EPI, 8-channel head coil, SENSE acceleration = 2.0), and axially oriented volumes were acquired using T2*-weighted echo-planar imaging (field of view: 256 × 256 mm, matrix size 64 × 64, repetition time = 2500 ms, echo time = 30 ms, flip angle = 75°). Resting-state scans were acquired for 5 min and 20 s. Children were asked to stay as still as possible and fixate on a centre cross. T1-weighted high-resolution anatomical images were acquired coronally (field of view 256 × 200 mm², matrix size 256 × 256, repetition time = 7.99 ms, echo time = 3.76 ms, flip angle = 8°, 1 mm isotropic voxels, slice thickness = 1 mm). These were used to generate age- and gender-matched symmetrical tissue priors.

Functional T2*-weighted images were preprocessed using statistical parametric mapping (SPM12; Wellcome Department of Imaging Neuroscience Group, London, UK; http://​www.​fil.​ion.​ucl.​ac.​uk/​spm). Images were slice-time corrected using the middle slice as reference slice and realigned relative to their mean. The high-resolution anatomical images were then co-registered to the functional images, segmented and normalized using a symmetrical, age- and gender-matched tissue prior generated with the Template-O-Matic toolbox [59]. The use of a symmetrical template prevents an additional introduction of anatomical asymmetries that might potentially interfere with functional asymmetries [60]. The normalization transformation was then applied to the functional images. Further steps included linear detrending at each voxel in the brain to correct for scanner drift, removal of nuisance variables such as the white matter (WM) and cerebrospinal fluid (CSF) using CompCor [61] (note that we did not use global signal regression (GSR) to avoid introduction of spurious anticorrelations in the data [62]) and six absolute and six differential motion parameters, spatial smoothing (6-mm full width at half maximum (FWHM)), and temporal band-pass filtering constraining the frequency window to 0.01–0.1 Hz. To minimise the confounding influence of micromovement, we computed the average framewise displacement (FD) (based on the median due to a non-normal distribution of movement) according to Power et al. [63] and excluded any participant with a z score of >2.58 [64].

Mostofsky et al. [22] identified the right hemisphere (RH) and left hemisphere (LH) circuits involved in motor execution during a finger-sequencing paradigm in both children with ASC and neurotypical controls between 8 and 12 years. They identified regions of interest by single-group, whole-brain random effects analyses across both groups by executing one-sample t tests on the individual subject’s right- and left-rest contrast images. Individual within-group whole-brain activation analyses were then run for the two groups separately identifying peak activations by group. Here, peak coordinates were derived from typical children and served as centres of seeds.

Based on the ‘left hemisphere dysfunction’ theory of autism, which states that neural impairments are lateralized to the LH in individuals with ASC, we selected peak coordinates from this LH motor circuit (see Table 2). The LH motor circuit (right-sided movement) included the left sensorimotor cortex (SMC), left thalamus (TH), left putamen (PUT), bilateral supplementary motor area (SMA)-rostral (SMA-r), bilateral SMA-dorsal (SMA-d) and right anterior cerebellum (AC) seeds [4].

Six-millimetre-radius 3D seeds were then created using the SPM toolbox PickAtlas (http://​fmri.​wfubmc.​edu/​software/​PickAtlas) with the peak coordinates as the centre of spheres. All seeds were flipped across the midline (x = 0) in order to obtain a set of homotopic RH and LH seeds (see Fig. 1). Time series were then extracted from the seed regions of interest, and pairwise Pearson’s correlations were performed for the LH and RH circuits separately. Correlation coefficients were transformed to a normal distribution via Fisher’s z transform. Standardized correlations between all seed pairs were then averaged to obtain mean connectivity in the motor circuit in each hemisphere (RH connectivity (R _con); LH connectivity (L _con)) using MATLAB. For the assessment of functional lateralization, we calculated a laterality index (LI) for each pairwise connection and for the overall mean network connectivity by subtracting the connections in the LH from homotopic connections in the right hemisphere: R _con − L _con. Contrary to the usually applied formula for the calculation of structural asymmetry indices ((R − L)/(R + L)), here we did not include a denominator as functional connectivity measures comprise both positive and negative values, adversely influencing the index values [50].

To test whether atypical lateralization is specific to the motor network and its underlying lateralized functions, we selected two control networks: the default mode network (DMN) and a visual network (VN). DMN seeds were based on peak seed regions of the task-negative network identified by Fox et al. [65] using resting-state connectivity. Seeds were located in the cerebellum, posterior cingulate cortex (PCC), medial prefrontal cortex (MPFC), retro-splenial cortex (RSC), lateral parietal cortex (LP), superior frontal gyrus (SFG), inferior frontal gyrus (IFG) and parahippocampal gyrus (PHG). The VN seeds were based on visual seeds applied by Yeo et al. [66] and were comprised of the central and peripheral primary visual cortex (V1_c and V1_p), the central and peripheral regions near the visual area V3_v (V3_pv and V3_cv) and the extrastriate regions of the central and peripheral visual subnetworks (ExP and ExC).

To test whether atypical lateralization was related to function, we conducted one-tailed (based on our hypothesis that stronger rightward lateralization would be related to more symptoms and poorer performance) Pearson’s correlations controlling for scanner, age and average FD within individuals with ASC. For individual connectivity pairs, the Global-Mean_network was additionally included as a covariate. Motor LIs showing group differences were correlated with (a) the repetitive behaviour subscores of the ADI-R (ADI-C) and ADOS (ADOS-D) and (b) the total PANESS score, total gait score (composite measure of speed, overflow and dysrhythmia during gaits and stations) and total timed score (composite measure of speed, overflow and dysrhythmia during all timed movements). Level of significance was set at p = 0.05, as all correlation analyses were targeted. All statistical analyses were carried out in SPSS (version 21, SPSS Inc.).

Based on previous reports showing a linkage between motor and social development [69, 70] and the hypothesis that motor deficits might underpin some of the core social-communicative symptoms in ASC [71], an additional, exploratory analysis was conducted posing the question whether lateralization of motor circuit connectivity was related to core social and communicative symptoms, with stronger rightward lateralization being related to more deficits as measured by the ADOS-A, ADOS-B, ADI-A and ADI-B.

Four control individuals and one individual with ASC were excluded due to z values of >2.58 on average FD. One additional individual with ASC was excluded due to z values of >2.58 on more than three laterality measures. The final sample consisted of 42 individuals with ASC (34 males, 8 females) and 76 controls (54 males, 22 females). After exclusion, the two groups were matched on age (t(62.51) = −0.092, p = 0.927), handedness (U = 1590.5, z = −0.032, p = 0.974), sex (χ ² (1) = 0.925, p = 0.336), average FD (U = 1481, z = −0.646, p = 0.518) and PRI (t(116) = 0.768, p = 0.444). There were, however, significant differences in FIQ (t(64.87) = 3.136, p = 0.003) and VCI (t(70.10) = 3.353, p = 0.001). Individuals with ASC had significantly poorer performance on the total score of the PANESS (F(1,115) = 45.081, p < 0.001), and on the two sub-measures: total gait (F(1,115) = 36.804, p < 0.001) and total timed (F(1,115) = 32.339, p < 0.001).

There was a significant difference for the LI of mean motor circuit connectivity between the ASC and control group (F(1,113) = 6.814, p = 0.010; Cohen’s d = 0.506), with children with ASC being more strongly rightward lateralized (see Fig. 2a). The result remained significant when controlling for FIQ (F(1,112) = 6.094, p = 0.015; Cohen’s d = 0.479). This difference was driven by both rightward over- and leftward under-connectivity in mean motor connectivity in children with ASC (group difference R _con F(1,113) = 1.110, p = 0.294; group difference L _con F(1,113) = 0.536, p = 0.466). A one-sample t test revealed that children with ASC were rightward lateralized (t(41) = 2.390, p = 0.022), whereas control children showed a symmetrical organization (t(75) = −1.255, p = 0.213). Individual LIs of ‘motor’ connections were not significantly different or did not survive correction for multiple comparisons (see Table 3). After stratifying analyses by sex, the result was trending in males only (F(1,83) = 3.920, p = 0.051; Cohen’s d = 0.438; FIQ F(1,82) = 2.512, p = 0.117; Cohen’s d = 0.351) and not significant in females (F(1,25) = 1.311, p = 0.263; Cohen’s d = 0.489; FIQ F(1,24) = 3.701, p = 0.066; Cohen’s d = 0.822) (for a distribution of values across sex, see Fig. 2b).

The LI of total mean motor connectivity was positively associated with the total PANESS scores (r = 0.455, p = 0.002), total gait subscores (r = 0.434, p = 0.003) and total timed subscores (r = 0.357, p = 0.013) (see Table 4). In each case, stronger rightward lateralization was associated with worse performance on the motor task (see Fig. 3a–c). There were no associations with either the ADOS-D (r = 0.117, p = 0.238) or ADI-C (r = −0.026, p = 0.438). There were no significant associations between the LI of mean motor connectivity and the ADOS-A (r = 0.127, p = 0.220), ADOS-B (r = 0.217, p = 0.093), ADI-A (r = 0.065, p = 0.348) and ADI-B (r = −0.075, p = 0.325). Correlations with the PANESS were not significant in typical children (total PANESS score r = −0.059, p = 0.310; total gait subscores r = 0.094, p = 0.215; total timed subscores r = −0.132, p = 0.133).

There was no significant difference in the LI of mean VN connectivity (F(1,113) = 0.467; p = 0.496) between individuals with ASC and controls. Individual LIs of VN connections were not significantly different (for details, see Additional file 1).

This study is the first to report a relationship between autistic symptoms and hemispheric lateralization outside the language domain in individuals with ASC; namely, the more rightward asymmetry of the functional correspondence within the motor circuit, the poorer motor performance on three different measures of the PANESS. Both Nielsen et al. [46], who reported atypical lateralization in a language-related network, and Cardinale et al. [45], who reported rightward lateralization of the motor network, did not find such brain-behavioural association. Here, we highlight the importance of their findings by showing functional relevance in children with ASC. Lateralization has been linked to fine motor skills, primarily in relation to handedness. Interestingly, here, we showed that, in addition to fine motor skills, gross motor performances such as gait, balance, timing and sequencing of movements were also associated with atypical hemispheric specialization in children with ASC. Therefore, rightward lateralization may underlie gross motor deficits and atypical gait commonly found in individuals with ASC.

Leftward lateralization of the same circuit has previously been shown to be associated with better motor performance in typically developing children [4]; however, we did not replicate this result here. It remains to be established whether this is due to differing sample characteristics and methods (note that Barber et al. used a different registration approach and laterality index) and whether lateralization of this network constitutes the biological underpinnings for typical motor development or for motor deficits in ASC specifically.

Lateralization in the motor system is present in both structure and function, although it is generally less pronounced compared to the language system lateralization, which would explain why typical children show a subtle shift only towards left here. Structurally, this is characterized by a deeper left central sulcus, increased leftward neuropil in BA4 [76] and expansion of the left hand motor cortex in right-handers [77]. Guye et al. [78] show that whole-brain functional connectivity is more extensive with the left than with the right primary motor cortex. Left premotor and parietal regions are more strongly implicated in higher order actions [79]. Also, the planning of complex, sequential movements [79‐81], bimanual coordination [82, 83] and response selection [84] have been attributed to the left hemisphere, which is interesting considering the problems individuals with ASC experience with fine motor skills [85] and planned sequencing of actions [86]. In contrast, sensory processing and spatial aspects of sensorimotor actions have been reported to show right hemisphere dominance [87‐89]. It has been suggested that posture and limb position [90], as well as the use of proprioceptive feedback [91, 92], are preferentially lateralized to the right hemisphere in right-handed individuals. Interestingly, we have previously reported that individuals with ASC rely to a much greater extent on proprioceptive feedback during motor learning, which is related to more severe motor impairments and other social symptoms [93]. Atypical rightward lateralization of motor circuits might constitute a neural mechanism contributing to this finding.

Reports of disturbance in establishing typical patterns of hemispheric specialization in ASC date back to the consistently replicated observation of increased rates of left-handedness among individuals with ASC [94‐101]. Left-handers activate the right premotor cortex during contra- and ipsilateral finger movements [102] and have increased intrasulcal surface in the right precentral gyrus [76]. Interestingly, our results are not driven by differences in handedness, as participants were matched on a measure of handedness and were right-handed. Still, it will be interesting to explore whether the observed pattern is even more pronounced in left-handed individuals with ASC.

Motor deficits are present in at least 80 % of children with ASC [28, 103, 104] comprising some of the earliest signs, such as delays in motor milestones [105, 106]. Later on in development, motor-related deficits persist and become more apparent, with deficits in gross and fine motor coordination [19, 47, 107, 108], atypical gait and posture [109, 110] and particularly difficulty with performance of skilled gestures (i.e. praxis) [18, 27, 28].

In line with this, it has been argued that core social and communicative symptoms in autism might be subserved by the same neural systems underpinning motor-related dysfunction [111, 112]. In fact, there is a close relationship between motor and social-communicative behaviours: motor development is the prerequisite for speech, articulation and engagement in the social world through directing attention, grasping and sharing things, and approaching, imitating and responding to others. In line with this, Iverson and Braddock [113] have shown that children with language deficits have poorer fine and gross motor skills. At the same time, production and understanding of actions, such as performance of sequences of actions and following verbal commands, are dependent on the development of language skills. Thus, atypical lateralization in related cognitive domains such as language might be influenced by the same developmental mechanisms [114, 115]. Notably, here, we did not find a significant association between atypical motor circuit lateralization and social-communicative sub-measures of either the ADOS or ADI. Given that other studies find such relationship [116], future research needs to examine a potential association with more specific measures of atypical social development in ASC and explore the link between motor and social networks directly at the neuronal level.

Here, we report atypical laterality of the motor network in children with ASC. Given that ASC is a neurodevelopmental condition, it remains to be established whether the observed pattern is stable across the lifespan. Many studies show temporal test-retest reliability and reproducibility of functional resting-state networks [117‐119]. It has been argued that neural coupling configurations are dynamic and transient over time with fluctuations in connection strengths [120, 121]. As arousal, sleep [122], conscious [123], cognitive [124] and emotional [125] states constitute influencing factors, only replication in independent, longitudinal samples can confirm the stability of the reported findings. Age-related changes have been described in a range of different networks, with decreasing connectivity representing increasing hemispheric specialization and increasing connectivity representing increasing functional cooperation [126]. The same authors also reported not only linear but also quadratic and cubic age effects on changes of functional connectivity, emphasizing the need for longitudinal samples to fully explore the stability and plausible developmental trend of the reported findings.

One limitation concerns the heterogeneity within ASC. Children with ASC frequently have other comorbid conditions, such as ADHD, which is another condition associated with atypical lateralization [127]. Also, Mahajan et al. [128] showed that individuals with ASC with and without ADHD show differences in motor cortex laterality with children with ASC without ADHD exhibiting rightward increases in M1, whereas children with ASC and comorbid ADHD showing left-lateralized increases. Medication exposure is another possible confound which might influence connectivity strength [129‐131]. Overall, a large portion of individuals with ASC have psychiatric comorbidities during their lifespan (ADHD 28–44 %; anxiety 42–56 %; depression 12–70 %; oppositional defiant disorder 16–28 %; [132]) and around 70 % of children with ASC use some form of psychoactive medication [133, 134], making any sample that excludes these individuals less representative of the autistic population.

It remains to be established whether these results are restricted to lateralized left hemisphere functions or whether right hemisphere functions such as visuospatial abilities are also affected. Cardinale et al. [45] report rightward asymmetries of more widespread networks including those underlying visuospatial attention, but no other studies have looked at hemispheric specialization of visuospatial information processing in ASC. Future research should explore whether stronger rightward lateralization might underpin superior visual attention and weak central coherence in ASC and whether this potentially differentiates between individuals with ASC with and without enhanced visuospatial processing abilities.

Heterogeneity within the autism spectrum is the biggest challenge in the current research. It will be of interest to explore whether atypical motor network lateralization is characteristic of an as-yet undefined subgroup within ASC, such as language delay, for example, [60]. Other factors constituting heterogeneity include handedness, sex or IQ. We previously showed that handedness can be a marker of rightward lateralization of structural connectivity of regions connecting sensorimotor cortical areas [135]. Whether this also applies to functional lateralization remains to be established. As for sex, our findings here were most pronounced when including both males and females. Sex-stratified analyses show an overlapping functional organization in males and females with ASC regarding lateralization of the motor network. In general, few studies have looked at sex differences in the laterality of resting-state connectivity. Tian et al. [136] reported that males show more locally efficient right hemisphere networks, whereas females have more locally efficient left hemisphere networks. Zuo et al. [126] reported that homotopic resting-state connectivity in the dorsolateral prefrontal cortex and the amygdala show opposite developmental trajectories in males and females. How functional connectivity and its lateralization further differ across genders in ASC awaits investigation. Regarding IQ, our sample was only matched on perceptual reasoning IQ; however, additional analyses showed that results remained largely unaffected by full-scale IQ (i.e. when controlling for FIQ). As motor delay is mostly evident in low-functioning individuals [137, 138], it will be important to explore whether this result is more pronounced in this sub-population with ASC.

The present findings were based on patterns observed in the so-called resting state, and it will be interesting to explore whether atypical rightward lateralization remains robust under conditions when the motor execution network is active, as well as whether they are more pronounced in particular parcels underlying the motor homunculus. Lastly, here, we focused on atypical intrahemispheric connectivity. However, many studies have shown that atypical interhemispheric connectivity (such as reduced corpus callosum size) contributes to atypical connectivity in ASC [139‐141]. Efficient information processing across the whole brain substantially depends on interhemispheric integration, which may be impaired in ASC [142]. Herbert et al. [143] show that only intrahemispheric connectivity is impaired in ASC, whereas others demonstrate that intra- and interhemispheric connectivities are both affected [141, 144]. Particularly, the right hemisphere regions involved in visuospatial processing have been suggested to rely on more integrative, interhemispheric processing [5]. Thus, future studies need to explore whether atypical lateralization in ASC extends to cross-hemispheric networks, particularly for functions such as visuospatial skills.