Cortico-basal ganglia white matter microstructure is linked to restricted repetitive behavior in autism spectrum disorder

Autism spectrum disorder (ASD) includes two diagnostic domains: (1) social and communication deficits and (2) restricted repetitive behavior (RRB). RRB refers to multiple categories of repeating patterns of behavior that often occur with high frequency, have no clear function, interfere with appropriate behavior, and can sometimes result in bodily harm. Although many research and intervention efforts in ASD have targeted social and communication deficits, there has been considerably less focus on the RRB domain. Although behavioral interventions have some efficacy in treating RRB [1‐3], pharmacological interventions have no demonstrated efficacy for treating RRB in ASD, but rather are used primarily to treat associated problems (e.g., aggressive behavior) [4, 5]. This lack of efficacious treatments is largely due to an incomplete understanding of the neural circuitry mediating RRB in ASD. There is a pressing need to elucidate how RRB relates to differences in brain structure and function in ASD.

There has been increasing focus on hypotheses regarding brain connectivity in ASD, and how disrupted connectivity patterns relate to the two diagnostic domains. Investigations of connectivity differences in ASD have largely utilized magnetic resonance imaging (MRI) to assess structural and functional connectivity. Functional MRI and diffusion tensor imaging (DTI) in ASD both generally support reduced connectivity within major brain networks, though there is also evidence of increased connectivity between networks [6‐10]. Although connectivity differences in ASD have frequently been studied in the context of social and communication deficits [11], far fewer studies have focused on altered connectivity as it pertains to RRB. [12]

Scalar measures from DTI provide insight into tissue microstructure and can be used to characterize structural aspects of brain connectivity. These scalar measures are confounded by partial volume effects from free-water (FW), such as cerebrospinal fluid. Using a two-compartment model, the unique contribution of FW can be quantified and other scalar measures adjusted accordingly [13]. The FW measure itself has already demonstrated significant value in a variety of neurodegenerative disorders and it is hypothesized that abnormal FW values are associated with atrophy and/or inflammation [14‐19]. Despite the widespread use of DTI to study connectivity in ASD, FW quantification and correction have had limited application in this population [20, 21]. Many of the brain regions and white matter tracts hypothesized to be involved in the expression of RRB in ASD are proximal to the ventricular system, and as such, FW quantification may be of great value to DTI research in ASD.

In this study, we assessed gray matter volume and white matter microstructure in the basal ganglia and cerebellum in individuals with ASD compared to typically developing (TD) participants. We focused on the basal ganglia and cerebellum because they have been linked to RRB and ASD in prior work, yet these relationships remain poorly understood [12, 22‐24]. We further investigated whether these measures differed between males and females with ASD, and whether they correlated with clinical measures of RRB. Current estimates suggest ASD is approximately four times more prevalent in males than females [25], and some imaging studies have suggested that the sex differences in brain morphology and connectivity in ASD are unlike those observed in TD individuals [26‐31]. There is also mixed evidence of sex differences in the expression of RRB in ASD [32‐36]. It remains unknown whether possible sex differences in the expression of RRB relate to sex differences in brain morphology or microstructure. In the present study, we used clinical and neuroimaging data available from the National Institute of Mental Health Data Archive (NDA) to investigate interactions between ASD diagnosis and biological sex on regional gray matter volumes in the basal ganglia and cerebellum, as well as microstructure in white matter pathways of the basal ganglia and cerebellum. We further investigated the relationship of these neuroimaging measures to clinical measures of RRB.

There has been relatively limited focus on morphological and connectivity differences associated with RRB in ASD. An approach frequently employed to assess brain-behavior relationships to RRB in ASD has been correlation of regional gray matter volumes with clinical scales [56‐62]. Other work has used fMRI to assess activation or functional connectivity as it relates to RRB [63‐71]. Fewer studies, however, have used diffusion imaging to assess white matter correlates of RRB [64, 66, 72‐74]. Adding complexity to this line of inquiry, there is mixed evidence regarding sex differences in expression of RRB in ASD [32‐36]. Relatively few studies have examined sex differences in ASD using diffusion imaging, and none of these studies investigated the relationship of such findings with RRB [28, 29, 75, 76]. In this study we assessed measures of RRB, regional gray matter volumes and white matter microstructure in the basal ganglia and cerebellum in a large cohort of ASD and TD individuals, with nearly equal numbers of males and females in each group. We also evaluated whether biological sex interacted with ASD diagnosis on these neuroimaging outcome measures, and their relationship to RRB. To our knowledge, this work is the first to assess microstructure of intra-basal ganglia white matter pathways (i.e., SN to putamen, GPe to STN) in ASD and how such pathways relate to expression of RRB. This study is also among the first to quantify FW or implement free-water correction of DTI data in ASD [20, 21].

We observed higher levels of RRB in individuals with ASD, as expected. Overall patterns of RRB were similar between males and females with ASD, as the ADI-R section C and RBS-R total score showed no significant sex differences. There was, however, a significant diagnosis by sex interaction for the restricted interests subscale of the RBS-R. This specific finding regarding restricted interests is in line with other studies that have used the RBS-R to investigate sex differences in the expression of RRB in ASD [34, 36]. We also observed that within TD participants, males showed higher scores on the restricted interests subscale, but this effect was less pronounced than in ASD participants.

Our neuroimaging findings suggest that the gray matter volumes in the basal ganglia and cerebellum, adjusted for total brain volume, do not distinguish children and adolescents with ASD from their TD counterparts. We found no interaction between diagnosis and sex on volume of structures in the basal ganglia and cerebellum, although there were significant sex effects that applied equally to both ASD and TD groups. Previously reported volumetric differences in ASD and their relation to RRB may have been biased by inclusion of only males or very small numbers of females, as well as smaller sample sizes [56‐59]. In a larger study of 472 individuals with ASD (54 females), Turner et al. reported that individuals with ASD showed significant volumetric enlargement in the pallidum and lateral ventricles [77]. That study, however, did not investigate potential interactions between diagnosis and sex, but their results were identical when replicated in the male-only group. In a similar study of subcortical volume in 539 individuals with ASD (67 females), Zhang et al. found no effect of diagnostic group for any of the structures evaluated, nor any interaction between diagnosis and sex [30]. It is worth noting that both of those large studies included adults, whereas our study sample included only children and adolescents. There was a significant negative correlation between volume of cerebellar lobule VI (L) and RBS-R total score, an observation which matches Rojas et al. [58], where there was no volumetric difference between ASD and TD individuals in cerebellar lobule VI (L), but a significant negative correlation between repetitive behavior and volume of lobule VI (L) was reported.

In contrast, we found that microstructure of basal ganglia white matter tracts was significantly different in children and adolescents with ASD compared to TD participants. We found that individuals with ASD had lower FA_T in the DLPFC to caudate (L) tract, but there was no effect of sex, nor interaction between diagnosis and sex. We also found that individuals with ASD had higher FW in the DLPFC to caudate tracts (L and R), as well as tracts from SN to putamen (L) and GPe to STN (L), with no effect of sex nor interaction between diagnosis and sex. Lower FA_T in these tracts, excluding the GPe to STN (L) tract, were correlated with higher scores in at least one subscale of the RBS-R. Higher FW in DLPFC to caudate tracts (L and R) also correlated with higher RBS-R restricted interest scores. Although RBS-R total score and many RBS-R subscales were correlated with white matter microstructure, it is worth noting that no significant correlations were found between self-injury or compulsive behavior subscales with either FA_T or FW for any of the tracts examined.

The supplementary analysis of uncorrected FA values showed significant group differences in DLPFC to caudate tracts (L and R). Our objective in including this supplementary analysis is to provide an example of the different conclusions that may be arrived at when applying free-water quantification and correction to diffusion MRI data. In some of the evaluated tracts, such as GPe to STN, uncorrected FA values were nearly 40% greater than corresponding free-water corrected FA_T values. The relationship between FA and FA_T is determined on a voxel-by-voxel basis and is dependent on the FW fraction [13]. To illustrate the relationship between FA and FA_T in the context of this manuscript, we also provide correlations between tract-average FA and FA_T in Additional file 1: Table S5. We believe it is especially relevant that the FW metric itself showed greater sensitivity at identifying significant differences between ASD and TD groups, which was observed in both DLPFC to caudate (L and R) tracts, as well as the in basal ganglia tracts SN to putamen (L) and GPe to STN (L).

In order to determine the specificity of these brain-behavior relationships to RRB we also performed correlations with the SCQ, which captures social and communication deficits seen in ASD, but also includes some items pertinent to RRB. There were no significant correlations between SCQ score and regional gray matter volumes in the basal ganglia and cerebellum, nor FA_T in the white matter tracts investigated. We found that higher SCQ scores correlated with higher FW in the DLPFC to caudate (L and R), SN to putamen (L), and GPe to STN (L) tracts. The supplementary analysis of uncorrected FA showed that DLPFC to caudate (L and R) was significantly correlated with nearly all RBS-R items as well as the SCQ, but no correlations were observed in other tracts. Considering RBS-R and SCQ findings together, evaluating uncorrected FA may lead to the conclusion that these white matter tracts reflect global severity of both RRB and social deficits. However, following free-water correction, FA_T was more specific to RRB and revealed novel involvement of basal ganglia pathways in the expression of RRB.

The DLPFC, which includes middle frontal gyrus and part of superior frontal gyrus [78‐80], connects to the basal ganglia via the caudate [81]. Connections from DLPFC to the caudate are thought to play an important role in response inhibition [82]. Aberrant microstructure in the DLPFC to caudate white matter has been shown in other clinical populations with deficits in response inhibition, such as attention-deficit hyperactivity disorder [83] and obsessive–compulsive disorder [84]. Rojas et al. [58] found evidence for altered FA in white matter near the middle frontal gyrus (MFG) in ASD using a voxel-wise approach, but no prior tract-based study has identified aberrant microstructure in white matter connecting the DLPFC to caudate. A previous study in males with ASD found functional connectivity deficits between MFG and caudate, which was further correlated with RRB scores, but found no difference in FA for caudate tracts identified with tractography [64]. Another tractography study in males with ASD found lower FA in putamen tracts, but not in caudate tracts [66].

Quantification of free-water or corresponding correction of DTI scalars has had limited application in ASD research [20, 21]. It is possible that prior evaluations of basal ganglia white matter in ASD using FA as an outcome measure were impacted by partial volume effects of extracellular free-water. This consideration is especially relevant for white matter tracts proximal to ventricles, such as the DLPFC to caudate tract. When we investigated FA_T of the DLPFC to caudate tract at slice-level, we found that the slices which had significantly lower FA_T in ASD were those most proximal to the ventricles (Fig. 3). The majority of these same DLPFC to caudate slices also had significantly higher FW in ASD (Fig. 4). We did not observe a main effect of sex, nor an interaction between diagnosis and sex, for either FA_T or FW in the DLPFC to caudate tract. FA_T in the DLPFC to caudate tract was negatively correlated with RBS-R total score and multiple subscales, whereas FW was positively correlated with the restricted interests RBS-R subscale. The supplementary analysis of uncorrected FA also showed significant group differences in ASD for the DLPFC to caudate tracts, and significant correlations for RBS-R and SCQ scores. Together these findings suggest altered microstructure of the DLPFC to caudate tract in ASD, and that higher rates of RRB are associated with greater disruption in this pathway.

The added value of FW imaging can also be seen in white matter pathways within the basal ganglia. Individuals with ASD had higher FW than TD individuals in the SN to putamen tract. Although we did not observe a group difference in FA_T for this SN to putamen tract, we found that FA_T was negatively correlated with the RBS-R restricted interests subscale. Projections from substantia nigra pars compacta (SNpc) to putamen provide dopaminergic tone to the striatum, and recurrent projections from putamen to SNpc provide regulatory feedback. Also connecting these regions, efferents from the putamen to substantia nigra pars reticulata (SNpr) are part the direct basal ganglia pathway. The SN target mask used in this study did not differentiate between the SNpr and SNpc. Consequently, we did not have the ability to differentiate whether the voxels represented in this tract were comprised of direct pathway GABAergic projections from putamen to SNpr, dopaminergic projections from SNpc to putamen, striosomal GABAergic recurrents from putamen to SNpc, or some combination of these.

Due to the multiple distinct white matter populations likely represented in the SN to putamen tract, there are multiple possible interpretations of the present findings for the SN to putamen tract. One possible interpretation is that there is impaired microstructure of the direct pathway in ASD associated with higher levels of RRB. Several theories surrounding RRB postulate an imbalance between direct and indirect pathways [85‐87]. Prior studies do not provide evidence of direct pathway dysfunction associated with RRB in ASD, whereas the indirect pathway has been demonstrated to be an important mediator of RRB in work from animal models [87‐93]. Another possible interpretation is that the present findings reflect impaired nigrostriatal dopamine tone, due to projections from SNpc to putamen or striosomal recurrent projections from the putamen to SNpc. We find this interpretation more plausible, as nigrostriatal dopamine abnormalities have been hypothesized to underlie RRB in ASD [94]. Moreover, dopamine dysregulation has been linked to RRB in individuals with Parkinson’s disease who receive dopamine replacement therapy, and is thought to relate to impaired activity of the indirect pathway [95, 96].

The GPe to STN tract (L) also had higher FW in individuals with ASD compared to TD participants. FW in the GPe to STN tracts did not correlate with RBS-R scores, but was found to have a positive correlation with SCQ scores. Although we did not observe a significant effect of diagnosis on FA_T in GPe to STN tracts, there was a significant negative correlation between FA_T in GPe to STN (R) and the RBS-R subscale for sameness behavior, which was further supported by trend level correlations with RBS-R total score and subscale for ritualistic behavior (see Table 6). Connections between the GPe and STN are specific to the indirect basal ganglia pathway, a pathway shown in animal studies to play a critical role in RRB [87‐93]. Our findings in this tract support a role for the indirect pathway in ASD, as well as on the severity of RRB and social deficits as indexed by the RBS-R and SCQ, respectively.

A significant main effect of sex on FA_T was observed in the putamen to SN tracts (L and R), as well as GPe to STN tracts (L and R). In all of these tracts, males had higher FA_T than females. The greater FA_T observed in males for these basal ganglia pathways is consistent with a prior DTI investigation of sex differences in TD individuals during childhood and adolescence [97]. We did not observe an effect of sex on FW in any of the evaluated pathways. Importantly, there was no interaction between diagnosis and sex on either FA_T or FW for any of the investigated pathways, nor in the clinical scales evaluated in this study. Thus, the main effect of sex was not further explored in this study, as it applied to both ASD and TD populations equally. Nevertheless, other studies suggest that there may be sex differences in brain morphology or connectivity in ASD and it is important that future work in this area continue to evaluate sex as a factor [28, 98, 99].

The lack of significant diffusion imaging findings in cerebellar white matter tracts may be partially related to demographic differences in the study population compared to other similar work. Cheung et al. [73] previously reported a significant negative correlation between RRB and FA in cerebellar white matter. Their investigation included a smaller sample of 13 individuals with ASD, only 1 of which was female, and utilized the ADI-R section C, which yielded no significant correlations in the present study. Regarding age, work has shown that children with ASD show dynamic changes in cerebellar white matter tracts across childhood that differ from their TD counterparts [100]. Moreover, altered FA in cerebellar pathways in infancy and toddlerhood has been associated with the later development of repetitive behaviors and autism. Wolff et al. [101] observed associations between RRB and FA alterations in cerebellar but not basal ganglia white matter tracts, leading them to speculate that these brain regions may play different roles in the development of repetitive behavior over time. The basal ganglia and cerebellum are tightly interconnected, both functionally and structurally, and correspondingly their influence on the development of motor and other behaviors are thought to be intertwined [102]. Thus, the older children in this study may represent a later developmental stage in which repetitive behaviors are more dependent on basal ganglia rather than cerebellar white matter.

Altered FA (or FA_T) is often considered an indicator of structural connectivity, reflecting factors such as aberrant myelination, but may also correspond to differences in other micro-structural properties (e.g., microtubule density). FW is thought to capture partial volume effects from extracellular space, as the diffusion-weighting time in common imaging sequences is such that FW signal most likely originates from spaces larger than a few tens of microns [103]. In the context of the results from this study, increased FW in ASD may reflect microstructural properties such as reduced density of glial cells in these white matter tracts [104] or neuroinflammation [19]. In support of findings from the present study, research using neurite orientation and dispersion imaging (NODDI) identified higher isotropic volume fraction in white matter of individuals with ASD, which was similarly interpreted as increased extracellular free-water [105]. Brain development in ASD is thought to be affected by “over-pruning” and increased FW could also reflect this process [106]. In conditions such as Parkinson’s disease, FW is also thought to reflect atrophic neurodegeneration [13‐16], but we do not suspect neurodegeneration to be driving the present findings from children and adolescents with ASD.

This study was performed on a deidentified dataset from the NDA, and thus we did not control data collection methods. In our data analysis we covaried for scanner type to account for small differences in imaging acquisition parameters, but information regarding the testing site from which each participant’s data originated was unavailable. Thus, this study cannot account for potential variability due to testing site. The white matter pathways evaluated in this study were chosen due to their hypothesized relevance to RRB. Differences between ASD and TD individuals that may occur outside basal ganglia and cerebellar pathways, and their potential relationship to RRB, were not examined. The diffusion MRI data analyzed in this retrospective study were acquired with a single diffusion-weighted shell (b = 1000 s/mm²). It has been demonstrated that when applying a two-compartment model to single-shell diffusion MRI data, changes in FW may be difficult to disentangle from tissue mean-diffusivity changes, and that multi-shell diffusion data offers improved FW estimates compared to single-shell data [107]. However, when only single-shell data are available, as was the case for this dataset, recent work has shown that applying a two-compartment model (i.e., free-water) provides increased signal-to-noise ratio and greater sensitivity than fitting a single compartment model to single shell data [108]. Although FA_T and FW capture aspects of tissue microstructure, the analyses performed here cannot determine the specific microstructural underpinnings of those differences; post-mortem experiments in clinical populations or work in animal models may elucidate how these metrics correspond to microstructural alterations in ASD. Lastly, although mean IQ score in the ASD group was significantly lower than mean IQ of the TD group, individuals in the ASD group had IQ in the normal range. We did not covary for IQ in this study, because it has been suggested that IQ is inappropriate to include as a covariate in studies of neurodevelopmental disorders, as differences in IQ are often part of the phenomena of the condition being studied [109]. It is unknown whether the findings from this study extend to individuals with ASD that have lower IQ scores.

In this study we used neuroimaging data available from the NDA to assess gray matter volume and white matter microstructure of the basal ganglia and cerebellum in ASD. These novel findings demonstrate that cortico-basal ganglia white matter microstructure is altered in ASD and linked to RRB. FW in cortico-basal ganglia and intra-basal ganglia white matter was more sensitive to group differences in ASD, whereas cortico-basal ganglia FA_T was more closely linked to RRB. In contrast, we found no significant difference for basal ganglia or cerebellar gray matter volumes in ASD. Sex-related differences in brain volume and microstructure were present in both ASD and TD groups and did not interact with diagnosis. Prior diffusion imaging investigations of white matter in ASD may have been impacted by partial volume effects of extracellular free-water, especially in basal ganglia pathways proximal to the ventricles. Future diffusion imaging investigations in ASD may benefit from quantification of FW and corresponding correction of FA to account for partial volume effects in order to better understand how basal ganglia white matter is affected in ASD, and how such measures are associated with expression or attenuation of RRB.