The effect of age on vertex-based measures of the grey-white matter tissue contrast in autism spectrum disorder

Autism spectrum disorder (ASD) is a complex neurodevelopmental condition characterized by deficits in social communication, social reciprocity, and repetitive/stereotypic behaviour [1]. There is strong evidence to suggest that these core symptoms are accompanied by differences in grey matter (GM) neuroanatomy and white matter (WM) connectivity [2], which typically manifest during early infancy [3, 4]. Despite the large number of existing neuroimaging studies, however, the neurobiological mechanisms that drive the atypical development of the brain in ASD remain poorly understood.

To date, most neuroimaging studies examining atypical brain development in ASD have focused on measures of brain volume [5‐7] and its two constituent components cortical thickness [8] and surface area [9, 10]. More recently, however, the attention of structural neuroimaging studies is shifting towards examining the grey-white matter boundary, as histological evidence suggests that the grey-white matter tissue contrast may be regionally less well defined (i.e. less distinct) in ASD [11]. Such ‘blurring’ of the grey-white matter transition zone seems to be caused by the presence of supernumerary neurons beneath the cortical plate, which—in turn—may result from migration deficits or failed apoptosis in the subplate region [12]. This finding also agrees with genetic investigations linking the aetiology of ASD to atypical neuronal proliferation, migration, and maturation [13, 14]. For stratification purposes, and to capture aspects of ASD neuropathology that may be more closely linked to aetiological factors, it is therefore important to also investigate neuroimaging measures that map onto these particular characteristics of the cortical microstructure in vivo.

With this aim in mind, we recently examined the contrast between grey and white matter (GWC) across different cortical layers in a sample of males and females with ASD and typically developing (TD) controls [15]. We found that the GWC was significantly reduced in ASD, particularly at the grey-white matter boundary, and in many brain regions that have previously been linked to autistic symptoms and traits [16]. Our in vivo finding of a reduced GWC is also consistent with prior postmortem reports of a less well-defined grey-white matter boundary in ASD [11, 12]. However, based on tissue contrast alone, it is not possible to disentangle whether the observed between-group effects are driven by (1) differences in grey matter cytoarchitecture, as suggested by the above histological studies, or by (2) local variations in myelin content. For instance, a recent neuroimaging study of typical ageing, examining a sample of healthy adults (with an age range of 20–84 years), suggests that the GWC typically declines with increasing age and most likely reflects local (i.e. region-dependent) age-related changes of myelin integrity in the superficial WM [17]. Thus, by studying the GWC in ASD across different developmental stages, it may be possible to gain in vivo insights into neurobiological processes that (1) should be completed around birth (e.g. migration deficits), (2) end during early childhood (e.g. apoptosis), and (3) that are ongoing (e.g. myelination). Here, we examined age-related changes in GWC in ASD individuals compared to TD controls during childhood and adolescence. In addition to between-group differences in GWC, the present study investigated age-by-group interactions in a cross-sectional sample of male individuals with ASD and matched TD controls using a spatially unbiased ‘vertex-wise’ approach (i.e. not restricted to regions of interest). We expected the differences in the contrast to be age-dependent (i.e. there are significant age × group interactions), which would suggest that differences observed during postnatal brain development are not exclusively driven by atypical grey matter cytoarchitecture.

Furthermore, it has previously been shown that the trajectory of brain maturation for different morphological features is complex and cannot adequately be captured by linear effects alone. For example, the trajectory of total brain volume seems to be U-shaped with an increase in volume during early childhood, a peak during adolescence, and a subsequent decline in volume [18]. There are also studies to suggest that there is considerable regional variation in the complexity of the normal developmental trajectory of cortical thickness, for example, which includes cubic, quadratic, and linear effects [19]. When examining age effects, it is therefore important to establish linear as well as non-linear effects, in order to adequately model the neurodevelopmental trajectory. While the typical neurodevelopmental trajectories are well established for measures of brain volume or cortical thickness, there is currently no comparable data for vertex-based measures of GWC. In the present study, we therefore examined linear, quadratic, and cubic effects of age in order to model the complex trajectory of the GWC in children and young adults between 7 and 25 years of age.

In the present study, we examined between-group differences in cross-sectional age-related trajectories of GWC in ASD and neurotypical controls across childhood and early adulthood (from 7 to 25 years) using a spatially unbiased vertex-wise approach. We first established that the developmental trajectory of GWC is complex in many areas of the brain and included linear as well as non-linear (i.e. quadratic) effects of age. Moreover, we found that while ASD individuals had significantly reduced GWC overall, these differences were age-dependent, with the most prominent decreases in GWC occurring during childhood. This is of importance as our findings suggest that differences in GWC in ASD are unlikely to reflect atypical grey matter cytoarchitecture alone, which is typically set around birth, but may also represent age-related variability in the white matter architecture (i.e. differences in myelination, axonal density, and diameter, etc.). Measures of GWC might thus be considered an age-sensitive in vivo marker for atypical neurodevelopment in ASD. Our finding of significantly reduced GWC in children and adolescents with ASD extends our previous neuroimaging study examining the GWC in adults with the condition, which concluded in suggesting that the tissue contrast between cortical grey and white matter may be less well defined in ASD [15].

Moreover, our study agrees with previous post-mortem reports suggesting that the boundary between cortical layer VI and the underlying white matter may be more ‘indistinct’ in ASD. This indistinct boundary may be due to increased ‘dispersion’ of neuronal cells across the grey-white matter interface [11, 12]. In turn, supernumerary neurons beneath the cortical plate may then arise as a consequence of disrupted migratory processes during prenatal brain development and/or atypical development and resolution of the cortical subplate (e.g. overproduction of subplate neurons or reduced apoptosis) [32]. Additionally, the cortical suplate plays a crucial role in the formation of the early intra- and extra-cortical neurocircuitry and contributes to the guidance and targeting of thalamocortical axons [33].

Significant reductions in GWC were found in several regions across the cortex, most of which have previously been associated with symptoms characteristic for ASD. More specifically, the medial and dorsolateral prefrontal cortices (mPFC and DLPFC) are integral parts of the so-called social and emotional brain, which encompasses a set of brain regions involved in wider aspects of social cognition and emotional processing [34, 35]. ASD-related neuroanatomical variation in these regions has also been linked to deficits in theory of mind [36], face processing [37], and various other aspects of impaired social cognition, for example, self-referential cognition and empathy [38]. In addition, some of our identified clusters (mPFC and precuneus) are also an integral component of the so-called default mode network (DMN), which characterizes a wider network of brain regions showing decreased activity during cognitive tasks and increased activity when the brain is ‘at rest’ [39]. In ASD, the DMN has been reported to be among the most disrupted functional networks, and this disrupted intrinsic DMN organization (e.g. in terms of functional connectivity patterns) seems to be associated with social deficits in children and adults with ASD [40]. Further significant reductions in GWC were found in occipital regions, which on a functional level have been associated with communication deficits and social reciprocity [2].

In many brain regions where we found a significant main effect of group in GWC, we also observed significant linear and quadratic age-by-group interactions. This implies that the between-group differences in these regions are age-dependent and caused by an atypical developmental trajectory of GWC in the ASD individuals. More specifically, while the GWC in TD controls declined consistently from 7 to 25 years of age, the cross-sectional age-related trajectory in ASD was significantly decreased relative to the normative trajectory during early childhood, followed by a period of no, or small, differences between the ages of 15 and 23 years. This early age-related reduction in tissue contrast was most prominent in temporal and prefrontal regions, which are also the latest ones to mature during typical development [41]. Given previous evidence to suggest that the GWC declines significantly as part of the typical ageing process [17, 42], and based on our previous results of a reduced GWC in adults with ASD [15], it is likely that the GWC also declines more rapidly across the remaining life-span (i.e. after the age of 23 years). However, future research is needed to test this hypothesis directly, using samples with a wider age range (i.e. 25 years plus). Taken together, our study suggests that the developmental trajectory of GWC in the ASD brain not only differs quantitatively from the trajectory in TD controls, but also qualitatively (i.e. in terms of its shape), and particularly during childhood. In turn, this implies that the GWC in ASD may be mediated via different neurobiological mechanisms as compared to TD controls.

Notably, our results show a lateralization towards the right hemisphere, i.e. group differences in GWC were mostly located in the right hemisphere while the left hemisphere seems to be relatively unimpaired. Previous studies have yielded highly heterogeneous findings concerning the lateralization of structural and functional abnormalities in the brain in ASD (e.g. [43]). On the functional level, studies demonstrate that the right hemisphere in particular seems to play a crucial role in mediating several autistic core symptoms, such as communication [44] and theory of mind deficits [45]. Furthermore, in a study by Dapretto et al. [46], ASD individuals showed no activation of the right hemisphere mirror neuron system (MNS) during an emotion recognition and imitation task, i.e. the right pars opercularis showed significantly greater activation in typically developing children than in children with ASD. Activity in the right pars opercularis was also negatively correlated with symptom severity measured by ADOS and ADI-R. These previous reports of a right hemispheric involvement of the brain in mediating ASD symptomatology are thus in line with our findings of more significant reductions in GWC predominantly in the right hemisphere.

Little is, however, currently known about the neurobiological mechanisms that underpin variability in GWC. In general, the T1-weighted signal, which constitutes the basis of the GWC, is heavily influenced by the structure and density of axonal myelin [47, 48], as well as non-architectural components such as iron deposition and water content [49, 50]. Out of these potential candidates, studies examining cortical ageing in the TD brain show that the age-related decline in GWC is foremost related to reduced signal intensities in the superficial white matter [17] and reduced intracortical myelin content as measured by the ratio between T1w/T2w image contrast [51]. Thus, the most prominent biological candidate influencing the GWC may be the degree of myelin in the superficial WM under the cortical mantle, which mostly contains short association and U-shaped fibers [17, 42, 52]. Deficits of short association fibers have been reported previously [53] and may hence contribute to the atypical GWC observed in our study.

In addition, we examined whether the differences in GWC were driven by differences in absolute tissue intensities within the grey or white matter. In many regions with significantly reduced GWC, ASD individuals also showed significantly decreased WMI sampled at 1 mm below the grey-white matter boundary. However, there were no differences in GMI (sampled at 30% CT and at the grey-white matter boundary) compared to TD controls. Analogue to the trajectories of GWC, differences in WMI seem to be most prominent during childhood and become less pronounced during adolescence and early adulthood. This finding is in agreement with previous voxel-based-morphometry (VBM) studies in ASD that compare white matter intensity using a whole-volume approach [5, 6, 54]. Evidence for general white matter abnormalities in ASD is further supported by DTI studies applying techniques such as tract-based spatial statistics (TBSS) [55]. For example, a study by Shukla et al. [56] examined atypicalities in the trajectories of white matter development in a sample of 9- to 20-year-old ASD individuals and TD controls. Here, ASD individuals showed less orientational coherence (i.e. fractional anisotropy) and stronger water diffusion (i.e. mean diffusivity) in many of the most prominent white matter fiber tracts in the brain compared to controls [56]. These maturational differences, however, diminished from childhood to adolescence and are thus in agreement with our finding of a delayed white matter maturation (based on tissue intensities) during early childhood. Our results are thus consistent with previous publications examining cortical white matter employing different methodological frameworks and spatial scales.

Last, our findings should be interpreted in the light of a number of limitations given the data and methods presented. First, we employed a cross-sectional study design to examine age-related differences in GWC associated with ASD. Thus, the resulting age-related trajectories were based on inter- rather than intra-individual variability in GWC. Future studies are required to replicate our findings in longitudinal samples, which would provide a more accurate characterization of developmental trajectories based on repeated measures acquired in the same set of individuals. Second, our sample only included right-handed males with ASD in the high-functioning range of the spectrum. It therefore remains to be established whether our findings generalize to other (sub)groups on the autism spectrum (e.g. left-handed individuals, females with ASD, or individuals with intellectual disability). Furthermore, our study examined age-related changes in GWC by sampling tissue intensities around the grey-white matter boundary, even though this boundary may be ‘blurred’ (i.e. less well distinct) in ASD as suggested by histological evidence [12]. In this histological study, however, such microstructural blurring occured up to 500 micrometers underneath the grey-white matter transition zone [12], which would result in a maximal displacement of the boundary of 0.5 mm. As we are sampling GMI at 30% CT, which roughly equals between 0.56 and 1.5 mm into the cortical mantle depending on CT variability across the cortex (see Additional file 2), and WMI at 1 mm into the white matter, we can be certain that our sampling points remain located within the grey or white matter even in case of a maximal boundary displacement. However, while our model can accommodate such potential displacements in terms of intensity sampling, we are unable to unambiguously allocate sampling points to specific cortical layers given the restraints with regard to the current resolution of structural MRI images (see also [17]). Similarly, while surface-based mapping allows for morphometric inferences on a sub-millimeter scale, the derived grey-white matter tissue intensity values, and hence the GWC, remain dependent on the native spatial resolution of the T1-weighted images (i.e. 1 mm isotropic). Thus, partial volume effects and/or the ability to clearly delineate the grey-white matter boundary may affect GWC values. However, both of these factors are expected to affect both groups equally, and our findings of significant between-group differences and age-by-group interactions cannot be fully explained by these limitations.

Using 3-Tesla structural MRI, we provide evidence that the tissue contrast at the grey-white matter boundary is regionally reduced in the brain of individuals with autism spectrum disorder. Compared to healthy controls, this neuroanatomical difference seems to be most prominent during childhood and early adolescence. Taken together, while future research is required to identify the specific neurobiological mechanisms underpinning the atypical development of tissue contrast in ASD, the findings presented in the present study suggest that the differences in GWC previously reported in adults with the condition [15] do not reflect differences in the grey matter cytoarchitecture alone. Instead, our study suggests that the differences in GWC are dynamic (i.e. ongoing) across the human life-span and are thus most likely caused by a combination of factors that include many of the grey matter atypicalities highlighted by the histological studies above [11, 12], but also by perturbations to the formation of the axonal neurocircuitry and subsequent myelination within the superficial white matter.