Left Hemisphere Diffusivity of the Arcuate Fasciculus: Influences of Autism Spectrum Disorder and Language Impairment

Discussion

There were 2 main findings. First, main effects of both ASD and also LI were found on MD of the left hemisphere arcuate fasciculus, with both ASD and LI associated with elevated MD. At first, this finding appears to reflect a shared ASD and LI biologic substrate. However, a second finding showed a main effect of ASD on the axial, but not radial, component of diffusivity, and a main effect of LI on the radial, but not axial, component of diffusivity. As both AD and RD elevations could contribute to the observed MD elevations, this trend suggests that though both ASD and LI are associated with microstructural abnormalities in the left hemisphere arcuate fasciculus, the nature of these microstructural abnormalities is somewhat distinct.

Diffusion imaging studies in ASD generally report elevated diffusivity and decreased FA in multiple brain areas, though the nature of arcuate fasciculus abnormalities is still in debate, as reviewed by Travers et al.²¹ Even less is known about the white matter integrity of children with SLI. Verhoeven et al¹ compared DTI parameters of the superior longitudinal fasciculus in children with SLI and children with ASD+LI with healthy control participants. Contrary to present findings and previously reported ASD studies,^8,21 the only significant finding in Verhoeven et al¹ was decreased FA in SLI. Discrepancies between this study and our present study may be related to the approach by Verhoeven et al,¹ namely to evaluate the entire superior longitudinal fasciculi tract rather than its subregions. Nonetheless, decreased FA is commonly associated with elevated MD; thus, apparent differences among studies may reflect differing sensitivities of these commonly used eigenvalue constructs rather than different underlying biologic processes. Furthermore, FA reflects a composite combination of eigenvalues perhaps obscuring opposing individual eigenvalue phenomena.

An additional finding emerges from attempting to understand the nature of the LI and ASD white matter abnormalities, with AD and RD analyzed separately (Figs 3⇑–5). In our current study, AD was associated with ASD and RD with LI. That is, in both SLI and ASD+LI, elevated RD was observed, reinforcing the notion of similar anatomic substrates underlying LI in both populations. A main effect of ASD (and not LI) on AD, as well as the significant ASD by LI interaction in AD, however, suggests an additional contribution to a microstructural anomaly in ASD+LI (above and beyond that in SLI). Thus, in short, although ASD+LI and SLI may share a microstructural anomaly of elevated RD, they differ in AD. As Fig 5 illustrates, this mechanism leads to a “doubly” elevated MD in the ASD+LI cohort.

• Download figure
• Open in new tab
• Download powerpoint

Fig 5.

A schematic of the diffusion changes accompanying ASD and LI. Mean diffusivity was elevated in LI and ASD. Children with ASD had elevated AD (right boxes) and, thus, elevated MD. Patients with LI (bottom boxes) had elevated MD because of increased RD. Patients with ASD and LI had the highest MD values because of combined effects of both elevated RD and AD.

Consistent with the general linear model main effects, linguistic abilities (reflected in the CELF-4 Core Language Index) correlated with left hemisphere arcuate fasciculus RD when assessed across all participants. In addition, an association with AD was also resolved. However, this relationship between language skills and AD (not predicted by the general linear model analyses) was most likely driven by the fact that more than half of the LI participants also have ASD and, thus, elevated AD. Indeed, the relationship between CELF-4 Core Language Index scores and AD did not persist when ASD symptom severity was included in the model. Conversely, the relationship between linguistic skills and RD strengthened when measures of ASD were included in the model. However, none of the associations remained significant when individual subpopulations were considered, though there was a nonsignificant trend (P = .08) in the association between MD and CELF-4 Core Language Index when evaluated over all children with LI (SLI and ASD+LI groups pooled). As an analogy, Social Responsiveness Scale scores correlated positively with AD and were unrelated to RD when evaluated across all participants, but again, associations were not significant for individual subpopulations, perhaps reflecting the need for yet larger sample sizes in each group.

We encountered limitations comparing our current results with the previous literature. A review of DTI studies in ASD found that only approximately one-third of studies reported AD and RD findings.²¹ Fletcher et al²³ reported elevated MD and RD in the left arcuate fasciculus of 10 adolescents with ASD, and no differences in FA or AD. However, the wide range of participants' linguistic skills (composite CELF-3 scores ranged from 50–111 and were significantly lower than those of an age-matched control group) makes it difficult to distinguish the separate effects of ASD and LI on diffusivity. In a similar fashion, the limited data available on SLI offer limited insight into the effects of LI on white matter microstructural integrity. Verhoeven et al¹ did report a positive correlation between the Word Classes Receptive subtest of the CELF-4 and mean FA in the superior longitudinal fasciculus of adolescents with SLI. However, no such correlations were found in the ASD+LI group, and the authors did not report AD or RD.

The biologic interpretation of these microstructural anomalies remains speculative. It is tempting to consider elevated RD as reflecting immaturely bundled axons (with greater interaxonal space for RD). Increases in AD are perhaps less intuitive. However, in white matter with fiber crossings, alterations to the pattern of these crossing fibers or, indeed, the RD of component fibers can both manifest as increased AD.²⁴ In mouse models, decreased AD and increased RD have been associated with acute axon and myelin damage, respectively.^25⇓–27 However, these findings may not directly translate to humans, particularly in cases of chronic white matter microstructural damage. Pathologic increases of AD have also been reported. In a small study of adults with acute optic neuritis, Naismith et al²⁸ reported decreased AD within 1 month of onset, but there was substantially increased AD relative to control participants 1 year later. Indeed, widespread increased AD has been demonstrated in Alzheimer disease,²⁹ and in the right frontal tracts of patients with Parkinson disease.³⁰ Moreover, RD and AD seem to be far less specific in areas of chronic microstructural damage. In a study of diffusion in the spinal cord of patients with multiple sclerosis, Klawiter et al³¹ reported associations between RD and both levels of demyelination and axonal damage, but no relationship between axonal damage and AD. Although DTI is sensitive to alterations to white matter microstructure, the technique alone cannot provide a specific mechanism for the observed changes.

Finally, although our present findings are intriguing, a limitation of our study was that small sample sizes prevented the inclusion of handedness and sex as factors or inclusion/exclusion criteria. Furthermore, although the sample size evaluated in our study is a non-meager 73, it is clear that fine-grained microstructural characterization of both ASD and LI substrates, common and distinct, will require much larger samples, perhaps a need served by emerging multi-institutional data sharing initiatives such as the National Database for Autism Research (http://ndar.nih.gov/).