N-acetyl aspartate in autism spectrum disorders: regional effects and relationship to fMRI activation

Abstract

Rapid progress in our understanding of macrostructural abnormalities in autism spectrum disorders (ASD) has occurred in recent years. However, the relationship between the integrity of neural tissue and neural function has not been previously investigated. Single-voxel proton magnetic resonance spectroscopy and functional magnetic resonance imaging of an executive functioning task was obtained in 13 high functioning adolescents and adults with ASD and 13 age-matched controls. The ASD group showed significant reductions in N-acetyl aspartate (NAA) in all brain regions combined and a specific reduction in left frontal cortex compared to controls. Regression analyses revealed a significant group interaction effect between frontal and cerebellar NAA. In addition, a significant positive semi-partial correlation between left frontal lobe NAA and frontal lobe functional activation was found in the ASD group. These findings suggest that widespread neuronal dysfunction is present in high functioning individuals with ASD. Hypothesized developmental links between frontal and cerebellar vermis neural abnormalities were supported, in that impaired neuronal functioning in the vermis was associated with impaired neuronal functioning in the frontal lobes in the ASD group. Furthermore, this study provided the first direct evidence of the relationship between abnormal functional activation in prefrontal cortex and neuronal dysfunction in ASD.

Figure 1

Voxel locations in the left frontal lobe, left parietal lobe, right cerebellar hemisphere, bilateral cerebellar vermis, and bilateral occipital cortex in the Talairach coordinate system.

Figure 2

Representative MRS spectrum and functional activation map from one ASD and one control participant. Each selected participant’s NAA concentration (metabolic chemical shift = 2.02 ppm) was closest to his respective group’s mean value. The red line in the MRS spectrum represents the LC-model fit for the underlying raw data, the black line is the least squares fit baseline, and the dotted line is the 0 line. Each structural anatomical image was warped into Talairach space. MRS voxel placement (dimension=14×14×14 mm, Talairach coordinates = −43, 13, 29) is designated with a black square. FMRI activation depicts significant clusters of activation in the left frontal lobe region of interest at p < .05, one-tailed, corrected. The color scale represents percent signal change during the letter fluency task compared to the control task. Note that data from the entire cluster of FMRI activation was included in the correlational analyses, not only the activation that overlapped with the MRS voxel.

Figure 3

Between group differences in NAA. Concentration of NAA refers to the marginal mean for each group, evaluated at age 23.8 years. Error bars represent the pooled standard error. The ASD group had a significantly lower concentration of NAA (p = .0431) in the left frontal lobe. Between group differences were not significant in the other regions (p > .05). Overall concentration of NAA, averaged across brain regions was also significantly lower in the ASD group (p < .005).

Figure 4

Interregional relationships by group. The relationship between the concentration of NAA in the left frontal lobe and the cerebellar vermis was significantly dependent on group (p = .013). Nonsignificant linear trends (p < .10) were found in each group. In contrast, no significant interaction, or interregional correlations were found between the vermis and the occipital cortex. This suggests that the relationship between the vermis and the frontal lobe does not reflect a nonspecific, generalized effect.

Figure 5

Data points represent raw data. A significant, semi-partial correlation was found in the ASD group between concentration of NAA in the left frontal lobe and the average percent signal change of voxels with significant activation during the letter fluency task, controlling for age and FSIQ. Only ASD participants with significant functional activation (n = 10) were included in this analysis. This relationship was not observed in the control group.