White matter connectivity in children with autism spectrum disorders: a tract-based spatial statistics study

Autistic Disorder, Pervasive Developmental Disorder-Not Otherwise Specified and Asperger Syndrome (referred to hereafter as Autistic Spectrum Disorders, -ASD-) are a group of neurodevelopmental pathologies characterized by a triad of early symptoms, namely social impairments, communication impairments and rigid and repetitive behaviours. There is increasing evidence that the heterogeneity and complexity of core and associated ASD symptoms can be explained as a fundamental impairment in brain connectivity [1–3]. In this perspective, the advent of Diffusion Tensor Imaging (DTI), a recent noninvasive magnetic-resonance-based method specifically sensitive to the presence of subtle white matter (WM) abnormalities, has allowed for a better understanding of ASD neuroanatomical underpinnings. DTI is based on measurement of water diffusion through cellular compartments in vivo [4]. In particular, compared to more isotropic movement of water in grey matter, WM is characterized by anisotropic water diffusion, referable to the anatomical configuration of structured axonal membranes and myelin sheaths [5]. Several measures can be extracted from DTI analyses, among which are fractional anisotropy (FA), and mean diffusivity (MD). FA is considered a valid measure of WM structural integrity, sensitive to anomalies in axonal density, diameter, myelination, interaxonal spacing and coherence of directional alignment of fibers within WM tracts [6–8], while MD represents a directionally independent measure of the average diffusivity that reflects the degree of myelination, interstitial space and axonal density [9]. A schematic distinction of DTI methods of analysis includes: 1. Region-of-interest (ROI) approach and 2. Whole-brain analysis. Whole-brain analysis investigates the entire brain without a priori ROI selection and thus could be of particular relevance to ASD, in which WM tracts implicated in the etiopathogenesis of the disorder are diffuse and not completely clarified. Among whole-brain approaches, a further innovative method of analysis is tract-based spatial statistics (TBSS) [10]. This analysis method provides some advantages with respect to traditional voxel-based analysis (VBA), since it does not require smoothing and it allows for a higher spatial comparability.

In the last decade, several papers based on DTI have been published, difficult to compare with each other, depending on substantial differences not only as to the clinical and demographic characteristics of the samples, but also as to the methods of analysis. Notably, an age-related pattern of structural changes in the ASD brain is a well-known finding [11]. Specifically, the neuroanatomical differences between ASD patients and typical controls appear to be more prominent in younger subjects and to attenuate with increasing age. This issue can partly explain the conflicting results of ASD imaging data. In particular, almost all the DTI studies on ASD adolescents and adults agree on a FA reduction with respect to typical controls.

To date, only few voxel-based DTI reports have been performed to investigate whole brain WM abnormalities in young children with ASD. Ben Bashat and co-workers [12] first examined a young age group (range 1.8-3.3 years) of ASD subjects, revealing FA increase in several brain areas, with particular involvement of the left and anterior regions. Subsequent reports in ASD early childhood age range highlight a picture of concomitant FA decrease and increase, with various anatomic locations [13, 14] or of reduced FA only [15]. More recently, a limited number of studies involving young children with ASD have specifically investigated the structural connectivity through the TBSS approach [16, 17]. The two above mentioned studies showed contradictory results, explained, at least in part, by differences in IQ as well as in other clinical features of patients. Specifically, they are characterized by reduced FA in right uncinate fasciculus, left arcuate fasciculus, right cingulum and corpus callosum [16] and increased FA in genu and midbody of the corpus callosum, in left superior longitudinal fasciculus and in bilateral cingulum [17].

As far as the correlation between clinical data and DTI indices, few papers have examined the relationship between WM microstructure integrity and language abilities of ASD patients. Impairments in language and communication skills represent a core diagnostic feature of ASD. In fact, ASD expressive language abilities are highly heterogeneous and may range from nearly absent (nonverbal patients) to a fluent speech production. However, deficits in one or more linguistic domains (phonology, syntax, semantics and pragmatics) are always present in these subjects [18]. A previous study found an inverse correlation between FA values and communication impairment ADI-R scores in fronto-striatal areas and posterior corpus callosum [14], whereas others reported no association between measures of structural connectivity and language functioning [13, 19].

In this study, the WM integrity of ASD children is investigated by using DTI with TBSS analysis method and tractography. In particular, the purpose of this research is three-fold: first, to examine possible brain region differences in the WM microstructure integrity between ASD children and typically developing controls (TD); second, to analyse tract-specific differences between ASD group and TD; third, to evaluate whether a relationship between productive language skills and indexes of WM abnormalities in language-related pathways is present in ASD subjects.

In this study, WM integrity in ASD children was investigated following two approaches: the TBSS analysis and the tractography method.

The results provide three key findings: 1) widespread FA increase in WM TBSS of ASD patients, especially evident in the corpus callosum, the cingulum, the arcuate fasciculus (AF) and the internal capsula; 2) significant differences in the corpus callosum, the cingulum and minor differences in the AF, as revealed by the tract analysis; 3) significant correlation between expressive language abilities and tract specific measures of AF, corpus callosum and cingulum.

The first result agrees with previous whole brain DTI works by Ben Bashat and colleagues [12, 17], but extends the stage at which FA is increased in ASD patients into an older age. The regression analysis of FA vs. age shows that the whole brain FA increase of our sample was clearly attributable to the subgroup of younger patients, while the older one reported a reduction in FA compared with the TD group. The trends are regionally dependent, since the developmental trend of FA in the corpus callosum crosses atapproximately 7–8 years, while the FA of the cingulum is constantly increased with respect to the control group. It is possible that the corpus callosum, the largest commissural WM tract in the human brain, could influence the global FA and hide the FA trend of other small tracts (e.g. the cingulum).

Most DTI papers on adolescents or adults with ASD show reduced FA in several brain regions, including the corpus callosum [38, 39], areas located between the right middle and inferior frontal gyri and in the right centrum semiovale, as well as motor and premotor areas [40] and the superior temporal gyrus [41]. More recently, Shukla and colleagues [42], using the TBSS analysis, found a reduced FA in numerous WM tracts. Moreover, in a study [43] performed on adolescents with ASD, an imbalance of FA was found, as some areas showed an FA increase, while others a FA decrease.

These contradictory results could reflect a change of the trend of FA in young children with respects to adolescents and adults and may be linked to the modification in ASD brain volume, which has an accelerated growth in the first years of life and then shifts to an abnormally slow growth [44–46]. However, the age at which the FA increase is no longer evident varies across studies: Wolff et al. [47] demonstrated in a longitudinal investigation that, by 24 months of age, a decreased FA is evident across all fiber tracts. Conversely, a pattern of FA increase is still present in some WM tracts of children with a mean age of 8.7 years [13], 9.3 years [14], and 13.7 years [43]. In fact, the weak concordance among DTI investigations is largely ascribable to the wide heterogeneity both of ASD clinical phenotype and of methods of analysis, which makes results difficult to compare with each other. Furthermore, experimental artifacts and confounding factors should be taken into account in order to enhance the reliability and comparability of DTI results [48].

Caution is required in interpreting high FA levels, also because this diffusivity change could represent an index of various WM microstructural alterations, like increased myelination, axon size and density, path geometry, and the presence of crossing fiber pathway. To date, the recently developed TBSS analysis has been applied to a limited number of ASD investigations on young autistic children [16, 17]. Among the advantages, the TBSS approach offers the possibility to choose as a template the most representative FA map among all the maps of the subjects involved in the study, instead of an external template such as the Talairach or MNI standard spaces. This benefit is particularly important when the study involves young children, since most of the above mentioned standard templates are based on adults.

Second, the reconstruction and the tractographic analysis of the corpus callosum, the cingulum and the AF revealed significant differences in these WM tracts. The corpus callosum is the largest inter-hemispheric tract in the brain and it is thought to be involved in emotional and social functioning, as well as in higher cognitive processes, such as decoding non-literal meaning, affective prosody, and understanding humour [49–51]. Several studies performed on adolescents and adults with ASD reported reduced FA in the corpus callosum [38–40, 52, 53]. Conversely, an increased FA in the corpus callosum was found when young children with ASD have been investigated [12, 17]. In accordance with these latter findings, we confirmed the FA increase in the corpus callosum of young children with ASD. Moreover, a significant increase in the mean length of streamlines in ASD with respect to controls was found.

The cingulate cortex is involved in tasks related to higher-level cognitive processes, like empathic cognition, social behaviour and pain perception in which is more likely to be hypoactivated in ASD subjects [54, 55]. Structural abnormalities in the cingulate area have been also demonstrated in DTI studies. Specifically, a FA reduction in the anterior cingulate cortex was found in children as well as in adolescents with ASD [40, 52], whilst an increase in the number of streamlines bilaterally within the cingulum was present in early-adults with ASD [56]. More recently, an opposite result was obtained, with a FA increase in the left cingulum of ASD young children [17]. Our results confirmed and extended bilaterally the FA increase within the cingulum. In addition, a significant increase in the length of streamlines of the cingulum in ASD with respect to controls was detected.

The increase in the length of streamlines of both the cingulum and the corpus callosum in the ASD group contrasts with the excessive short-range connectivity hypothesis suggested by several studies [57–59]. On the other hand, it is consistent with two reports of the same research team [16, 60], in which the long-range association fibers were significantly longer in the frontal lobe [60], and particularly in the uncinate fasciculus, the AF and the corpus callosum of the ASD sample [16]. While the role of the increased fiber length could be reconductable to a compensatory function with respect to the impaired integration abilities of the ASD brain, the pathogenic mechanism may be ascribed to altered serotonin levels [60]. In fact, several previous studies have indicated that serotonin acts as a neurotrophic factor in axonal outgrowth during development [61, 62]. In addition, recent findings point to high serotonin levels in the crucial stages of fetal brain development as a risk-factor for ASD [63].

Conversely, we did not find significant differences in the mean length of streamlines within the AF, although the values are higher in the ASD group with respect to control (Table 3), especially in the left hemisphere.

The AF is a fiber bundle related to language, a function always impaired in ASD subjects, at least from the qualitative point of view. To the best of our knowledge, this is the only study in autism in which the AF is dissected using a three-ROIs approach that allows reconstructing the three tracts of the AF. A previous investigation on ASD adults reported decreased value of FA in the right AF [39], studies on children with ASD described an increased FA and MD bilaterally in the AF [12, 17], while another report [16] found an increased MD in the right AF, but a decreased FA in the left AF in ASD group. Our result of an increased MD in the AF is in agreement with the aforementioned study. We found also an increased FA in the TBSS investigation, but not in the tractography analysis, probably referable to the differences between the two techniques. In fact, while the TBSS makes a comparison voxel to voxel on the complete WM skeleton, the tractography is based on the evaluation of the mean value of the invariant indices across each tract: in this way differences in specific areas of the tract might be hidden. Moreover, the tractographic approach can be vexed by the degree of the reproducibility that varies according to WM tract size: while it is possible to obtain a high reproducibility with thick and big tracts (e.g. the corpus callosum), it is harder with thin tracts like the ones that form the AF. The three-ROIs approach for the dissection of the AF allows identifying which specific sub-tracts are compromised in ASD. In particular, we found an increased MD within the indirect segments of the right AF. The MD is a measure directly related to interstitial space between axonal fibers; therefore, the increase in his value could be due to reduced neural or glial cell packing or cell size, as well as decreased water exchange rate between the intra- and extracellular compartments.

The result of increased MD in short associative tracts in ASD subjects with respect to controls is consistent with previously reported DTI studies [38–41, 60]. In particular, an analysis of microstructural organization of the AF in ASD patients [19] highlights a significant increase of MD in the ASD left hemisphere.

Conversely, the present study as well as Kumar’s [16] work, localize significant differences in the right hemisphere. However, a comparative discussion between our and previous results [19] is not possible, as differences in the subject selection and tractography methodology are present. First of all, we selected young children and we also included in the sample low functioning and non-verbal individuals with ASD, while the above-mentioned investigators use a high-functioning verbal sample of adolescents. Moreover, these authors only examined the direct segment of the AF, whereas we also considered the indirect segments, in which structural abnormalities (number of streamlines and MD) are particularly prominent.

As far as the correlation between expressive language and tractographic indices, the following findings were found: a) a positive correlation with length of streamlines in the indirect anterior segment of the AF bilaterally and b) a negative correlation with MD and parallel diffusivity of the left indirect anterior segment of the AF. A similar result emerges from a recent study [64], which reveals a significant negative correlation between the MD of AF -considered as a unique tract- and clinical assessment of language function. Furthermore, we identified which are the sub-tracts of the AF related to language impairment.

According to a study of Catani et al. [65], the indirect pathway of the AF appears to relate to semantically-based language functions (e.g. auditory comprehension and vocalization of semantic content), while the direct pathway relates to phonologically-based language functions (e.g. automatic repetition). Considering this assumption, the structural abnormalities in the language pathway in ASD patients could be related to impairment in semantic-based functions. However, this hypothesis should be confirmed with a better characterization of the linguistic profile in the sample subjects. In fact, in the present study, the language score assigned to the subjects arises from an evaluation of the global expressive language ability, without differentiating among the specific domains (i.e. phonology, syntax, semantics and pragmatics). Future research will be addressed to the identification of specific brain-behaviour abnormalities in language-related fiber tracts.

There are several limitations to our study that should be noted. First, the sample size was relatively small, which might reduce the power of the statistical significance and generalization of the findings. Second, the group of ASD children we have studied reflects the epidemiological characteristics of autistic population both as for the gender distribution (with a skewed for males), that for the level of intellectual functioning (with a frequent mental retardation comorbidity): thus patients and controls were not matched for the sex and the NVIQ. Future investigations, in which separate analyses of data for large subgroups of ASD patients will be conducted (e.g. males and females, high-functioning and low-functioning), are therefore warranted, in order to highlight WM microstructural differences, similarities and distinctive features of each subgroup. Third, subjects included present a wide age range, from 2 to 11 years of age. In order to overcome the latter limitation, (a) we have excluded toddlers under 2 years of age, since changes in DTI indices are age-related [66] and occur, for the most part, within the first 24 months [67]; (b) we have searched for an high age comparability between patients and controls, not only in terms of arithmetic mean, but also of SD and range. Notwithstanding the problematic issues raised above, this study adds to the limited existing literature on young children with ASD. A widespread increased of FA and several tract-specific abnormalities in structures linked to ASD behavioural deficits (e.g. cingulum, corpus callosum and AF) were detected. Moreover, we identified connectivity modifications related to language functioning in the ASD sample. Further studies in which subgroups of ASD subjects will be better characterized according to their language abilities, will allow to identify which structural connectivity abnormalities are related to specific language impairments.