Several studies over the last decade have implicated atypical white matter (WM) microstructure in the neurobiology of autism spectrum disorder (ASD; Travers et al., 2012; Aoki et al., 2013; Hoppenbrouwers et al., 2014), a behaviorally characterized, multidimensional neurodevelopmental disorder with varying degrees of symptom severity (American Psychiatric Association, 2013). Previous investigations of WM microstructure in ASD using various diffusion imaging techniques typically show reduced fractional anisotropy (FA) and increased mean diffusivity (MD) in many WM tracts, including the corpus callosum (CC), inferior longitudinal fasciculus (ILF), superior longitudinal fasciculus (SLF), and the arcuate and uncinate fasciculi (AF and UF, respectively). There are, however, inconsistencies regarding these findings (see Travers et al., 2012; Aoki et al., 2013; Hoppenbrouwers et al., 2014 for comprehensive reviews and a meta-analysis), and the nature of WM organization in ASD remains unclear.

Methodological inconsistencies and limitations in diffusion imaging are likely to account for some of the variability observed in the previous literature. Many diffusion tensor imaging (DTI) approaches are unable to reconcile fiber orientation or the presence of multiple fibers within the one voxel (referred to as crossing fibers). These factors have a substantial impact upon approximations, and therefore the interpretability of findings. Advances in the field of diffusion-weighted imaging (DWI) research have resulted in the development of novel approaches that can manage the complexity of such data. Constrained spherical deconvolution (CSD), for example, can estimate fiber orientation at a voxel level (Tournier et al., 2007), and has been suggested to provide more accurate estimates of WM tracts compared to other commonly used DTI techniques (Farquharson et al., 2013). Within the ASD literature, CSD has confirmed WM aberration identified using tract-based spatial statistics (TBSS; Roine et al., 2015), and has revealed WM aberration in tracts linking regions of reduced functional connectivity (McGrath et al., 2013).

Fixel based analysis (FBA) builds on the properties of CSD by accounting for multiple fibers within single voxels, referred to as fixels (Raffelt et al., 2015). Given the tract specific nature of FBA, the metrics used to describe WM structure go beyond the traditional voxel-based metrics of DTI analysis. Rather, FBA provides indices of microscopic fiber density (FD; intra-axonal volume, thought to reflect the density of a population of fibers within voxel), macroscopic fiber cross-section (FC; an estimate of fiber bundle diameter derived from the FD metric), and fiber density and cross-section (FDC; a combined measure of FD and FC; Raffelt D. A. et al., 2017). This technique has been applied to investigate WM in preterm infants (Pannek et al., 2018; Pecheva et al., 2019) and to understand WM development and function in healthy individuals (Genc et al., 2018; Bleker et al., 2019; Mizuguchi et al., 2019), as well as to understand WM aberration in various mental health and medical conditions (Gajamange et al., 2018; Grazioplene et al., 2018; Mito et al., 2018; Mu et al., 2018; Lyon et al., 2019; Dimond et al., 2019; Feshki et al., 2019).

To our knowledge, only one study to date has used FBA to investigate WM aberration in ASD. Dimond et al. (2019) applied FBA to compare WM microstructure between 25 (four females) adolescents/young adults (aged 14–20) with ASD and matched healthy controls (HC; six females). The authors report reduced FD within the CC, the inferior fronto-occipital fasciculus bilaterally, as well as the right AF and UF in the ASD sample compared to HC. Global WM FD was also reported to be reduced in ASD. Importantly, no differences in WM microstructure were observed using a voxel-based approach, highlighting the comparative fidelity of FBA. Furthermore, a negative relationship was identified between social impairment [as measured by the Social Responsiveness Scale–2nd Edition (SRS-2)] and FD within the CC among the ASD group, an association not identified in the HC group, demonstrating the potential clinical relevance of understanding WM structure among this population.

Another factor that may contribute to the inconsistencies seen within the ASD DWI literature, and the ASD literature more broadly, is the failure to consider the role of biological sex on the inherent heterogeneity of ASD, an area of growing interest and importance in the field (Kirkovski et al., 2013; Lai et al., 2015, 2017; Mandy and Lai, 2017). Though limited, the available literature investigating the role of biological sex in ASD does provide evidence to suggest that the structural neuroanatomy (Bloss and Courchesne, 2007; Schumann et al., 2009, 2010; Nordahl et al., 2011; Irimia et al., 2017; Zeestraten et al., 2017) and functional neurobiology (Holt et al., 2014; Alaerts et al., 2016; Kirkovski et al., 2016; Ypma et al., 2016) of ASD differ between males and females with ASD, and beyond expected sexual dimorphisms. Findings from diffusion studies are variable. Work from our group (Kirkovski et al., 2015), as well as others (Beacher et al., 2012), do not indicate sex differences in WM microstructure between those with ASD and HC. Conversely, other studies provide evidence for sex-mediated WM aberration within the frontal (Zeestraten et al., 2017) and temporal (Irimia et al., 2017) tracts in ASD, as well as the CC (Lei et al., 2019). All of these studies, however, used older, voxel-based techniques, and therefore issues associated with multiple and crossing fibers remain unresolved.

In the present study, we sought to replicate and expand upon the findings of Dimond et al. (2019) by using an adult sample and investigating sex differences. First, in line with the findings of Dimond et al. (2019), it was hypothesized that FD will be reduced in adults with ASD compared to matched controls. Regardless of the outcome of this primary analysis, we then sought to explore the effects of biological sex on the WM microstructure among this population. A stratification approach was explicitly adopted to avoid the possible masking of effects of biological sex, as has been previously demonstrated (Holt et al., 2014; Kirkovski et al., 2016; Ecker, 2017), and in line with the previous diffusion MRI studies of sex differences in ASD (Kirkovski et al., 2015; Lei et al., 2019). All analyses were conducted using a whole-brain approach (rather than a priori region of interest) to ensure that any relevant sites or tracts could be detected.

This study implemented FBA, a DWI technique capable of accounting for fiber volume and orientation within single voxels and hence addressing important limitations of older DTI approaches, to investigate micro- and macrostructural aberration among adults with ASD. This is the first study to investigate the role of biological sex using this method. Specifically, in line with previous investigations of biological sex and the neuropathophysiology associated with ASD (Holt et al., 2014; Kirkovski et al., 2015, 2016, 2018; Lei et al., 2019), data were stratified by sex and indices of WM micro- and macrostructure were compared within sex. For data such as that which is presented in the present study, this approach has been shown to reveal effects that might be masked using statistical models that do not adopt a stratification approach. While not statistically significant, there was a trend indicative of reduced FD within the CC, at the posterior midbody/isthmus, in the ASD group compared to age, sex, and IQ matched HC participants. This was replicated between males, but not females when the data were stratified by sex; again, however, this result did not reach statistical significance. Finally, comparing the female sub-samples, FDC at the CC (posterior midbody/isthmus) was significantly reduced for females with ASD compared to their age and IQ matched female counterparts. There was also a non-significant trend level difference to suggest reduced FC in the genu of the CC for the female ASD group compared to matched HC participants. The reverse contrasts did not reveal any areas showing increased FD, FC, or FDC in the ASD group when comparing the whole sample, nor when data were stratified by sex. A vital point when considering these results is that our overall sample size is relatively comparable to that presented in the only other published FBA investigation of ASD (Dimond et al., 2019), and therefore the subtlety of our results, particularly concerning our trend level findings when comparing the entire sample, might reflect upon, and inform, two pertinent avenues of ASD research.

Firstly, as expected, we present evidence in support of the notion that the pattern of neuropathophysiology observed among those with ASD differs between males and females when compared to matched counterparts. Our sample comprised substantially more females than that of Dimond et al. (2019), and therefore we must consider the extent to which this might affect the results observed between the studies. Here, the only observable group difference to reach statistical significance was between the female sub-groups, which might indicate greater neuropathophysiology among females with ASD. To date, research investigating sex differences in WM microstructure in ASD is scarce and inconsistent (Beacher et al., 2012; Kirkovski et al., 2015; Irimia et al., 2017; Zeestraten et al., 2017; Lei et al., 2019). Our findings, however, do align with that of Lei et al. (2019), who report no significant group differences between ASD and control groups when males are females were included, nor any difference between the male-only subsamples. Interestingly, the authors report reduced FA in females with ASD at the cingulate, inferior fronto-occipital fasciculus, ILF, SLF, UF, anterior thalamic radiation, corticospinal tract, forceps major, and the forceps minor (Lei et al., 2019). Concerning sex differences in the neurobiology of ASD more broadly, it was previously considered that females with ASD may have more severe and widespread neuropathophysiology, despite seemingly less severe behavioral profiles (Kirkovski et al., 2013). Indeed, this is supported by early volumetric studies of ASD (Bloss and Courchesne, 2007; Schumann et al., 2009, 2010). Furthermore, in considering our finding of reduced FDC in females with ASD compared to their unaffected counterparts, we must consider this theory as well as the FDC metric itself. As described above, FDC combines measures of WM micro- (FD) and macro- (FC) structure, neither of which alone revealed significant differences between our female subgroups. This approach, which is arguably more sensitive than FD or FC alone, indicates total intra-axonal volume changes (Raffelt D. A. et al., 2017; Mito et al., 2018). Further corroborating the aforementioned theory of greater neurobiological aberration among affected females, the lack of difference on our FDC measure between the male sub-groups might indicate that the main source of WM abnormality among this sub-sample was micro-structural, in line with the observed trends, and therefore combining this with FC, a macrostructural measure, to provide the FDC metric did not yield any significant differences between groups.

The findings presented here, along with those of Lei et al. (2019), highlight the benefit of sex-stratification when investigating sex differences in ASD. Such an approach has been taken in several studies investigating sex differences in the neuropathophysiology of ASD (Holt et al., 2014; Kirkovski et al., 2015, 2016, 2018; Lei et al., 2019), the majority of which identified sex-related differences that were masked within whole group analyses. Ecker, 2017 discussed the importance of stratification approaches in such research, highlighting that not only would such an approach be beneficial in reducing confounds associated with the phenotypic heterogeneity of ASD, but also the common situation where included males far outnumber included females.

Second, while inconsistent with our hypotheses, the results presented here are, for the most part, in line with the findings of our previous study investigating WM pathology in ASD (Kirkovski et al., 2015). In our earlier study, we found no group differences in DTI based measures of WM microstructure. The results presented in this study, as well as our previous work with a largely overlapping sample, might be reflective of some degree of age-related “normalization” of the neuropathophysiology observed in ASD, a phenomenon identified in brain volume research in this population over a decade ago (Courchesne, 2004; Courchesne et al., 2011; Redcay and Courchesne, 2005). Indeed, more recent reviews highlight the implications of age on research investigating brain volume and morphometry, structural integrity, and function among ASD populations (Duerden et al., 2012; Pua et al., 2017). Of particular relevance to this study, there is evidence in the ASD literature to indicate a relationship between age and tensor-based measures of WM microstructure (FA, MD, axial and radial diffusivity), highlighting the importance of age when interpreting such data (Bakhtiari et al., 2012; Kleinhans et al., 2012; Karahanoğlu et al., 2018). Karahanoğlu et al. (2018) report a switch in the directionality of this relationship, whereby in younger participants (late childhood/adolescence) FA and axial diffusivity were higher in the ASD compared to control group, while among the older sub-sample (young adulthood) the opposite pattern was observed. When splitting their samples by age, two separate studies (Ameis et al., 2011; Bakhtiari et al., 2012) report no differences between the older samples. Importantly, both samples were younger than those presented here. There are, however, many reports of altered WM structure among adults with ASD (see Travers et al., 2012; Aoki et al., 2013; Hoppenbrouwers et al., 2014), including the present study. Indeed, we see an overlap between the areas implicated in our study and that of Lei et al. (2019). One important difference between these two studies is the age of participants, with the sample presented by Lei et al. (2019) comprising children, adolescents, and young adults (4–21 years), while the present study comprises only adults. In considering the effects of age on WM structure, as described above, the widespread aberration observed in the younger sample “may normalize” to some extent in older samples, whereby only areas that were perhaps more aberrant remain detectable among older samples.

Concerning the role of the CC and the pattern of observed results within this study, the CC is one of the most heavily implicated brain regions in research using various diffusion protocols to investigate WM in ASD (Travers et al., 2012; Hoppenbrouwers et al., 2014), and was a tract identified as showing reduced FD among ASD participants in the only other FBA comparison of individuals with and without ASD by Dimond et al. (2019). The CC is typically described in terms of five sections, including, from anterior to posterior, the rostrum, genu, body, isthmus, and splenium, to better characterize locations along the tract. Here, the areas of the CC most indictive of having WM aberration included the body and isthmus, located between the body and splenium, with some indication of the genu being affected among the female ASD subsample. Our findings are in line with those reported by Sui et al. (2018), who, using diffusion kurtosis imaging (DKI), demonstrated that individuals with ASD showed reduced f_axon, a DKI metric indicative of intra-axonal water, in segment 3, 4, and 5 of the CC, corresponding with the midbody and isthmus, as well as the splenium. Further, in their meta-analysis of CC area, Frazier and Hardan, 2009 report that the greatest reduction of CC size in ASD can be observed in the midbody. A recent post-mortem investigation of nine ASD brains demonstrated overall reductions in the total number of axons across all parcellations of the CC (Wegiel et al., 2018), again supporting the use of FBA for investigating WM in ASD.

Given that the CC is the largest commissural WM bundle connecting the left and right hemispheres of the brain, the interhemispheric signal transfer is likely to be compromised in ASD (Valenti et al., 2020). Another consideration in terms of the implications of these findings is the functional relevance of the fiber projections stemming from these areas. It is well established that the posterior midbody projects to somatosensory regions, while the isthmus includes temporal and parietal projections (de Lacoste et al., 1985; Hofer and Frahm, 2006). Concerning the genu, also implicated by Dimond et al. (2019), fibers within this section of the CC cross through the forceps minor, a region implicated by Lei et al. (2019), and one that has projections to the frontal cortex. Therefore, the CC segments implicated in the present study project WM fibers to regions involved in the two core domains of autistic symptomatology; namely, social communication and repetitive behaviors/restricted interests (which, under DSM-5, now includes sensory sensitivities (American Psychiatric Association, 2013). Regarding the former, temporoparietal regions are strongly implicated in the cognitive processes involved in social understanding (Patriquin et al., 2016), while sensory difficulties experienced by affected individuals have been attributed to functional deficits within the somatosensory cortex (Lajiness-O’Neill et al., 2014). Indeed, Kana et al. (2014) address the potential relationship between structural and functional neuropathophysiology in ASD by combining task-based fMRI and DTI to investigate the structural and functional neural correlates of theory of mind (ToM), a cognitive model of social understanding, in ASD. As expected, blood oxygen level-dependent (BOLD) response was reduced at the temporoparietal junction during ToM in the ASD group, and FA in the WM underlying this region was also reduced.

In further considering the potential role and influence of biological sex on the structural and functional neuropathophysiology in ASD, two functional magnetic resonance imaging (fMRI) studies exploring sex differences in social cognition revealed the expected pattern of atypical neurobiological processes among males with ASD, but not females (Holt et al., 2014; Kirkovski et al., 2016). There are two important factors to consider here. First, while the aforementioned effects at a functional level may be attributed to behavioral or learning processes, the extent to which structural abnormalities, such as those presented in this paper, are implicated must be considered. Concerning the fMRI studies described above, along with the knowledge that isthmus projections reach temporal and parietal areas, we demonstrate a plausible link between structural and functional neuropathophysiological sex differences in ASD that warrants further investigation.

Despite using a cutting-edge approach to investigate a largely overlooked area of ASD (i.e., the role of biological sex), the present study is not without limitation. Though our sample was well matched for age, sex, and IQ, all enabling stratification of data and investigation of the role of biological sex between those with and without ASD, our stratified samples are small, hence reducing statistical power. An important consideration, too, is that there are limitations regarding the interpretations that can be made based on such analytical approaches (Nieuwenhuis et al., 2011). While we do not discount the value of testing such data for interaction effects, in considering our small sample, as well as the effects observed in similar research in this area (Holt et al., 2014; Kirkovski et al., 2015, 2016, 2018; Ecker, 2017; Lei et al., 2019), we assert that our approach has merit. Further, while some of our samples reported taking psychotropic medication, given our small sample size we were not able to account for this in the analysis (although Dimond et al., 2019 report that medication did not have any substantial impacts on their findings). Therefore, these findings should be considered preliminary, and we encourage future studies of ASD to also consider the role of biological sex. It is also important to acknowledge that many factors beyond biological sex, and the scope of this paper, contribute to the heterogeneous nature of ASD, and should also be considered.

In conclusion, we applied FBA to investigate WM micro- and macrostructure among adults with ASD, without intellectual disability, and also explored the influence of biological sex. We provide evidence for micro-and macrostructural aberration among adult females with ASD, but not for males. Our findings on the role of biological sex are of great interest and importance to the ASD literature. While the pattern of results was convergent to some extent, we show that males and females with ASD differ in terms of the strength of the aberration observed and the underlying structural properties. In considering that the vast majority of research in ASD is heavily biased towards the male ASD population, the ramifications of the historical approach of researchers excluding (actively or otherwise) affected females can be detrimental to our understanding of the condition, and therefore our ability to accurately and appropriately identify, understand, and treat those affected. It is imperative that future research further investigates sex differences in the neurobiological basis of ASD, and we strongly encourage such research to adopt a stratification approach.

The datasets generated for this study are available on request to the corresponding author.

The studies involving human participants were reviewed and approved by Alfred Human Research Ethics Committee, The Alfred, Melbourne, VIC, Australia. The patients/participants provided their written informed consent to participate in this study.

MK contributed to the study design, data acquisition and analysis, interpretation of results, and preparation of the manuscript. IF and CH contributed to data analysis, interpretation of the results and preparation of the manuscript. PD and TF contributed to the interpretation of the results, and preparation of the manuscript. SR contributed to the study design and data acquisition. PF and PE oversaw the entire study, including conceptualization, analysis, and manuscript preparation. All authors contributed to the article and approved the submitted version.

PF has received equipment for research from MagVenture A/S, Medtronic Limited, Neuronetics and Brainsway Limited and funding for research from Neuronetics. He is on scientific advisory boards for Bionomics Limited and LivaNova and is a founder of TMS Clinics Australia. The remaining authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.