The BG play a critical role in various brain functions, including movement and emotion regulation59. Damage or altered cortico-striatal white matter (WM) connections in ASD can lead to motor stereotypies and repetitive behaviors, possibly due to abnormal motor firing in the BG59,60. Higher-level functions such as executive control, motor coordination, and emotional regulation are mainly managed by the prefrontal cortex, which includes the dorsolateral prefrontal cortex (DLPFC). The DLPFC is crucial also for social functions, cognitive flexibility, and deception55,61,62,63.
A recent automatic MRI technique has enabled reliable quantification of PVS in the brain64,65, allowing us to correlate PVS metrics with ASD symptoms and the dysfunctional brain networks underlying them. In this study, we mapped PVS in the WM and BG regions of children with ASD, investigating how these PVS changes relate to the presence or absence of specific symptoms. We also analyzed the spatial patterns of PVS alterations to identify potential areas of vascular or glymphatic dysfunction, which may contribute to ASD symptoms. Since PVS volume is influenced by the size of the region, we measured the volume of each brain area and controlled for this in our analysis, ensuring that differences in PVS size were not confounded by regional volumetric differences.
In our global and lobar PVS analyses, we found that ASD clinical severity was directly and significantly related to the number of WM-PVS in both the left and right hemispheres. Additionally, we observed that age was inversely correlated with PVS number and volume in the occipital lobes bilaterally. With the exception of the occipital regions (Table 1), PVS differences did not coincide with significant differences in corresponding regional volumes, suggesting that the overall severity of ASD is linked to higher PVS counts and volumes, independent of lobar or global brain volume increases. The inverse relationship in the occipital lobes, however, remains unclear and may represent a spurious finding that requires further replication.
We also found that PVS counts and volumes in the BG were significantly higher in males (Fig. 1; Table 1).
On the other hand, we observed that impairments in expressive language were directly and significantly related not only to ePVS but also to increased WM volume in both hemispheres and the BG region (Table 2). This suggests that language delays or absence may be associated with both increased brain volume (with more severe language impairment linked to larger WM volumes) and PVS enlargement. These findings support the hypothesis that an excess of CSF volume contributes to PVS enlargement and dysfunction of the glymphatic system9,26,27,45.
Regarding sensory overload, we identified a significant relationship between PVS number in the right BG region and PVS volume in the entire BG region, particularly on the right side. Interestingly, ePVS in the BG were not associated with increased volumes of BG regions, suggesting that hypersensitivity may specifically correlate with PVS changes in the BG, independent of overall BG volume.
Lastly, we found that motor stereotypies were directly related to the number of PVS in the left insular region, as well as to the increased WM volume in the same region (Table 2). Therefore, this association may be spurious and should be interpreted with caution.
Regarding the PVS analysis across WM and BG parcellations, more detailed results are provided in the supplementary tables, while significant findings are shown in Fig. 4 as gradient-colored areas. We observed that different ASD symptoms are linked to specific PVS locations in both the right and left hemispheres (Fig. 4). However, the most notable finding is that both the right and left rostral middle frontal areas are associated with the severity of all symptom categories (Fig. 5). This suggests that increased PVS in these areas could be a key neuroimaging feature of severe autism.
The rostral middle frontal gyrus, along with the superior frontal gyrus, forms the DLPFC, a brain region extensively studied in schizophrenia and working memory65,66. Recently, the DLPFC has also gained attention in autism research, particularly as it relates to cognitive, emotional, and social development, as well as flexibility and deception61,63. Previous studies have reported altered synapse numbers and disrupted network connectivity in all cortical layers of the prefrontal cortex in post-mortem brains of individuals with ASD67,68.
Our results are consistent with fMRI studies that show reduced functional connectivity between the prefrontal cortex and other cortical and subcortical regions during cognitive tasks63. Although cautiously, this may offer a new perspective on the clinical dysfunction seen in ASD. Specifically, deficits in planning and response inhibition, which lead to stereotypical behaviors, might be linked to under-connectivity in cortico-cortical and cortico-subcortical networks originating from the DLPFC55,56,60,61,62. In younger children with ASD, this region appears to suffer from an excess of ePVS52.
The maternal immune activation (MIA) model suggests that women who contract infections during early pregnancy have an increased risk of giving birth to offspring with autism-like disorders69. A recent MIA experiment with rhesus monkeys showed that offspring exposed to MIA had altered neuronal dendritic branching in the DLPFC compared to unexposed controls, highlighting the neurodevelopmental impact of MIA on primate brain structures, particularly the DLPFC70.
How might PVS enlargement affect neurodevelopment? Proper regulation of the glymphatic system (GS) is essential for maintaining the balance of brain electrolytes, molecules, neurotrophic factors, and the removal of waste. The GS primarily transports these “neurofluids” through periarterial and perivenular spaces, basal cisterns, and the subarachnoid space. CSF moves from the periarterial system into the interstitial space through specific water channels (aquaporin-4) on astrocytic end feet. Disruption in CSF reabsorption across PVS may impair waste clearance from the brain. In neurodegenerative diseases as well as in aging and cerebral small vessel disease, a dysregulated aquaporin-4 expression due to inflammation or ischemic injury can create a cycle of poor waste clearance, inflammatory mediator accumulation, beta-amyloid and tau-protein buildup, and glial cell activation27,28,29,31,71. This chain of events can contribute to synaptic changes and secondary neurodegeneration.
We recognize that our study has certain limitations and methodological concerns.
One limitation is the relatively small sample size, and particularly the hospital-based case selection. This approach may skew the analysis towards more severe cases of autism, particularly those requiring MRI due to additional neurological or syndromic symptoms. Furthermore, because of the small sample and the exploratory nature of the study, we did not adjust for multiple comparisons. Therefore, our findings should be viewed as preliminary, guiding future hypothesis-driven research. Larger sample sizes and confirmatory analyses will allow for more robust statistical control.
Another key limitation is the absence of a control group of typically developing children, which makes it difficult to assess the true magnitude of PVS dilation in autism. Additionally, 14% of the children who underwent MRI scans before the age of 3 were later diagnosed with ADHD, which typically emerges after the age of 316. These comorbid cases reflect later developmental diagnoses rather than concurrent conditions at the time of the MRI. It is important to note that both the early MRI and the later ADHD diagnosis likely point to a more severe ASD phenotype. Early assessments are often driven by significant clinical concerns. This raises the possibility that undiagnosed or emerging ADHD in our sample could influence the results. Longitudinal studies are needed to explore the potential role of DLPFC dysfunction in ADHD symptoms among children with ASD.
Another limitation is the large amount of data analyzed, which could raise concerns about the casual association of symptoms with the measurements. One might be concerned that some observed associations are random or biologically implausible. However, in all our analyses, the relationship between increased ePVS and worsening ASD symptoms was consistently unidirectional—the greater the ePVS, the more severe the symptoms. We never observed values that contradicted this trend, even by chance. Thus, we believe our findings are clear and merit attention.
Finally, there are some specific findings that warrant further discussion.
One issue is the inverse relationship between PVS burden and age in the occipital lobe, which contrasts with the pattern seen in normal aging31,32,33,34,35,36,37. However, unlike adults, where neurological damage occurs after brain maturation, the expansion of extra-axial CSF volume and PVS enlargement is an early phenomenon in ASD, with transient characteristics9,26,27,45. As previously observed, the occipital lobe plays a crucial role in processing visual information during the first year of life, which is essential for binocular depth perception. In contrast, regions responsible for higher cognition continue to mature into adolescence72. A recent review on fetal functional connectivity has shown that adult-like functional networks, including the occipital/visual network, are present at birth73. We believe that early neurodevelopmental disturbances may alter the volume of the occipital lobe, at least temporarily.
A second issue is the spatial specificity of ASD symptoms, with global severity being more strongly associated with the right hemisphere than the left. Abnormal lateralization of brain function is observed in various neuropsychiatric disorders. Atypical lateralization of language regions, with increased right hemisphere activation, is common in ASD74,75. This may reflect a compensatory mechanism for insufficient left hemisphere development of language networks75. Additionally, symptoms such as absent interoception and high pain thresholds are often linked to right hemisphere dysfunction in ASD56,57,58. Other studies have also associated right hemisphere abnormalities with stereotypies, socialization difficulties, and non-verbal deficits in ASD54. However, lateralization patterns in ASD are highly individualized, and generalizations are difficult to make76.
In our cohort, the most severe symptoms (stereotypies, hypersensitivity, and language issues) were linked to greater PVS alterations in the left hemisphere and left pallidum. However, when considering global clinical severity (Fig. 4A), the right hemisphere showed more ePVS involvement, particularly in the right precuneus and cingulate regions.
In the largest cohort to date77, nearly 3000 subjects with ASD showed widespread leftward cortical reductions. While left lateralization in autism is a well-established finding77, several studies report small effect sizes due to factors like sample heterogeneity, underpowered analyses, and confounding variables such as sex, age, symptom profiles, and co-occurring neurodevelopmental disorders76,77. These same factors are present in our hospital-based cohort.
In conclusion, our study suggests that ePVS in the white matter underlying both right and left rostral middle frontal regions is linked to severe ASD cases and to the severity of specific symptoms, including verbal dysfunction, stereotypies, and hypersensitivity. Excessive PVS enlargement in these prefrontal subregions could lead to hypoactivity of the DLPFC, contributing to ASD symptoms. These findings may have clinical implications, potentially supporting new therapeutic approaches such as non-invasive transcranial DLPFC stimulation78.
