Our goal was to investigate the association between social communication performance and the cerebellar morphology in a large multicenter transdiagnostic pediatric cohort. We did not identify significant associations with social communication performance using linear mixed models. However, in the CCA analysis, we found a significant association between the canonical clinical variate (including SRS-SCI subscales and FSIQ) and the canonical cerebellar neuroanatomical variate.
Our results suggest that the cerebellar structure is associated with social cognition and FSIQ, but that there is no one-to-one relationship between a sub-region of the cerebellum and social performance. The current findings add to a growing body of literature on the role of the cerebellum in social cognition and general cognitive abilities. Alterations in the cerebellar structure could lead to atypicalities in multiple types of cognition and disrupt cerebello-thalamo-cortical pathways. Interestingly, Wang et al. (Wang et al., 2014) suggest that an insult in the cerebellum during sensitive periods of brain development might be causal and have a distant effect on cortical structure.
The postero-superior “cognitive” cerebellum has been implicated in several psychiatric disorders. For example, there is now strong evidence (Laidi et al., 2019, 2015; Moberget et al., 2019) that the volume of the posterior cerebellum is reduced in schizophrenia. In the field of autism, despite initial studies in small samples (D’Mello et al., 2015; Saitoh and Courchesne, 1998) suggesting that there might be atypicalities in this region, recent larger studies (Traut et al., 2018, Laidi et al., 2022) found no compelling evidence for volumetric alterations of the cerebellum. These studies typically rely on case / control comparisons. However, autism is a heterogeneous construct with many co-occurring conditions such as ADHD, mood disorders, anxiety or learning disabilities (Lai et al., 2019). These factors might explain why to date case-control studies have failed to identify reliable biomarkers in autism. A recent study from our group (Laidi et al., 2022), found no evidence for difference of cerebellar anatomy when comparing typically developing subjects and individuals with autism.
To overcome these limits, a dimensional approach is key to better understand the relationship between the brain anatomy and symptoms of neurodevelopmental disorders. The HBN cohort is a transdiagnostic pediatric cohort that allows this type of analysis (Alexander et al., 2017). The inclusion criteria are very broad: approximately half of the participants are suffering from a neurodevelopmental disorder, 20% from anxiety or depression and 10% did not have a diagnosis after the clinical evaluation. Our goal was to understand, in this pediatric population, how social communication performance might be related to cerebellar structure.
We tested two linear models for each region of interest. In the first one, we considered age, sex, site and ICV as covariates / cofactors and in the second one we added in addition interactions between the social communication performance and sex, age, ICV and IQ. While the model without interactions revealed no significant results, we found significant results in the second model. However, the significant interaction terms varied depending on the region of the cerebellum, suggesting that a single model would not have been appropriate for all regions of interest. Lastly, we found high collinearity bias, underlining the limits of linear mixed models to study brain-behavior association.
To overcome these limits, we conducted a canonical correlation analysis to study the association between cerebellar neuroanatomy and social communication performance in a multivariate fashion with a CCA model (Wang et al., 2020). Our goal was to explore in a single analysis the complex relationship between clinical and anatomical features. The CCA model allowed us to include both SRS social subscales and FSIQ in the canonical clinical variate, instead of treating FSIQ as a confound. We found a significant canonical correlation between the canonical clinical variate (including SRS-SCI subscales and FSIQ) and the canonical cerebellar neuroanatomical variate. The canonical clinical variate loadings revealed a predominant effect of the SRS social motivation subscale and the FSIQ. Autism and intellectual disability are frequently associated conditions (Saito et al., 2020) and mice models of autism, such as SHANK or CTNAP, are strongly associated with intellectual deficiency (Kazdoba et al., 2015). This suggests from a neuroanatomical perspective a close imbrication between social and general cognition, assessed by FSIQ.
Our study has several strengths. First, we tested a specific hypothesis in a large multicentric and transdiagnostic pediatric cohort and found that linear models might not be ideal to capture the association between cerebellar anatomy and social communication performance. Instead, CCA allowed us to unravel a more complex association between social communication, FSIQ and the cerebellar structure. Our approach might be relevant to explore the association between other specific brain circuits or regions and clinical features, and not only in brain wide association studies (Marek et al., 2022). Second, because we decided to focus on a specific brain region, we were able to perform a careful visual inspection to ensure the quality of the cerebellar parcellation. We decided to perform the cerebellar parcellation with the CERES pipeline. This automated parcellation method is well validated, has been successfully applied to healthy populations (Coupé et al., 2017), psychiatric disorders (Laidi et al., 2022, 2019) and outperforms other cerebellar parcellation methods (Carass et al., 2018). To the best of our knowledge, this study is to date the largest to investigate the association between the cerebellar morphology and symptoms related to autism. Third, we performed a thorough and careful visual inspection of all parcellation outcomes and T1 data. After quality check, we excluded 559 images based on motion artifacts movements (n = 280) or parcellation errors (n = 279). This suggests the importance of performing a visual quality inspection of all data, even in large datasets. Given that excluded subjects tend to be younger males and with a slight increase of symptoms compared to other subjects, not performing a careful quality check could lead to spurious results.
Our work opens several perspectives. Our results were the results of an analysis in HBN, a large transdiagnostic developmental cohort. Our study focuses on structural MRI. However, studying the functional connectivity of the regions of the posterior-cerebellum might help to gain more insight on how the cerebellum is interacting with the cerebrum in social cognition. However, one should be cautious on how to interpret the functional connectivity findings, since functional connectivity analyses and the common MNI template rely on the assumption that there are no strong differences in the structural morphology. Atypicalities in cerebellar functional connectivity might be only related to changes in the structural anatomy of this region. Stoodley et al. (Stoodley et al., 2017) studied the functional connectivity of the cerebellum in animal models and humans, showing modifications of cerebellar-cerebral cortex connectivity when delivering transcranial direct current stimulation (tDCS) of lobule VII and suggesting a clinical improvement of social abilities after such a stimulation. Defining specific cerebellar regions altered in social and communication symptoms may therefore lead to new therapeutic strategies focused on accurate transcranial stimulation.
To conclude, we found a canonical correlation between social communication performance and cerebellar neuroanatomy. Our work suggests the interest of CCA models to investigate the link between cerebellar structure and dimensions of psychopathology and adds to the body of literature on the role of the cerebellum in social and general cognitive functions from a dimensional perspective.