The Cerebellum’s Role in Mental Illness: New Insights From Genetic Studies

Individuals with 22q11.2 deletion syndrome, which is caused by a copy number variant (CNV) that occurs in 1 in every several thousand live births, are well known to have a substantially increased risk for autism, schizophrenia, and other neurodevelopmental disorders. While 22q11.2 deletion syndrome is readily diagnosed through genetic testing, the intermediate biological mechanisms through which the genetic lesion predisposes individuals to psychiatric risk remain unknown. The recent study by Fung et al. (1) in Biological Psychiatry: Global Open Science associated regional differences in cerebellar volumes with both 22q11.2 deletions and insertions, the latter of which have received relatively scant attention in clinical and brain imaging studies despite comparable prevalence. In turn, region-specific variation in cerebellar volumes has been associated with numerous clinical phenotypes, including cognitive, social, and psychosis spectrum outcomes (1). Their work serves as an informative example of the “genetics-first” approach, wherein rare but highly penetrant variants illuminate biology that could underlie common psychiatric syndromes.

This approach has yielded important dividends in other areas of medicine. For example, in the early 2000s, genetic studies of familial hypercholesterolemia associated gain-of-function mutations in the PCSK9 gene with greater risk of early-onset heart disease and loss-of-function mutations in the same gene with lower risk. These and other findings prompted the development of a novel class of biological therapies, anti-PCSK9 monoclonal antibodies, which were approved in 2015 to lower low-density lipoprotein cholesterol in patients with poor responses to statins. Notably, the gain-of-function mutations in PCSK9 contribute to fewer than 1% of familial hypercholesterolemia cases, but PCSK9 inhibitors are broadly effective, including among individuals without gain-of-function mutations (2). Similarly, while only a small minority of patients with schizophrenia or autism have 22q11.2 copy number variants, the hope is that broadly tractable intervention strategies can emerge from biological studies of these select individuals.

A key area of difference with respect to 22q11.2 is that, unlike PCSK9 mutations, which occur within a single gene, the CNV in 22q11.2 impacts approximately 30 genes. These genes perform an array of cellular and molecular functions, any of which, if disrupted by the mutation, might influence neuropsychiatric risk (3). While parsing the genomic complexity of this region—i.e., understanding which gene or genes may contribute to mental illness risk—remains a challenging problem, Fung et al. (1) approach the question of mechanism at a different level of biology, namely, where in the brain might 22q11.2 mutations predispose to psychiatric symptoms?

In turning to the cerebellum, they focused on a part of the brain that has historically been underappreciated (and, some even might say, neglected) in the structural and functional neuroimaging literature. The cerebellar cortex is vastly expanded in humans compared with nonhuman primates and contains the highest neuronal density in the brain (4). Until the late 20th century, the cerebellum’s primary role was attributed to fine-tuning the planning and execution of motor function. Because the cerebellum lacks monosynaptic connections with the cerebral cortex, conventional tract tracing failed to detect its cortical fingerprints.

Functional neuroimaging played a key part in understanding the cerebellum’s substantially broader roles, including emotional and cognitive processes. A seminal paper by Buckner et al. (5) recapitulated more than a dozen distinct cerebral networks—including those thought to mediate executive function, attention, salience detection, and the default network, in addition to motor and sensory networks within the cerebellar cortex. Redundancy of these network representations in the cerebellum, with mirrored anterior and posterior topography, hints at additional complexity in corticocerebellar interplay.

This work echoed sophisticated anterograde tracing conducted by Schmahmann and others, which provided an anatomical substrate for polysynaptic connections between the cerebellum and diffuse neocortical regions (6). Schmahmann was also among the first to recognize the potential contributions of the cerebellum to cognitive and affective modulation and, relatedly, to psychopathology, terming such effects “cognitive dysmetria.” Just as it serves to modify the “gain” of balance and posture—coordinating across motor, premotor, and other regions that control motor output based on sensory input—the cerebellum may play a parallel role in fine-tuning distributed cognitive and affective networks through error correction and prediction. Unsurprisingly, then, structural and functional differences in the cerebellum have been associated with risk for a host of psychiatric disorders across the life span (7).

However, technical challenges have posed a key limitation in developing a more nuanced understanding of how the cerebellum contributes to mental illness risk. The cerebellar cortex is approximately one-third the thickness of the cerebral cortex and is organized into tightly wrapped folds that support a massive surface area (equivalent to approximately 78% of the entire neocortex). While improved computational methods have enabled measurement of other subfield volumes in subcortical regions (e.g., the hippocampus and amygdala), segmenting cerebellar subfields requires even more specialized approaches at standard (e.g., 3T) field strengths. Fung et al. deployed a state-of-the-art automated parcellation of cerebellar subfield volumes using ACAPULCO (Automatic Cerebellum Anatomical Parcellation using U-Net with locally constrained optimization), a deep learning-based software package that has been validated in studies of both children and adults. This approach enabled them to localize specific areas of volume loss in the cerebellum that tracked with 22q11.2 CNV status, social and cognitive impairment, and, in some regions (e.g., lobule VIII), both.

Interestingly, reduced volume in lobule VIII, which contains somatomotor and attention network territories, was also recently linked to psychopathology risk in a population-level study of early adolescence. In an investigation of neuroimaging, genomic, and psychopathology data from the ongoing ABCD (Adolescent Brain Cognitive Development) Study, we associated reduced lobule VIII volumes with increased externalizing psychopathology (8). In our analysis, the cerebellum emerged as a region of interest not because of a particular a priori hypothesis, but rather because, among all brain regions, it showed the highest expression for a newly identified set of neurodevelopmental genes that predicted dimensional symptoms. Furthermore, using developmental transcriptome data, we demonstrated that, within the cerebellum, such gene expression peaks during prenatal life.

These findings collectively point to the importance of not only spatial but also temporal specificity in the ways by which genes expressed in the cerebellum are relevant to psychopathology risk. Ongoing work is contributing spatial and temporal data at an even higher level of resolution. Using single-cell transcriptional profiling, Kozareva et al. recently elaborated a vast diversity of cell types across the mouse cerebellar cortex, with regionally and temporally specific patterning (9). Transcription differed widely within canonical cell types (e.g., Purkinje, granule, and Golgi cells), with corresponding differences in electrical properties. Comparison with the human postmortem brain indicated conservation of many of these patterns, and full-scale efforts to build a transcriptomic atlas of the human cerebellum are currently underway.

Bearing such spatial and temporal complexity in mind, it is important to approach triangulation of genetic, neuroimaging, and behavioral data with some degree of humility. For example, Fung et al. (1) (understandably) did not approach the question of whether structural variation in the cerebellum mediates the effects of genotype on behavior. While the appeal of such analyses seems obvious, i.e., establishing that measurable elements of brain biology bridge underlying genomic variation with symptoms—it is much more often true in imaging-genetics studies such as this one that mediation is not possible to detect. In addition to recognizing the dynamic nature of gene expression, it is important to note that the genetic bases of variation in brain structure and behavior may be comparably complex and only partially overlapping. To expect straightforward, directional linkages among these sets of variables may be asking too much of the data, especially in relatively small studies.

That said, a complete understanding of the interplay of risk-conferring genes and cerebellar structure and function is certainly not a prerequisite for developing novel treatment strategies. While work such as Fung et al.’s (1) is helping the field take important steps toward resolving these molecular mechanisms, region-specific targeting of the cerebellum itself is on the near horizon. For example, a recent study by Rohira et al. used magnetic resonance imaging-based neuronavigation to improve individual-level targeting of transcranial magnetic stimulation of the cerebellum in healthy volunteers, achieving lobule-specific changes in electrical activity (10). The use of such noninvasive brain stimulation methods holds promise for precisely altering cerebellar network function in individuals with psychiatric illness, regardless of its genetic underpinnings.