The cerebellum is a neuron-rich structure residing at the base of the brain, connecting with the cerebrum, the brainstem and the spinal cord. Long recognized for its central role in movement control and coordination, the cerebellum is increasingly implicated in cognitive functions, including social behavior, reward circuitry, emotional and language processing (Carta et al., 2019; Wagner and Luo, 2020; Stoodley et al., 2021). A range of neurodevelopmental disorders, such as Joubert syndrome and autism spectrum disorders (ASD), arises from disruptions in cerebellar development of varying severity. Unraveling how genetic disruptions affect cerebellar morphogenesis is crucial. To gain insights into developmental origins of brain disorders, we conducted an extensive literature and database search to identify gene deficiencies that affect cerebellar fissure formation in mouse models. Through such an analysis, this review offers unique insights into how cerebellar fissure morphogenesis is controlled by cellular signaling programs. Moreover, we discuss the potential of small molecule compounds that modulate signaling pathways for reversing cerebellar developmental disorders, particularly hypoplasia of the vermal lobules VI and VII associated with ASD.

The cerebellum consists of two hemispheres connected by a narrow midline region called the vermis. The foliation pattern is symmetrical to the midline and gives rise to 10 lobules I–X that run perpendicular to the anterior–posterior axis along the vermis (Sillitoe and Joyner, 2007; Farini et al., 2021) (Figure 1A, upper panel). The cerebellar cortex has a laminar organization, i.e., the molecular layer (ML), Purkinje cell layer (PCL), and granule cell layer (GCL), going from the outer to inner direction (Figure 1A, lower panel). Nearly 99% of cerebellar neurons are granule cells (GCs). During development, GC precursors (GCPs) proliferate at the external granule layer (EGL), which is an outermost and transient layer (Figure 1B). After mitosis, GCs migrate inward to populate the GCL (for the detail, see below). The somata of Purkinje cells (PCs) form a monolayer called PCL. The molecular layer (ML) is largely cell-free but contains neuronal microcircuits, including GC axons, PC dendrites, and their synapses, as well as interneurons and glia (Apps and Hawkes, 2009) (Figure 1B). In the ML, axons of GCs ascend from the soma, forming parallel fibers which have extensive connections with dendrites of PCs. Besides, two major excitatory afferents, mossy fibers and climbing fibers, from other brain areas and spinal cord terminate in GCs (Legué et al., 2016). Embedded within the white matter under the cerebellar cortex, the deep cerebellar nuclei primarily receive afferents from the GABAergic PCs and generate the principal outputs of the cerebellum (Figures 1A,B). Precise organization of neurites and their connectivities constitute the cerebellar circuitry (Legué et al., 2016).

The cerebellum processes information from functionally diverse regions of the cerebral cortex to its support motor, cognitive, and affective functions (Stoodley et al., 2012). An early study reported that birth date-related PC clustering may correlate with functional compartmentalization along the mediolateral axis in the adult cerebellum (Hashimoto and Mikoshiba, 2003). Task-based neuroimaging studies have more recently revealed distinct functional territories of the cerebellum, i.e., sensorimotor (lobules II–VI, VIIIB), vestibular (lobules IX–X), oculomotor (lobules VI–VII, IX–X), visual (lobule VI), and auditory (lobules V–VI) zones (Stoodley and Schmahmann, 2018; Raymond and Medina, 2018) (Figure 1A). Based on spatiotemporal transcriptomics, it is now evident that the cerebellum exhibits regional specialization physiological and anatomical properties across subregions (i.e., folia) and neuronal subtypes (Cerminara et al., 2015; Kozareva et al., 2021; Lanore et al., 2021; Sepp et al., 2024). As a prime example of non-uniform microcircuitry, the longitudinal zebrin II+/− stripes in PCs reveal molecular heterogeneity that restricts specific splicing variant expression that influences PC plasticity, firing properties, and input–output connections (Lin et al., 2020). Therefore, cerebellar development is a highly regulated and genetically influenced process, rather than a uniform one. Nevertheless, more research is needed to determine how molecular factors align with functional specialization.

Gene deficiencies affect cerebellar morphogenesis to varying degrees. Certain developmental defects contribute to hypoplasia of the vermian lobules VI–VII, a feature associated with ASD (Courchesne et al., 1988). Developmentally, lobules VI and VII expand significantly in the lateral hemisphere forming crus I in the rodent cerebellum, the homologous region of human crus I/II implicated in cognitive and visuomotor functions (Sugihara, 2018). Notably, lobules VI-VII exhibited a delayed developmental timeline in comparison with that of other lobules (Legué et al., 2016). Functional topography has revealed that lobules VI and VII are embryologically and phylogenetically distinct from the anterior lobules I–V (Courchesne et al., 1988). Tract-tracing studies revealed reciprocal connection from this posterior vermal region to associative and paralimbic cortices, providing anatomical substrate for cognitive functions (Kelly and Strick, 2003). Of note, lobule VII accounts for 47.70% gray matters of the human cerebellar volume (Diedrichsen et al., 2009). In vivo electrophysiological studies in the mouse cerebellum underscored an interplay of intrinsic cell properties and input–output profiles underpinning a zonal distribution of various cerebellar functions. For example, mossy fiber burst inputs to GCs and PC firing rates are characteristically different for lobules VI–VII versus X. This difference in input–output regularity may support the neural processing required by distinct tasks each subregion is involved in: an oscillation-based communication with cerebral cortex (lobules VI–VII) versus an always-on mechanism for vestibular functions (lobule X) (Witter and De Zeeuw, 2015).

The pattern of vermis foliation is generally conserved across mammalian species. The human cerebellum begins to develop at gestational week 4 and ends around 2–3 years after birth (Carletti and Rossi, 2008; Iskusnykh and Chizhikov, 2022). In mice, cerebellar development begins at embryonic day (E) 9 and ends in the third postnatal week. The cerebellar primordium emerges in the roof of the fourth ventricle, comprising two primary germinal zones, i.e., the ventricular zone (VZ) and the rhombic lip (RL) (Leto et al., 2016) (Figure 2A, E12.5). VZ progenitor cells produce GABAergic PCs and Bergmann glia precursors. The RL produces glutamatergic GC precursors (GCPs). Foliation occurs concurrently with the formation of the cortical cell layers. Fissure formation is initiated around E16.5 when GCP expands at the outermost EGL, a transient secondary germinal zone, and move inwards to the inner EGL (Figure 2A, E18.5 and Figure 2B). GCP proliferation peaks between postnatal day 5 and 8 (P5-8). Subsequently, postmitotic GCs undergoes radial migration from the EGL across the PCL to populate the internal granule layer (IGL) (Figure 2B). The radial glia in the VZ transforms into Bergmann glia during E14.5-E18.5. Bergmann glial cells are essential for the folding of the cerebellar surfaces (Sudarov and Joyner, 2007; Leung and Li, 2018). As the anchoring points, Bergmann glia located next to the PC layer emit their radial fibers toward the pial surface for GC migration (Rahimi-Balaei et al., 2018) (Figure 2B). GC maturation and migration complete by P15, and proliferative EGL disappears (Figure 2A, P15) (Carletti and Rossi, 2008; Iskusnykh and Chizhikov, 2022). Notably, the kinetics of GC accumulation varies across different zones of the cerebellum. Maximal GC production is delayed in the lobules VI and VII compared to other zones (Legué et al., 2016). Whether such a delay results in the formation of the fissure (namely the intercrural fissure, see below) between lobules VI and VII particularly sensitive to certain cellular signals or neurotransmitters remains as an intriguing question.

Cerebellar development is orchestrated through a program of transcriptional regulation and signaling pathways. Recently, single-cell transcriptomic profiling has identified the expression of nearly 200 cell-type specific transcription factors in the developing cerebellum and determined their respective regulon activities (Sepp et al., 2024). On the basis of mutant-induced phenotypic changes (see Section 6), this review discusses signaling pathways essential for cerebellar development. As early as E8.5, Wnt1 drives the expression of FGF8, a morphogen that specifies hindbrain and controls the onset of cerebellar development (Spassky et al., 2008). Bone morphogenetic proteins (BMPs) function before the formation of the cerebellar primordium to control stem cell specification in the anterior rhombic lip (Tong et al., 2015). Notch signaling preserves a pool of neural progenitor cells in an undifferentiated state. However, Notch levels differ between daughter cells after cell division. Although high levels of Notch maintain the progenitor fate, the intermediate and low levels of Notch activities may, respectively, generate inhibitory and excitatory neurons (Zhang et al., 2021). From E17.5 to postnatal days, PCs express Sonic hedgehog (Shh) to promote GCP proliferation in the EGL (Wallace, 1999). Reciprocally, Reelin released by GCPs disperses PC clusters into the monolayer (Miyata et al., 1997). Dysregulation of these cellular signaling pathways contributes to a range of cerebellar disorders (see below).

Cerebellar malformations underpin a spectrum of debilitating disorders, from ataxia to autism. Accumulating animal-based evidence highlights loss of function genes and defective signaling that give rise to abnormal cellular patterning and foliation defects underlying disease progression. In this review, we describe several cerebellar developmental disorders and their associated susceptibility genes that have been confirmed in animal studies.

Over the past three decades, studies of mouse disease models have provided substantial information regarding cerebellar development and implication for human disorders associated with cerebellar malformations (Manto and Jissendi, 2012; Butts et al., 2014; Haldipur and Millen, 2019). Moreover, our understanding of cerebellar morphogenesis may be sped up by unexpected discoveries of genes associated with disease phenotypes. A comprehensive database has compiled a total of 543 mouse genes critical for cerebellar development and 630 human mutant loci associated with cerebellar phenotypes (Ramirez et al., 2022). Recent studies using single-cell RNA-seq of the cerebellum across developmental stages and species have revealed its cellular architecture, evolutionary differences, and insights into cerebellar diseases (Cerminara et al., 2015; Carter et al., 2018; Haldipur et al., 2019; Aldinger et al., 2021). To date, we still lack a complete understanding of the mechanisms governing cerebellar morphogenesis and how their dysregulation contributes to cerebellar disorders.

To improve our understanding of cerebellar foliation, we collected mouse genes that have been reported in cerebellar foliation from Mouse Genome Informatics (MGI)¹ and PubMed.² In the MGI database, 69 genes are categorized as having abnormal cerebellar fissure morphology or lobule formation, or with reduced or absent foliation in the cerebellum (Table 1, MGI). A PubMed search for “cerebellar hypoplasia” identified 52 genes whose knockout reduces cerebellum size and 19 genes whose knockout impairs intercrural fissure formation (Table 1, PubMed). Intriguingly, only 15 genes from this search were identified in the above categories of MGI (Table 1, footnotes). PubMed also revealed that approximately 30 genes whose knockout does not significantly affect cerebellar morphology, but several of them affect PC arborization or function, such as Bcl7a, Pten and Tsc2 (Tsai et al., 2012; Cupolillo et al., 2016; Wischhof et al., 2017).

Having identified genes essential for cerebellar foliation, we turn to the molecular signaling pathways that affect cerebellar cytoarchitecture. From the 125 genes identified in our MGI and PubMed analyses pertaining to cerebellar hypoplasia and intercrural fissure defects, 28 emerge as key players in developmental signaling, neurotransmission, and cell survival (Table 2). Their disruption profoundly alters cerebellar morphology, often with striking specificity. Here, we explore how core pathways like BMP, Wnt, and Shh, alongside subtle modulators like BDNF, orchestrate foliation and reveal vulnerabilities in developmental disorders (Table 2).

Since ASD is associated with abnormalities in vermal lobules VI and VII, identifying the key pathways that shape the intercrural fissure is essential. Although BDNF has been implicated in the formation of the intercrural fissure, strong evidence has been lacking. Strikingly, conventional knockout of two paralogous copies of Rbm4 gene results in the loss of the intercrural fissure and a significant reduction in BDNF levels (Tsai et al., 2023). RBM4 is an alternative splicing regulator (Markus and Morris, 2009). RNA-seq analysis of Rbm4 knockout brains revealed intron retention in Hsf1, which encodes a transcriptional activator for Bdnf (Shen et al., 2024). Intron retention leads to downregulation of HSF1 protein, and hence BDNF reduction. Prenatal re-expression of HSF1 in Rbm4 knockout brains restored BDNF levels and the intercrural fissure. Similar results were obtained with prenatal supplementation of 7,8-dihydroxyflavone, a TrkB agonist (Tsai et al., 2023), indicating that BDNF plays a crucial role in the formation of the intercrural fissure. Moreover, activation of N-methyl-D-aspartate (NMDA) receptors induces a kinase cascade involving Ca²⁺/Calmodulin-dependent protein kinase II (CaMKII) and SR protein kinase 1 (SRPK1), leading to phosphorylation of RBM4. Phosphorylated RBM4 translocates into the nucleus, where it promotes Hsf1 intron excision (Figure 4). Such stimulus-activated splicing mechanism is in line with the report that neuronal activation promotes the removal of retained introns (Mazille et al., 2022). It is noteworthy that, unlike Hsf1 intron excision upon NMDA stimulation, acute stress induces nuclear translocation of HSF1 for Bdnf transactivation in the hippocampus (Franks et al., 2023). Thus, developmental cues and cellular stress activate HSF1 via different mechanisms. Taken altogether, BDNF plays a crucial role in intercrural fissure formation. Further studies are needed to determine why this fissure is especially sensitive to BDNF signaling and whether BDNF is crucial for functional circuit formation particularly in the central cerebellar vermis during development.

Mouse model with cerebellar hypoplasia can recapitulate key features of human cerebellar disorders (Section 5), making them a viable tool for developing interventions against neurodevelopmental disorders. For example, Joubert syndrome mouse model with defective Wnt signaling exhibit cerebellar midline fusion that can be partially reversed with lithium treatment, an agonist of Wnt signaling (Lancaster et al., 2011). Moreover, a mutation of the glia cell-line derived neurotrophic factor Ret gene causes cerebellar hypoplasia in mice that mimics Down’s syndrome. Such a mutation impairs Shh-mediated development of GCs and glial fibers, and a Smo agonists can rescue these neuronal defects (Ohgami et al., 2021). Mice with mutant methyl-CpG-binding protein 2 gene provide a Rett syndrome model and exhibit deficient BDNF–TrkB activity. A small molecule TrkB agonist, LM22A-4, can alleviate the motor learning deficits in these mice (Medeiros et al., 2024). Cerebellar BDNF expression is reduced in postmortem SCA6 human tissues (Takahashi et al., 2012). Consistent with this finding, reduced TrkB–BDNF signaling is evident in the early disease stage of a SCA6 mouse model (SCA6^84Q/84Q). Prolonged administration of 7,8-dihydroxyflavone improved ataxic phenotypes and PC firing rate (Cook et al., 2022). As described above, prenatal administration of 7,8-dihydroxyflavone restored cerebellar development and motor learning in Rbm4 knockout mice (Tsai et al., 2023). Given the effects of HSF1 overexpression in Rbm4 knockout mice (Shen et al., 2024), using small-molecule compounds to activate HSF1 for intervention is possible. HSF1 is targeted by multiple stress-induced signaling cascades (Hooper et al., 2016). Tanespimycin (17-AAG), a derivative of the antibiotic geldanamycin, can de-repress HSF1 from sequestration by its molecular chaperone HSP90 (Chen et al., 2014). 17-AAG can restore synaptic protein levels such as PSD95 and BDNF in Alzheimer’s disease models (Chen et al., 2014). Therefore, it may be possible to treat developmentally disordered brains with low BDNF, such as those with Rbm4 knockout, with 17-AAG or similar functional molecules. These findings suggest that pharmacological restoration of signaling activities may be useful for treating developmental disorders in the future.

This review summarizes key developmental signaling pathways and neuromodulators involved in cerebellar development. Dysregulation of these pathways results in cerebellar malformation, ranging from hypoplasia to local foliation defects. It is noteworthy that BDNF plays a pivotal role in shaping the intercrural fissure between lobules VI-VII—a structure implicated in ASD. Prenatal restoration of BDNF biogenesis or signaling can prevent cerebellar deficits in BDNF deficient mouse models. Additionally, it is possible to reverse other BDNF-deficient-caused deficits in the cerebellum through activation of TrkB. Unraveling how these pathways converge across species and disorders promises deeper insights into cerebellar morphogenesis and innovative therapies for neurodevelopmental challenges.