Clinical brain stimulation studies have provided indications of a role for the cerebellum in regulating emotions as well as anxiety. A single session of repetitive, high-frequency transcranial magnetic stimulation (rTMS) of the medial cerebellum elevated positive mood in healthy volunteers (Schutter et al., 2003), whereas in a follow-up study, slow-frequency rTMS of the cerebellar vermis in healthy volunteers amplified negative mood (Schutter and van Honk, 2009). Although Schutter et al. (2003) acknowledged a lack of objectivity in the quantification of mood changes in their pilot study and provided a number of alternative interpretations for their observations, they nonetheless concluded that these studies collectively indicated the possible involvement of the cerebellum in regulating emotions. Additionally, chronic cerebellar stimulation reportedly led to a reduction in “tension-anxiety” in 63% of neurologically-impaired patients and mood improvement in 38% of these patients, although 8% of patients reported worsened “tension-anxiety” after treatment (Riklan et al., 1977). Thus, the cerebellum may be a valuable target for mood and anxiety-related therapies, in particular for mood disorders (reviewed in Lupo et al., 2019) as well as fear and anxiety-related disorders (reviewed in Moreno-Rius, 2018). However, our lack of understanding of how the cerebellum may affect emotions, particularly anxiety, is a major obstacle to improving therapeutic options.

Changes in cerebellar activity and connectivity have been associated with anxiety disorders in several clinical studies (Engel et al., 2009; Moreno-Rius, 2018). Most of this work is related to SAD. SAD patients exhibit increased connectivity between the cerebellum and the amygdala (Liao et al., 2010), as well as the posterior cingulate (Doruyter et al., 2016). Patients suffering from SAD have higher activity in both their cerebellum and amygdala in response to stimuli that induce social anxiety (Evans et al., 2008). Further, the volume of the left cerebellum of SAD patients is increased in comparison to healthy controls (Talati et al., 2013). A more compelling implication of the cerebellum in SAD has come from the work of Cassimjee et al. (2010), who detected decreases in left cerebellar hemisphere volume upon treatment with a selective-serotonin reuptake inhibitor (SSRI) and consequent reduction in SAD symptoms. This reversion of cerebellar volume in SAD patients correlated with the reduction in SAD symptoms, suggesting possible cerebellar involvement in the manifestation of social anxiety symptoms. Conversely, Talati et al. (2015) found an increase in the volume of the right cerebellar hemisphere volume upon SSRI treatment and associated improvement in SAD symptoms. The discrepancy in the effects of anxiolytic drug treatment on cerebellar volume may arise from different parts of the cerebellum being measured in these two studies: SAD patients have higher resting perfusion of their right cerebellum and decreased resting perfusion in their left cerebellum (Furmark et al., 2002; Warwick et al., 2008). These opposing alterations in the left and right cerebellar hemispheres of SAD patients can potentially explain the discrepancy between the findings of Cassimjee et al. (2010) and Talati et al. (2015). Further, different SSRIs were utilized in these studies: the former employed escitalopram, while the latter used paroxetine. It is possible that the varying mechanisms of action and side effects of these two drugs could have yielded contrasting observations. Nevertheless, it is evident that these studies collectively emphasize that parts of the cerebellum may be differentially engaged during anxiety, a topic discussed at length below.

The cerebellum has also been implicated in other types of anxiety disorders. Connectivity between the cerebellum and the basolateral and centromedial amygdala reportedly is increased in GAD patients, compared to healthy controls (Etkin et al., 2009; Roy et al., 2013). In addition, PD patients exhibit higher levels of cerebellar glucose utilization (Sakai et al., 2005), an effect that is reversed by treatments that reduce their PD symptoms (Prasko et al., 2004). Cerebellar changes are also associated with other mental disorders that are closely related to anxiety disorders, such as Post Traumatic Stress Disorder (Bonne et al., 2003; Ke et al., 2016) and Obsessive-Compulsive Disorder (Xu et al., 2019; Zhang H. et al., 2019; Kashyap et al., 2021). Although these conditions are considered to differ from anxiety disorders in the Diagnostic and Statistical Manual of Mental Disorders, Fifth Edition (DSM-5), their pathophysiology nonetheless can include an anxiety component (American Psychiatric Association, 2022). Taken together, it is likely that the cerebellum plays a role in the manifestation or progression of an array of clinical anxiety disorders.

Although there is progress in clinical efforts to identify the possible involvement of the cerebellum in anxiety disorders, these must be viewed as merely a promising start. Most clinical studies performed to date have relatively small sample groups, a recurring issue, especially in the field of clinical neuroimaging (Szucs and Ioannidis, 2020). Meta analyses and longitudinal studies on the relationship between cerebellar changes and anxiety disorders are sorely lacking. These shortcomings can be attributed to a lack of attention to the cerebellum, specifically in regional cerebral blood flow studies where the cerebellum is often used as a reference region due to assumptions of its irrelevance to psychiatric disorders (Bonne et al., 2003). Moreover, further effort is required to standardize the clinical measures of anxiety and its symptoms, to allow quantifiable research output and better interpretability of data (Baker et al., 2019). These systematic changes in clinical studies, though daunting, will be necessary to advance our understanding of the role of the human cerebellum in anxiety disorders.

Additionally, despite the ability of such clinical studies to implicate the cerebellum in anxiety, they cannot determine whether cerebellar abnormalities produce the pathophysiology of anxiety, or whether these abnormalities result from anxiety. To understand the possible causal relationships between cerebellar changes and anxiety, it is essential to turn to experimentation in animal models of anxiety. When considering such work, it is important to acknowledge that studies of emotions based on animal models must rely on behavioral and physiological proxies of their mental state, which necessarily limits the interpretations that can be derived from such studies. Despite such limitations, animal experiments have greatly expanded our knowledge of the role of the cerebellum in anxiety.

The first suggestion of a possible role for the cerebellum in emotional behavior came from cerebellar lesion experiments (Chambers and Sprague, 1955). The first experimental evidence for cerebellar involvement in anxiety behavior came later: Supple et al. (1987) reported that lesions of the cerebellar vermis of rats (Figure 1A) exhibited greater anxiety behavior, indicated by an aversion to the central area of an open field (thigmotaxis). Numerous subsequent cerebellar lesion studies have supported this finding. For example, Bobée et al. (2000) observed increases in the number of entries and time spent in open arms of the elevated-plus maze (EPM) in young vermis-lesioned rats (Figure 1B), also indicating an anxiolytic effect of vermal lesions. Collectively, such studies established a plausible role for the cerebellum in anxiety behavior.

Another indication of the possible involvement of the cerebellum in anxiety behavior comes from work done on the Lurcher mutant mouse, a cerebellar disease model. Lurcher mice have a mutation in the δ2 glutamatergic subunit, which is a receptor for molecules (cerebellins) involved in synapse formation (Uemura et al., 2010). Mutation of this receptor leads to the complete loss of Purkinje cells in the cerebellar cortex. This, in turn, yields retrograde degeneration of Purkinje cell afferents, namely cerebellar granule cells and molecular layer interneurons, as well as projection neurons from the inferior olive (Vogel et al., 2007). In addition to the motor problems indicated by their name, Lurcher mice display “behavioral disinhibition”, evident as them spending more time in the open arms of the EPM compared to non-mutant mice (Monnier and Lalonde, 1995; Hilber et al., 2004; Lorivel et al., 2010). This apparent reduction in anxiety levels is not caused by motor deficits, because Lurcher mice spend more time in the open arms of the maze and enter the open arms as frequently as, or perhaps more frequently, than control mice. Furthermore, although it is also reported that Lurcher mice have higher stress levels due to the overactivation of their hypothalamic-pituitary-adrenal (HPA) axis, experimentally reducing blood corticosterone levels does not abolish the low anxiety state. This indicates that the anxiolytic behavior displayed by Lurcher mice may arise from disruptions in the neural circuitry involved in regulating anxiety behavior (Lorivel et al., 2010).

Studies on spinocerebellar ataxia (SCA) rodent models also support a cerebellar role in anxiety behavior. SCA models are generally characterized by abnormal expansion of CAG repeats in genes that lead to an ataxic pathology due to neurodegeneration in the cerebellum, often in Purkinje cells (Boy et al., 2009, 2010; Kelp et al., 2013; Asher et al., 2021; Bohne et al., 2022). Anxiety behavior is one of the most studied behaviors amongst the various non-sensorimotor impairments also observed in SCA models, beyond their characteristic ataxic phenotype. Many SCA models exhibit anxiolytic behavior (Boy et al., 2009), although anxiogenic behavior is also often observed (Kelp et al., 2013). Interestingly, in certain SCA models, both abnormalities in anxiety behavior could be observed by varying the expression pattern of the mutant gene, or by simply modifying the aversiveness of the external environment. For instance, Asher et al. (2021) demonstrated that SCA 1 knock-in mice display anxiogenic traits, whereas Purkinje cell-specific SCA 1 knock-in mice exhibit anxiolytic behavior. Additionally, Bohne et al. (2022) showed that in their SCA 6 mouse model, reduced anxiety was observed in low-threat conditions, but these mice displayed enhanced anxious defensive behaviors in contexts with higher aversiveness. This suggests a possible miscalculation of threat probability due to cerebellar pathology. Collectively, the SCA rodent models provide further support for a possible cerebellar involvement in regulating anxiety behavior.

The identification of lobule VII as a functional locus for anxiety behavior aligns with current hypotheses about the topographical organization of the cerebellum. One hypothesis postulates that the posterior cerebellum modulates non-motor behavior (Stoodley and Schmahmann, 2010, 2018), while another suggests that the vermis regulates emotion (Timmann et al., 2010; Klein et al., 2016). The module hypothesis posits that the cerebellar functions are spatially organized in “modules”, areas organized according to their connectivity and outputs (Apps et al., 2018). These modules could extend across multiple lobules (Fujita et al., 2020). Notably, each of the vermal lobules I-IX comprises a multitude of modules, and each lobule could be involved in multiple functions. Lobule VII, located in the posterior vermis, is anatomically connected to functional circuits that are involved in threat, behavioral flexibility, arousal, sensory-motor gating, and cardiorespiratory functions (Fujita et al., 2020), which are vital in processing and producing defensive responses, such as anxiety. It is possible that cerebellar areas in addition to lobule VII could be part of an anxiety module, due to the abundant anatomical connections between the cerebellum and other anxiety-related brain regions. However, we do know that perturbation of parts of lobules IV/V, as well as lobules VI, VIII, Crus I, and II, does not evoke anxiety behavior (Figure 1; Badura et al., 2018; Chao et al., 2021).

The cerebellar vermis is known as the “limbic cerebellum” because of its numerous anatomical connections to the rest of the limbic system (Blatt et al., 2013; Reeber et al., 2013). These include brain regions that are known to mediate anxiety, such as the periaqueductal gray (PAG; Snider and Maiti, 1976) and dorsal raphe nuclei (DRN; Pierce et al., 1977). These connections provide a means for the cerebellum, in particular lobule VII, to communicate with the rest of the brain anxiety circuitry. Additionally, the well-established bidirectional connection between the cerebellum and hypothalamus (Dietrichs, 1984; Dietrichs and Haines, 1986; Supple, 1993; Kondo et al., 1998; Cavdar et al., 2018) could possibly underlie a cerebellar autonomic modulation during periods of fear or anxiety. Neurophysiological and neuroimaging studies support a cerebellar role in regulating visceral activities, such as gastrointestinal motor control and cardiovascular modulation (Zhu et al., 2006; Schutter, 2012; Ito and Nisimaru, 2014). This pathway provides a direct anatomical link between the cerebellum and the HPA axis that is critical for the physiological responses of fear and anxiety, such as heart rate, breathing rate, and stress hormone regulation. Because lobule VII includes a module governing cardiorespiratory functions (Fujita et al., 2020), it is possible that this anatomical connection could also involve lobule VII in anxiety-related physiological responses.

Alteration of MLIs activity in lobule VII bidirectionally controls anxiety behavior. As discussed above, chemogenetic inhibition of these MLIs has an anxiolytic effect (Badura et al., 2018), while their optogenetic activation has an anxiogenic effect (Figure 2B). Additionally, cerebellar cortical GCs also play a role in cerebellar-related anxiety behavior. Rudolph et al. (2020) demonstrated that genetic deletion of the δ GABA_A receptor subunit in GCs, to eliminate GABAergic tonic inhibition and make GCs hyperexcitable, increases anxiety levels.

PCs are the sole output neurons of the cerebellar cortex (Ito, 2006). Because MLIs and GCs provide direct input to PCs, PCs must relay anxiety information from the cerebellar cortex to the DCN. Indeed, mice display anxiogenic behavior when Shank2, an excitatory postsynaptic scaffolding protein, is knocked out of PCs, an effect not observed in global Shank2 knockout mice which instead exhibit autistic-like behaviors (Ha et al., 2016). Shank2 clusters were found to colocalize extensively with the excitatory synaptic vesicular glutamate transporter 1 (VGluT1), which is expressed selectively in parallel fibers as opposed to climbing fibers (Hioki et al., 2003), indicating postsynaptic localization of Shank2 at the PF-PC synapses This is associated with a decrease in the frequency of spontaneous excitatory synaptic currents in PCs, suggesting a net reduction of GC excitatory input into PCs that affects anxiety behavior. Thus, both inhibitory and excitatory inputs into PCs, as well as their effect on PC output, seem to play a significant role in anxiety behavior processing within the cerebellar cortex.

The effects of these cell-level effects on cerebellar cortex circuity are complex. The results above suggest that increasing PC activity—via MLI inhibition and consequent PC disinhibition—will decrease anxiety levels. Conversely, decreasing PC activity—via MLI activation or decreasing excitatory input into PCs—should increase anxiety levels. However, there are other experimental observations that do not fit into this simple scheme. For example, GC hyperexcitability should increase PC activity, yet it increases anxiety instead of the expected decrease (Rudolph et al., 2020). The complex nature of neuronal connectivity and interactions upstream of PCs presumably account for such contradictions. Despite having an overall inhibitory effect on PCs, MLIs also inhibit each other via chemical synapses and excite each other through electrical synapses (Kim et al., 2014; Rieubland et al., 2014). Moreover, while having an overall excitatory effect on PCs, GCs also excite interneurons—MLIs as well as Golgi cells—which in turn would inhibit GCs and decrease PC output (Dieudonne, 1998; Dugué et al., 2005; Kim and Augustine, 2021). Therefore, manipulation of GC and MLI activity will not simply translate to changes in PC activity.

In addition, PCs also directly receive or are indirectly modulated by an array of afferents originating from outside the cerebellar cortex, including mossy fibers, climbing fibers, and various neuromodulator fibers (Bloedel, 1973; Longley et al., 2021; Canton-Josh et al., 2022). Different PC firing patterns—simple spikes evoked by mossy fiber input to parallel fibers or complex spikes resulting from climbing fiber excitation of PCs—can encode different functions (Popa et al., 2019). Because mossy fibers innervate granule cells (Ito, 2006), which are known to modulate cerebellar-related anxiety behavior (as discussed above), it is likely that mossy fiber input into the cerebellum could also be involved in cerebellar anxiety. It remains unknown whether climbing fiber inputs onto PCs could be involved in relaying anxiety information into the cerebellum.

Taken together, it is likely that complex cerebellar circuit effects upstream of PCs are involved in producing the anxiety behavioral response. How these upstream effects are relayed onto the Purkinje cells, and ultimately to the DCN, and thence translated into anxiety behavior, remains unknown. Direct experimental control of PC activity, to bypass upstream neurons in the cerebellar cortical circuitry, would help to clarify the role of PCs and the entire cerebellar cortex in anxiety.