Dystonia presents with phenotypic and etiologic heterogeneity. Considered the third most common movement disorder, “dystonia” does not comprise a single disease or symptom, but rather describes an array of disorders sharing overlapping behavioral outcomes. While different forms of dystonia express unique etiologies, substantial evidence implicates the cerebellum as a major node in the underlying network disruptions [1–4]. Dysfunction of Purkinje cells and the cerebellar nuclei, the primary outputs of the cerebellar cortex and cerebellum, respectively, are implicated in both hereditary and idiopathic forms of dystonia [4–7]. Importantly, while cerebellar dysfunction is sufficient to induce dystonia in animal models, therapies addressing cerebellar dysfunction can modulate dystonia and reduce motor symptom severity. One such therapy, cerebellar deep brain stimulation (DBS), has been used to effectively reduce motor symptom severity in both mouse models [4] and human patients [8, 9], further suggesting a critical cerebellar involvement in the etiology of dystonia. However, nonmotor behaviors are also relevant to the cerebellum and to dystonia.

Along with its known role in regulating motor function, increasing evidence shows that the cerebellum also serves as a key brain region in the control of a variety of nonmotor behaviors such as cognitive and emotional processing [10], associative learning [11], and reward expectation [11, 12]. Recent work also suggests that the cerebellum may play a role in sleep-related behaviors. Purkinje cells and cerebellar nuclei neurons have been found to display sleep-dependent activity, increasing their firing during NREM (non-rapid eye movement) sleep [13–16]. In addition, lesioning of the cerebellar vermis has been shown to impair sleep, suggesting that normal cerebellar circuitry and activity are important for maintaining sleep rhythms [17, 18]. Numerous studies have shown that cerebellar disruptions form the basis for the many comorbidities of motor disorders [19–22]. However, the role of the cerebellum and its circuit components in sleep regulation have not been studied thoroughly in dystonia. It is known that dystonic patients demonstrate increased sleep latency and REM (rapid eye movement) latency, and in some cases, persistent involuntary muscle contractions during sleep [23–25]. It is also noted that therapeutics which successfully alleviate the motor symptoms of dystonia appear to have little to no effect on the sleep disruptions [26]. The importance of addressing sleep dysfunction is becoming increasingly apparent in society, as sleep disruptions can significantly impact quality of life, driving many subsequent comorbidities [27, 28]. Disrupted sleep is also associated with impaired motor learning/function [29–31], as synaptic activity is normalized during sleep [32]. Together, the mounting evidence inspires a compelling model in which sleep and motor dysfunction in dystonia comprise two halves of a self-propelling cycle. It is possible that cerebellar dysfunction drives both the more commonly appreciated motor abnormalities and the nonmotor sleep disruptions, although how they emerge needs to be systematically resolved.

It remains unclear whether dystonic motor dysfunctions persist during all stages of sleep, and consistently across different manifestations of the disease; recent evidence from a survey of cervical dystonia patients suggests that, in cases of idiopathic cervical dystonia, it does [24]. The lack of clarity on which factors (motor dysfunction and/or cerebellar dysfunction) drive sleep dysfunction in dystonia highlights the significance of this knowledge gap. To investigate the relationship between cerebellar dysfunction, motor dysfunction, and sleep, here we used a constitutively active Cre/lox-p system to drive the deletion of the Vglut2 gene in afferent neurons that project excitatory fibers that ultimately communicate with the Purkinje cells. Vglut2 was deleted using the Ptf1a and Pdx1 gene regulatory elements to spatially drive Cre expression: the resulting mice had the genotypes Ptf1a ^Cre ;Vglut2 ^fx/fx and Pdx1 ^Cre ;Vglut2 ^fx/fx . The Ptf1a and Pdx1 genes are expressed in the excitatory neurons of the inferior olive, a region of the brainstem, which projects afferent fibers to the cerebellar cortex and terminate as excitatory climbing fibers [4, 33]. However, Ptf1a expression occurs in a wider distribution across the inferior olive relative to Pdx1, and Pdx1 is also expressed in mossy fiber afferent neurons (Lackey et al., 2023 in preparation). Therefore, the silencing of excitatory climbing fiber synapses occurs with differential coverage in mice with Ptf1a- versus Pdx1-driven Cre expression. Only Ptf1a ^Cre ;Vglut2 ^fx/fx mice present with severe motor dysfunction involving twisting of the torso and hyperextensions of the back and limbs [4], while Pdx1 ^Cre ;Vglut2 ^fx/fx adult mice show only subtle dystonic behaviors [34]. Thus, the use of both mouse models allows us to deliberately silence excitatory olivocerebellar synapses while also providing us with an opportunity to query varying severities of cerebellar dysfunction.

In this work, we report that, despite their different dystonia-related motor severities, Pdx1 ^Cre ;Vglut2 ^fx/fx and Ptf1a ^Cre ;Vglut2 ^fx/fx mice display similarly impaired sleep physiology and circadian rhythms. Both mutants display highly disrupted sleep, spending greater time awake and in NREM sleep at the expense of REM sleep. Furthermore, we found that both the Pdx1 ^Cre ;Vglut2 ^fx/fx and Ptf1a ^Cre ;Vglut2 ^fx/fx mutant mice display increased latency to reach REM, which is similar to what is observed in human patients with dystonia [23]. Intriguingly, only mice with overt dystonic motor behaviors (Ptf1a ^Cre ;Vglut2 ^fx/fx ) show differences in ECoG spectral power frequency, particularly in the latter half of the time spent asleep. We also found that circadian activity rhythms remain unchanged across all groups of mice, and that the circadian “master clock” remains ostensibly unaffected by our circuit manipulation. Our work demonstrates that aberrant cerebellar activity dually disrupts motor function and sleep, paving the way for improved future therapeutics that may be able to simultaneously address both motor and sleep dysfunction in the context of motor disease.

We genetically dissected the interaction between sleep impairments and cerebellar-initiated motor impairments in two mouse models of dystonia. Altogether, the results from this work provide insight into the unique sleep, ECoG, and EMG disturbances observed in our mouse models of cerebellar circuit dysfunction (Figure 6A). We found that sleep impairments, a common nonmotor symptom in human dystonia, occur in Ptf1a ^Cre ;Vglut2 ^fx/fx and Pdx1 ^Cre ;Vglut2 ^fx/fx mouse models of cerebellar miswiring, with and without severe dystonia-related motor dysfunctions. We show that both groups of mutant mice display an increase in the length of wake bouts, increased NREM and more frequent NREM bouts, and decreased REM and less frequent REM bouts (Figure 6A). While existing studies on sleep quality in dystonia patients is limited, our results are striking in that they reflect patterns of sleep deficits observed in dystonia patients [23–25]. We also highlight our finding that motor activity in Ptf1a ^Cre ;Vglut2 ^fx/fx mice remains elevated in all stages of sleep, even during REM. This result is particularly intriguing. Existing studies are split, some suggesting that abnormal muscle activity in dystonic patients disappears during sleep [23] while other indicate that it might persist during sleep [24]. On the one hand, it is possible that our results indicate that dysfunction of the mechanisms involved in synaptic renormalization are affected in dystonia, which are believed to occur during sleep and mediate muscle recovery and atonia during sleep [32, 49, 50]. On the other hand, as a reconciling interpretation, the work we presented here could also suggest that cerebellar dysfunction, in the presence or absence of dystonic motor dysfunction, is sufficient to drive nonmotor impairments in sleep in mouse models of dystonia (Figure 6B). Existing knowledge suggests that the cerebellum and its circuit components (namely, the Purkinje cells and cerebellar nuclei neurons) are a key node in dystonia [4, 51]. Therefore, our results may point to the cerebellum as a central dystonia locus, which could help to anchor future studies on the development of therapies that can address motor and nonmotor (sleep) dysfunction in dystonia.

Evidence from case-control studies in patients with cervical dystonia suggests that human patients with dystonia exhibit distinct abnormalities in their circadian rhythms, including fatigue and excessive daytime sleepiness [25, 26, 52]. We sought to understand if the clinical symptoms of dystonia that are relevant to sleep were observed in our two mouse models that exhibit cerebellar dysfunction, with and without motor deficits. Our results show that both Pdx1 ^Cre ;Vglut2 ^fx/fx and Ptf1a ^Cre ;Vglut2 ^fx/fx mice display normal circadian timing of behavior (Figures 2C–E,H), suggesting that cerebellar dysfunction and motor deficits do not impact overall circadian behavior. Although these data are intriguing, given the substantial sleep impairments experienced by these mice, this finding is in line with work in mice with cerebellar ataxia, which also show normal circadian wheel-running behavior [41]. To make sure that our genetic manipulation in each model did not disrupt activity in the SCN master clock, we verified in our mice existing work, which states that the SCN is 95% GABAergic in its neuronal identity [43]. In accordance with our in situ hybridization results (Supplementary Figure S3), this may indicate that while the cerebellum is involved in the regulation of sleep, its role in circadian timekeeping is limited, at least in the current context. Indeed, while numerous projections exist between the cerebellum and the major circadian centers of the brain, including the hypothalamus, locus coeruleus, and pedunculopontine nucleus, direct projections between the cerebellum and the SCN master clock are lacking [53, 54]. It is possible then that cerebellar access to circadian processes is tightly regulated and restricted to sleep rather than overall activity rhythms. In this case, the fatigue and excessive daytime sleepiness experienced by patients with dystonia may be attributed to lack of sleep rather than aberrant circadian timekeeping.

While dystonia is commonly considered a network disorder in humans [1], our genetic manipulation attempts to recreate the circuit-wide defects through a mechanism that initiates the dystonia by precisely blocking glutamatergic signaling in the cerebellum. Although both Pdx1 and Ptf1a are expressed in several brain regions, including in the hypothalamus, neither Pdx1 nor Ptf1a is expressed in the SCN, the circadian master clock of the brain [33, 55]. Furthermore, as discussed before, over 95% of the cells in the suprachiasmatic nucleus are GABAergic [43], which further suggests that our genetic manipulation does not extend to directly affect the master clock. Indeed, our analysis of Vglut2 mRNA expression showed that Vglut2 expression in the suprachiasmatic nucleus is sparse (Supplementary Figure S3). As suggested by existing work, these regions instead heavily express Vgat, indicative of using primarily GABAergic signaling. Therefore, it is possible that human dystonia patients do experience malfunctioning circadian rhythmicity, but our model is unable to capture this specific aspect of heterogeneity in dystonia network dysfunction.

While overall circadian rhythm patterns remained unchanged in our mutant mice, we did observe a difference in siesta time for Ptf1a ^Cre ;Vglut2 ^fx/fx mice. The siesta period is a brief bout of sleep during the active period [42]; it represents an important output of the circadian systems’ sleep regulation process [42, 56]. Thus, it serves as an additional marker of typical circadian rhythmicity. We note that the siesta period in the Ptf1a ^Cre ;Vglut2 ^fx/fx mice does appear more pronounced and is significantly longer by 7–9 min (Figure 2I). However, given the low overall activity profile of the Ptf1a ^Cre ;Vglut2 ^fx/fx mice, it is difficult to determine whether this increase in siesta time is circadian in origin, or if it arises as a result of the decreased overall activity, making the accurate calculation of siesta onset/offset difficult. Recent work does suggest that daily timing of the siesta is under control of the SCN [57], which we have determined to be spared from our genetic manipulation. Together, these data further support the finding that circadian rhythmicity is largely unaffected in Ptf1a ^Cre ;Vglut2 ^fx/fx mice and also in the Pdx1 ^Cre ;Vglut2 ^fx/fx model of dystonia.

Human studies suggest that alterations in sleep efficiency and sleep latency are prevalent in dystonia patients [19, 23, 24, 26, 52, 53]. We found that changes in sleep/wake dynamics are of particular interest, as they begin to explain the specific ways in which sleep deficits arise in the disease. Our results further suggest that REM disruption may be one of the primary sleep deficits encountered by our mouse models. We found not only that our mutants spend less time in REM, but that this impairment is complemented by an increase in both wake and NREM sleep (Figures 3G–I). Though mice sleep in bouts of 120–180 s per full sleep cycle [58] and not in long consolidated bouts like humans, they do follow similar sleep stage patterns (Figure 3E). Despite REM representing the lightest sleep stage, it is typically preceded by NREM sleep [59]. Therefore, the increase in NREM sleep combined with decrease in REM sleep suggests that the sleep deficits in our mice specifically result from involuntary waking during NREM sleep. This is further evidenced by our EMG results for the Ptf1a ^Cre ;Vglut2 ^fx/fx mice, which display elevated cervical EMG power in all sleep states (Supplementary Figure S1). As mice must pass through NREM again before entering REM as they start a new sleep cycle, this prolongs the time spent in NREM while decreasing the time spent in REM. Further work needs to be conducted in order to connect our findings in mice to human patients with dystonia, as an equivalent result in humans could potentially explain the reported symptoms of daytime fatigue. Even if the total sleep time is similar between dystonic and non-dystonic patients, the quality of sleep is still being affected, as proportions of NREM versus REM during the sleeping phase are equally as important as overall time spent asleep versus awake [59].

Given the cerebellums’ known projections to/from a variety of cortical regions involved not only in sleep regulation, but also regulation of specific sleep stages (NREM and REM) [53, 54, 60], it was not unsurprising to find that sleep-stage specific deficits exist in both Pdx1 ^Cre ;Vglut2 ^fx/fx and Ptf1a ^Cre ;Vglut2 ^fx/fx mice. Other groups have found that dystonia patients [23, 26] and mouse models of motor dysfunction [61] present with sleep-stage specific deficits. Our findings of increased average wake bout length (Figure 4D), increased number of NREM bouts (Figure 4G) and decreased number of REM bouts (Figure 4E) specifically highlight and further reinforce our main findings of impairments in overall sleep stage timing. These results highlight the existence of significant REM-related sleep deficits. This is further reflected in our results of increased latency to reach REM for both Pdx1 ^Cre ;Vglut2 ^fx/fx and Ptf1a ^Cre ;Vglut2 ^fx/fx mice (Figure 4I). For Ptf1a ^Cre ;Vglut2 ^fx/fx mice, the motor dysfunction may partially explain this result. REM-related sleep impairments are typically accompanied by some form of motor dysfunction [24, 26, 61], as the canonical mechanisms of muscle atonia during REM are disrupted [62]. However, since our Pdx1 ^Cre ;Vglut2 ^fx/fx mouse model does not display overt motor dysfunction, but still displays the same wake, NREM, and particularly REM-related deficits, motor dysfunction may not be the sole culprit for impaired sleep. In this case, cerebellar dysfunction may also be to blame. Indeed, the cerebellum itself and many regions receiving direct cerebellar innervation are known to be involved in sleep regulation or control sleep-dependent behaviors, particularly REM regulation [53]. The locus coeruleus regulates NREM/REM intensity [63] while sending and receiving dense projections to cerebellar Purkinje cells and the cerebellar nuclei [64–66]. The pedunculopontine nucleus is a known regulator of REM sleep [67] and also sends/receives inputs between cerebellum and the basal ganglia [68]. Therefore, it is possible that the cerebellar malfunctions in the Pdx1 ^Cre ;Vglut2 ^fx/fx and Ptf1a ^Cre ;Vglut2 ^fx/fx mice directly and indirectly influence REM latency through intermediary regions such as the pedunculopontine nucleus or the locus coeruleus, or even other regions, all of which play a role in the regulation of REM sleep and receive/send direct innervation from the cerebellum [54, 60, 63]. The circuit pathways mediating the direct versus indirect effects on sleep were not resolved in the current work. Ultimately, the impaired sleep dynamics further reinforces our results, which suggest that Pdx1 ^Cre ;Vglut2 ^fx/fx and Ptf1a ^Cre ;Vglut2 ^fx/fx mice experience interrupted sleep cycles, cutting REM sleep short or missing it entirely and re-starting subsequent sleep cycles from NREM sleep.

Our ECoG spectral activity results are of particular interest from a mechanistic view, as they may not only explain the factors underlying sleep deficits in Pdx1 ^Cre ;Vglut2 ^fx/fx and Ptf1a ^Cre ;Vglut2 ^fx/fx mice but may also serve as additional “biomarkers” that differentiate each group based on their degree of cerebellar dysfunction. For instance, we observed an increase in delta power for Ptf1a ^Cre ;Vglut2 ^fx/fx but not Pdx1 ^Cre ;Vglut2 ^fx/fx mice, which occurs predominantly in the latter stages of recording (Figures 5C, D). The increase for Ptf1a ^Cre ;Vglut2 ^fx/fx is in agreement with recent work suggesting that higher delta power is associated with arousal and sleep impairment, particularly in the context of obstructive sleep apnea which involves both motor dysfunction and cerebellar dysfunction [69]. However, as delta power in Pdx1 ^Cre ;Vglut2 ^fx/fx is also elevated relative to control mice and is not significantly different from Ptf1a ^Cre ;Vglut2 ^fx/fx , there stands the possibility that the lack of significance stems from the “intermediate” phenotype of Pdx1 ^Cre ;Vglut2 ^fx/fx mice. We have shown that Pdx1 ^Cre ;Vglut2 ^fx/fx mice lack overt motor dysfunction (Figures 1H, I; Supplementary Figure S2, Supplementary Video S1), and that the extent of Vglut2 deletion in the cerebellar cortex, at least with respect to the climbing fibers, is less extensive relative to the Ptf1a ^Cre ;Vglut2 ^fx/fx mice (Supplementary Figure S1). Therefore, if we consider each mutant group as a model for different cerebellar/motor disorders of varying intensity, we expect to see such differences in spectral power despite similar sleep deficits. In this case, spectral differences may represent “biomarkers” of disease severity. It is known that changes in delta power can differ across individuals with different diseases even if all individuals display poor sleep [47]; it is possible then that Pdx1 ^Cre ;Vglut2 ^fx/fx and Ptf1a ^Cre ;Vglut2 ^fx/fx mice, even with an overlap in the genetic manipulations, indeed represent different manifestations of dystonic motor disease. Ultimately, this is evident with the difference in observed dystonic motor phenotypes between the two groups. It should be noted though, the measurement of sleep-related spectral difference between the Pdx1 ^Cre ;Vglut2 ^fx/fx and Ptf1a ^Cre ;Vglut2 ^fx/fx mice could still intersect with the induced alterations in the motor program, as movement patterns are known to impact ECoG spectral activity [70].

Additional patterns of significantly different spectral power for Ptf1a ^Cre ;Vglut2 ^fx/fx but not Pdx1 ^Cre ;Vglut2 ^fx/fx mice are seen for beta power, indicative of alert wakefulness (Figures 5I, J). While beta power is typically increased in patients with primary insomnia, previous work has shown that decreased beta activity is also associated with poor sleep quality, particularly in patients with obstructive sleep apneas [69]. The relationship between the cerebellum and breathing is well-established and may provide a fruitful avenue for further research in the context of dystonia. The cerebellum is known to be involved in both the rhythmicity of breathing and in regulating air hunger [71]; both mechanisms are known to play a role in obstructive sleep apneas [19, 72]. It is possible that Ptf1a ^Cre ;Vglut2 ^fx/fx mice, with overt motor dysfunction, have some degree of sleep apnea behavior, which could contribute to their observed sleep impairment. Interestingly, cortical gamma power was only significantly changed during the middle of the recording period, for both the Pdx1 ^Cre ;Vglut2 ^fx/fx and the Ptf1a ^Cre ;Vglut2 ^fx/fx mutant mice (Figure 5L). While gamma oscillations are typically associated with working memory and attention, human and mouse EEG/ECoG studies have found that gamma oscillations occur spontaneously during REM and NREM sleep [73, 74]. This may explain why the overall gamma power is not significantly lower for either the Pdx1 ^Cre ;Vglut2 ^fx/fx or Ptf1a ^Cre ;Vglut2 ^fx/fx mice, yet it does reach the threshold for significance during “mid-recording”. Ptf1a ^Cre ;Vglut2 ^fx/fx and Pdx1 ^Cre ;Vglut2 ^fx/fx mice exhibit an increase in NREM and a decrease in REM sleep, and gamma oscillations spontaneously occur during both stages; changes in gamma activity may be effectively “canceled out” due to the opposing NREM and REM dynamics. While these observed changes in spectral frequency oscillations uncover some of the potential mechanisms driving the observed changes in sleep in our Ptf1a ^Cre ;Vglut2 ^fx/fx and Pdx1 ^Cre ;Vglut2 ^fx/fx mice, we note that attributing sleep disruption to specific directional changes in any one frequency band is difficult. Both positive and negative changes in average power, in any frequency band, can be associated with various disease states, and notably with disordered sleep [46, 47]. Therefore, here, we highlight the presence of a change in spectral frequency power as an indicator of fractured sleep homeostasis in our mouse circuit models of dystonia, without differentiating the specific directionality of the change in spectral frequency power.

Our findings build upon existing evidence from both human patients and mouse models of motor disease demonstrating that sleep impairments are a common nonmotor symptom in dystonia. Previous work has been unable to distinguish between dystonia-dependent versus independent sleep dysfunction, particularly in the context of dystonic motor dysfunction. Importantly, our results suggest a model in which cerebellar dysfunction alone (Figure 6B), without overt dystonic motor phenotypes, can drive sleep deficits. This may be an indication of a broader set of network dysfunctions in dystonia, with the cerebellum located at the center of multiple disease symptoms.