Dystonia is the third most common movement disorder in the United States, manifesting as abnormal, disabling postures in children, adolescents, and adults [1]. Dystonia is classified as either isolated (“primary”) or acquired (“secondary”) and can be inherited (e.g., DYT1) or caused by injury (e.g., stroke). A dominantly inherited mutation ( ∆ GAG or ∆ E) in the ubiquitously expressed TOR1A gene causes DYT1 dystonia, the most prevalent form of inherited dystonia. TorsinA, the protein encoded by TOR1A, is highly expressed in postnatal brain development, with prominent levels detected between 8 and 35 weeks in the cerebellum, midbrain, basal ganglia, and hippocampus [2]. TorsinA continues to be expressed in the human striatum from 3 to 7 years of age, corresponding to a time of increased myelination, synaptic densities, and structural maturation [3]. Due to the location and timing of its expression, torsinA is hypothesized to play an important role in synaptogenesis [2, 4, 5]. However, incomplete disease penetrance (∼30%) and a variable age-of-onset (3–70 years) complicates studying the consequences of torsinA loss in DYT1 dystonia [6, 7].

One way to overcome this challenge is by identifying cell types directly involved in DYT1 dystonia pathophysiology. The oral medications prescribed to patients following initial diagnoses—anticholinergics, benzodiazepines, levodopa, and antiepileptics—broadly alter cholinergic, GABAergic, dopaminergic, and glutamatergic cell activity and hint at possible contributors [8]. Trihexyphenidyl, an anticholinergic drug, is among the most effective treatments for DYT1 dystonia, especially in pediatric patients who exhibit greater tolerance to higher doses [8, 9]. Considered in conjunction with human neuroimaging studies that show an age-dependent disruption of striatal vesicular acetylcholine transporter (VAChT) expression in DYT1 patients [10], cholinergic neurons appear important early in dystonia pathophysiology. Whether these two phenotypes are physiologically connected is unclear but altered VAChT expression and cholinergic hyperactivity are not mutually exclusive given the functional redundancy of vesicular proteins in striatal cholinergic cells and co-expression of neurotransmitters [11–13]. Comprising 1%–3% of the cell population, cholinergic interneurons use acetylcholine (or glutamate) to modulate striatal circuit functions and motor control. Acetylcholine is loaded into pre-synaptic vesicles by VAChT and VGLUT3 and its release counterbalances the effects of dopamine on neuronal excitability and plasticity [11, 14]. During striatal development, acetylcholine modulates dopamine release to direct medium spiny neuron maturation and glutamatergic receptor expression [15, 16]. Early dopamine loss stunts medium spiny neuron growth, increases excitability, and impairs behavior [15, 17]. While the direct impact of cholinergic dysfunction on behavior has not been investigated in the juvenile striatum, it is likely that motor functions are compromised because cholinergic maturation overlaps with the development of locomotor-related movements and their loss or misfiring contributes to the motor deficits found in Parkinson’s disease, Huntington’s disease, and Tourette Syndrome patients [18–20].

A more thorough understanding of striatal cholinergic interneuron development may inform DYT1 origins and targeted therapies. In this mini review, I focus on the neurodevelopmental disorder of DYT1 dystonia, where I examine the evidence supporting cholinergic dysfunction in its pathophysiology. I then discuss what is known about striatal cholinergic interneuron development and compare its timeline to that of striatal maturation, with the intention of highlighting periods of motor circuit vulnerability applicable to dystonia. I will end considering maladaptation and implications of abnormal cholinergic cell development on DYT1 penetrance and treatments.

Human patients and clinical observations have historically been used to characterize DYT1 dystonia etiology [21, 22]. Now, preclinical models support more in-depth investigations into DYT1 pathophysiology, including the identification of developmental alterations. Genetically engineered DYT1 mice generated through overexpressing human ∆ E-torsinA (“hMT”), conditionally deleting torsinA from forebrain GABAergic and cholinergic neurons (“Dlx-CKO”) or the entire central nervous system (“N-SKI”), and knocking in inducible (“Tor1a ^i-ΔGAG/+ ”) or non-inducible (“Tor1a ^ΔGAG/+ ”) ∆ GAG-torsinA mutations are a few models that have provided opportunities to probe dystonia circuitry during development [23–27]. While Dlx-CKO, N-SKI, and Tor1a ^i-ΔGAG/+ models display early-onset motor and postural deficits at postnatal (P) days 15 and P21, Tor1a ^ΔGAG/+ reports synaptic plasticity defects from P15 to P26 [24, 26–28]. P14 hMT mice exhibit imbalanced striatal dopaminergic and cholinergic signaling and potentially behavioral deficits, although quantification of the visualized motor changes are needed for corroboration [23, 29]. Complementary studies in cell culture furthermore identified perinuclear ubiquitin accumulation, nuclear pore defects, and nuclear membrane abnormalities in developing torsinA-deficient (Dlx-CKO, Tor1a ^−/− , Tor1a ^ΔGAG/ΔGAG ) neurons [25–27, 30–32]. Interestingly, nuclear pore clustering and membrane abnormalities, but not perinuclear aggregation, persist with neuronal maturation [25, 30, 32]. The contributions each molecular event makes to DYT1 pathogenesis is unclear but are hypothesized to interfere with synaptic efficacy through affecting the organization and stabilization of synaptic proteins or the maturation of spines and dendritic trees [33–35]. Anatomical and behavioral analyses support neurodevelopmental synaptic dysfunction, with morphological and plasticity changes being observed in torsinA-deficient medium spiny neurons as early as P14 and behavioral alterations resolving with torsinA restoration by P21 [28, 36, 37]. Understanding the developmental events and key players in circuit formation may be vital to advancing our knowledge and treatment of DYT1 dystonia.

The plasticity alterations—enhanced long-term potentiation (LTP) and absent long-term depression (LTD)—previously discovered in postnatal DYT1 mice are restricted to the striatum, occurring specifically at corticostriatal synapses [28, 38]. Accompanying impaired plasticity is reduced dopamine binding to D1 and D2 receptors, which is critical for LTP and LTD induction [39, 40]. Not all DYT1 model mice exhibit impaired D1 and D2 receptor activity [41], lending support to the idea that input from other cells, such as striatal cholinergic interneurons, may be responsible for altering plasticity. The striatal cholinergic interneuron pathology that has been reported in DYT1 models with altered plasticity are cell degeneration, increased firing, and enhanced acetylcholine tone [29, 38, 42–47]. To determine whether hypercholinergic phenotypes contribute to DYT1 pathophysiology, antimuscarinics were administered before measuring plasticity and behavior [24, 38, 48]. Drugs, such as trihexyphenidyl, subdue cholinergic activity, improved motor symptoms, and normalized medium spiny neuron LTP and LTD in different DYT1 mouse models [24, 49]. These data call attention to the important role cholinergic neurons may play in DYT1 pathogenesis and unveil possible pharmacological mechanisms (i.e., plasticity restoration) for antimuscarinics in dystonia patients [50, 51].

From the above pharmacological and electrophysiological studies, it remains unclear whether the functional alterations found in striatal cholinergic cells are a cause of dystonia or consequence of torsinA mutations. Chemogenetic and transgenic mouse studies more directly test cholinergic interneuron involvement. Using Designer Receptors Exclusively Activated by Designer Drugs (DREADDs) under the choline acetyltransferase (ChAT) promoter, Gemperli et al., (2022) found that chronic striatal cholinergic excitation incited dystonic behavior [52]. Abnormal motor behavior and responses to dopamine (D2/D3) receptor activation in the striatum also developed after deleting exons 3-4 of torsinA only in cholinergic cells using mice expressing Cre recombinase under the ChAT promoter [53]. Together, these cell-specific manipulations hint at a causal role of cholinergic interneurons in DYT1 dystonia, as do early intervention studies. Given that trihexyphenidyl is effective in a subset of children and that altered synaptic plasticity is found in some mouse models within 3 weeks of life, striatal cholinergic interneurons may contribute to DYT1 behavioral onset [8, 23]. Evidence suggesting early cholinergic involvement include: 1) striatal cholinergic interneurons degenerate just prior to motor symptoms manifesting in juvenile Dlx-CKO mice [24], 2) developing striatal Dlx-CKO cholinergic interneurons exhibit altered synaptic gene expression [37], and 3) striatal cholinergic interneurons demonstrate abnormal excitatory dopamine receptor responses in juvenile hMT mice [29]. Interestingly, these findings occur at P14-15, ages corresponding to neural and behavioral maturation in mice and approximately early adolescence and the average age-of-DYT1-onset in humans [8, 54, 55]. While fundamental developmental events in the striatum—including, rostral-to-caudal organization, cell identity, and striosome and matrix compartmentalization—have been extensively characterized from embryogenesis through P14 [56–60], the developmental trajectories of individual cholinergic neurons are less clear. One reason being the presumption of a structurally normal nervous system in DYT1 dystonia had stalled efforts to track anatomic changes [61–65]. Now, with advances in neuroimaging, microstructural changes throughout the adult nervous system have been identified [66, 67]. Detailed morphometric analyses in mature DYT1 tissue reveal enlarged striatal cholinergic cell bodies, postulated to result from altered connectivity [24, 68]. Repeating these experiments in developing DYT1 tissue will contribute to making a timeline of changes in dystonia.

Nearly all striatal cholinergic neurons are Gbx2-expressing cells born in the medial ganglionic eminence and preoptic area, that then undergo tangential migration [69]. A small proportion of striatal cholinergic neurons co-expressing Nkx2.1 and Zic4 are born in and have migrated from the septum [70]. Cholinergic interneurons are among the first striatal cell types to be born, between embryonic days (E) 12 and E17 in rat (E11.5-E14.5 in mouse; [57, 58]). In contrast, striatal medium spiny neurons are born as late as P0-P5 in rat [71–74]. By the end of the first postnatal week (P0-P6), striatal cholinergic neurons are postmitotic, express acetylcholinesterase (AChE), release acetylcholine, and start maturing morphologically [69–71, 75–78], raising the possibility of these cells playing a crucial role in circuit assembly. Providing further support for this hypothesis are the concurrent timelines for striatal cholinergic synaptogenesis, morphogenesis, and locomotor maturation [54, 79]. Excitatory afferents enter the striatum starting at E12, then significantly increase in density through the second postnatal week (Figures 1A, B; [80, 82]). Intrastriatal GABAergic connections form later, starting at P0-P4, with a density that remains constant throughout postnatal development (Figures 1A, B; [80]). The fluctuation in density of striatal excitatory inputs coincides cholinergic morphogenesis and developing locomotor behavior (Figures 1C, D). In rodents, excitatory innervation onto striatal cholinergic interneurons increases as locomotion increases and the dendritic arbor grows in size and complexity (Figures 1C, D; P7-13 [81, 83]) More specifically, movement begins at P8, when all striatal cell types are present and cell densities are comparable to that of adults [72]. Although circuit components are present and operational by P8, neuronal maturation is incomplete and locomotion is unrefined (Figures 1B, C; [54, 72, 74]). At P15, this changes when medium spiny and cholinergic neuronal counts and morphology become adult-like [72, 75, 81, 84]. Simultaneous excitatory synapse pruning and dendritic remodeling in the third postnatal week (P14-P20) overlaps with the acquisition of activity-dependent synaptic plasticity properties and the transformation of gait into that of adult rodents (Figures 1C, D; [79, 85, 86]). During the fourth postnatal week (P21-P27), gait maturation is complete, while striatal cholinergic morphology and connectivity undergoes fine-tuning (Figures 1C, D; [54, 85, 86]).

The final striatal circuit is achieved in mice approximately 1 month after birth (P28+), when medium spiny neuron spine density and excitability normalizes and cholinergic interneurons lose their globose shape, larger somata, and complex branching in favor of irregular shapes, fewer bifurcations, and dendrites spanning up to 1 mm (Figure 1D; [75, 79, 81, 87]). Cortical, thalamic, and dopaminergic afferents are also specifically organized on mature cholinergic interneurons. Corticostriatal synapses are the sparsest, and predominantly form on the distal dendrites of cholinergic interneurons [87–90]. In contrast, thalamostriatal synapses are the most abundant, concentrated on both the proximal dendrites and somata [87, 88, 90]. Dopaminergic afferents activate D1/D5 and D2 receptors, which are similarly localized throughout the cell, dispersed across the cholinergic somata and neurites [91–94]. The spatial distribution and quantity of each excitatory synapse type is not known for striatal cholinergic interneurons during postnatal development, but likely differs from adults, when dendritic remodeling and synaptic pruning is complete. Characterizing synaptic inputs onto striatal cholinergic interneurons throughout postnatal development may be important for understanding the pathogenesis of neurodevelopmental disorders, including DYT1 dystonia, especially given that torsinA is intensely expressed in striatal cholinergic interneurons from P14 to P21 and its functions have been implicated in the secretory pathway and synaptogenesis [37, 95, 96].

Altogether, several developmental processes must be coordinated within the mouse striatum during the first 4 weeks of life to establish a functional, mature circuit and execute movement. Whether striatal cholinergic synaptogenesis and morphogenesis directs behavioral changes remains to be elucidated, but their embryonic or conditional loss from the forebrain supports this prospect [97, 98]. Additionally, some DYT1 models (e.g., [24, 26, 27]) develop motor symptoms (P14-P21) as cholinergic neuron firing and morphology matures (Figures 1C,D; [79, 81]). During normal striatal development, cholinergic interneurons fire with increased frequency, pacemaking, and less irregularity from P14 to P18, as excitatory synaptic input and dendritic branching becomes refined [81, 99]. Prolonged synaptogenesis or impaired pruning resulting in aberrant excitatory connections onto striatal cholinergic interneurons may explain why some DYT1 models have enhanced acetylcholine tone or increased striatal cholinergic neuron excitability (e.g., [24, 29, 42, 44, 46]) and are responsive to anticholinergic interventions [24, 49]. Not only are cholinergic interneurons particularly sensitive to changes in excitatory input [88, 100, 101], but they also start making connections themselves, onto neighboring cells from P15 to P21, as evidenced by a drastic increase in background ChAT staining [75]. Therefore, increased cholinergic excitation would have downstream effects on medium spiny neuron function, with one effect being increased synchronization [102]. Human dystonia patients exhibit neuronal synchronization in the globus pallidus internus (GPi) due to burst-firing propagating from the striatum [102–106]. Burst-firing and neuronal synchronization is presumed to underlie dystonic postures because behavior improves when either are alleviated through GPi-targeted deep brain stimulation (GPi-DBS) or anticholinergics [106–109].

Here, I considered the development of striatal cholinergic interneurons and their possible role in DYT1 pathogenesis. Not only is striatal cholinergic interneuron dysfunction detected in several DYT1 models, but cholinergic receptor expression is altered in patients and anticholinergic medications are among the leading treatment options to manage their motor symptoms [8, 10, 23, 24, 43, 44, 110]. However, not knowing the role of cholinergic interneurons in dystonia limits our understanding of antimuscarinic-based therapies and their efficacy in DYT1 patients. Given that DYT1 has neurodevelopmental origins and cholinergic interneurons are among the first striatal cell type to be born and reach functional maturity, it is possible that cholinergic interneurons are fundamentally involved in setting up a dysfunctional striatal circuit. This phenomenon is found in other motor regions and neurodevelopmental disorder models. For instance, Purkinje cells are among the first cell type to be born in the cerebellum (∼E10.5–E12.5), and instructs the development of granule cells, which constitute approximately 99% of the cerebellum via neurochemical cues [111–113]. Altering Purkinje cell functioning through pharmacological or genetic manipulations disrupts interactions with granule cells and leads to aberrant cerebellar circuit development and motor abnormalities [114–118]. In the postnatal striatum, cholinergic interneurons could direct innervation by excitatory afferents, which in turn prompts medium spiny neuron maturation and supports motor learning and execution. Cholinergic interneurons are in a position to coordinate rudimentary connections with excitatory projections because 1) cortical and dopaminergic afferents enter the striatum when cholinergic neurons are postmitotic and active, but medium spiny neurons are still being born [71, 74, 77]; 2) Cholinergic neurons act as “gatekeepers” and control excitatory inputs onto medium spiny neurons [119, 120]; 3) Cholinergic survival is unaffected by early dopaminergic or cortical denervation [121]. Indeed, cholinergic interneurons modulate the development of nigrostriatal pathways through expressing sonic hedgehog and TrkA, a nerve growth factor receptor [122, 123]. Both sonic hedgehog and TrkA signaling in the postnatal striatum maintains cholinergic cell numbers, which is required for nigrostriatal connectivity [122]. Through proper nigrostriatal connectivity, dopamine is released and guides medium spiny neuron maturation [15, 16, 81]. TrkA expression is downregulated in adult DYT1 patient and Dlx-CKO mouse model tissue [24], suggesting cholinergic dysfunction in dystonia. If TrkA signaling is dysregulated earlier, striatal connectivity could be impacted.

While altered striatal cholinergic interneuron instruction is possible, it is more likely that the striatal circuit develops normally but later becomes abnormal. Evidence for this hypothesis includes neonatal DYT1 mice exhibiting normal reflex behaviors [24]. When reflex behaviors develop normally, it suggests that underlying neural substrates are present and properly integrated into the motor circuit [54]. Later dysfunction of these neural substrates due to genetic insults, environmental stresses, or a combination of both then contributes to the development of abnormal behaviors [8, 124, 125]. In the Dlx-CKO mouse model where neonatal, but not postnatal, behaviors are preserved [24], the introduction of sensory inputs into the motor circuit may contribute to this shift. In the striatum, sensory inputs start influencing circuit function from P15 on, corresponding to a time of cortico-striatal plasticity changes [126] and locomotor behavior refinement [54]. Both medium spiny neurons and striatal cholinergic interneurons play integral roles in sensorimotor processing through their integration of dopaminergic and glutamatergic inputs [127]. This process is impaired in the hMT model, where plasticity or D2-receptor signaling defects were found in developing medium spiny neurons and striatal cholinergic interneurons [28, 29]. While similar observations have been reported in adult Tor1a ^ΔGAG/+ , Tor1a ^+/− , and muscarinic receptor (M₂/M₄) double knockout mice [38, 128, 129], determining how reproducible these electrophysiological phenotypes are in development requires earlier timepoints. Increased investigation into distinguishing the contributions of medium spiny neurons from striatal cholinergic interneurons in sensorimotor processing during development and across models are also necessary.

Interestingly, what makes developing striatal cholinergic interneurons individually equipped to integrate sensory inputs (e.g., their D2 and corticotropin releasing factor type 1 (CRF-R1) receptors), could also provide an explanation for the incomplete penetrance of DYT1 dystonia [79, 101]. Activation of D2 and CRF-R1 receptors enable reward as well as alters striatal cholinergic interneuron activity specifically in response to environmental cues [130, 131]. One prevailing hypothesis for the incomplete penetrance of DYT1 is that loss of torsinA primes the circuit for dysfunction, but symptoms only manifest if a secondary, external insult is applied [125]. Since striatal cholinergic interneurons normally express torsinA and can directly change their firing in response to environmental cues, then this two-hit hypothesis is plausible, despite not knowing the triggering sensory event(s). Another possibility is that DYT1 dystonia manifests when there is maladaptation or a lack of compensation. Focal dystonia is a prime example of maladaptation, where abnormal connectivity between the basal ganglia, cerebellum, and cortex develops as a result of repeatedly engaging muscles in prolonged motor tasks, such as writing or playing an instrument [132]. Unlike focal dystonia, maladaptation in DYT1 could result from changes occurring during development because peak torsinA expression coincides with periods of synaptogenesis in maturing striatal cells. Indeed, both striatal cholinergic interneurons and medium spiny neurons show evidence of increased synaptic inputs through exhibiting larger somata, increased membrane capacitances, and more spines [24, 37]. Alternatively, DYT1 dystonia could manifest due to a lack of compensation, which is supported by animal model studies. TorsinB is structurally and functionally similar to torsinA and can rescue cellular and behavioral phenotypes in Dlx-CKO mice, when overexpressed [133–136]. Reducing torsinB levels in Dlx-CKO animal models worsens DYT1 phenotypes, indicating that loss of torsinA drives dysfunction while torsinB attempts to normalize dysfunction [133]. Despite an early loss of torsinA (from ∼E11), the timing of torsinB expression remains the same ( ≥ P14; [107, 108]). The emergence of motor symptoms during peak torsinB expression reveal that the extent of compensation depends significantly on torsin function interchangeability and timing.

Questions remain regarding how involved striatal cholinergic interneurons are in DYT1 pathophysiology, including: Are cholinergic neurons main contributors to dystonic postures? Are striatal cholinergic interneurons aberrantly connected? Are observations detailing striatal cholinergic dysfunction in adults conserved during development? Major limitations to answering these questions is a lack of behavioral reproducibility and consistency among cellular phenotypes across preclinical DYT1 models. Standardizing experimental readouts and testing various ages can help, but the lack of reproducibility also stems from the biological and technical requirements of different genetic models and an unclear understanding of torsinA mutations in cholinergic neurons. For example, some studies suggest that striatal cholinergic excitation alone is not sufficient to elicit dystonia-like behaviors [137, 138], but the duration of striatal cholinergic excitation may be imperative for behavioral manifestation [52]. As for our understanding of pathogenic torsinA mutations, they are hypothesized to result in a loss-of-function or dominant-negative effect. Not knowing which mechanism(s) torsinA mutations utilize—or which under what circumstances—affects how we develop models and interpret data. Accumulating evidence in human cells and DYT1 knockin (Tor1a^ΔGAG/ΔGAG, Tor1a^ΔGAG/+ ) mice suggests that the ∆ E mutation results in a loss-of-protein-function [25, 27, 139]. Not only has torsinA been found to have a significantly reduced half-life, but the protein also exhibits impaired binding to enzymatic cofactors, LAP1 and LULL1, and undergoes premature degradation culminating in decreased protein levels [140–144]. Despite having strong evidence to support torsinA loss-of-function, mutant torsinA ( ∆ E) also has been shown to act in a dominant-negative manner by mislocalizing and suppressing wild-type torsinA activity [139, 142, 145]. Studies that have dissected the molecular mechanisms of torsinA mutations have not directly tested whether these change between cell types, or more specifically whether loss-of-function or dominant-negative mutant protein behavior is conserved in cholinergic neurons. Several germline (e.g., Tor1a ^−/− ) and conditional (e.g., Dlx-CKO, N-SKI) DYT1 models cannot be used to answer this question because they have been engineered using directed, loss-of-function mutations, but mice heterozygous for murine ∆ E-torsinA (e.g., Tor1a ^i-ΔGAG/+ , Tor1a ^ΔGAG/+ ) or overexpressing human ∆ E-torsinA (e.g., hMT) can be [24, 53, 146].

Overall better functional characterizations will improve our understanding of dystonia, the cells contributing to its pathophysiology, and the clinical relevance of basic DYT1 research. This includes providing a timeline of the physiological changes surrounding dystonia onset because currently, there is no consensus on how early pathogenesis begins. Already, there have been advances in elucidating torsinA roles in nuclear-cytoplasmic communication [32, 147] and uniting behavior from preclinical and clinical work, including the finding that both human patients and dystonia animal models exhibit variable leg adduction [52]. Increased knowledge will inevitably guide the development of more effective therapies by improving target specificity and timing.