DYT-TOR1A (DYT1) is a dominantly inherited dystonia characterized by early onset involuntary abnormal movements and postures (1, 2). TorsinA resides in the endoplasmic reticulum and nuclear envelope lumen, where interaction with cofactors LAP1 and LULL1 promote its ATPase activity (3-9). The DYT-TOR1A disease mutation deletes a single glutamic acid (ΔE) (2), impairing torsinA function (5,8,10-12). The natural history of DYT-TOR1A suggests that processes occurring during development are particularly important for disease pathogenesis (reviewed in (13)) and findings in mouse models suggest that the functions of the Tor1a encoded protein torsinA are essential during a developmental critical period but dispensable in adult animals (14). Several CNS developmental processes are altered by torsinA loss of function. In animal models, Tor1a deletion or Tor1a ^ΔE knock-in disrupts nuclear envelope structure (15, 16) and alters nuclear pore distribution and function (17, 18) during a postnatal CNS developmental period in which neuronal nuclear pore complex biogenesis and insertion is upregulated (19). TorsinA dysfunction causes deficits in secretory processing, protein quality control, and translational control (20-26) and alters synapse formation (27-30), all of which potentially contribute to the altered synaptic plasticity identified in dystonia (31-34). The mechanisms underlying synaptic changes in torsinA deficient neurons are not defined.

Multiple animal models of torsinA dysfunction display aberrant corticostriatal plasticity, including enhanced long term potentiation and decreased synaptic inhibition (24,35-37). Abnormal cholinergic signaling contributes to disrupted plasticity in some DYT-TOR1A models (36) and antimuscarinic compounds improve disease features in some people with DYT-TOR1A (38). Altered striatal plasticity is an early pathophysiological feature. Knock-in mice expressing the Tor1a ^ΔE/+ disease mutation exhibit premature long term potentiation, impaired long term depression, and increased AMPA receptor abundance in corticostriatal synapses during early striatal development (29). This converging evidence suggests that diminished inhibitory synaptic function (39) and dysfunction of striatal cholinergic interneurons (ChI) (40) are drivers of dystonia and suggest that processes occurring during development or maturation are critical for dystonia pathogenesis.

To mechanistically explore the relationship between torsinA loss-of-function and synaptic and behavioral change, we modeled DYT-TOR1A dystonia by conditionally deleting torsinA in forebrain inhibitory and cholinergic neurons using Dlx5/6-Cre (41) (“Dlx-CKO mice”). TorsinA is thus deleted from all neurons in the striatum, globus pallidus, reticular thalamic nucleus, and basal forebrain, and from inhibitory interneurons in the cortex and hippocampus in Dlx-CKO mice. Like the natural history of DYT-TOR1A, Dlx-CKO mice exhibit motor dysfunction beginning as juveniles, which worsens with increased handling and is responsive to antimuscarinic treatment (41). During the same juvenile period, a subpopulation of ChI in the dorsolateral striatum selectively degenerate. These findings suggest that Dlx-CKO mice model a link between cholinergic and motor dysfunction (42) believed important in human DYT-TOR1A dystonia (43).

To probe the mechanisms by which torsinA loss alters ChI and surrounding cell types during striatal maturation, we conducted RNAseq analyses on maturing Dlx-CKO striatal ChI somas or surrounding striatum tissue (mainly comprised of spiny projection neuron somas, with a small proportion of GABAergic interneuron somas, glia, and neural processes) isolated using laser microdissection. Striatal cholinergic and non-cholinergic enriched samples from control vs. Dlx-CKO identified a core set of genes enriched in secretory pathway and synaptic function. We further demonstrate abnormal synaptic structure in Dlx-CKO striatum with Golgi-Cox staining of spiny projection neurons. This study identifies a role for torsinA within the secretory pathway and implicates abnormal synaptic structure in the torsinA deficient striatum.

These studies identify a core set of differentially expressed genes in the striatum of torsinA conditional knockout mice during postnatal CNS maturation. Despite the previously reported divergent phenotype between cell types (cholinergic neurodegeneration vs. GABAergic neuron survival (41)), both ChI soma and non-cholinergic striatum samples demonstrated a discrete set of gene expression changes consistent with the role of torsinA in the secretory pathway. Striatum samples also displayed expression changes of genes regulating synaptic transmission and an upregulation of activity-dependent immediate early genes. Consistent with our RNAseq analyses, striatal spiny projection neurons in adult mice demonstrated significantly higher spine density, suggesting that surviving striatal neurons exhibit increased excitability during striatal maturation and increased afferent inputs in adulthood.

We isolated either ChAT-GFP+ ChI somas or GFP negative striatal tissue containing mainly spiny projection neuron somas, as well as GABAergic interneuron somas, glia, and neural projections using laser microdissection (see Figure 1A for a summary of the laser microdissection approach). The ChI soma samples were therefore highly enriched in a single cell type, while striatum samples contained mostly GABAergic neurons in a mixture of cell types and compartments, mainly comprising spiny projection neurons. This is reflected in our analyses as a cholinergic marker enrichment of 193.7 fold vs. GABAergic marker enrichment of up to 2.8 fold. Differential expression analyses of both sample types were overrepresented for genes encoding factors that are secreted to the extracellular space. Several neuropeptides were overrepresented in Dlx-CKO ChI soma samples, including Pdyn (upregulated), Vip, Npy, Cartpt, and Sst (downregulated). These factors were previously found to be enriched in GABAergic striatal neurons (56-61), but our enrichment protocol may have enabled measurement of sparse neuropeptide expression. The differential expression of genes encoding extracellular proteins and neuropeptides are consistent with a central role of torsinA in the secretory pathway (21, 22), as suggested by its localization in the endoplasmic reticulum lumen (3, 62).

Dlx-CKO striatum samples demonstrated a suite of gene expression differences consistent with a structural or functional change in striatal synapses. Whether the synaptic changes of GABAergic neurons reflect intrinsic responses to torsinA deficiency or a compensation consequent to neighboring cholinergic neurodegeneration remains unknown. Striatal cholinergic signaling matures postnatally and begins to dynamically regulate the synaptic activity of other striatal neurons as skilled motor function develops (63). The second postnatal week (when samples were collected in this study) is a maturational period during which corticostriatal synaptogenesis and spinogenesis begins and progresses (reviewed in (64)) as activity induced factors shape the connectivity of striatal neurons (65, 66). Several differentially expressed genes identified in this study modulate striatal spiny projection neuron spine density. The nuclear receptor Nr4a1 (upregulated 1.83 fold in Dlx-CKO) is enriched in spiny projection neurons (67), where its activity-induced expression alters spine density as part of a transcriptional program that regulates density and distribution of dendritic spines (68, 69) and promotes spiny projection neuron maturation (67). Npas4 (upregulated 2.51 fold) is a transcription factor that regulates GABAergic synaptic function (70) and is important for synaptic formation, function and ongoing plasticity (71). Knockdown of Npas4 reduces dendritic spine density on D1 receptor-expressing spiny projection neurons (72). Expression of the cytoskeleton associated protein Arc (upregulated 2-fold in Dlx-CKO striatum) increases spine density in vivo (73, 74). IGF-1 (upregulated 2.19-fold in Dlx-CKO striatum) administration rescues spine density (75) or spine motility (76) in Mecp2 mutant mice and knockdown of IGF-1 decreases spine density of purkinje cells (77). The upregulation of these factors during striatal maturation is consistent with changes to synaptic structure, as evidenced by significantly increased spine density of Golgi-Cox-stained spiny projection neurons in the present study.

Our differential expression analyses also suggest functional synaptic changes in Dlx-CKO mice. Doc2g (upregulated 2.24 fold in Dlx-CKO) is a member of the DOC2 family of proteins that modulates spontaneous synaptic transmission (78). Knockdown of DOC2 proteins triggers excitatory synaptic scaling without altering action potential dependent activity (79). Cnih3 (downregulated 1.84 fold in Dlx-CKO) is an AMPA receptor auxiliary subunit that functions in the endoplasmic reticulum and remains associated with the AMPA receptor complex at the synapse (80). CNIH3 regulates AMPA receptor trafficking and gating properties by determining the subunit composition of heteromeric AMPA receptors (81) and controlling the export of AMPA receptors from the endoplasmic reticulum (82). The structure of the interface between CNIH3 and AMPA receptors suggests that lipids play a role in the assembly of these complexes (83). The endoplasmic reticulum localization of CNIH3 and its interplay with lipids in complex with AMPA receptors suggests that it could be one link between torsinA function and the synaptic plasticity differences observed in animal models (24,35-37) and in people with dystonia (31-34). The synapse-related gene expression changes identified in torsinA null striatal neurons during maturation may therefore contribute to long lasting enhancement of spiny projection neuron synaptic structure and function.

To our knowledge, spine density has not been assessed previously in torsinA null mice. Heterozygous Tor1a ^ΔE/+ mice have reduced spiny projection neuron spine density at P26 (29), but no difference at P60 (29), consistent with other spine density studies in adult Tor1a ^ΔE/+ mice (27, 84). Spine density on distal dendrites of cerebellar purkinje neurons is reduced in 3 month old Tor1a ^ΔE/+ animals (28). However, motor behavior is not altered in these mice (85).

Surprisingly, despite glial enrichment, Gfap (encoding Glial Fibrillary Acidic Protein) was upregulated in both ChI soma and non-cholinergic striatum samples of Dlx-CKO mice. ChI soma sample Gfap expression could reflect “contamination” with adjacent or (synapsed) astrocytes, as increased neuronal activity increases expression of glial Gfap (86). However, astrogliosis is not observed in Dlx-CKO striatum (41). Neurons can express Gfap in neurodegenerative disease (87), but we observed robust Gfap expression in both control and Dlx-CKO samples. Some neuronal Gfap expression is observed in the normal mouse brain ((47); brainrnaseq.org). Fate mapping studies demonstrate that Gfap-expressing progenitors give rise to some neurons, including in the striatum (88), suggesting that we may be observing physiological ChI expression of Gfap during striatal maturation.

Six genes were differentially expressed in both ChI soma and non-cholinergic striatum samples. Fos, Pdlim3, and Pdyn were all upregulated to similar extents in both sample types, suggesting that these genes could represent common responses to torsinA loss of function or striatal circuit changes. In contrast, Ptgds, Tuba1c, and Gfap were downregulated in ChI somas, but upregulated in non-cholinergic striatum, suggesting a role in differential vulnerability of striatal neurons to cell death or cell type specific responses to torsinA loss of function. Tuba1c reduction (35.71 fold decreased in ChI) may reflect microtubule disruption or active degeneration of ChI, while its increase in non-cholinergic striatum (6.98 fold increased) could reflect compensatory neurite outgrowth or axon elongation in surviving cells (89). Only a single tubulin isoform was altered in this study, suggesting that torsinA loss of function caused a highly specific change rather than broad disruption of microtubule structure. Microtubule dynamics contribute to dendritic spine development, morphology, and synaptic plasticity (90-93). Increased Tuba1c expression may therefore reflect or contribute to the spine density increases we observed in Dlx-CKO spiny projection neurons.

Ptgds encodes lipocalin type prostaglandin D2 synthase, which catalyzes the conversion of prostaglandin H2 to the neuromodulatory prostaglandin D2 in the brain (94-96). Prostaglandin D2 is neuroprotective in contexts such as hypoxia-ischemic injuries, excitotoxicity, and oxidative stress (97,98,99,100,101). Prostaglandin D2 synthase (also called β-trace) itself is a neuroprotective chaperone that inhibits Aβ aggregation (102, 103), and alterations to its expression may be a biomarker of several neurological disorders (104). In the present study, Ptgds was 14.08 fold decreased in ChI soma and 6.56 fold increased in non-cholinergic striatum. Ptgds upregulation could contribute to the selective survival of non-cholinergic neurons in the striatum of Dlx-CKO mice. Further investigations would be required to determine whether this association is causative.

This study supports a developmental role for torsinA in the secretory pathway and demonstrates abnormal synaptic development in the torsinA deficient striatum. These transcriptomic datasets are freely available as a resource for future hypothesis driven work exploring the consequences of torsinA loss for striatal structure and function.