TorsinA dysfunction causes persistent neuronal nuclear pore defects

Abstract

A critical challenge to deciphering the pathophysiology of neurodevelopmental disease is identifying which of the myriad abnormalities that emerge during CNS maturation persist to contribute to long-term brain dysfunction. Childhood-onset dystonia caused by a loss-of-function mutation in the AAA+ protein torsinA exemplifies this challenge. Neurons lacking torsinA develop transient nuclear envelope (NE) malformations during CNS maturation, but no NE defects are described in mature torsinA null neurons. We find that during postnatal CNS maturation torsinA null neurons develop mislocalized and dysfunctional nuclear pore complexes (NPC) that lack NUP358, normally added late in NPC biogenesis. SUN1, a torsinA-related molecule implicated in interphase NPC biogenesis, also exhibits localization abnormalities. Whereas SUN1 and associated nuclear membrane abnormalities resolve in juvenile mice, NPC defects persist into adulthood. These findings support a role for torsinA function in NPC biogenesis during neuronal maturation and implicate altered NPC function in dystonia pathophysiology.

Introduction

DYT1 dystonia is a neurodevelopmental disease manifesting as involuntary twisting movements (1). The illness is caused by a loss-of-function (LOF) mutation in the TOR1A gene encoding the AAA+ protein torsinA (2,3). Symptoms typically emerge during childhood-adolescence (∼8–16 years of age), progress for many years, and then plateau and remain fixed for life (4). This pattern implicates torsinA in molecular events essential for normal CNS maturation. Neuroimaging and clinical electrophysiological studies suggest that there are circuit level structural and functional abnormalities in adult patients (5–7). It is unknown, however, if the effects of the DYT1 mutation are limited to transiently disrupted neurodevelopment selectively causing hodological abnormalities, or whether there are persistent molecular defects in torsinA deficient neurons.

TorsinA is localized to the endoplasmic reticular (ER)/nuclear envelope (NE) endomembrane space (2,8–10) and is implicated in the regulation of protein trafficking, quality control (11–16), and NE function (8–10,17–21). High levels of protein synthesis and nucleocytoplasmic transport are required during synaptogenesis and circuit maturation (22,23), making it plausible that disruption of one or more of these processes could contribute to aberrant CNS development or persistent molecular defects disrupting neuronal function.

Prior work demonstrates that torsinA loss-of-function causes abnormal evaginations of the neuronal inner nuclear membrane (termed ‘buds’) that occur exclusively during a critical period of postnatal CNS maturation (3,21,24). These buds contain ubiquitin and nuclear pore proteins (25,26). Ubiquitin accumulation was originally described as occurring selectively within the NE buds of degenerating neurons (25), whereas NE budding occurs ubiquitously in torsinA null neurons (3,21,24). This dissociation raises the question of what, if any, abnormalities occur in the majority of cells as NE buds form or consequent to their resolution. Indeed, no NE defects have been described in the CNS of adult torsinA mouse mutants. Loss of the C. elegans torsin homolog OOC-5 disrupts nuclear pore localization and function (27), linking torsinA function to nuclear pore biology. No data demonstrate nuclear pore alterations in vertebrate disease models, however.

Here, we probe the link between perinuclear ubiquitin accumulation and additional NE abnormalities during postnatal CNS maturation and adulthood in a DYT1 rodent model (28). In contrast to previous work, we find that perinuclear ubiquitin accumulation occurs in nearly all torsinA deficient neurons and, like NE buds, is present exclusively during a discrete neurodevelopmental period in juvenile animals. In juvenile mice, accumulation of perinuclear ubiquitin is accompanied by abnormal clustering of nuclear pore complexes (NPC). These clustered NPCs appear to represent a stalled intermediate in NPC assembly, as many lack NUP358, a nucleoporin added at a late step in NPC biogenesis. Consistent with this possibility, neurons with clustered NPC also exhibit abnormalities in molecules essential for normal NPC biogenesis and function, including SUN1, laminA/C, and Ran GTPase. Whereas the NPC defects persist into adulthood, all other NE abnormalities examined resolve. These findings identify aberrant nuclear membrane ubiquitin accumulation and nuclear pore abnormalities as core features of torsinA deficiency in neurons that otherwise appear normal into adulthood. The nuclear pore defects define a molecular neurodevelopmental event that, like disease symptoms, emerges during juvenile development and is permanent, and suggests that nuclear pore dysfunction should be further pursued as a potential contributor to dystonia pathogenesis and pathophysiology.

Results

To begin to define the relationship between NE abnormalities and CNS dysfunction in dystonia, we pursued a neuropathological analysis of the Dlx-CKO model of DYT1 dystonia at postnatal day 14 (P14), when NE buds are present and just prior to the onset of motor abnormalities (21,28). Striking abnormal accumulations of perinuclear ubiquitin were observed in all torsinA null brain regions (Fig. 1A and B, Supplementary Material, Fig. S1). Ubiquitin accumulations were evident using both pan-ubiquitin (Supplementary Material, Fig. S1N) and polyubiquitin (K48 linkage) antibodies (Fig. 1A). The highest densities of affected cells were in the reticular thalamic (RT) nucleus, ventral forebrain, and cortex (Supplementary Material, Fig. S1A–H, M). The Dlx5/6-Cre transgene expresses selectively within forebrain GABAergic and cholinergic neurons. Despite the fact that only striatal cholinergic interneurons degenerate in this model, ubiquitin accumulation occurred exclusively in GABAergic neurons; no abnormalities of ubiquitin were observed in cholinergic neurons or in cells not expressing Cre recombinase (Supplementary Material, Fig. S1I–L). Ubiquitin accumulations were present in all layers of the cortex and in all GABAergic neuronal subtypes, but were more predominant in somatostatin (SST) immunoreactive neurons (Supplementary Material, Fig. S2). No ubiquitin abnormalities were observed in any cell type in P0 mice (data not shown). These results demonstrate that abnormal accumulation of neuronal perinuclear ubiquitin is a much more frequent consequence of torsinA LOF in maturing neurons than previously appreciated (25).

Figure 1.

Maturing torsinA deficient neurons accumulate perinuclear ubiquitin and develop abnormally localized nuclear pores in vivo. (A,B) UbiquitinK48 immunohistochemistry demonstrates perinuclear spheres in maximum intensity projection images of z-stacks, and (A’, B’) ‘rings’ in single optical slices of Dlx-CKO cortex. (C,D) UbiquitinK48 and mab414 immunohistochemistry demonstrate abnormal localization of nuclear pores in P14 cortex. (C’, D’) Optical slices show that ubiquitin is not colocalized with mab414 immunoreactivity. Scale bars = 10 μm.

Prior work implicates a role for torsinA in modulating ER protein quality control pathways, abnormalities of which are frequently associated with ubiquitin accumulation (11,12,15,16,25,29). We therefore explored whether the CNS of Dlx-CKO mice exhibited abnormalities of the unfolded protein response (UPR), integrated stress response (ISR) or ER-associated protein degradation (ERAD). As these homeostatic mechanisms are tightly controlled transcriptionally, we assessed mRNA in tissue from P14 ventral forebrain, when this region exhibits a high density of cells with perinuclear ubiquitin accumulations. Although Tor1a null cells are interspersed with some Cre recombinase non-expressing cells in this region, Tor1a mRNA was decreased 3-fold in Dlx-CKO samples compared with control (t₈= 3.696, P = 0.006), suggesting acceptable dilution effects. Nevertheless, there were no significant changes in mRNAs for any of the 84 UPR, ISR, or ERAD factors examined (Supplementary Material, Fig. S3). Three genes encoding ER membrane spanning proteins were upregulated approximately 2 fold, but did not reach statistical significance (Creb3l3, 2.02 fold, adjusted P = 0.32; Ern2, 2.04 fold, adjusted P = 0.99; Atf6b, 2.45 fold, adjusted P = 0.99; t-test with Holm-Sidak correction; Supplementary Material, Fig. S3A). Immunohistochemical studies of CrebH (the protein encoded by Creb3l3) and isoforms of Atf6 and Ire1 did not exhibit any abnormalities in localization or immunoreactivity in cells containing abnormal accumulations of perinuclear ubiquitin (Supplementary Material, Fig. S3B–G). The widespread and robust perinuclear ubiquitin accumulation observed in this model therefore appears independent of overt changes in the UPR, ISR, or ERAD pathways.

NE buds contain ubiquitin and nuclear pores (25,26) and NE buds and nuclear pore abnormalities occur in torsin deficient C. elegans (27). Based on these facts, we hypothesized that perinuclear ubiquitin accumulation would be accompanied by NPC abnormalities. Indeed, neurons with perinuclear ubiquitin accumulation exhibited striking abnormalities in NPC distribution in Dlx-CKO P14 tissue (assessed with FG-repeat recognizing mAb414 antibody; Fig. 1C and D). Control brains and Dlx-CKO neurons lacking ubiquitin abnormalities exhibited the expected smooth perinuclear distribution of mAb414 signal. In contrast, cells with abnormal ubiquitin accumulation showed large perinuclear punctae of mAb414 signal, with areas of NE adjacent to these punctae devoid of normal immunoreactivity (Fig. 1C). Ubiquitin accumulations were not colocalized with nuclear pore immunoreactivity. NPC disruption occurred in all brain regions containing cells with perinuclear ubiquitin accumulation (Supplementary Material, Fig. S4).

The presence of abnormal perinuclear ubiquitin and NPC clustering selectively in torsinA null neurons suggested that cell autonomous mechanisms provoke these abnormalities. To explore this possibility, we tested whether perinuclear ubiquitin accumulation and NPC disruption could be recapitulated in vitro. Robust perinuclear ubiquitin accumulation and nuclear pore complex mislocalization occurred in primary cortical neurons from Dlx-CKO mice (Fig. 2A–F). Similar to our findings in vivo, ubiquitin accumulation is not present initially, but emerges with neural maturation. Consistent with the expression pattern of the Dlx5/6-Cre transgene, ubiquitin accumulation occurred exclusively within GABAergic neurons (comprising approximately 20% of the primary neurons, Fig. 2G–I), and was not observed in wild type, Dlx5/6-Cre control, or Tor1a^+/− control cultures. These data establish Dlx-CKO primary cortical cultures as an in vitro system to dissect the mechanisms and consequences of torsinA LOF on the neuronal nuclear membrane. Further, they demonstrate that the ubiquitin and nuclear pore abnormalities reflect a cell autonomous requirement for torsinA during neural maturation.

Figure 2.

Maturing torsinA deficient neurons accumulate perinuclear ubiquitin and develop abnormally localized nuclear pores in vitro. (A–F) UbiquitinK48 and nuclear pore immunohistochemistry (mAb414) in DIV10 primary cortical culture. (G–I) UbiquitinK48 and GAD67 immunohistochemistry demonstrates the presence of ubiquitin accumulations exclusively in GABAergic neurons in Dlx-CKO primary cortical culture.

The mislocalization of NPCs suggests disrupted NPC function. Active nuclear import and export requires a nuclear to cytoplasmic gradient of the small GTPase Ran (30). To begin to examine NPC function, we evaluated the distribution of Ran in control or torsinA null neurons containing mislocalized NPCs and perinuclear ubiquitin accumulation. Primary neurons from control animals and Dlx-CKO neurons lacking perinuclear ubiquitin exhibited the normal, predominately nuclear distribution of Ran (Fig. 3A–H). In contrast, Dlx-CKO neurons containing perinuclear ubiquitin accumulation demonstrated a striking redistribution of cytoplasmic Ran (Fig. 3D). Within these cells, the majority of the cell soma lacked Ran immunoreactivity, with the remaining fluorescent signal present in abnormal jagged accumulations surrounding the nucleus (Fig. 3D’, D’’). This pattern of Ran redistribution was present in the majority of cells containing ubiquitin accumulations (Fig. 3I) and extended beyond the boundaries of the nucleus, as assessed by DAPI. Because the abnormal perinuclear accumulations of Ran extended into the cytoplasm (Fig. 3D’), the ratio of nuclear to cytoplasmic Ran was only slightly increased (Fig. 3J and K). Nevertheless, the abnormal distribution of Ran is striking and suggests that nuclear transport is altered in cells containing mislocalized NPC and perinuclear ubiquitin.

Figure 3.

Ran GTPase is mislocalized in torsinA deficient neurons. (A–H) UbiquitinK48, MAP2, and Ran immunohistochemistry in DIV10 primary cortical neurons from control or Dlx-CKO mice. Dashed lines represent the parameters used to delineate cell soma and nucleus for signal intensity measurements. (D’, D’) Long exposure demonstrates redistribution of cytoplasmic Ran signal from the normal diffuse cytoplasmic distribution (right) to jagged perinuclear in a ubiquitin ring positive neuron (arrow head). (I) Blinded analysis of Ran immunoreactivity demonstrates the percent of neurons with jagged perinuclear Ran signal. (J) Frequency histogram of nuclear to cytoplasmic Ran signal (Control n = 186 neurons, Dlx-Cre negative [no ring] n = 224 neurons, Dlx-CKO with ring n = 212). (I) Nuclear to cytoplasmic ratio of Ran. Scale bar = 25 μm.

The mislocalization of NPC and altered Ran-GTPase distribution suggest abnormal nuclear transport. To determine if nuclear transport is altered in neurons lacking torsinA, we examined the nuclear to cytoplasmic (N/C) ratio of a fluorescent nuclear transport reporter containing a strong nuclear export signal and weaker nuclear localization signal [NLS-mCherry-NES; (31)] in constitutive Tor1a^+/+, Tor1a^+/−, and Tor1a^−/− primary neurons. Neurons transfected with NLS-mCherry-NES exhibited a range of phenotypes along a continuum of nuclear to cytoplasmic signal, with most wild type cells exhibiting a higher fluorescence intensity in the cytoplasm (Fig. 4E–H). In contrast, Tor1a^−/− neurons (Fig. 4A–D) contained a higher ratio of mCherry signal in the nucleus, and on average, a higher N/C ratio compared with control and heterozygous neurons (Fig. 4I;F_2,4=48.3, P = 0.0016). While all genotypes exhibited a range of N/C, Tor1a null neurons had significantly more neurons with a high N/C ratio (Fig. 4J). Furthermore, a small subset of torsinA null neurons contained a jagged perinuclear mCherry signal reminiscent of the RAN immunostaining. Considered together with other findings, these results suggest that neurons lacking torsinA exhibit abnormal nuclear export, but enhanced nuclear import could also contribute to the altered N/C ratio.

Figure 4.

Nuclear transport is altered in torsinA deficient neurons. (A–H) NLS-mCherry-NES, DAPI, mAb414, and ubiquitin-K48 immunofluorescence in DIV8 primary cortical neurons from constitutive Tor1a+/+ and Tor1a^−/− mice. (I) Average nuclear to cytoplasmic (N/C) mCherry fluorescence intensity. (J) Frequency histogram of N/C mCherry in Tor1a+/ + (n = 108 neurons), Tor1a^+/− (n = 153 neurons), and Tor1a^−/− (n = 213 neurons).

Interphase nuclear pore biogenesis and insertion occurs in defined stages and requires Ran-GTP (32), suggesting that these processes may be altered in torsinA deficient neurons. Early NPC assembly intermediate structures contain specific early-incorporated nuclear and cytoplasmic ring component nucleoporins, but not late-recruited cytoplasmic filament nucleoporins (33,34). The mAb414 antibody recognizes several of these nucleoporins, including the early incorporated Nup153, and late-incorporated Nup358, which are both FG-nucleoporins (35–37). To begin to explore whether the NPC mislocalizations discovered with mAb414 reflect abnormal assembly intermediates, we examined Nup153 and Nup358 in primary cortical neurons from Dlx-CKO mice. Double immunostaining revealed distinct distributions of the early-incorporated nucleoporin Nup153 vs the late-recruited nucleoporin Nup358 (Fig. 5). Nup153 exhibited the same pattern of large perinuclear clusters as mAb414 staining, with adjacent areas devoid of normal immunoreactivity in cells containing perinuclear ubiquitin accumulations (Fig. 5A–C). In contrast, Nup358 was not generally clustered, mostly did not colocalize with Nup153 clusters, and exhibited significantly decreased fluorescence intensity compared with primary neurons from control animals or Dlx-CKO neurons without perinuclear ubiquitin accumulation (Figs 5A–D and 6; assessed by 2 observers blinded to genotype and antibodies). The dissociation between Nup153 and Nup358 demonstrates that not all FG-nucleoporins exhibit similar behavior, and suggests that early assembled Nups may be more prone to abnormal clustering. Neither nucleoporin colocalized with perinuclear ubiquitin accumulations (Fig. 6). The mislocalization of Nup153, reduced co-localization and intensity of Nup358 suggest that NPCs may be structurally and functionally altered; Nup358 (also termed RanBP2) binds Ran GTPase and promotes GTP hydrolysis, which is required for active nuclear transport (38–41). Considered together, these findings suggest that NPC accumulations may be incompletely assembled NPC intermediates with abnormal localization and function.

Figure 5.

TorsinA deficient neurons accumulate incompletely assembled NPCs and exhibit abnormalities in proteins required for interphase NPC assembly and localization. (A, B) Maximum intensity projection images from Z-stacks of Ubiquitin K48, Nup153, and Nup358 immunohistochemistry. (A’, B’) Single optical slices demonstrate Nup153 clustering and reduction of Nup358 intensity. (C) Percentage of cells containing nucleoporin clustering. (D) Nuclear rim intensity of Nup358 immunohistochemistry. * indicates significant difference from control groups (One way ANOVA with Dunnett’s multiple comparisons test. F_2,3 = 58.92, P=0.0039). (E–H) LaminA/C staining demonstrates dysmorphic nuclear lamina in Dlx-CKO neurons. (I–L) Overall sun1 localization is normal, but small perinuclear punctae are colocalized with ubiquitin in a subset of ubiquitin accumulating neurons from Dlx-CKO mice. Scale bars = 10 μm.

Figure 6.

Nup153 clusters do not colocalize with Nup358. (A–F) Maximum intensity projections of Z-stacks (left) and optical slices (right) of Nup153, Nup358, and ubiquitin in a single Dlx-CKO neuronal nucleus demonstrate lack of Nup358 within Nup153 clusters, and no colocalization of ubiquitin and nucleoporins. circles indicate Nup153 clusters in z-stacks and slices. (G–I) Optical slice of single control neuronal nucleus. Scale bar = 10 μm.

Normal nuclear pore localization requires the interaction of multiple NE proteins of the inner nuclear membrane (INM) and lamina (42–47). To begin to identify candidate molecules that contribute to NPC abnormalities in torsinA deficient neurons, we screened known participants in this pathway using the in vitro primary cortical model system. This analysis demonstrated relatively specific abnormalities in torsinA-related molecules. Lamins interact with the torsinA binding partner, LAP1 (17,48,49). LaminA/C staining was less intense and abnormally shaped in Dlx-CKO-derived neurons (Fig. 5E–H). The INM protein SUN1 participates in the abnormal NE accumulation of DYT1 mutant torsinA and functionally associates with nuclear pore proteins (18,45,47). SUN1 immunostaining exhibited small punctae that colocalized with ubiquitin punctae in a subset of Dlx-CKO neurons (Fig. 5I–L). In contrast, distribution of an INM protein with no known links to torsinA, the laminB receptor, appeared normal (Supplementary Material, Fig. S5). Similarly, the distribution of the outer nuclear membrane protein nesprin-2-giant, which is not required for normal nuclear pore complex distribution (45,47), appeared normal in Dlx-CKO neurons (Supplementary Material, Fig. S5).

To explore whether these in vitro phenotypes reflect in vivo abnormalities we assessed several of these NE proteins in CNS tissue derived from P14 Dlx-CKO mice. SUN1 and laminA/C both exhibited abnormalities in vivo mimicking the in vitro findings. SUN1 formed large immunoreactive punctae colocalized with ubiquitin accumulations (Fig. 7A and B). LaminA/C fully surrounded the nucleus, but instead of exhibiting a characteristically smooth and well demarcated boundary, formed an indistinct and jagged border of immunostaining around the nucleus (Fig. 7C and D). Immunostaining for laminB, which is involved in nuclear pore complex assembly (50), was also misshapen (Supplementary Material, Fig. S6). In contrast to these abnormalities, the distribution of laminB receptor appeared normal in P14 tissue (data not shown). Considered together with the in vitro findings, these data indicate that torsinA deficiency does not cause a global disruption of nuclear envelope proteins. Rather, torsinA deficiency appears to disrupt a relatively selective group of proteins related to NPC assembly and localization.

Figure 7.

Sun1 and laminA/C are abnormally localized in P14 Dlx-CKO brain. (A,B) Sun1 and ubiquitinK48 immunohistochemistry demonstrate large sun1 positive punctae colocalized with ubiquitin accumulations in Dlx-CKO and Control P14 cortex. (C,D) LaminA/C and ubiquitinK48 immunohistochemistry in Dlx-CKO and Control P14 cortex. Scale bars = 10 μm.

Ubiquitin accumulation has been linked to selective neurodegeneration in other DYT1 dystonia models (25,51), and NPC abnormalities have been associated with neurodegeneration in amyotrophic lateral sclerosis/frontotemporal dementia (52–54) and Huntington disease (55,56). We therefore sought to determine whether neurons that exhibit nuclear pore and ubiquitin abnormalities degenerate in the Dlx-CKO model. We began by characterizing the temporal profile of ubiquitin abnormalities in defined neuronal populations using unbiased stereological cell counting. Perinuclear accumulation of ubiquitin began as early as P7 in the RT nucleus and zona incerta (ZI), and reached maximal frequency by P14 (Fig. 8A–D). These abnormalities were largely resolved by P28 and absent at P70 (Fig. 8A–D). Ubiquitin accumulation in the cerebral cortex exhibited a similar temporal profile (Fig. 8E–H), which was mimicked in primary cortical neurons in vitro (Supplementary Material, Fig. S7). To determine if loss of perinuclear ubiquitin accumulation at later ages reflected neurodegeneration, we stained for apoptotic markers previously associated with torsinA-mediated neurodegeneration (cleaved caspase 3; CC3) (25,28) and quantified the number of (PV+) RT neurons using unbiased stereological cell counting. Although 89.6% (±0.83 SEM, n = 4) of PV+ cells in the RT contained ubiquitin accumulation at P14 (Fig. 8I–N), there was fewer than one CC3+ cell per section in the RT at any developmental time point observed (Fig. 8O) and no significant difference in the total number of PV+ cells at 6 months, as assessed by unbiased stereological cell counting (Fig. 8P). Considered together with prior stereologic cell counts demonstrating normal numbers of PV+ and SST+ GABAergic cell populations in cortex and striatum of Dlx-CKO mice (28), and the lack of ubiquitin accumulation or nuclear pore mislocalization in cholinergic neurons (which degenerate in this model), these data dissociate ubiquitin accumulation from neurodegeneration. Rather, NE-localized ubiquitin accumulation appears to be a common and robust feature of many populations of maturing torsinA null neurons, and is not restricted to degenerating neurons.

Figure 8.

TorsinA deficient neurons display ubiquitin abnormalities occur during a discrete neurodevelopmental window in vivo. (A–H) UbiquitinK48 immunohistochemistry demonstrates the presence of perinuclear ubiquitin accumulation in the reticular thalamus (RT), zona incerta (ZI), and cortex of Dlx-CKO mice from P7-P28, but absence beyond P70, and no accumulation in control mice. (I–N) 89.6% (±0.83 SEM, n = 4) of parvalbumin positive neurons in the RT contain perinuclear ubiquitin accumulation at P14. (O) Fewer than one RT cell per section expressed cleaved caspase-3 from P7-P28 in control and Dlx-CKO mice (No difference between control and Dlx-CKO. Two-way ANOVA Interaction: F3,21 = 0.35, P=0.7; Age: F3,21 = 1.48, P= 0.24; Genotype: F1,21 = 0.33, P=0.57). (P) Stereological cell counts of PV+ neurons in the RT demonstrate normal RT cell numbers at P168 (Welch’s t-test: t3.19 = 1.842, P=0.15). Scale bars = 50 μm.

Abnormal ubiquitin accumulation and nuclear membrane budding both occur selectively during a discrete period of CNS maturation in Dlx-CKO mice, during the same period when the number of nuclear pores in mouse cortical neurons greatly increases (57). In contrast, many NPC components are long lived in mature neurons (58,59). As a whole, NPCs do not disassemble except during nuclear envelope breakdown during mitosis and do not turn over (or turn over extremely slowly) after mitosis (60,61). This raises the possibility that developmental defects in NPC may be long lasting. We explored this possibility by examining brain tissue from 3-month-old Dlx-CKO and control mice, long after the disappearance of ubiquitin abnormalities. We focused on SST+ neurons at 3 months because they make up the majority of ubiquitin ring positive cortical cells at P14 (Supplementary Material, Fig. S2). Despite the lack of ubiquitin accumulation at this age, widespread nuclear pore abnormalities remained (Fig. 9). An analysis by observers blinded to genotype demonstrated that 89.4% (±2.6SEM, n = 4) of SST+ neurons contained abnormal nuclear pore clustering at 3 months, compared with 3.8% (±0.7SEM, n = 4) of control SST+ neurons. In contrast, we observed no difference in SUN1 or laminA/C immunostaining in brain tissue from 3-month-old Dlx-CKO and control mice (data not shown). Disordered nuclear pore immunoreactivity was present as late as 7 months of age, with a significantly greater number of cortical neurons exhibiting clustered nuclear pores as compared with littermate controls (Supplementary Material, Fig. S8). Persistent abnormal NPC abnormalities in torsinA deficient neurons indicate that nuclear pore dysfunction could contribute to dystonia pathogenesis and pathophysiology.

Figure 9.

Neuronal nuclear pore defects persist into adulthood in the torsinA deficient CNS. (A,B) UbiquitinK48, mab414, and somatostatin immunohistochemistry in cortex from 3-month-old Dlx-CKO and control mice. (C) Significantly more somatostatin+ cortical neurons exhibit disordered, clustered nuclear pores in Dlx-CKO than control mice at 3 months (t-test: t6 = 31.54, P<0.0001). Scale bars = 10 μm.

Discussion

Our studies identify novel abnormalities of torsinA deficient neurons that are pervasive and persistent and emerge during juvenile CNS development, the time when abnormal movements manifest in torsinA mutant rodents and DYT1 subjects. Prior work linked torsinA-related perinuclear ubiquitin accumulation to selective neurodegeneration. In contrast, these new findings clearly dissociate NE ubiquitin accumulation from neurodegeneration and identify these changes as a core feature of torsinA deficiency during neural maturation. NPC mislocalization accompanies the onset of abnormal perinuclear ubiquitin, but in contrast to the ubiquitin and related abnormalities, appears to be a permanent defect of torsinA deficient neurons. The cause and relationship between ubiquitin accumulation and NPC mislocalization are unknown, but the coincident onset and close localization of these abnormalities (together with NE buds) suggest a potential link. Maladaptive neural plasticity is implicated in the pathophysiology of many forms of isolated and complex dystonia (62,63) and long-lasting changes in synaptic efficacy require translocation of transcription factors through nuclear pores into the nucleus (23,64,65). The novel observation of permanent NPC abnormalities in torsinA deficient neurons may therefore provide a molecular link between the consequences of torsinA deficiency and a neurophysiological abnormality that contributes to dystonia.

These results are consistent with prior work establishing a temporal requirement for torsinA function in neurons (21), and suggest that processes occurring during CNS maturation may provoke abnormalities relevant to disease pathogenesis. The marked defects in nuclear pores, abnormalities of Ran GTPase distribution and gradients, and the altered nuclear to cytoplasmic ratio of the NLS-mCherry-NES nuclear transport reporter suggest that abnormalities of nucleocytoplasmic transport, a process critical for circuit development, synaptic plasticity and function, and cell survival (66–70), may contribute to torsinA-related neural dysfunction. Considered with prior work (3,19,21,25–27), our findings further implicate NE disruption as a contributor to neuronal and circuit dysfunction in dystonia.

The emergence of nuclear pore and perinuclear ubiquitin abnormalities during postnatal circuit maturation raises the possibility that developmental processes regulating synaptogenesis and plasticity may trigger this phenotype. TorsinA-related INM buds form during a similar time frame, and the physical association between ubiquitin, nuclear pores and INM buds suggest that these abnormalities reflect dysregulation of a common process (21,25,26). A strong candidate process is the insertion of nuclear pores during interphase. The number of nuclear pores in mouse cortical neurons increases by up to 5-fold during the first two postnatal weeks (57), indicating that the process is robustly upregulated during the period that torsinA-related NE abnormalities emerge. This upregulation likely reflects a high demand on maturing neurons to synthesize a diverse array of proteins as they integrate into circuits. Indeed, high demand for new protein synthesis has been identified as a trigger for interphase nuclear pore insertion (71,72).

De novo nuclear pore assembly and insertion occurs in stages, with early intermediate structures containing nuclear and cytoplasmic ring components, followed by later incorporation of cytoplasmic filament nucleoporins (32,34). Nup153 is required for interphase NPC assembly, and mediates the recruitment of the structural Nup107-Nup160 complex to NPC assembly sites (33). Nup107 is incorporated early in assembly intermediates, while the cytoplasmic filament nucleoporin Nup358 is incorporated at a later step of NPC maturation (34). Ran GTPase function is required for normal NE formation (73), regulates Nup153 membrane interactions (33) and NPC insertion into intact nuclei (32). Ran binding to Nup358 (also known as RanBP2) promotes its interaction with the GTPase activating protein RanGAP1, thereby facilitating the GTP hydrolysis required for nuclear import (30,38,74,75). The absence of Nup358 from NPC clusters in torsinA null neurons suggests that these NPC accumulations are abnormally localized NPC assembly intermediates. Alternatively, initial NPC formation may be normal, followed by later loss of Nup358 expression. The overall reduction in Nup358 expression may drive the abnormal cytoplasmic Ran redistribution from diffuse somatic to jagged perinuclear signal (Fig. 3). These findings suggest strongly that nuclear transport is abnormal in torsinA deficient neurons. The altered distribution of the NLS-mCherry-NES nuclear transport reporter in torsinA null neurons implicates abnormalities in nuclear export or enhanced nuclear import. However, considering the Ran redistribution and reduced Nup358, it is more likely that torsinA null neurons exhibit diminished nuclear export.

TorsinA is a membrane-localized AAA+ protein, ideally situating it to participate in the processes required for the formation, insertion or surveillance of newly assembled NPCs. The striking abnormalities of Ran, SUN1, and lamins identified in torsinA deficient neurons likely contribute to the observed nucleoporin abnormalities. Lamins are required for normal distribution of nuclear pore complexes (42–44), and interact with the torsinA binding partner, LAP1 (17,48,49). There are known connections between SUN1 and torsinA, and SUN1 is required for the abnormal NE accumulation of DYT1 mutant torsinA (18). SUN1 forms functional associations with nucleoporins, links lamins and nuclear pores, is critical for the normal distribution of NPCs (45,46), and is required for interphase nuclear pore complex assembly (47). Indeed, depletion of SUN1 causes nucleoporin clustering (45) similar to that observed in torsinA deficient neurons. These findings suggest a central role for SUN1 in torsinA-related nuclear pore defects, either through its role in the assembly or insertion of nuclear pores during interphase, or in nucleoporin localization.

Studies in yeast demonstrate the existence of a surveillance mechanism that monitors nuclear pore assembly and is utilized to remove defective intermediates (76,77). Protein quality control mechanisms, also identified in yeast, exist to target mislocalized nuclear membrane proteins for degradation by INM-localized E3 ligases (78,79). No torsinA homologue exists in yeast, and these processes have not yet been demonstrated in mammalian cells. Our data do not demonstrate ubiquitination of FG-nucleoporins, but it is possible that other nucleoporins could be ubiquitinated (80). Future studies will be necessary to determine a potential role for torsinA in the targeting of defective nuclear pores in mammalian cells, including in neurons.

TorsinA function is important during CNS development and maturation (21,25,28,81), but the functional ramifications of torsinA LOF during this period remain unclear. At least two mechanisms could drive long lasting functional deficits following torsinA-mediated developmental disruption. Developmental abnormalities may disrupt circuit wiring, possibly leading to an aberrant circuit comprised of intact, functional neurons. Alternatively, developmental processes may lead to long-lasting deficits in neuronal function, causing ongoing dysfunction. These possibilities are not mutually exclusive. The neurodevelopmental onset of NE ubiquitin accumulations associated with persistent NPC mislocalization and dysfunction may represent a link between NE dysfunction and circuit defects observed in dystonia.

Materials and Methods

Animals

Dlx5/6-Cre conditional Tor1a null mice were generated, maintained, and genotyped as described previously (28). Age matched littermate male and female mice were used for all experiments, including four possible genotypes: Tor1a^flx/+ (WT), Tor1a^flx/− (Flx control), Cre⁺Tor1a^flx/+ (Cre control), and Cre⁺Tor1a^flx/- (Dlx-Tor1a CKO).

Primary neuronal culture

P0 mouse pups were sacrificed via decapitation and brains were removed and placed in ice cold 0.01M phosphate buffered saline. The meninges were removed, and the cerebral cortex was dissected, placed into a microcentrifuge tube, and cut into small pieces. Cortices were incubated in 2 µg/ml papain (BrainBits) and 5 µl DNase I (Worthington Biochemical) for 30 min at 37 °C. 500 µl BrainPhys Neuronal Medium (Stemcell Technologies) was added, cortices were triturated three times through wide bore P1000 pipet tips, and centrifuged at 200g for 5 min. Supernatants were discarded, and pellets were reconstituted in BrainPhys medium, triturated and centrifuged twice more. Neurons were resuspended in BrainPhys Neuronal medium containing SM1 supplement (1: 50; Stemcell Technologies), glutamax (1: 100; ThermoFisher Scientific), and gentamicin (1: 1000; ThermoFisher Scientific), and plated on polyethylenimine-coated (100 µg/ml; Polysciences) coverslips in 24-well plates. Seeded neurons were incubated in 5% CO₂ at 37 °C, and half of the volume of media was replaced every third day. Following growth, neurons were fixed with 4% paraformaldehyde in phosphate buffer for 10 min at room temperature, and then stored in phosphate buffered saline until immunostaining. NLS-mCherry-NES (pDN160) reporter [Addgene plasmid #72660; (31)] was transfected into DIV7 neurons using DNA-In-Neuro Transfection Reagent (MTI Globalstem) in BrainBits transfection medium for 5.5 hrs, and then coverslips were returned to BrainPhys Neuronal medium overnight. Cells were fixed at DIV8 as described above.

Immunohistochemistry and imaging

Mice were deeply anesthetized with a lethal dose of ketamine/xylazine and underwent transcardial perfusion, brain removal, fixation, cryoprotection, and sectioning as previously described (28). Fixed primary neuron coverslips or free-floating 40 μm brain sections were immunostained using the protocol described in (28) (see Table 1 for primary and secondary antibody details). Brain sections were mounted onto Superfrost Plus microscope slides (Fisher Scientific) and cover slipped with prolong gold antifade medium with DAPI (ThermoFisher Scientific). Primary neuron coverslips were mounted onto colorfrost microscope slides (Fisher Scientific) with prolong gold antifade medium with DAPI.

Table 1.

Antibodies used for immunohistochemistry

Level	Antigen	Host	Conjugate	Dilution	Source
Primary	Atf6α	Mouse	–	1: 200	Santa Cruz sc-166659
Primary	Atf6β	Rabbit	–	1: 200	Novus NBP1-91687
Primary	ChAT	Goat	–	1: 100	Millipore AB144P
Primary	CrebH	Mouse	–	1: 50	Santa Cruz sc-377332
Primary	GABA	Mouse	–	1: 300	Sigma A0310
Primary	GAD67	Mouse	–	1: 1000	Millipore mab5406
Primary	Ire1α	Rabbit	–	1: 50	Santa Cruz sc-20790
Primary	Ire1β	Goat	–	1: 50	Santa Cruz sc-10511
Primary	LaminA/C	Mouse	–	1: 200	Cell Signaling 4777
Primary	LaminB1	Rabbit	–	1: 500	Abcam ab16048
Primary	LaminB Receptor	Rabbit	–	1: 800	Abcam ab122919
Primary	mab414 (NPC Proteins)	Mouse	–	1: 800	Abcam ab24609
Primary	MAP2	Guinea Pig	–	1: 1000	Synaptic Systems 188004
Primary	Nesprin2Giant	Rabbit	–	1: 200	Gregg Gunderson Laboratory
Primary	Nup153	Mouse	–	1: 300	Abcam ab96462
Primary	Nup358/RanBP2	Rabbit	–	1: 100	EMD Millipore ABN1385
Primary	Ran	Mouse	–	1: 200	BD Biosciences 610341
Primary	Parvalbumin	Mouse	–	1: 500	Swant 235
Primary	Somatostatin	Rabbit	–	1: 500	Abcam ab103790
Primary	Sun1	Rabbit	–	1: 250	Abcam ab124770
Primary	Ubiquitin	Rabbit	–	1: 500	Dako Z0458
Primary	UbiquitinK48	Rabbit	Ax488	1: 100	Millipore 05-1307-AF488
Secondary	Anti-Goat	Donkey	Ax555	1: 800	ThermoFisher A-21432
Secondary	Anti-Mouse	Donkey	Ax488	1: 800	ThermoFisher A-21206
Secondary	Anti-Mouse	Donkey	Ax555	1: 800	ThermoFisher A-31570
Secondary	Anti-Mouse	Donkey	Ax647	1: 500	ThermoFisher A-31571
Secondary	Anti-Rabbit	Donkey	Ax488	1: 800	ThermoFisher A-21206
Secondary	Anti-Rabbit	Donkey	Ax555	1: 800	ThermoFisher A-31572
Secondary	Anti-Rabbit	Donkey	Ax647	1: 800	ThermoFisher A-31573
Secondary	Anti-Guinea Pig	Goat	Ax555	1: 800	ThermoFisher A-21435

Brain sections and primary neurons were imaged under epifluorescence using a Zeiss AxioImager M2 microscope with Apotome. All images were taken using either a 20x (Zeiss 20x/0.5 EC Plan NeoFluar) or 63x (Zeiss 63x/1.40 Oil Plan Apo) objective with apotome structured illumination. 17 μm (for tissue) or 5 μm (for primary neurons) z-stacks were taken at 63× (1 μm steps between images), and maximum intensity projections were produced. Staining, imaging, and analysis were performed in parallel for all replicates.

Unbiased stereological cell counting was performed using the optical fractionator probe in StereoInvestigator software (MBF Bioscience) as previously described (28). 50×50 μm counting frames in 300×300 μm grids with 14 μm Dissectors and 1.5 μm guard zones were used for counts of cells containing perinuclear ubiquitin accumulation. 75×75 μm counting frames in 200×200 μm grids with 14 μm dissectors and 1.5 μm guard zones were set for counts of PV+ cells in the reticular thalamus. Cell counts were normalized to the volume of tissue measured.

Ran and NLS-mCherry-NES nuclear/cytoplasmic ratios were calculated as described in (54), using ImageJ to calculate mean pixel density per μ² from 63x apotome images of a single plane through the nucleus of individual neurons containing normal ubiquitin distribution or perinuclear ubiquitin rings. DAPI was used to outline the nucleus and MAP2 or mCherry delineated the cell soma.

Quantitative real-time PCR

Littermate male and female mice were sacrificed with a lethal dose of ketamine/xylazine, followed by decapitation and brain removal. Brains were flash frozen in dry ice-chilled isopentane and sectioned into 1mm slabs. Regions containing ventral forebrain were dissected with a razor blade, immediately homogenized in buffer RLT containing β-mercaptoethanol (Qiagen RNeasy Micro Kit), and RNA extraction was performed according to manufacturer’s instructions. RNA content and integrity was assessed using an Agilent 2200 Tapestation, and samples were frozen at −80 °C overnight. 100ng RNA was used as starting material for each biological replicate, and first-strand cDNA synthesis and cDNA target template preamplification were performed according to the RT² profiler PCR array handbook (Qiagen). 25  µl of ‘PCR components mix’ (2x RT² SYBR Green mastermix, cDNA, and RNase-free water) were loaded into 96 well plates containing pre-loaded primers for the mouse unfolded protein response RT² profiler array (including genomic DNA controls, reverse transcription controls, and positive PCR controls; Qiagen #PAMM-089Z), sealed with optical thin walled caps, and real-time PCR was performed in a BioRad MyIQ thermal cycler (cycling conditions according to Qiagen RT² profiler PCR array handbook). Baseline and thresholding values were defined such that the positive PCR controls exhibited a threshold cycle (C_T) of 20 ± 1 for all arrays, and C_T values were exported to excel. Data were analysed using the SABiosciences RT² Profiler PCR Array Data Analysis software version 3.5 (http://pcrdataanalysis.sabiosciences.com/pcr/arrayanalysis.php; date last accessed November 30, 2017). Quality control checks for reproducibility, reverse transcription efficiency, and genomic DNA contamination were performed. ΔC_T values were calculated for samples passing quality control by normalizing the raw data for each gene of interest to the arithmetic mean of five housekeeping genes. Fold change values between control and Dlx-Tor1a CKO groups were calculated using the 2^^-(ΔΔC_T⁾ method.

Statistics

Data are reported as mean ± SEM. Student’s t-tests, one-way ANOVA, and two-way ANOVAs were performed using Graphpad Prism software (version 7), and post hoc Sidak’s or Bonferroni’s multiple comparisons tests were performed when significant main effects were observed (P < 0.05).

Supplementary Material

Supplementary Material is available at HMG online.

Acknowledgements

We thank Roger L. Albin and members of the Dauer lab for helpful discussion and critical reviews of the manuscript.

Conflict of Interest statement. None declared.

Funding

National Institute of Neurological Disorders and Stroke (RO1NS077730) and Tyler’s Hope for a Dystonia Cure to WTD, and Parkinson’s Foundation-American Parkinson Disease Association Summer Student Fellowship (PF-APDA-SFW-1750) to COR.