TorsinA and torsion dystonia: Unraveling the architecture of the nuclear envelope

The nuclear envelope (NE) is the membrane structure that forms the boundary of the nucleus in eukaryotes (1). Not only does the NE control the trafficking of macromolecules between the nucleus and the cytoplasm (2), but it also provides an anchoring site for chromosomes at the nuclear periphery (3). During the past several years, mutations in certain NE proteins have been shown to cause a diversity of human diseases, including muscular dystrophy, neuropathy, lipodystrophy, and the premature aging condition progeria (1, 4, 5). The list continues to expand, and recent work from Goodchild and Dauer (6) and work from Naismith et al. (7), which was published in a recent issue of PNAS, have indicated that torsinA, the product of the DYT1 gene that is mutated in early-onset torsion dystonia, functions in the NE, where it may cause this disease. Moreover, these papers raise the provocative possibility that torsinA may be involved in connecting the NE to the cytoplasmic cytoskeleton.

The NE is a specialized subdomain of the endoplasmic reticulum (ER) (Fig. 1A). It consists of inner and outer nuclear membranes that are joined at the nuclear pore complexes (NPCs), the channels for molecular transport across the NE (2). The outer nuclear membrane is morphologically continuous with the more peripheral ER, providing continuity between the perinuclear luminal space and the lumen of the peripheral ER (Fig. 1A). The inner nuclear membrane is lined by the nuclear lamina, a filamentous protein meshwork that consists mainly of a polymer of lamins, proteins of the intermediate filament family (4). The lamina also contains a number of less abundant components, including peripheral and integral membrane proteins of the inner nuclear membrane that are anchored at the lamina (8, 9). Lamins A, B1, B2, and C are the four major isotypes expressed in mammalian cells (4). Most diseases of the NE that have been described are caused by mutations in the gene encoding lamin A and its splice variant lamin C. Although lamins A/C are expressed in most differentiated cells, disease-causing mutations usually affect only a limited number of cell types.

Fig. 1.

Schematic model of the NE and potential functions of torsinA. (A) Schematic diagram of the NE. ONM, outer nuclear membrane; INM, inner nuclear membrane; NPC, nuclear pore complex. Examples of peripheral and integral membrane proteins associated with the lamin polymer (dark brown) are shown. (B) Model for interaction between the INM and ONM and association of cytoskeletal filaments with the NE. INM and ONM may be linked through the luminal domains of transmembrane proteins (on the left). Microtubules and actin filaments are proposed to interact with transmembrane protein(s) of the ONM, which in turn may associate with lamina-associated transmembrane proteins of the INM via the luminal domains (on the right; ref. 22). TorsinA may regulate these luminal interactions between INM and ONM proteins.

TorsinA and Early-Onset Torsion Dystonia

Early-onset torsion dystonia, the most severe form of hereditary dystonia, is characterized by sustained muscle contractions that induce twisting and repetitive movements in the legs and/or arms. It is an autosomal dominant disease that typically develops between the ages of 9 and 13 years and appears in ≈30% of the individuals who inherit a mutant DYT1 gene. Although the disease is caused by defects in the central nervous system, no overt structural abnormalities of neurons have been observed. The DYT1 gene product torsinA (10) is a member of the AAA+ superfamily of ATPases (11). AAA+ proteins comprise a superfamily of mechanoenzymes with a remarkably diverse set of functions (12, 13), including membrane trafficking, organelle biogenesis, proteosome function, and microtubule regulation. AAA+ ATPases commonly form hexameric rings and use the energy of ATP binding and hydrolysis to effect conformational changes in their substrates, often leading to dissociation of stable protein assemblies (12, 13).

TorsinA is the only AAA+ ATPase known to reside in the lumen of the ER (Fig. 1). Most cases of torsion dystonia are caused by deletion of a single glutamic acid residue at position 302 or 303 of torsinA (ΔE302/E303) (14). When overexpressed in cultured cells, the ΔE302/E303 mutant becomes strikingly concentrated at the NE (6, 7, 15), in addition to accumulating in cytoplasmic membrane bodies (16, 17). Moreover, patient fibroblasts heterozygous for the ΔE302/E303 mutant also show significant accumulation of torsinA at the NE relative to controls (6).

Goodchild and Dauer (6) and Naismith et al. (7) have dissected the molecular basis for this NE accumulation by analyzing the subcellular localization of point mutants of torsinA. The mutations, which targeted the highly conserved Walker A and B motifs in the ATP-binding site, were designed to mimic mutations in other AAA+ family members that block ATP binding and hydrolysis (18–20). One mutant contained a lysine to alanine substitution at amino acid position 108 (K108A) in the Walker A motif and was predicted to block ATP binding. This mutant was localized throughout the ER, similar to wild-type torsinA (6, 7). A second mutant contained a glutamic acid to glutamine substitution at position 171 (E171Q) in the Walker B motif and was predicted to support ATP binding but not hydrolysis. Because AAA+ ATPases generally are tightly associated with their substrates when ATP bound (12, 13), this mutant was predicted to act as a substrate “trap.” Strikingly, the E171Q mutant strongly accumulated at the NE (6, 7). This result argues that major substrate(s) of torsinA reside in the NE lumen. Presumably, wild-type torsinA is not highly concentrated at the NE at steady state due to its continuous cycling between substrate-bound and free forms, the latter of which would rapidly equilibrate between the NE and peripheral ER (Fig. 1A).

Cellular Clues to Disease Etiology

The ΔE302/E303 mutation occurs in a region adjacent to the core ATPase domain and also may promote an ATP-stabilized form of torsinA, as predicted for the E171Q mutant, because a double mutant containing both K108A and ΔE302/E303 no longer accumulates at the NE (7). The ΔE302/E303 mutant protein also promotes the accumulation of wild-type torsinA at the NE (6), suggesting that the mutant can form heterooligomers with the wild-type protein. This can explain why the DYT1 mutation causing torsion dystonia is dominant.

How could accumulation of mutant torsinA at the NE functionally compromise motor neurons and cause torsion dystonia? Potential insight into this question comes from detailed morphological analysis of cells expressing the mutant torsinA proteins. The E171Q and ΔE302/E303 torsinA mutants were observed to concentrate in regions of the NE lumen adjacent to the nuclear lamina rather than at the NPCs (6, 7). Moreover, the NE of cells overexpressing E171Q revealed several types of conspicuous ultrastructural aberrations, including an abnormally close apposition of inner and outer nuclear membranes, gross dilation of the perinuclear lumen, and herniation of both inner and outer membranes (7). Intriguingly, these changes in NE structure are reminiscent of the architectural defects in the NE that are seen in cells expressing disease-causing mutants of lamin A and in cells derived from mice containing a homozygous null mutation for the lamin A/C gene (1, 5, 21).

Based on the abnormalities of NE structure that occur in cells expressing mutant lamins A/C or lacking these proteins altogether, a number of mechanisms have been proposed for the appearance of the disease phenotype in certain cells types (1, 5). The mutations could compromise the mechanical stability of the nucleus, modify gene expression due to altered chromosome attachment to the NE, or cause changes in the interaction of cytoskeletal components with the NE. Considering the effects of mutant torsinA on NE structure, it is plausible that early-onset torsion dystonia arises by similar mechanisms. Clearly, identifying the substrates of torsinA is of paramount importance for answering this question.

The Emerging Cytoskeletal Connection

The relatively uniform spacing of ≈50 nm between the inner and outer nuclear membranes potentially could be maintained by interactions between the luminal domains of transmembrane proteins of the inner and outer membranes (Fig. 1B). Because AAA+ ATPases frequently act as molecular chaperones to break or form protein interactions (12, 13), torsinA could act in the lumen of the perinuclear space to modulate these interactions and, when functionally impaired, could lead to significant structural aberrations in the NE as observed (7). Interactions between transmembrane proteins of the inner and outer membrane also could be fundamental to the connection of nuclei to the cytoplasmic cytoskeleton (Fig. 1B and ref. 22).

The actin and tubulin cytoskeletons have long been appreciated to have an important role in the positioning and migration of nuclei in cells (22, 23), but until recently the molecular basis for the interactions of these filament systems with the nucleus has been obscure. Because the nuclear lamina provides a mechanically rigid scaffold for the NE (4), cytoskeletal attachments to the nucleus are expected to involve connections to the lamina, most likely via transmembrane proteins of the outer and inner nuclear membranes (22).

A flurry of recent work has identified NE proteins involved in interactions with actin filaments and microtubules. The protein ANC-1 is an apparent transmembrane protein of the outer nuclear membrane implicated in anchoring nuclei to the actin cytoskeleton in hypodermal syncytial cells of Caenorhabditis elegans (24). The association of ANC-1 with the NE requires UNC-84, which is suggested to be a lamin-binding inner membrane protein (22, 25). Interestingly, ANC-1 has mammalian homologues (called Syne1/2 and other names; ref. 22), as does UNC-84, which shares the conserved “Sun” domain with two mammalian NE proteins (22). Moreover, in C. elegans, the attachment of the microtubule-binding motor dynein to the NE, which is implicated in centrosome anchoring to the nucleus, also involves a Sun domain protein (26). An involvement of torsinA in cytoskeletal interactions with the nucleus is suggested by the finding that mutations in the gene for OOC-5, a NE-enriched torsinA homolog in C. elegans, impair nuclear rotation in the early embryo (27). Thus, one of the major questions that emerges from the recent studies on torsinA is whether the association of the cytoskeleton with the nucleus is altered in cells expressing the mutant protein. If such changes occur, this might affect signal transduction to the nucleus and gene expression and, as a consequence, may lead to disease.

Notwithstanding a potential role of the NE in the etiology of torsion dystonia, it remains possible that the disease results from altered torsinA interactions with substrates in the peripheral ER. Moreover, recent evidence suggests that a fraction of torsinA is localized outside the lumen of the ER in the cytosol and interacts with kinesin (28), raising another potential disease mechanism. Whatever the molecular basis for early-onset torsion dystonia, an analysis of the substrates and mode of action of torsinA at the NE promises to yield a wealth of new insight into the functional organization of the NE, which is just beginning to be unraveled.