LAP1 has been consistently identified as a NE-localized binding partner of torA [12, 14, 15, 20], and it appears that LAP1 is required for NE-localized torA activity . However, we find that the NE localization of both torA-WT and torA-ΔE require torA-motifs that are unimportant for the LAP1 interaction. Furthermore, we show that SUN1 depletion removes torA-WT and torA-ΔE from the NE, while LAP1 is not required for torA-ΔE to concentrate in the NE. Therefore, although LAP1 is the confirmed binding partner for ATP-bound torA, LAP1 is neither the sole NE-localized torA interacting protein nor the partner of disease-associated torA-ΔE.
The multiple differences between torA-E171Q and torA-ΔE reveal that distinct mechanisms underlie their superficially similar ability to concentrate in the NE lumen [13, 15]. While it appears that the E171Q mutation traps torA in an ATP-bound state, the mechanism by which ΔE affects torA remains unclear. A previous genetic analysis found that torA-ΔE cannot compensate for torA loss, therefore revealing that this isoform is hypoactive or inactive . Biochemical analysis has also demonstrated that torA-ΔE may fail to bind ATP, or perhaps fail to undergo a structural rearrangement on ATP-binding , either of which may explain why torA-ΔE cannot form the ATP-dependent interaction with LAP1 or LULL1 [13, 15]. Despite these inhibitory effects, our data now highlight that ΔE does not generate a 'dead' torA isoform that is incapable of all binding interactions. Instead, the abnormal association between torA-ΔE and SUN1 suggests that ΔE 'traps' torA in a specific conformation that is distinct to the ATP-bound state. This finding also provides further evidence that the DYT1 ΔE mutation has a gain-of-function activity, in addition to producing the previously characterized impairment of torA function.
Our data indicate that torA associates with the SUN1 LINC complex component, and that the DYT1 mutation abnormally promotes or stabilizes this interaction. SUN1 is an inner nuclear membrane component of the LINC complex that couples the nuclear interior to cytoskeletal networks. SUN1 has a nucleoplasmic domain that mediates interaction with lamins, an extended membrane spanning hydrophobic region, and an approximately 50 kD luminal domain that appears to mediate both SUN-protein multimerization and interaction with KASH-domains of outer nuclear membrane Nesprin proteins [30–33]. There is loss-of-function evidence for functional redundancy between SUN1 and the similarly widely expressed and homologous SUN2 [22, 23]. It is therefore surprising to find that SUN1 is specifically required for torA localization in the NE. We cannot rule out that SUN1 is selectively required because NIH-3T3 cells predominantly rely on SUN1, or that SUN1 siRNA more efficiently depletes SUN1 from the NE. However, there are reported differences in SUN1 and SUN2 characteristics [20, 22, 24, 34], and our finding that SUN1 is selectively important for torA localization provides further evidence that these proteins have distinct, as well as overlapping cellular roles.
We do not resolve whether torA interacts directly or indirectly with SUN1. However, there is evidence against the possibility that SUN1 loss causes general NE disruption that removes torA by a highly indirect mechanism. Several previous studies have shown that NE structure and LINC complex function are largely normal in the absence of SUN1, and that combined SUN1 and SUN2 loss is required to perturb NE morphology [23, 30, 35, 36]. Consistent with these findings, we observe that the NE-localization of LAP1-associated torA-E171Q is undisturbed by SUN1 loss, and that other inner nuclear membrane proteins are normally localized in cells lacking SUN1 (not shown). Furthermore, we also find that torA-ΔE colocalizes with SUN1 in puncta that lie outside of the NE. Thus, while it is possible that torA indirectly associates with SUN1, any such indirect interaction would be mediated by a protein or proteins that are also tightly coupled to SUN1. Unfortunately, we were unable to biochemically verify that a direct or indirect interaction exists between torA and SUN1. Like many inner nuclear membrane proteins, SUN1 solubilization requires that ionic detergents disrupt binding to the nuclear lamina and chromatin . This treatment necessarily also impairs luminal interactions and a failure to detect biochemical association does not preclude that SUN1 and torA interact in vivo. Furthermore, despite our negative findings with anti-GFP immunoprecipitation of (GFP)torA isoforms, one previous report described that nesprin KASH-domains co-immunoprecipitate with anti-torA antibodies , which provides general support for an interaction between torA and a LINC complex component. However, our finding that Nesprin2 depletion increases, rather than decreases, the amount of NE-localized torA-ΔE, indicates that this previous result reflects an indirect, rather than direct, interaction between torA and nesprins. It is unclear why Nesprin2 loss increases torA-ΔE localization in the NE. However, an increased number of torA-ΔE binding sites, perhaps caused by compensatory upregulation of other LINC components, could account for both our observations and the previous association between torA and the nesprin KASH-domain.
Our findings support a model where torA interacts with at least two different NE proteins . The existence of additional NE-localized partners suggests that torA either operates on multiple substrates and/or that some identified torA interacting partners have a regulatory function and are not subject to torA AAA+ activity. Finding that the SUN1-dependent localization requires a putative substrate interaction residue [26, 28], Y147, suggests that SUN1 or the LINC complex is a substrate affected by torA AAA+ activity. Surprisingly, this also suggests that Y147A destabilizes an interaction that is distinct to the ATP-bound torA state that associates with LAP1 and LULL1. This insensitivity of LAP1 and LULL1 binding raises the possibility that these proteins are not torA substrates. Furthermore, these data also suggest that torA could operate using an atypical biochemical AAA+ mechanism; a hypothesis that is supported by the presence of a non-canonical nucleotide binding motif in the torA AAA+ domain that appears to convey preferential binding to ADP, rather than ATP .
The possibility that torA activity modifies the LINC complex is also supported by two previous reports. In one study, torA loss appeared to remove the Nesprin3 LINC complex component from the NE, suggesting that torA normally maintains intact LINC complexes . In contrast, a separate study made the reverse observation, and found that overexpression of the LULL1 torA-binding partner appeared to induce a NE-localized torA activity that removed SUN2 and Nesprin2 . This LULL1-activated torA was strongly concentrated in the NE of cells that lacked SUN2 and Nesprin2, which supports our finding that SUN1 is important for the NE retention of torA. Furthermore, since AAA+ activity often disassembles otherwise stable protein complexes [6, 38], this data is also consistent with torA dissociating SUN1 binding interactions to release SUN1-associated proteins from the NE lumen. The relationship between the LULL1-overexpression paradigm and physiological torA function is unclear, and there are some AAA+ proteins that promote protein complex formation or remodeling, rather than disassembly . Therefore, a difference between the actions of physiological levels of torA activity, versus overactive torA enzymes, could explain the discrepancy between these two studies.
While our study emphases a relationship between torA and the LINC complex, it is also clear that LAP1 has a central role in torA function and, notably, LAP1 and torA gene knock-out result in the same NE membrane abnormalities . Although this association does not resolve whether LAP1 activity is upstream (an essential torA regulator/co-factor) or downstream (a substrate) of torA activity, it is clear that LAP1 is the partner of the ATP-bound form of torA that is typically the substrate-associated state of a AAA+ protein [13, 15]. There are examples of promiscuous AAA+ enzymes that operate on several distinct substrate proteins, and it is possible that torA is a multi-functional AAA+ protein that operates on both LAP1 and the LINC complex. Furthermore, since adapter proteins often regulate the substrate selection of multifunctional AAA+ proteins , the importance of Y147, residues 22-40, and the effect of ΔE, could be explained if these motifs affect torA interaction with an adapter that promotes the LINC complex association over LAP1 binding.
Our study is performed using cells that also express torA. It is therefore possible that these endogenous torA-WT subunits play a role in the association between (GFP)torA-ΔE and SUN1. Nevertheless, mixed torA-ΔE and torA-WT expression also occurs in DYT1 dystonia, and it is demonstrated that torA abnormally concentrates in the NE of DYT1 dystonia cells, as well as neurons ectopically expressing torA-ΔE [10, 29]. Thus, our findings suggest that the LINC complex of DYT1 dystonia neurons is abnormally associated with torA-ΔE. There are several mechanisms by which an abnormal association between torA-ΔE and SUN1 could negatively impact cell function. Firstly, torA-ΔE occupation of binding sites could prevent functional torA enzymes from accessing the LINC complex, and therefore inhibit torA activity in the event that this normally modifies the LINC complex. Secondly, there is evidence that torA-WT and torA-ΔE co-oligomerize, and that the abnormal SUN1-association of torA-ΔE is conferred to co-expressed torA-WT . This could result in torA-WT sequestration away from other molecular targets of torA AAA+ activity. A third possibility is that abnormal association with torA-ΔE directly impacts LINC complex assembly or function, independent of torA-WT activity. These mechanisms are not mutually exclusive, and it is possible that a combination of such loss-of-function and/or gain-of-function torA-ΔE actions underlie why DYT1 dystonia is dominant and the torA-ΔE generating mutation is the only identified cause of this disease.
Our study did not identify grossly abnormal LINC complex component localization in cells transfected to express (GFP)torA-ΔE. However, our assessment did not examine whether physiological torA-ΔE expression impacts LINC complex components as these are utilized during particular cellular behaviors, such as nuclear movement [23, 24]. It is unlikely that torA-ΔE expression completely ablates LINC complex activity, as genetic deletion of the complex in mice results in severe neurodevelopmental abnormalities , compared with undetectable or limited neuropathology in human DYT1 dystonia patients and heterozygous torA-ΔE expressing mice . However, given the importance of the LINC complex for nervous system development , it is conceivable that torA-ΔE driven abnormalities in LINC complex mediated events impact neuronal development to generate the circuit abnormalities that are thought to underlie the debilitating twisting movements of DYT1 dystonia .