Nuclear export of misfolded SOD1 mediated by a normally buried NES-like sequence reduces proteotoxicity in the nucleus

Essential revisions:

1) A major conclusion of the study is that nuclear SOD1 is more toxic than cytosolic SOD1, suggesting that nuclear export by the mechanism described here is protective. However, the experiments to demonstrate this are not optimal, and some controls are missing. The authors have enforced wild type-like localization of mutant SOD1 by addition of a NLS. The consequence is that this rather abundant and over-expressed protein is constitutively imported and exported endlessly, thereby imposing a greater burden on the nuclear export machinery than would ordinarily be the case in the absence of a NLS. Thus, the increased toxicity might well be due to this (or to the appended NLS somehow having a toxic effect), and not necessarily to SOD1 localization per se. The more suitable and precise experiment is to determine whether L42Q (which is in fact the normal variant in mouse and not expected to disrupt folding by itself) changes the toxicity of G85R or other SOD1 mutants. By doing this, normal wild type-like localization is restored without adding any extra burden to the cell's trafficking pathways. In both the NLS and L42Q experiments, it is crucial to include wild type SOD1 as a control. This would ensure that any effects on toxicity are mutant-specific. These experiments in cultured mammalian cells are straightforward and the authors have all the tools to do it in a reasonable span of time. The referees and editor acknowledge that the analogous experiments in worms would take longer, but it is sufficiently important to wait for such results even if it exceeds the two-month revision timeline.

We agree with the reviewers that enforced wild type-like localization of mutant SOD1 by addition of a NLS could change it into a nuclear-cytoplasmic shuttling protein, which may impose a greater burden on the nuclear export machinery and thus may alter the toxicity of mutant SOD1. Thanks to the reviewers for suggesting the more suitable and precise experiment, that is, to use the L42Q substitution to restore mutant SOD1 nuclear localization instead of amending a NLS. Using this approach, we have confirmed that, as predicted, L42Q substitution restored SOD1^G85R nuclear localization (Figure 6A). Using an anti-NLP immunoprecipitation approach (under native conditions), we demonstrated that the L42Q substitution does not cause exposure of the NES-like sequence in wild type SOD1 (Figure 6B). L42Q substitution also did not change the amount of SOD1^G85R immunoprecipitated by anti-NLP, suggesting that it does not alter SOD1^G85R structure (Figure 6B). We also showed that the toxicity of SOD1^L42Q to NSC34 cells is similar to that of wild type SOD1 (Figure 6C). These results strongly support that L42Q substitution has minimal effects on SOD1 folding. Therefore, SOD1^G85R/L42Q differs from SOD1^G85R only in its ability to be exported from the nucleus. In sum, these results set the stage for further testing the toxicity of misfolded nuclear SOD1 by comparing SOD1^G85R (misfolded, exported) versus SOD1^G85R/L42Q (misfolded, retained) in a C. elegans model of ALS.

We generated several new transgenic C. elegans lines: SOD1^G85R/L42Q and the control SOD1^L42Q. Three independent lines each for SOD1 variant were made, and these animals were compared with SOD1^G85R and wild type SOD1 lines using several approaches, including subcellular localization, effects on locomotion (body bending), and survival. All of the SOD1 proteins were tagged with YFP for direct observation by fluorescence microscopy. We found that SOD1^L42Q behaves similarly to wild type SOD1 for all aspects examined, which further confirms the L42Q substitution does not cause misfolding of SOD1 (Figures 7–9). As observed in cultured cells, SOD1^G85R was cleared from the nuclei of motor neurons in both L1 and adult C. elegans (Figure 7A,B). In contrast, SOD1^G85R/L42Q was localized in both the cytoplasm and the nuclei in motor neurons in L1 C. elegans (Figure 7B). In adult C. elegans, the nuclear localization of SOD1^G85R/L42Q was only observed in a fraction of the neurons (Figure 7B). A possible explanation is that cytoplasmic aggregation of the mutant protein prevents its entry into the nucleus. Indeed, we found that in adult C. elegans, like SOD1^G85R, SOD1^G85R/L42Q formed large cytoplasmic aggregates (Figure 7B). Biochemical characterization demonstrated that SOD1^G85R and SOD1^G85R/L42Q lines express the transgene at similar levels and also form similar amount of insoluble aggregates (Figure 7C). These results support that SOD1^G85R/L42Q differs from SOD1^G85R only in its nuclear localization.

We obtained two lines of evidence suggesting that nuclear localization of SOD1^G85R is toxic to neurons in C. elegans. First, three lines of SOD1^G85R/L42Q worms all exhibited more severe defects in locomotion compared to SOD1^G85R animals (Figure 7D). As controls, three lines of SOD1^L42QC. elegans showed body bending rates similar to their wild type counterparts (Figure 7D). Second, there was a strong effect on survival for all three SOD1^G85R/L42Q lines, where lifespan decreased compared to the SOD1^G85R line (Figure 8A). As controls, all three SOD1^L42Q lines were found to have similar survival curves as the wild type SOD1 line (Figure 8A). In search for the potential cause of death, we found that about 70% of SOD1^G85R/L42Q animals had egg-laying defects with an average of 4.2 eggs/worm hatched inside mothers, known as “bagging”. The bagging defect for SOD1^G85R/L42Q was significantly worse than was found for SOD1^G85R animals (Figure 9). About 17% of SOD1^G85R animals had an average of 2.4 eggs/worm hatched inside mothers (Figure 9). It is likely that loss of muscle function caused by severe motor neuron degeneration is responsible for these bagging phenotypes in SOD1^G85R/L42Q animals. Future studies are warranted to understand the mechanisms underlying the bagging phenotype observed for SOD1^G85R/L42Q animals.

These results confirm that restoring nuclear localization of SOD1^G85R, via an L42Q substitution, can cause additional toxicity, whereas nuclear export of misfolded SOD1 has a protective effect to the cells. Because the potential complex effects associated with the amended NLS, as pointed out by the reviewers, we have removed all data using NLS-tagged SOD1^G85R in this revised manuscript.

2) The experiment to suggest that nucleocytoplasmic transport is affected by SOD1 mutants (Figure 6E) is incomplete and rather preliminary. The observation of steady state differences in polyA FISH signal is insufficient to conclude an effect on transport, since other explanations such as altered mRNA processing or polyadenylation are possible. If the authors wish to make conclusions about transport, then the authors need to measure transport more directly. Furthermore, a key control in any experiments aimed at elucidating toxicity mechanisms would be to ensure that the effect is due to misfolded SOD1, and not just a highly-expressed NES-containing protein. Thus, it seems important to test whether their mCherry-P1 and mCherry-P2 constructs have any effect on mRNA localization (the prediction being that neither would have any effect in the authors' model, but P1 would have an effect in the alternative model above). The referees acknowledge that properly elucidating the effect on nucleocytoplasmic transport as a potential contributor to toxicity would require extensive studies. Thus, if the authors are not prepared to pursue such investigations for this paper, the recommendation is to remove Figure 6E and F and only speculate in the Discussion about toxicity mechanisms.

We agree with the reviewers that the data to suggest that nucleocytoplasmic transport is affected by SOD1 mutants (Figure 6E) was rather preliminary. As suggested by the reviewers, we have removed this data and briefly discussed this possibility in the Discussion section.

3) An important unanswered question is whether the exposure of NES by SOD1 mutations is an intrinsic property of the mutant protein, or is the result of an unfolding activity from the cell. A key experiment that may distinguish between these possibilities is to determine whether recombinant SOD1 purified from E. coli has the NES exposed in a mutant-specific manner, or if this is only a feature observed in the cellular context. Such an experiment would provide greater insight into the mechanism of NES exposure.

As discussed with the editor, we have clarified how the experiments were done for Figure 3B and stated that GST-tagged SOD1 mutants, but not wt SOD1, interact with recombinant CRM1.

Disruption of the NES-consensus in SOD1 mutants also abolished their interaction with CRM1 (Figure 3B). As suggested by the editor, we have now added new data showing that the interaction between CRM1 and SOD1 mutants is RanGTP-dependent using recombinant proteins (Figure 3C). These results favor a model in which the exposure of NES by SOD1 mutations is an intrinsic property of the mutant protein.