Dear Dr. Droppelmann,
thank you for your interest in our manuscript. In the following I will try to answer your questions:

"I like your paper, specially for the controversial aspects.
I have some comments and questions about the results:
1) In the figure 1 said that two CTF are analyzed but is not specify in the results. Is used the 206-424 or the 174-414? or both?"

CTF 206-414 was used in chick, CTF 174-414 was used in flies. The reason is simple. We had generated CTF174-414 based on an educated guess, and generated the TDP-43CTF transgenic flies based on this fragment. Later, CTF 206-414 was reported to represent the dominant truncation fragment in ALS patients. From this point onwards, we consequently used this particular fragment, such as the eventual chick experiments. We assume that the near absence of toxicity of CTF observed in both systems is due to the principal equivalence of the two versions. Thus, so far, we did not generate new transgenic fly lines based on the CTF 206-414 version that may more accurately reflect the CTF accumulating in ALS/FTLD.

"
2) Figure 2. Why the levels of endogenous TDP-43 are so low in the transfected cells (ex. 2b or 2c) compared with unstranfected control (fig.2a)? different laser intensities? different gain? is not specified in the text.
"

You are right that the levels of endogenous TDP-43 look lower in cells that have been transfected compared to untransfected cells. However, this is not the case. We have used less exposure time to visualize transfected cells, since the TDP-43 antibody also recognizes transfected TDP-43 and similar exposure times would have caused the images of transfected cells to be oversaturated.

"3) Figure 5Bii. The statistics analysis is not clear. If each condition is compared with the vector bar I don't believe that the difference between CTF/NTF and vector are not significant. But this does not affect the interpretation of the data. However, in 5Cii the statics looks according the graph.
"

For each construct MN loss (%) was the ratio between Isl1/2+ cells in the motor column of control (non-electroporated) and the electroporated hemicord. “Significance” was calculated based on the thus determined MN loss (%) for each TDP-43 construct compared to the MN loss (%) obtained with the vector control. Since due to technical reasons DNA electroporation by itself was observed to trigger a small hike in MN loss on the electroporated side, this more accurately reflected ‘true’ MN loss due to TDP-43 activity than by calculating the significance based on comparing electroporated vs. non-electroporated side for each construct.
NTF/CTF showed slightly significance calculated thus by t-test; but due to the specific experimental setup we assume that the differences are not suficient to relate this to any specific differences in the activity of these constructs compared to say “FFLL”. We conclude that the calculated significances for NTF/CTF, but not FFLL are due to the small ‘experiment-to-experiment variation’ inherent to the method based on transient transfection. Such subtle differences can only be meaningfully resolved by assuring near-equalized and stable expression, such as achieved by single-site insertion transgenes.

"4) Why in the figure 3 the co-localization in Drosophila between nuclear TDP-43 and DNA look so poor?
"

In the presented confocal sections, not all cells have the same level of TDP-43 expression (according to specification in development), this is why some nuclei do not stain as robust for TDP-43. The co-localization with DNA is obvious, although might be even better seen by looking at 3D data of the entire confocal Z-stack. Compared to cytoplasmically localized TDP-43[∆NLS], however, we think the difference is obvious. In addition, we provide additional pictures in figure S1. Herein, in register with the large cell and nucleus size of the salivary gland cells co-localization becomes more obvious.

"About the conclusions:
I don't know if the title of the paper is completely demonstrated. An important part of the conclusions are based in the work with FFLL mutant. But these mutations do not eliminate completely the RNA binding activity of TDP-43. In fact, in the figure 1B can be observed a low RNA binding activity with (UG)12. Has been demonstrated that several mutants with low bound to (UG)12 have still a significant binding activity with other RNA targets. I missed in the paper a deletion (total or partial) of RNA binding domain to support your main finding."

You are right, we do not include a construct which selectively eliminates the activity by both RRMs. We mention precisely this point in the manuscript: ”At present, however, we cannot strictly exclude that TDP-43[FFLL] may retain residual RNA/DNA binding activity exerted by the remaining first RRM, which may underlay the slightly higher toxicity of TDP-43[FFLL] compared to TDP-43[CTF]. However, since “CTF” lacks RRM1 and is not toxic, this in essence supports our findings with “FFLL”; especially since RRM1 seems to account for the vast majority of (UG)12 RNA binding. Moreover, our data also suggests that not only RNA-binding is required to facilitate toxicity, since “NTF” (which is nuclear localized, possesses both RRMs, but selectively lacks the C-terminal segment) did not display any detectable toxicity when expressed in chick motor neurons. We think it is likely that neurotoxicity exerted by TDP-43 requires both, RRM-mediated RNA binding activity, plus the C-terminal segment - and that this may reflect the necessity to couple RNA-bound TDP-43 to other proteins relevant at least for over-expression-driven toxicity.

"Personally, I think that the most important implication of the paper is not discussed. I think that these results suggest that the mutations of TDP-43 are not the cause of ALS. Figure 4a clearly supports this hypothesis. You show clearly that RNA-binding activity and a complete protein is necessary for the toxic effect of TDP-43. But, at the same time you show that the most toxic protein is the TDP-43wt, some mutations in fact have an important protector effect compared with the wt. This is according with the hypothesis that TDP-43 mutants are only polymorphism observed in the population and are not related with the cause of the pathology. May be, actually, they could protect of the toxic effects of the mis-regulated TDP-43 in ALS."

This is an intriguing and of course provocative interpretation of the results presented by us. However, we think more work has to be done to finally clarify this issue. We think our data shows that most of the model systems used so far (or the way in which they are widely used) to infer impacts of ALS-linked mutations may not be suitable to finally solve this problem. Proper analysis of familial TDP-43 linked ALS cases will hopefully provide more convincing information then available at the moment. We think the ultimate acid test would be ‘knock-in’ experiments (e.g. in mouse) to introduce ALS-linked mutations into endogenous TDP-43 to unambiguously understand the role of these modifications in neurodegeneration in vivo.

I hope we satisfactorily answered all of your questions and comments, and thank you for your insightful suggestions.
Sincerely,
Aaron Voigt
Till Marquardt
Jörg B. Schulz