Cryo-EM observation of the amyloid key structure of polymorphic TDP-43 amyloid fibrils

The transactive response DNA-binding protein-43 (TDP-43) is a multi-facet protein involved in phase separation, RNA-binding, and alternative splicing. In the context of neurodegenerative diseases, abnormal aggregation of TDP-43 has been linked to amyotrophic lateral sclerosis and frontotemporal lobar degeneration through the aggregation of its C-terminal domain. Here, we report a cryo-electron microscopy (cryo-EM)-based structural characterization of TDP-43 fibrils obtained from the full-length protein. We find that the fibrils are polymorphic and contain three different amyloid structures. The structures differ in the number and relative orientation of the protofilaments, although they share a similar fold containing an amyloid key motif. The observed fibril structures differ from previously described conformations of TDP-43 fibrils and help to better understand the structural landscape of the amyloid fibril structures derived from this protein.

Major comments: 1.The quality of the cryo-EM maps should be improved.The PDB validation reports show regions of continuous density running along the helical axis, instead of clearly separated molecular layers, which indicates that the maps may have converged on local minima.Guidance on how to spot and overcome this problem is provided in (Scheres. 2020. Acta Cryst D76, 94-101).In addition, the intermediate resolutions of the maps (~4 Å) may have led to inaccuracies in model building.At present, a relatively modest number of initial particles is used (26,881), suggesting that additional data may improve the maps.The quality of the maps may also be improved by additional rounds of 3D classification, CTF refinement and Bayesian polishing.
2. Are the authors sure that they are dealing with full-length TDP-43, as claimed in the manuscript?Data on the characterisation of the recombinant TDP-43 protein preparations should be shown.At a minimum, this should include chromatograms from the SEC step, as well as SDS-PAGE of the preparations before and after aggregation.This is particularly vital for TDP-43, because it is prone to proteolytic cleavage and aggregation during purification, as well as proteolytic cleavage during in vitro aggregation assays.
3. Please deposit the raw cryo-EM movies to EMPIAR.
Minor comments: 1.The centrifugation step used to pellet fibrils (13,000 g for 5 min) may result in many fibrils being discarded in the supernatant.For example, TDP-43 fibrils were found in the supernatant following centrifugation at 27,000 g for 10 min in (Arseni et al. 2022 Nature 601, 139-143).The authors should acknowledge this possibility.
2. Line 55, 'Amyloid fibril samples are typically polymorphic and contain multiple fibril morphologies,' does not apply for brain-derived fibril samples, which are often homogeneous.The authors should clarify that this refers to fibril samples generated in vitro.

Line 83, '
The observed fibril structures are constructed from similarly folded fibril proteins,' and line 121, '...differ in the number of the fibril protein stacks...' are potentially confusing.Do these sentences refer to protofibrils?Please clarify. 5. Line 84, '...exhibiting an amyloid key motif.'This is a central point of the manuscript.Therefore, it would be helpful to define this motif here, or even earlier in the manuscript, rather than waiting until the Results and Discussion sections.
6. Line 140, 'The abundance of the two structural forms varies in the mixed fibrils almost continuously from one to zero.' Does this refer to the proportion of mixed fibrils 0-100%?7. Line 149, 'The three fibril structures contain essentially the same fibril protein conformation that is formed by 45 residues (Gly304 to Gly348) from the LC domain.'Is there any evidence that the flanking residues to the N-and C-terminus of this region may be structured?It would be helpful for the reader to comment on this.8. Line 159, 'An unusual structural feature of the fibril core is the absence of buried ion pairs.'This statement needs to be qualified.There are many amyloid fibril structures that lack salt bridges.9. Line 231, 'Finally, we noted that our structure and all previous in vitro formed structures from TDP-43 differ from a fibril morphology that was purified from a patient.'This study purified TDP-43 fibrils from two patients with ALS and FTLD-TDP Type B, and two brain regions, prefrontal cortex and motor cortex.This should be corrected.10.Line 240, 'Therefore, the currently available data on TDP-43 fibrils show that in vitro formed fibrils do not necessarily match the structure of ex vivo fibrils.'It would also be helpful for the reader to point out that the sequence that forms the structured part of the fibril also does not necessarily match (G304-G348 here vs. G281-Q360 in ALS and FTLD).
11. Line 243 '...but that more complex and more difficult to establish spontaneously.'Should this read, '...but that is more complex and more difficult to establish spontaneously.' ?12. Line 244 'One reason to establish these specific folds is the higher biological stability of ex vivo fibrils compared with most in vitro formed fibrils.'This claim needs to be supported by data or references.
13.The model of the rotator/shaker used during the aggregation of TDP-43 should be stated to help the reader interpret a rotation of 40 rpm.
14.The program/method used to carry out global refinements of the atomic coordinates, including refining B-factors (i.e.Phenix or REFMAC) should be specified in the Methods.15.In Figure 1a, it would be helpful to add the position of the NLS Reviewer #2 (Remarks to the Author): This short manuscript describes cryo-EM determined three polymorphic structures of amyloid fibrils formed from the recombinant full-length TDP-43.The resolution of these structures (approx.4 A) is somewhat below typical resolution of cryo-EM structures previously reported for most other amyloid fibrils.Nevertheless, it is clear that the present structures are fundamentally different (both with regard to the size of the amyloid core and the overall fold) from those previously determined for the recombinant TDP-43 low complexity domain or TDP-43 fibrils extracted from patient brain.Even though high-resolution structures of fibrillar aggregates are always of interest, it is not obvious whether the present structures are of direct physiological/pathological relevance or just represent yet other polymorphs formed in the test tube (see also below).

Specific points:
1. Recombinant full-length TDP-43 is notoriously difficult to purify in a monomeric form.In this context, the authors need to clarify in the Methods section (which is somewhat cryptic in this regard) how the protein (which was applied onto size-exclusion column in the presence of a strong denaturant (6 M thiocyanate quinidine) was eluted from this column, and whether fibrils were formed in the absence or presence of the denaturant.Furthermore, the monomeric state of the purified protein needs to be documented.Those are important issues for the assessment of potential physiological relevance of these structure.2b -There are extra densities near the contact sites of two protofilaments in morphology 1-a and morphology 1-b.These can also be seen in the validation report provided by the authors.The authors may want to comment on these densities.2c -map-model FSC should also be provided.Furthermore, model resolution should be provided in SI table 1 or 2.

SI Figure
6.The authors provide a detailed analysis of beta-sheets formation (line 151-155, figure 4), which likely requires higher resolution model/map than the current results.Moreover, the beta-strand information was actually imposed by the author during the modeling (line 297-299), making it a circular argument.Minor Issue: Line 272-273 -"An initial 3D reference map was created using 3D initial model in Relion3.1……"The 3D initial model function in Relion3.1 does not have the function for generating helical models.The authors probably mean the inimodel2d program here?Sharma et al. used cryo-EM to analyze amyloid fibrils of TDP-43 that had been purified recombinantly and aggregated in vitro.They find that the fibrils form a mixture of ultrastructural polymorphs (i.e.fibrils with an essentially shared protofibril fold, but with a different protofibril number and/or protofibril packing interfaces).
Previous studies have used cryo-EM to analyze in vitro-generated amyloid fibrils of recombinant CTFs of TDP-43 (Cao et al. 2019. Nat Struct Mol Biol 26, 619-627;Li et al. 2021. Nat Commun 12, 1620;Kumar et al. 2023. Nat Neurosci 26, 983-996).Their structures, like those presented here, are different from the structures of TDP-43 fibrils composed of a mixture of full-length protein and CTFs that are found in the brains of individuals with ALS and FTLD (Arseni et al. 2022 Nature 601, 139-143).As such, the relevance of these in vitro-generated fibrils to disease is unclear.
Unlike previous reports of in vitro-generated fibrils, this study used a construct encoding fulllength TDP-43.Recombinant full-length TDP-43 has been shown to form fibrils in vitro (Kumar et al. 2023. Nat Neurosci 26, 983-996), but structures have not been determined.If the purified TDP-43 was indeed full-length (see major comment 1), this would represent an important advance over previous studies, by showing that fibrils formed of full-length TDP-43 can also be polymorphic.This work will be of interest to those working on amyloids and interested in their structural and biophysical properties.The figures are well-constructed and the manuscript is generally wellwritten (but see minor comments).However, major concerns with the quality of the protein preparations and cryo-EM maps (and resulting models) need to be addressed before publication.Please see below for more detailed comments.
Major comments: 1.The quality of the cryo-EM maps should be improved.The PDB validation reports show regions of continuous density running along the helical axis, instead of clearly separated molecular layers, which indicates that the maps may have converged on local minima.Guidance on how to spot and overcome this problem is provided in (Scheres.2020.Acta Cryst D76, 94-101).In addition, the intermediate resolutions of the maps (~4 Å) may have led to inaccuracies in model building.At present, a relatively modest number of initial particles is used (26,881), suggesting that additional data may improve the maps.The quality of the maps may also be improved by additional rounds of 3D classification, CTF refinement and Bayesian polishing.

Response:
We are grateful to this referee for thoroughly analyzing our data and for providing very helpful comments.While the collection of new data was not feasible for us within a reasonable time frame, we managed to further improved the resolution of all the density maps by additional rounds of 3D-auto refine, post processing and Bayesian polishing steps (see method section).The new resolutions range from 3.76 Å to 4.05 Å.More importantly, they now enable us to see the separation between the molecular layers, which also improved our model building.We have updated related figures accordingly.
2. Are the authors sure that they are dealing with full-length TDP-43, as claimed in the manuscript?Data on the characterisation of the recombinant TDP-43 protein preparations should be shown.At a minimum, this should include chromatograms from the SEC step, as well as SDS-PAGE of the preparations before and after aggregation.This is particularly vital for TDP-43, because it is prone to proteolytic cleavage and aggregation during purification, as well as proteolytic cleavage during in vitro aggregation assays.

Response:
We thank the reviewer for the suggestion.We have taken the feedback into consideration and made the necessary updates to the manuscript.We have now added gel stripes from different steps of the fibril preparation process, including before aggregation (= SEC elution) and after fibril formation, showing a protein size of approximately 43 kDa (see new SI fig.1a).In addition, we included the chromatogram from size exclusion chromatography column (see new SI Fig. 1b).Full-length TDP-43 elutes at approximately 15 ml on a Superdex 200 10/300 column (GE Healthcare).An elution maximum of 15 ml does not correspond to the expected molecular mass of TDP-43.Presumably this is due to interactions of the protein with the column.
We agree to the reviewer that full-length TDP-43 might undergo partial proteolytic cleavage during post-aggregation stages, leaving only a fraction of the C-terminal domain.Further studies would be necessary to investigate in greater detail the proteolytic aging of in vitro fibrils in various experimental storage conditions (temperature, nature of the buffer, pH) and how this process might impact the final fibril structural architecture.However, we believe that such work is clearly outside the scope of the current manuscript, which mainly deals with the structural assessment of the polymorphism of TDP-43 fibrils.To account for the reviewer's comment, we have added a sentence mentioning this possibility in the result section: "although it is possible that proteolytic truncation may occur" after fibril formation.

Response: Thank you this is done now.
Minor comments: 1.The centrifugation step used to pellet fibrils (13,000 g for 5 min) may result in many fibrils being discarded in the supernatant.For example, TDP-43 fibrils were found in the supernatant following centrifugation at 27,000 g for 10 min in (Arseni et al. 2022 Nature 601, 139-143).The authors should acknowledge this possibility.

Response:
We thank the reviewer for the comment, although we are not sure whether it might have arisen from a misunderstanding of our methodology.The centrifugation step that is mentioned by the referee was performed to collect insoluble TDP-43 protein at an intermediate step of our purification process (between Histrap and size exclusion chromatography).We do not know whether the insoluble TDP-43 collected by centrifugation shows amyloid structure.The precipitated protein was denatured in guanidine thiocyanate, purified as monomeric protein by size exclusion chromatography and allowed to fibrillate in the size exclusion chromatography buffer.This sample was then used for further analysis and it was used without centrifugation.
2. Line 55, 'Amyloid fibril samples are typically polymorphic and contain multiple fibril morphologies,' does not apply for brain-derived fibril samples, which are often homogeneous.The authors should clarify that this refers to fibril samples generated in vitro.Response: Thank you.In the present fibril, a fibril protein stack represents one protofilament.This is expressed in the introduction as "The fibril protein stacks define the fibril protofilaments (PFs), which represent filamentous substructures of amyloid fibrils." 5. Line 84, '...exhibiting an amyloid key motif.'This is a central point of the manuscript.Therefore, it would be helpful to define this motif here, or even earlier in the manuscript, rather than waiting until the Results and Discussion sections.

Response:
Thank you for pointing this out.It very much makes sense to define this motif in the introduction.We now added the following description to our introduction: "In several cases, there was evidence for a fibril protein fold that remotely resembles the Greek key structure in globular proteins.This motif of amyloid fibrils is termed an amyloid key.An amyloid key represents an arrangement of β-strands but in contrast to a Greek key, the strands interact through their side-chains and not through backbone hydrogen bonds."6. Line 140, 'The abundance of the two structural forms varies in the mixed fibrils almost continuously from one to zero.' Does this refer to the proportion of mixed fibrils 0-100%?
Response: Thank you.This statement means that mixed fibrils have variable percentages of segments corresponding to the structural forms 1a and 1b.We revised the criticized sentence, which now reads as: "'The abundance of the segments corresponding to the two structural forms (1a and 1b) varies almost continuously within mixed fibrils from one to zero".7. Line 149, 'The three fibril structures contain essentially the same fibril protein conformation that is formed by 45 residues (Gly304 to Gly348) from the LC domain.'Is there any evidence that the flanking residues to the N-and C-terminus of this region may be structured?It would be helpful for the reader to comment on this.
Response: Thank you.While there is evidence for extra density outside the fibril core, this density is very diffuse and we are unable to discern significantly ordered parts in this density, such as flanking residues.
8. Line 159, 'An unusual structural feature of the fibril core is the absence of buried ion pairs.'This statement needs to be qualified.There are many amyloid fibril structures that lack salt bridges.

Response:
We have revised the sentence to now read as: "The fibril core devoid of buried ion pairs".9. Line 231, 'Finally, we noted that our structure and all previous in vitro formed structures from TDP-43 differ from a fibril morphology that was purified from a patient.'This study purified TDP-43 fibrils from two patients with ALS and FTLD-TDP Type B, and two brain regions, prefrontal cortex and motor cortex.This should be corrected.
Response: Thank you.We have corrected it in the revised manuscript.We now write 'from patient tissue' instead of 'from a patient'.10.Line 240, 'Therefore, the currently available data on TDP-43 fibrils show that in vitro formed fibrils do not necessarily match the structure of ex vivo fibrils.'It would also be helpful for the reader to point out that the sequence that forms the structured part of the fibril also does not necessarily match (G304-G348 here vs. G281-Q360 in ALS and FTLD).

Response:
Thank you for the suggestion.We have added this information in manuscript.Now, we write "Therefore, the currently available data on TDP-43 fibrils show that in vitro formed fibrils do not necessarily match the structure of ex vivo fibrils in terms of composition and conformation of the fibril protein."11.Line 243 '...but that more complex and more difficult to establish spontaneously.'Should this read, '...but that is more complex and more difficult to establish spontaneously.' ?
Response: Thank you.We have now corrected this in revised version.
12. Line 244 'One reason to establish these specific folds is the higher biological stability of ex vivo fibrils compared with most in vitro formed fibrils.'This claim needs to be supported by data or references.

Response:
Thank you for this comment.We now added references to this statement.
13.The model of the rotator/shaker used during the aggregation of TDP-43 should be stated to help the reader interpret a rotation of 40 rpm.

2.
The authors should provide a figure that depicts more clearly the density maps.It is impossible to see any relevant details in the small panels of Fig.2b.3.SI Figure2-The folds of morphology 1-a and morphology 1-b appear to be mirrored images, suggesting that one of them may represent the top view and another one the bottom view?The authors should keep the orientation consistent.