Competition between inside-out unfolding and pathogenic aggregation in an amyloid-forming β-propeller

Studies of folded-to-misfolded transitions using model protein systems reveal a range of unfolding needed for exposure of amyloid-prone regions for subsequent fibrillization. Here, we probe the relationship between unfolding and aggregation for glaucoma-associated myocilin. Mutations within the olfactomedin domain of myocilin (OLF) cause a gain-of-function, namely cytotoxic intracellular aggregation, which hastens disease progression. Aggregation by wild-type OLF (OLFWT) competes with its chemical unfolding, but only below the threshold where OLF loses tertiary structure. Representative moderate (OLFD380A) and severe (OLFI499F) disease variants aggregate differently, with rates comparable to OLFWT in initial stages of unfolding, and variants adopt distinct partially folded structures seen along the OLFWT urea-unfolding pathway. Whether initiated with mutation or chemical perturbation, unfolding propagates outward to the propeller surface. In sum, for this large protein prone to amyloid formation, the requirement for a conformational change to promote amyloid fibrillization leads to direct competition between unfolding and aggregation.

The authors report molecular insights into the fibrillogenesis of the protein OLF and two variants.Their study shows that fibril formafion likely proceeds through the populafion of and intermediate conformafion with oligomerisafion ablated at high urea concentrafions where the fully unfolded state dominates the folding landscape.The studies are complimented with a considerable amount of NMR which further revels an inside-out unfolding transifion preceding fibre formafion.The authors have performed a lot of experiments and while I am no expert in the area in general I agree with the main conclusions but I have some of my own points/comments.I have been specifically asked by the editor to look at the HDX-MS aspects of the m/s so I'll start here.The authors are evaluafing isotope incorporafion directly and not using the more convenfional approach comparing the HDX behaviour of one protein by comparison to a reference sample.There is nothing wrong with this approach but I note that the authors have not taken any control measurements to account for back and forward exchange arfifacts.Back exchange artefacts in parficular can be considerable and lead to false interpretafion of experimental data if they haven't been accounted for.As things stand they cannot use the current data and if they want to include HDX-MS dynamics in the arficle they will have to perform new acquisifions and they have 2 opfions.First, reacquire the experimental data but with 3 replicates of back exchange controls and 3 replicates for forward exchange controls.Back exchange control samples are fully in-exchanged protein (the same as the experimental protein) acquired as a 15 second labelling experiment.In exchange can be performed by unfolding the protein in excess D2O and then refolding by dialysis against D2O or some equivalent using diafiltrafion.Alternafively the protein can be incubated in labelling buffer for several days at 37 degrees C this normally exchanges all the 1H for 2H.Ideally aliquots would be taken at several fimepoints during incubafion to see when the D-incorporafion has plateaued but 7-days incubafion is normally sufficient.Forward exchange controls consfitute a reference acquisifion on 1H protein but with the quench made in D2O so that the final D2O concentrafion in the quench is 50%.Correcfion of the RFU for each fimepoint and each pepfide is then made using the following: RFUcorr = (RFUobs.-RFUfwdexch.)/ (RFUbackexch.-RFUfwdexch.)Second, the authors could use the more convenfional difference data approach and in their case obtain experimental data for the WT and one of the variants and report deltaRFU.Note for this they would sfill have to take back exchange control data and correct the RFU, forward exchange correcfion is considered unnecessary for this type of data.Note the acquisifion of forward and back exchange control data cannot be taken after the fact.The control data must be taken in the same window as the experimental acquisifions.Further comments F2 A change colour key remove [urea] line F2 C The test suggests that the differences between 3 and 5 M urea are due to compefifion between refolding and aggregafion and at 5 M aggregafion wins.But at higher urea concentrafions could the protein just be kinefically trapped in a higher energy unfolded state?It's probably not, but the authors should indicate refolding fime so we can be sure the system is in a steady state.Also find it odd that the protein cannot refold.Once chaotrope is removed the protein is no longer unfolded or folded where is it the authors didn't state, does it precipitate or form fibres? -clarify.F3 B can the authors indicate on the figure the residues missing from the xtal structure for quick reference.Line 208 "This result emphasizes the central importance of the Ca2+ site in the structural integrity of the OLF propeller."What do the authors mean by this point can they clarify?Taken as read they mean the site not the ion and that loss of structural integrity at the site perturbs that whole structure.If ion is important in structural integrity the authors should demonstrate this with wild-type but with the addifion of a chelator to remove the calcium.Figure 7d is confusing.There should be 2 folding landscapes one for monomers and one for different types of mulfimers with both landscapes being linked by a common monomeric intermediate.The current image doesn't accurately represent the authors claim that the fibril form is accessed via an intermediate in the current representafion the fibril can be access from the fully unfolded state which is at odds with the experimental data.The fibril is also lower energy that the nafive monomer which I don't think is necessary true. 2 landscapes joined by a common intermediate would be much clearer.

Reviewer #3 (Remarks to the Author):
Saccuzzo et al. analyzed the structural dynamics and instabilifies of the olfactomedin (OLF) domain of myocilin.Mutafions in this 5-bladed beta-propeller protein are causal for early-onset hereditary openangle glaucoma, which is connected to amyloid aggregafion of OLF.It is therefore of upmost interest to understand the protein dynamics that leads from the folded protein to the amyloid state.To this end, the authors performed a number of NMR spectroscopy experiments as well as MD simulafions to resolve the protein dynamics under different condifions (causing protein unfolding at increasing urea concentrafions and by introducing disease-relevant point mutafions) and at different fime scales.The most important findings are that structural changes leading to unfolding start at the inside of the beta-propeller, which are encouraged by loss of the internal Ca2+ center and propagate outward to the surface of the propeller.However, complete unfolding as provoked by very high urea concentrafion do not lead to amyloid fibrillizafion.This is an interesfing study.The most excifing finding is that the unfolding starts in the interior and not exterior of the beta-propeller.
Based on my solid knowledge of NMR spectroscopy (though I am not an expert in it), I would say that the NMR experiments (and also H-D exchange mass spectrometry) were performed very carefully and at a high level of standards.
However, the MD simulafions are below standard in terms of quanfity and approach.First, the simulafions should be repeated at condifions mimicking the different experimental condifions by adding urea to the WT and introducing mutafions D380A and I499A.Moreover, either much longer (10 microsconds) or enhanced simulafions (REMD) should be conducted to observe structural changes that the could be related to the experimental observafions.I highly recommend to use another MD engine than NAMD as NMAD is too slow for systems below 500,000 atoms.Gromacs or Amber would be befter MD codes for this purpose.
Because of the rather short MD simulafions, the results on the protein dynamics presented in Fig. 6 are the weakest part of the study, as there is a missing link between the short-and long-fime scale dynamics.The aim (for a manuscript in Nat.Commun.)should be to provide a structural picture that shows how the unfolding starts.
Apart from this, I have some more specific comments: 1) l. 71: Introduce ER (even though it is a common abbreviafion) 2) Fig. 2: i) The color scale is not opfimal as the different shades of blue are difficult to disfinguish.Please change.
ii) Based on Fig. 2B the authors conclude that at high urea concentrafions complete unfolding wins and inhibits amyloid aggregafion.It is therefore speculated that the lafter is inifiated from parfially folded OLF.However, for Abeta -which is never folded in solufion at pH ~7 -it is known that high urea concentrafions inhibit amyloid aggregafion (see, e.g., Protein Sci.2004 Nov; 13(11): 2888-2898).If one understands amyloid aggregafion as a metafolding problem, it is understandable that this does not take place at high urea concentrafions.As long as the structure that finally leads to amyloid aggregafion is not fully determined, I find the authors' conclusions with regard to their observafions at high urea concentrafions not jusfified.
3) Fig. 3 needs to be improved.The residue labels in panel A are not or hardly readable.The size of the figure can be increased to make full use of the page width.The resolufion of the figures needs to be increased.And I am sure that also the size of the residue labels could be increased by 1 pt.4) Fig. 4: The meaning of the residues colored in gray in panel C needs to be explained.Please add spheres for Na+ and Ca2+ to the structures and lefters "A" to "E" at the outside of the blades.5) Fig. 5: Please add spheres for Na+ and Ca2+ to the structures, mark the respecfive mutafion site by a star (or alike), and add lefters "A" to "E" at the outside of the blades.6) p. 13: Please add in % how many of the chemical shifts remain almost unchanged upon mutafion (and how many cannot be determined).7) Fig. 6: The same kind of results should be provided for selected urea concentrafions and the mutants.This would befter allow to link protein dynamics with unfolding and amyloid aggregafion.See also my comment with regard to running further MD simulafions.

Reviewer #4 (Remarks to the Author):
This paper from the Lieberman lab reports Compefifion between inside-out unfolding and pathogenic aggregafion in an amyloid forming β-propeller.There are several strengths to the work, including unfolding events, that required for amyloid formafions.Authors used several techniques including NMR, X-rays, and molecular dynamics simulafions to give insight on the OLF domain folding.

Major comment:
• My major concern regarding the novel of this work.The Crystal Structure of the Myocilin Olfactomedin Domain has been solved (PDB: 4WXQ).It will be more interesfing if the authors can determine a novel structure of an intermediated species such as Oligomers or final aggregated form such as amyloid fibril structure of OLF domain.
• I am wandering if OLF domain is fibrilize in human pafient.Do the authors have any evidence that OLF present in human pafients as fibril species?
The authors should address the following comments and recommendafions on addifional crifical experiments to make for a stronger study.

Minor comments:
• The abstract is long, need to be shortened

Introducfion
• Missing reference at the end of this sentence.A recent addifion to the list of pathogenic proteins that form amyloid is the myocilin olfactomedin (OLF) domain.
• Missing reference at the end of this sentence.Many different single point mutafions, dispersed throughout the sequence, are the strongest genefic link to early onset open angle glaucoma.
• Remove figure 1 from introducfion, reference is enough.Figure 1, it is not structure solved in this paper and cannot be in the introducfion.
• At the end of introducfion, Authors required to bring the message of the quesfion need to be answered in this paper (this is not clear here).

Results:
• Authors reported that they monitored kinefics of amyloid formafion in the presence of varying concentrafions of urea (Figure 2A, S1B).It will be interested to see under Electron Microscope using negafive stain, the aggregafion products (Oligomers, fibrils, mature fibrils and amorphous aggregates).
• In Figure 1S, it looks like tube with 0 urea, has some aggregafion as well.Can author explain?
• In Figure 2C, CD spectra of the untreated sample doesn't seem folded well.I am wandering that the protein is not completely pure or folded after purificafion.Authors need to show CD spectrum starfing from 190 nm, possible there are more interesfing spectrum to show.
• The ThT signal, doesn't confirm that the protein can form amyloid fibrils, many ThT signal can also produce from amorphous protein aggregafion.Authors should use EM to prove that the OLF can form amyloid fibrils.
• Authors reported at urea concentrafions >2 M, the unfolded state is dominant and limits the ability to confidently map assignments and track CSP.I would suggest that Authors need to examine the conformafion state under EM first, and see if these conformafions are stable structure, they can dialysis the protein against the suitable buffer and recorded the data.It is also possible that at urea concentrafions >2 M, the OLF generate number of heterogenous conformafions that limit the assignments.
• Authors reported that this observafion may explain why D380A forms P3-like fibrils: if P3 residues are mobile and 214 disordered and not sequestered as in WT, they are available to template fibril formafion.There is no experimental evidence that show that D380A forms can form fibrils.
• Authors reported that they solved a 1.27 Å crystal structure of OLF (Figure 6A, Supp.Table S3).The new structure shares overall features with our previously reported ~2 Å resolufion structures.The resolufion is higher and befter but sfill similar to previous structure, 2 Å structure.I feel that there no need to solve previously published structure.It will be more interested to solve the same structure in the presence of ligand or inhibitors.
• Authors claim that OLF fibilizafion in many parts in the paper, especially in the discussion secfion (For OLF, 439 the fibrillizafion and chemical unfolding pathways are in significant compefifion at 37 °C 440 (Figure 7D).Author didn't show single experiment to show that OLF form fibrils. ThT is indicafion for aggregafion but no evidence for fibrils.

At the secfion of Expression and purificafion of OLF for NMR experiments.
Average yields from 1 L minimal media were 0.04 mg/L for OLFD380A, 0.02 mg/L for OLFI499F and 2.2 mg/L for WT. is it 0.02 mg/L correct?NMR required mM quanfity.
Can authors give a detailed purificafion protocol?The protein was expressed as fusing with MBP.Does all experiments perform after or before MBP cleavage?

1.
The proposed model for structure disruption is that Ca is lost first; this step needs to be further validated.There is evidence that Ca binding is diminished in D380A, consistent with mutation of this key Ca-liganding residue.Could there also be perturbations to sodium binding that impact the protein behaviour?Are WT and I499F fully metal bound?Additional information is needed on the effect of urea on metal binding, for WT, D380A, and I499F.For example, a denaturation curve monitored by CD and/or NMR, with more urea concentrations than in Figure 4 (and/or for different metal concentrations), might show evidence for multiple transitions, i.e. loss of metal binding followed by denaturation.It is important to include urea denaturation data for the mutants (as for WT in Figure 2D,C).
Thank you for these excellent points regarding missing direct evidence for Ca 2+ loss as an early step in OLF unfolding.We considered additional NMR experiments, but already at our lowest urea concentration for WT, 1 M, the Ca 2+ ligand D478 is one of just a few residues altered.CD is not sensitive enough to detect changes in tertiary structure as a function of urea, and fluorescence chelators we tested were quenched as a function of urea so to address this point we conducted ICP-OES on I499F, using the methods and facility as in PMID 23129764 (where WT and D380A were first reported).The results show that the Ca 2+ site in I499F is occupied at a molar ratio of 0.2:1 whereas WT OLF is occupied to the same extent as we report in prior publications (0.7-0.8:1, PMID 35831671 and 35831671).These results strengthen the conclusion that the Ca site is perturbed in a partially folded OLF that in principle remains competent for Ca 2+ binding (i.e.I499F).In sum, although there is ample evidence for perturbation of the Ca 2+ site, in our revision we toned down language that specifically mentions Ca 2+ release as a trigger for misfolding because we were unable to explicitly nail down the release of Ca 2+ in the urea-containing samples.

Manuscript Changes:
Abstract p.2: Whether initiated with mutation or chemical perturbation, unfolding propagates outward to the propeller surface propeller.
Results p.7, section "Partially folded moderate (OLF D380A ) and severe (OLF I499F ) disease variants..: Analysis of Ca 2+ levels in OLF I499F by inductively coupled plasma optical emission spectroscopy (ICP-OES) reveals ~20% occupancy compared to OLF WT under buffer conditions used in this study and elsewhere 23,26 (Supp.Table S2).Thus, despite still harboring all residues necessary for Ca 2+ chelation, the metal center in OLF I499F is not fully occupied and thus predominantly non-native.
Results p. 21, section "Unfolding from the inside out..": Interestingly, other calcium-coordinating ligands like D380 and N428 experience minimal perturbations in both the 1 and 2 M spectra.The most likely interpretation of this is that Ca 2+ binding are retained to some extent during urea-induced unfolding, a notion that is supported by MD simulations of OLF I499F and OLF D380A in which metal ions remain at least partially bound (Supp.Fig. S10) as well as by crystal structures of certain metal ligand stable variants tested (Supp.Fig. S11) in which a metal binds, but using an alternative ligand arrangement.

2.
The conclusion that amyloid aggregation competes with chemical unfolding but only below the threshold where OLF loses discrete tertiary structure needs clarification and further consideration.Since urea and mutations promote aggregation under conditions where the protein is presumably predominantly folded, how does this competition occur?Also, the prominent random coil signals in the spectrum of I499F (Figure 5) could be due to disruption of the monomer structure, as proposed, or be caused by protein self-association.Are the spectra of the various proteins constant with time or varying due to changing proportions of monomer and self-associated species?
We appreciate this comment.Our urea experiments suggest that when OLF WT is predominantly folded, the partially folded elements trigger the templated aggregation.It is the population of low-lying excited states that are accessed with urea or mutation, and these are aggregation prone due to accessing the APRs.Self-association cannot readily account for the molten globule-like spectrum for I499F.Even though this variant is prone to aggregation we did not see a reduction in intensity in the spectrum.When purified, I499F (and WT, D380A) elute from size exclusion as monomers.For WT OLF, we have evidence for only very weak dimerization (PMID 31484937), using crosslinking where the OLF concentrations used were similar to the NMR studies here.Thus, we do not think that oligomerization or self-association in a soluble state can account for what is observed here.

Manuscript changes:
Results p. 13 section "Analysis of the moderate disease mutant.." ..eight residues in the previously identified 19 APR stretch P3 (N428, A429, F430, I431, C433, L436, Y437, T438), experience CSP.Interestingly, P1 residues remain largely WT-like, indicating they retain a chemical environment that is similar to OLF WT .This observation may explain why D380A forms P3-like fibrils 19 : if P3 residues are mobile and disordered and not sequestered as in WT, they are available to template fibril formation.
Results p. 22 section "Unfolding from the inside out.." In summary, our data suggest a model that explains how OLF misfolding arises, from inside out by unfolding β-propeller (Figure 7C), which differentially exposes internal APRs on the pathway to fibrillization (Supp.Fig. S8).
Discussion p. 26 full paragraph

3.
The effects on protein structure and amyloid aggregation of D380A are proposed to resemble those of 1 M urea on WT, while the I499F effects resemble those of 2 M urea on WT.The authors hypothesize that OLF fibrillization proceeds via access to partially unfolded states.It may be that multiple partially unfolded states are promoted by the toroidal fold of OLF, which might initially unfold not just near the metal binding site but also in the region where the N and C termini associate, in a process that is less cooperative than for typical globular proteins.I recommend adding some discussion of these points to emphasize their significance.
We completely agree unfolding is more complicated than two-state unfolding, and this is emphasized throughout the manuscript.Notably, the N and C terminal molecular clasp is stabilized by the disulfide bond (Cys 245-Cys 433).

Manuscript Changes:
Discussion p. 24 "Even though we do not have complete resonance assignments in the molecular clasp, we can infer that the clasp may be somewhat more resistant to complete unfolding due to a disulfide bond between C433 on blade D and C245 on blade E, but likely experiences perturbations given that D478 within blade E is one of the early residues perturbed with 1M urea."

4.
The minor changes observed by NMR for WT over months might arise from various covalent changes, e.g.deamidation, oxidation, cleavage not apparent by SDS-PAGE.It is important to check for such modifications e.g. by mass spectrometry.
We recruited collaborators to conduct the suggested mass spectrometry analysis on the sample used for NMR ( 15 N labeled, now 1 year old) and a freshly purified WT OLF (unlabeled) sample for comparison.Chymotrypsin was used for digestion (instead of trypsin) to maximize overall sequence coverage.Oxidation and deamidation were observed at several sites but no clear differences were delineated, most differences between fresh and old sample less were than 10%.The only candidate for a new PTM is deamidation at N480, which was only detected in the old sample, but no conclusions can be drawn because it was measured at low abundance (1 PSM, 2.22% modified).In addition to mass spectrometry we looked again at the 1.3 Å crystal structure, which was solved from a 10 mo old crystal, and no modifications were observed to the domain in the structure.

Manuscript Changes:
Supp.Fig. S13 contains mass spectrometry data analysis.Supplemental data set included with the data.
Materials and Methods (p 36) have been added under section "Sample preparation and LC-MS/MS analysis for posttranslational modifications." Results p. 20 section "Dynamics of OLF WT .." Further, comparing the posttranslational modifications (PTMs), namely, deamidation and methionine oxidation, in the sample used for NMR that had been stored at 4 °C for over a year and freshly purified OLF WT (Supp.Fig. S13) reveals just one PTM detected exclusively in the old sample, deamidation at N480, but it was detected at very low abundance (2.22% mean modified %PSM).Eleven PTMs were detected in both samples.The extent of methionine oxidation or deamidation across these common sites in the two samples differ by < 10% in most cases except for N350, N450, and M494 where the difference is less than 17%.In sum, the extremely long half-life for the observed structural change cannot be reconciled by cleavage, degradation, or PTMs and suggests that OLF WT possesses two very similar structural states that are separated by a high energetic barrier.

5.
Minor points Residues prior to Gly244 were not modeled due to them being absent from available crystal structures.Note MD simulations have been updated per R3 are now presented in Supp.Fig. S10.
Reviewer #2 ..The authors have performed a lot of experiments and while I am no expert in the area in general I agree with the main conclusions but I have some of my own points/comments.
We thank Reviewer 2 for their key critical review of our HDX-MS data and other comments.

6.
I have been specifically asked by the editor to look at the HDX-MS aspects of the m/s so I'll start here.The authors are evaluating isotope incorporation directly and not using the more conventional approach comparing the HDX behaviour of one protein by comparison to a reference sample.There is nothing wrong with this approach but I note that the authors have not taken any control measurements to account for back and forward exchange artifacts.Back exchange artefacts in particular can be considerable and lead to false interpretation of experimental data if they haven't been accounted for.As things stand they cannot use the current data and if they want to include HDX-MS dynamics in the article they will have to perform new acquisitions and they have 2 options.First, reacquire the experimental data but with 3 replicates of back exchange controls and 3 replicates for forward exchange controls.Back exchange control samples are fully in-exchanged protein (the same as the experimental protein) acquired as a 15 second labelling experiment.In exchange can be performed by unfolding the protein in excess D2O and then refolding by dialysis against D2O or some equivalent using diafiltration.Alternatively the protein can be incubated in labelling buffer for several days at 37 degrees C this normally exchanges all the 1H for 2H.Ideally aliquots would be taken at several timepoints during incubation to see when the D-incorporation has plateaued but 7-days incubation is normally sufficient.Forward exchange controls constitute a reference acquisition on 1H protein but with the quench made in D2O so that the final D2O concentration in the quench is 50%.Correction of the RFU for each timepoint and each peptide is then made using the following: RFUcorr = (RFUobs.-RFUfwdexch.)/ (RFUbackexch.-RFUfwdexch.)Second, the authors could use the more conventional difference data approach and in their case obtain experimental data for the WT and one of the variants and report deltaRFU.Note for this they would still have to take back exchange control data and correct the RFU, forward exchange correction is considered unnecessary for this type of data.Note the acquisition of forward and back exchange control data cannot be taken after the fact.The control data must be taken in the same window as the experimental acquisitions.
Thank you so much for this detailed comment, indeed it would have been awful to have these improperly interpreted results in any published manuscript.In our revised manuscript, we restrict our comparison to relative differences across WT vs I499F vs D380A, not absolute differences.The HDX-MS results were edited out of an early version of the original manuscript but the experiments conducted within a short time frame in late 2019, under the same experimental conditions, instruments, etc.We discussed our data with Dr. Renhao Li, now an author, whose lab conducts HDX-MS on proteins on a frequent basis.We believe our interpretation of the HDX-MS data bolsters the initial observation that D380A and I499F adopt structures that are different from WT (which is how the story unfolded for us in real time, before we did any NMR).
We could not redo the experiments to take into consideration absolute back exchange because the core facility we used to do these experiments was taken off line after the pandemic.We ask for R2's understanding regarding this unfortunate complication.Our data have been submitted to ProteomeXchange (Accession PXD045520 Username: reviewer_pxd045520@ebi.ac.ukPassword: Y5LjuCZy).

Manuscript changes:
Results p. 7 section "Partially folded moderate (OLF D380A ) and severe (OLF I499F ) disease.." See new paragraph, the second in this section.
Materials and Methods p. 34 section HDX-MS.See amended paragraph starting with "HDX-MS was performed on OLF WT , OLF D380A , and OLF I499F on three separate days, within a two month period,.." Data availability p. 38 "The HDX-MS data have been deposited to the ProteomeXchange Consortium with the dataset identifier PXD045520."

7.
Further comments • F2 A change colour key remove [urea] line Fixed.
• F2 C The test suggests that the differences between 3 and 5 M urea are due to competition between refolding and aggregation and at 5 M aggregation wins.But at higher urea concentrations could the protein just be kinetically trapped in a higher energy unfolded state?It's probably not, but the authors should indicate refolding time so we can be sure the system is in a steady state.Also find it odd that the protein cannot refold.Once chaotrope is removed the protein is no longer unfolded or folded where is it the authors didn't state, does it precipitate or form fibres? -clarify.
Fig. 2C shows the refolding from 3 M and 5 M yields tertiary CD spectra that are both flatlined, indicating that what is left in solution is unfolded.The precipitated OLF resulting from dilution from 5 M is not recorded in the instrument.Regarding steady state-the refolding time, e.g.time between diluting out urea to CD measurements, was ~20 minutes for both 3M and 5M.OLF refolded from 3 M urea resembled native, folded OLF in that timeframe (Fig. 2C).For OLF refolded from 5 M urea, more protein crashed out of solution as more time passed, so measurements were taken at early time points where something was left in solution to measure.Regarding the finding that the protein cannot refold, this is not unusual; many proteins require molecular chaperones to fold.Only the simplest model systems are fully reversible.

Manuscript changes:
Results p. 6 section "Amyloid aggregation competes with unfolding": At room temperature, OLF WT incubated in 3M urea, where tertiary structure is no longer detected by near-UV circular dichroism (CD), can be refolded upon dilution to buffer lacking urea (Figure 2C) 20 .By contrast, when OLF WT incubated in 5M urea is diluted, significant precipitation is visible.Still, some OLF WT remains in solution, at high enough concentration to be measured by near-UV CD, and this species does not have detectable tertiary structure (Figure 2C).Thus, refolding OLF WT from urea is possible from the midpoint of unfolding but not at later stages.
Materials and Methods p. 31 section "OLF WT unfolding and refolding by CD: For unfolding, OLF was allowed to equilibrate in 3 M urea for at least 1 h at 4 °C prior to acquisition of CD spectra.For refolding, OLF was equilibrated in either 3 or 5 M urea for at least 15 m at room temperature before being diluted 10-fold with PGF buffer.Spectra were recorded within 20 m of diluting with PGF.

•
F3B can the authors indicate on the figure the residues missing from the xtal structure for quick reference.
These are now listed the text in the results section for WT assignments (specifically E230-S231, L236-T243 and K503-M504).

•
Line 208 "This result emphasizes the central importance of the Ca2+ site in the structural integrity of the OLF propeller."What do the authors mean by this point can they clarify?Taken as read they mean the site not the ion and that loss of structural integrity at the site perturbs that whole structure.If ion is important in structural integrity the authors should demonstrate this with wild-type but with the addition of a chelator to remove the calcium.
See comments in response to R1.
• Figure 7d is confusing.There should be 2 folding landscapes one for monomers and one for different types of multimers with both landscapes being linked by a common monomeric intermediate.The current image doesn't accurately represent the authors claim that the fibril form is accessed via an intermediate in the current representation the fibril can be access from the fully unfolded state which is at odds with the experimental data.The fibril is also lower energy that the native monomer which I don't think is necessary true. 2 landscapes joined by a common intermediate would be much clearer.
Thank you for noticing the error in our figure, it has been updated to better reflect our finding that aggregation from a chemically unfolded state does not lead to aggregation.The relative energies of folded and fibrillized protein have been left as-is as they are in agreement with literature whereby folded proteins are higher energy than fibrils (e.g.see Figure 1 in most review articles, e.g.PMID: 21776078, 23746257).

Reviewer #3
.. This is an interesting study.The most exciting finding is that the unfolding starts in the interior and not exterior of the beta-propeller.
We thank R3 for their overall positive impression of our study and excellent suggestions.

8.
Based on my solid knowledge of NMR spectroscopy (though I am not an expert in it), I would say that the NMR experiments (and also H-D exchange mass spectrometry) were performed very carefully and at a high level of standards.
HDX-MS was performed carefully but in the originally submitted manuscript, was inadvertently presented in a misleading way; see comment 6 (R2); this has been fixed by comparative studies between WT and I499F and D380A and its discussion moved to earlier in the MS.

9.
However, the MD simulations are below standard in terms of quantity and approach.First, the simulations should be repeated at conditions mimicking the different experimental conditions by adding urea to the WT and introducing mutations D380A and I499A.Moreover, either much longer (10 microseconds) or enhanced simulations (REMD) should be conducted to observe structural changes that the could be related to the experimental observations.I highly recommend to use another MD engine than NAMD as NMAD is too slow for systems below 500,000 atoms.Gromacs or Amber would be better MD codes for this purpose.
We repeated the simulations for the WT and added simulations for the D380A and I499F mutants (Supp.Fig. S10).These were run in triplicate for 10 μs each (90 μs in total).As recommended by the reviewer, we used Amber for these simulations.For comparison, we note that using the recommended settings for the CHARMM force field (most notably a 12-Å cutoff), Amber produced ~330 ns/day while NAMD3 on GPUs produced ~250 ns/day.

10.
Because of the rather short MD simulations, the results on the protein dynamics presented in Fig. 6 are the weakest part of the study, as there is a missing link between the short-and long-time scale dynamics.The aim (for a manuscript in Nat.Commun.)should be to provide a structural picture that shows how the unfolding starts.
Extending the simulations to 10 μs revealed additional subtle behaviors compared to the previous 1-μs runs, most notably increased RMSF in particular regions.The addition of simulations of the two mutants also allowed us to make further comparisons to experimental results.Since a connection between simulations results to longer-time behavior in experiment remains speculative, likely due to the complexity of the system, additional suggested experiments like adding urea, seem premature.

Manuscript changes:
Results p. 18 section "Dynamics of OLF WT from ps-month time scale": see new paragraph We added a new figure (Supp.Fig. S7) where the spectra are separated out.
• Fig. 2: ii) Based on Fig. 2B the authors conclude that at high urea concentrations complete unfolding wins and inhibits amyloid aggregation.It is therefore speculated that the latter is initiated from partially folded OLF.However, for Abeta -which is never folded in solution at pH ~7 -it is known that high urea concentrations inhibit amyloid aggregation (see, e.g., Protein Sci.2004 Nov; 13(11): 2888-2898).If one understands amyloid aggregation as a metafolding problem, it is understandable that this does not take place at high urea concentrations.As long as the structure that finally leads to amyloid aggregation is not fully determined, I find the authors' conclusions with regard to their observations at high urea concentrations not justified.
Thank you for this point.We address this with changes to the discussion.
Manuscript changes: Discussion p. 26: Although high urea may stabilize the unfolded state to prevent aggregation, as seen for intrinsically-disordered Aβ 60 , NMR spectra show that above the unfolding midpoint, OLF WT adopts a conformationally heterogeneous structural ensemble consistent with a molten globule-type ensemble (Figure 4A), suggesting that in principle, aggregation could still proceed beyond the midpoint of unfolding.Instead, our interpretation of the data point to the possibility that APRs in OLF cannot be fully unfolded for templated fibrillization to occur.
• Fig. 3 needs to be improved.The residue labels in panel A are not or hardly readable.The size of the figure can be increased to make full use of the page width.The resolution of the figures needs to be increased.And I am sure that also the size of the residue labels could be increased by 1 pt.
The labels have been made larger.We confirmed that this image was prepared at adequate resolution for the journal.If this manuscript is accepted for publication, we will ensure all figures are up to journal standards.

•
Fig. 4: The meaning of the residues colored in gray in panel C needs to be explained.Please add spheres for Na+ and Ca2+ to the structures and letters "A" to "E" at the outside of the blades.
• Fig. 5: Please add spheres for Na+ and Ca2+ to the structures, mark the respective mutation site by a star (or alike), and add letters "A" to "E" at the outside of the blades.
Grey residues (unassigned residues), spheres for the metals, and blade labels have been added to all figures.

•
p. 13: Please add in % how many of the chemical shifts remain almost unchanged upon mutation (and how many cannot be determined).

Manuscript changes:
Results (p 13) In total, 73 resonance assignments were confidently transferred from OLF WT to the OLF D380A spectrum which represents about 60% of the discreet OLF D380A resonances (see Methods).
Results (p 15) Specifically, of the 106 discreet NMR resonances in the OLF I499F spectrum, 51 resonance assignments could be confidently transferred.

•
Fig. 6: The same kind of results should be provided for selected urea concentrations and the mutants.This would better allow to link protein dynamics with unfolding and amyloid aggregation.See also my comment with regard to running further MD simulations.
We have added HDX-MS results for the mutants for comparison with WT.With this paper as the foundation, our long term goal is to link protein dynamics with unfolding and amyloid aggregation.We are pursuing this with experiments in the lab but this is outside the scope of this already-lengthy manuscript.

Reviewer #4
This paper from the Lieberman lab reports Competition between inside-out unfolding and pathogenic aggregation in an amyloid forming β-propeller.There are several strengths to the work, including unfolding events, that required for amyloid formations.Authors used several techniques including NMR, X-rays, and molecular dynamics simulations to give insight on the OLF domain folding.
We thank R4 for their overall positive impression of the system.
12. Major comment: My major concern regarding the novel of this work.The Crystal Structure of the Myocilin Olfactomedin Domain has been solved (PDB: 4WXQ).It will be more interesting if the authors can determine a novel structure of an intermediated species such as Oligomers or final aggregated form such as amyloid fibril structure of OLF domain.
We agree, but considering the continuum of the trajectory from folded protein (e.g.4WXQ) to final fibril includes partially folded monomer, we believe the contributions in this manuscript to be a critical piece of the puzzle.

13.
I am wandering if OLF domain is fibrilize in human patient.Do the authors have any evidence that OLF present in human patients as fibril species?
To our knowledge, there is just one histochemical study reporting ER-localized accumulation of Y437H mutant myocilin in primary trabecular meshwork cells from a donor eye (PMC6425711).Such donor eyes, with genotyping and complete family medical history, etc are exceedingly rare.As such, for 20 years the field has relied on studies of mutant myocilin from a variety of models.

14.
The authors should address the following comments and recommendations on additional critical experiments to make for a stronger study.

•
The abstract is long, need to be shortened The abstract is now 150 words per journal requirements • Missing reference at the end of this sentence.A recent addition to the list of pathogenic proteins that form amyloid is the myocilin olfactomedin (OLF) domain. done.
• Missing reference at the end of this sentence.Many different single point mutations, dispersed throughout the sequence, are the strongest genetic link to early onset open angle glaucoma.done.

•
Remove figure 1 from introduction, reference is enough.Figure 1, it is not structure solved in this paper and cannot be in the introduction.
Respectfully, we disagree with the suggestion.In order to introduce the elements of the protein structure germane to the study, we need to have an image of the (previously published) structure in the introduction.We anticipate that this will increase access of our study to non-specialists.

•
At the end of introduction, Authors required to bring the message of the question need to be answered in this paper (this is not clear here).
We apologize for the confusion, and have edited the manuscript accordingly.

Manuscript changes: Introduction, p. 4:
To better comprehend the proclivity of OLF toward amyloid aggregation, here we probed the molecular details of its misfolding and aggregation using structural and biophysical tools.We used both low-and high-resolution techniques to probe the mechanism by which OLF WT initially transitions from a folded to a partially unfolded, aggregation prone state, as well as the relationship between partially unfolded states adopted by OLF WT and those adopted by disease-causing mutants.

•
Authors reported that they monitored kinetics of amyloid formation in the presence of varying concentrations of urea (Figure 2A, S1B).It will be interested to see under Electron Microscope using negative stain, the aggregation products (Oligomers, fibrils, mature fibrils and amorphous aggregates) Representative images have been added as Supp.Fig. S1.In other studies e.g.PMC3946817, PMC3323732 we have shown fibril morphologies extensively by TEM and AFM.

•
In Figure 1S, it looks like tube with 0 urea, has some aggregation as well.Can author explain?This is just glare from the light in the picture, there was no (yellow tinged) pellet in the tube.

•
In Figure 2C, CD spectra of the untreated sample doesn't seem folded well.I am wandering that the protein is not completely pure or folded after purification.Authors need to show CD spectrum starting from 190 nm, possible there are more interesting spectrum to show.
CD spectra in the near-UV (~250 -320 nm) assesses protein tertiary structure, a fingerprint unique to each protein.Far-UV (~190 -300 nm) CD spectra, which assess secondary structure, are not included in this work due to the fact that we previously did not see any major differences in secondary structure among variants by far-UV (see PMC3946817).

•
The ThT signal, doesn't confirm that the protein can form amyloid fibrils, many ThT signal can also produce from amorphous protein aggregation.Authors should use EM to prove that the OLF can form amyloid fibrils We have shown morphologies extensively by TEM and AFM in prior publications, e.g.PMC3946817, PMC3323732.We have now added TEM images of fibrils grown in urea in Supp.Fig. S1.
• Authors reported at urea concentrations >2 M, the unfolded state is dominant and limits the ability to confidently map assignments and track CSP.I would suggest that Authors need to examine the conformation state under EM first, and see if these conformations are stable structure, they can dialysis the protein against the suitable buffer and recorded the data.It is also possible that at urea concentrations >2 M, the OLF generate number of heterogenous conformations that limit the assignments.
Current cryoEM methods have difficulty resolving "small" proteins.OLF is a 30 kDa protein, well below the size limit to resolve the conformational ensemble.Our data are most consistent with multiple unfolded states, as stated in the manuscript.
• Authors reported that this observation may explain why D380A forms P3-like fibrils: if P3 residues are mobile and 214 disordered and not sequestered as in WT, they are available to template fibril formation.There is no experimental evidence that show that D380A forms can form fibrils.We apologize for the confusion.We showed D380A forms circular fibrils via AFM in PMC3946817.The reference was added to the end of this sentence in the results section, "This observation may explain why D380A forms P3-like fibrils 20 "

•
Authors reported that they solved a 1.27 Å crystal structure of OLF (Figure 6A, Supp.Table S3).The new structure shares overall features with our previously reported ~2 Å resolution structures.The resolution is higher and better but still similar to previous structure, 2 Å structure.I feel that there no need to solve previously published structure.It will be more interested to solve the same structure in the presence of ligand or inhibitors.
The purpose of the near-atomic resolution structure is to allow us to get more accurate Bfactors, which help provide better insight into picosecond time frame dynamics.

•
Authors claim that OLF fibilization in many parts in the paper, especially in the discussion section (For OLF, 439 the fibrillization and chemical unfolding pathways are in significant competition at 37 °C 440 (Figure 7D).Author didn't show single experiment to show that OLF form fibrils. ThT is indication for aggregation but no evidence for fibrils.
We have shown morphologies extensively by TEM and AFM in prior publications, e.g.PMC3946817, PMC3323732.We have now added TEM images of fibrils grown in urea in Supp.Fig. S1D.

•
Average yields from 1 L minimal media were 0.04 mg/L for OLFD380A, 0.02 mg/L for OLFI499F and 2.2 mg/L for WT. is it 0.02 mg/L correct?NMR required mM quantity.
Indeed, the yield for I499F is that low and we struggled to produce this sample.60 g of cell paste (4 different growths and a week of lysing) was purified for a single NMR sample of low uM concentration.Modern NMR spectrometers and cryogenically cooled probes are compatible with this range of protein concentration.Our data collection was optimized for the lower concentration.
• Can authors give a detailed purification protocol?The protein was expressed as fusing with MBP.
Does all experiments perform after or before MBP cleavage?

•
teal indicates no significant CSP and the quantitative criteria for yellow or teal color.In the Figure 7B the yellow color indicates regions of OLF with minimal CSP where resonances were confidently assigned.The teal color indicates assigned WT-OLF resonances that were not able to be confidently transferred across all conditions listed (1 Updated to be more clear.• include explanation of color for bound metals.Metals are now labeled in figures • Many resonances for WT in Figure 3 may exhibit splitting (e.g.shoulders on T259, S393, I432, D454, maybe nearby minor peaks for G266), or these unlabelled minor signals may arise from residues that are not assigned.Please elaborate on this aspect (e.g. in the SI).This is an interesting point, one, unfortunately, that we cannot confidently resolve with our current data.Some of the shoulders are clearly other resonances (see G241 and T448).However for the unassigned resonances we are not confident.Nominally NMR dynamics experiments would illuminate this (ZZ-exchange or similar) but because WT-OLF exhibits unusual long timescale dynamics we are under non-equilibrium conditions limiting this application.We note that some resonances have shoulders which could indicate slow time-scale dynamics.However, the current level of assignments and data quality do not allow for confident assessment of the origins of the less intense resonances that are overlapping on some more intense resonances.Please include additional quantitative details for the principal component analysis of Figure 4 e.g.how many significant components were identified?TRENDNMR outputs multiple principal components.However as PC1 represents the most significant principal direction with the largest variance and in this experiment we were deliberately changing the urea concentration, we anticipate that only PC1 is relevant.Accordingly PC1 shows twostate behavior and generally agrees with the ThT kinetics experiments.PC1 represents 29% contribution of the data.This contribution is listed in the methods (p 31).• RMSF values are not shown for the (most mobile) N-terminal residues in Supp.Fig. S6.Please add information on these values (if too large, can state in the legend).
i) The color scale is not optimal as the different shades of blue are difficult to distinguish.Please change.