Stress-mediated aggregation of disease-associated proteins in amyloid bodies

The formation of protein aggregates is a hallmark of many neurodegenerative diseases and systemic amyloidoses. These disorders are associated with the fibrillation of a variety of proteins/peptides, which ultimately leads to cell toxicity and tissue damage. Understanding how amyloid aggregation occurs and developing compounds that impair this process is a major challenge in the health science community. Here, we demonstrate that pathogenic proteins associated with Alzheimer’s disease, diabetes, AL/AA amyloidosis, and amyotrophic lateral sclerosis can aggregate within stress-inducible physiological amyloid-based structures, termed amyloid bodies (A-bodies). Using a limited collection of small molecule inhibitors, we found that diclofenac could repress amyloid aggregation of the β-amyloid (1–42) in a cellular setting, despite having no effect in the classic Thioflavin T (ThT) in vitro fibrillation assay. Mapping the mechanism of the diclofenac-mediated repression indicated that dysregulation of cyclooxygenases and the prostaglandin synthesis pathway was potentially responsible for this effect. Together, this work suggests that the A-body machinery may be linked to a subset of pathological amyloidosis, and highlights the utility of this model system in the identification of new small molecules that could treat these debilitating diseases.

acidosis, and transcriptional/proteotoxic stress) [27][28][29][30] .These structures share many of the biophysical characteristics ascribed to pathological protein deposits, as they are dense, proteinase K-resistant, insoluble aggregates that possess immobile molecular constituents and a strong affinity for the amyloidophilic dyes Congo red and Thioflavin S 27 .However, unlike their disease-associated counterparts, A-body biogenesis is a rapid and reversible process that generates non-toxic physiological structures 27,31 .Upon stimulation, these biomolecular condensates displace nucleoli by sequestering a large and heterogeneous family of cellular proteins 27,29 .The distinct stressors aggregate both universal and unique molecular residents 29 , potentially enhancing survival under harsh environmental conditions by tailoring a metabolic response to each cellular insult.Proteins found in A-bodies under all conditions tested, such as cell division cycle protein 73 (CDC73), represent universal A-body residents, and are useful marker molecules of these subnuclear foci 29 .The presence of unique constituents, such as flap endonuclease 1 (heat shock), DNA methyltransferase 1 (hypoxia/acidosis) and anaphase-promoting complex subunit 2 (transcriptional/proteotoxic stress) is particularly fascinating 29 .This observation suggests that aggregation of a subset of proteins is catalyzed by stress-specific machinery, which mediates the recruitment and amyloid conversion of these proteins only under specific environmental conditions.A-bodies are also shown to be capable of recruiting and aggregating the Alzheimer's disease peptide β-amyloid 27 , suggesting that pathological amyloidogenesis may be linked to the dysregulation of processes associated with A-body formation.In this study, we assess the ability of various disease-associated proteins to aggregate within the A-bodies.We also explore the cellular pathways and mechanisms regulating this aggregation event.This data could provide useful insights into the origins of several amyloidogenic disorders and a framework for the development of a new screen platform for therapeutics.

Results
Amyloid disease-associated proteins aggregate within A-bodies.A-bodies represent a physiological site of protein aggregation, where cellular factors and local environmental conditions have created a setting that is conducive to the residents proteins adopting an amyloid-like conformation 27 .Here, we assess the ability of a variety of fibrillation-prone amyloidosis-associated proteins, to aggregate within A-bodies.Under normal growth conditions, GFP-tagged β-amyloid (1-42) (Alzheimer's disease), Tau (Alzheimer's/Parkinson's disease), α-synuclein (Parkinson's disease), amylin (type II diabetes), immunoglobin light-chain (AL amyloidosis), and serum amyloid A (AA amyloidosis) can be found throughout transfected cells (Fig. 1A,B), possessing highly soluble and mobile properties (Fig. 1C,D).However, in response to environmental stressors only a subset of the tested disease-associated proteins translocate to the nucleus, co-localize with the A-body marker molecule CDC73 29 (Fig. 1A,B), and adopt the hallmark insoluble and immobile characteristics of these amyloidlike structures (Fig. 1C,D).It is interesting to note that like physiological A-body constituents 29 , pathological proteins also possess stress-specific targeting properties.For example, β-amyloid (1-42) was found to aggregate under all known A-body-inducing stimuli, while serum amyloid A and amylin primarily localized to A-bodies under heat shock and hypoxic/acidotic conditions (Fig. 1A,B).Immunoglobulin light-chain was recruited and immobilized within the A-bodies of heat shock-, hypoxia/acidosis-, and a sub-population of transcriptional/ proteotoxic stress (TPS)-treated cells (Fig. 1A-D).However, the un-sequestered proteins maintained a mobile profile (Fig. 1D-bottom panel), highlighting that molecules found within these structures adopt the amyloidlike properties of this subnuclear domain, while the untargeted proteins remain highly mobile under the same environmental conditions.Many of the proteins described above aggregate with wild-type sequences, however, the pathology of other disease-associated proteins can be tied to fibrillation-promoting mutations.We next considered fused in sarcoma (FUS), TAR DNA-binding protein 43 (TDP-43), T-cell-restricted intracellular antigen-1 (TIA1), and superoxide dismutase 1 (SOD1), which often contain missense substitutions in ALS patients.Expression of wild-type FUS, TDP-43, and TIA1 in cells exposed to A-body-inducing stimuli failed to result in significant A-body targeting or protein immobilization (Fig. 2A-F), though wild-type SOD1 was weakly sequestered in a subset of heat shock-treated cells (Fig. 2E).Therefore, we used site-directed mutagenesis to introduce two common pathological amino acid substitutions into each protein and re-assessed their A-body targeting/aggregation potential.Familial ALS-associated SOD1 mutants (A4V and G93A) 13,15 were efficiently recruited in hypoxia/acidosis-and heat shock-induced A-bodies (Fig. 2D,E), whereas none of the disease-associated FUS (F521H and P525L), TDP-43 (A315T and M337V), or TIA1 (P362L and A381T) mutations altered the affinity of these proteins for this subnuclear domain (Fig. 2A-C,E).The mutant SOD1 proteins also adopted the characteristic insoluble and immobile properties seen by other A-body constituents, demonstrating that protein aggregation was occurring under these conditions (Fig. 2F,G).FUS and TIA1 mutant constructs were targeted to cytoplasmic foci during heat shock/acidotic treatments and peri-A-body caps during TPS exposure (Fig. 2A,C).These results align with published reports demonstrating the presence of these proteins in cytoplasmic stress granules and nucleolar caps upon exposure to comparable stimuli [32][33][34][35] .Together, these data demonstrate that a subset of pathological proteins can utilize stress-specific A-body assembly machinery to rapidly induce their aggregation, suggesting that components of this physiological pathway may be associated with the disease etiology of a subgroup of human amyloidoses.

A-body aggregation of pathological proteins can be pharmacologically impaired.
As A-bodies can capture known pathological proteins (Figs. 1, 2), we sought to uncover the molecular pathways regulating this process and determine whether this could be chemically impaired within a cellular setting.We focused on hypoxia/acidosis as the A-body-inducing stimulant, because the low pH (6.0) and oxygen (1%) environment we use mimics the effects of reduced blood flow seen in a stroke setting, which has been shown to promote Alzheimer's disease pathogenesis [36][37][38] .Using an established method for quantifying the efficiency of A-body targeting 28 , we first assessed whether broad inhibitors of transcription (actinomycin D), translation (cycloheximide), and kinase signaling (staurosporine) were capable of having an impact on β-amyloid (1-42) targeting (Fig. 3A,B).Our results indicate that repressing de novo gene expression, protein synthesis, and kinase activity had no significant impact on A-body recruitment of β-amyloid (1-42), suggesting that cellular factors required for amyloid aggregation are likely present prior to stress exposure.Moving forward, we selected a panel of small molecules that were either shown to directly impair in vitro fibrillation (myricetin, epigallocatechin-gallate: EGCG, and rosmarinic acid) [39][40][41][42][43] or epidemiologically reduce the risk of neurological disease (Zileuton, Vitamin K, and Diclofenac) [44][45][46] .Under the conditions tested, the in vitro fibrillation inhibitors all failed to repress A-body recruitment of β-amyloid (1-42) in a cellular setting (Fig. 3A,B), however, two of the chemicals associated with a diminished risk of neurological disease (Vitamin K3 and Diclofenac) significantly impaired targeting to A-bodies under hypoxic/acidotic conditions (Fig. 3A,B).To compare the results from a cellular setting to the established ThT in vitro fibrillation assay, we screened all nine compounds using this classic tube-based approach.Interestingly, there was no overlap in the small molecules that impaired aggregation in the cell-based (Fig. 3B) and in vitro Thioflavin T-based (Fig. 3C) assays.Actinomycin D and the fibrillation inhibitors myricetin, EGCG, and rosmarinic acid re-capitulated their previously observed in vitro repression of β-amyloid (1-42) aggregation (Fig. 3C: top and middle panel) [39][40][41][42][43] , while the cellular inhibitors of A-body recruitment (vitamin K3 and diclofenac) had no significant effect in the tube-based assay (Fig. 3C: bottom panel).These findings highlight the potential value of using a cellular model to assess aggregation, as in vitro inhibitors may not be effective under physiological conditions.
Next, we wanted to determine whether the strong diclofenac-mediated impairment was specific to the β-amyloid (1-42) peptide, or if it had a broad inhibitory effect on the targeting of physiological and pathological proteins.The gross formation of A-bodies was assessed using the amyloidophilic dye Congo red and the www.nature.com/scientificreports/cellular marker molecule CDC73, while the targeting efficiency of additional disease-associated protein (SOD [A4V] and immunoglobulin light chain) was also determined.Our results demonstrated that A-body formation remained largely intact in diclofenac-treated cells, as the Congo red signal and CDC73 targeting efficiency were not statistically reduced (Fig. 3D,E).Like β-amyloid (1-42), the A-body targeting of immunoglobulin light chain was impaired, though recruitment of the ALS mutant SOD1(A4V) was not statistically different in the presence or absence of diclofenac (Fig. 3E).Together, these results suggest that the diclofenac-mediated effect on β-amyloid (1-42) and immunoglobulin light chain A-body targeting may be indirect, through the modulation of a cellular pathway that has some specificity towards these disease-associated proteins.
The Cyclooxygenase (COX) pathway regulates recruitment of β-amyloid to A-bodies.To assess the putative mechanism of this diclofenac-mediated impairment, we co-treated cells with diclofenac and inhibitors of transcription, translation, or kinase activity.The results from this experiment suggest that the diclofenac-induced targeting impairment is not dependent upon the expression of new gene products or kinasedependent signalling cascades (Fig. 4A).As a non-steroidal anti-inflammatory drug, diclofenac has several onand off-target effects.It has been shown to impair cyclooxygenases (COX), acid-sensing ion channels, IKK-2, and the proteosome, while also inducing ER stress [47][48][49][50][51] .Thus, we tested inhibitors for each of these proteins/pathways and assessed whether they were also capable of repressing β-amyloid aggregation.While the ER-stressor (Thapsigargin), and inhibitors of acid-sensing ion channels (amiloride), IKK-2 (TPCA-1), and the proteosome (MG132) had no effect, the COX inhibitor (celecoxib) mimicked the repression observed by diclofenac treatment (Fig. 4B).Two additional COX inhibitors (SC560 and ibuprofen) also significantly repressed β-amyloid (1-42) targeting to A-bodies (Fig. 4C), further highlighting the role of prostaglandin synthases in protein aggregation.In MCF-7 cells, only COX1 is expressed at detectable levels (Fig. 4D), and the protein is responsible for converting the unsaturated fatty acid arachidonic acid into prostaglandins 52 .This suggests that either elevated arachidonic acid levels or a deficiency of prostaglandins may be linked to the blocking of β-amyloid (1-42) recruitment.The addition of prostaglandin E2 (PGE 2 ) to diclofenac-treated cells did not rescue A-body targeting, however, the application of arachidonic acid efficiently impaired β-amyloid (1-42) localization (Fig. 4E,F).This fatty acid also repressed fibrillation in a ThT and α-amyloid fibril dot-blot assays (Fig. 4G) 53 , while a lipid of similar length that is not utilized by prostaglandin synthases (arachidic acid), had no effect in the in vitro or cellular settings (Fig. 4E,F).Our observations indicate that treatment of hypoxic/acidotic cells with diclofenac and arachidonic acid leads to increased cytoplasmic and nuclear accumulation of β-amyloid (1-42) (Fig. 4F).Thus, we assessed whether these β-amyloid (1-42)-containing structures possessed the same amyloid-like characteristics as the A-body population using Amytracker, a live-cell amyloidogenic dye.As expected, A-bodies stained positively with Amytracker, but other nuclear and cytoplasmic β-amyloid (1-42)-containing structures were not amyloid-like in nature (Fig. 4H).Together, this data suggests that diclofenac can regulate β-amyloid (1-42) aggregation, potentially through an increase in unsaturated fatty acids levels.

Discussion
The cellular processes involved in amyloidogenic diseases are not well characterized, so to explore these pathways we used stress-inducible functional amyloids as a model system.In this report, we demonstrate that a subset of proteins associated with neurodegenerative and systemic amyloid disorders aggregate within A-bodies (Fig. 1), suggesting common mechanisms exist between physiological and pathological amyloid-based aggregation.Our observations indicate that components of this physiological pathway could contribute to disease progression, perhaps by the dysregulation of factors associated with A-body formation.However, the participation of the A-body pathway does not appear to be a universal characteristic of pathological amyloidogenesis, indicating that there is potentially a divergence in the pathogenic origins of the various aggregation-based disorders.For example, FUS and TIA1 mutants were not targeted to A-bodies under any of the stress treatments examined (Fig. 2), but did aggregate in what appear to be cytoplasmic stress granules 32,34,54,55 .Thus, the distinct characteristics and regulators of these two membrane-less organelles could be linked to different pathological conditions, and further study may lead to a stratification of the amyloidogenic disorders based on their phase transition pathway of origin.These observations also provide an exciting opportunity for translational research, as this work could be developed into a new cell-based approach for the identification or validation of compounds that impair pathological protein aggregation.Our small-scale pharmacological analysis of β-amyloid (1-42) inhibitors (Fig. 3) could serve as a proof-of-principle that this system may have utility as a new drug discovery platform.Here, we found that compounds known to impair amyloid fibrillation of disease-associated proteins in vitro [39][40][41][42][43]56,57 had no detectable effect on aggregation in a cellular setting. We suect this divergence may be attributed to factors including: short biological half-life or low membrane permeability, as myricetin, EGCG and rosmarinic acid are polar compounds.Additionally, the large GFP-tag associated with the pathological proteins in our cellbased setting could also alter the aggregation properties of a protein, leading to false negatives.However, what is unmistakably true is that the sole use of in vitro approaches would exclude the detection of small molecules that indirectly regulate amyloid aggregation.The anti-inflammatory drug diclofenac may be a prime example of this, as it has been linked to reduced risk of Alzheimer's disease 44,58 and shows a strong impairment of A-bodymediated β-amyloid (1-42) aggregation in this cell-based assay, however, it failed to supress in vitro fibrillation.Thus, a combinatory approach using in silico 59 , in vitro 60,61 , bacterial 62 , and model organism-based 63,64 screening platforms along with rapid validation in this A-body model may provide a holistic approach to optimize drug candidate selection.Furthermore, we envision this model could be refined to use cell lines that more closely resemble the site of plaque formation or examining the aggregation potential of additional pathological proteins that have been linked to the various amyloidogenic disorders.Overall, this system clearly fills an underserviced www.nature.com/scientificreports/niche, as the creation of drug discovery models for the study of amyloidogenic diseases is a pressing need in the therapeutic community 65 .
Despite the extracellular localization of some disease-associated plaques (e.g., β-amyloid (1-42) and amylin), the rapidity with which these peptides convert from a diffuse and soluble molecule (no treatment) to an immobile and insoluble aggregate (stress treatment) within the cell demonstrates that environmental conditions or aberrant exposure of these peptides to A-body regulators may play a role in transitioning to an amyloid state during disease progression.We envision that localized cell death may release these cellular factors into the extracellular milieu, which could decrease the nucleation lag-phase of proximal pathological peptides and enhance fibrillation kinetics.Thus, a better understanding of cellular factors regulating this stress-response pathway may identify novel genetic risk factors associated with amyloid diseases, while exploitation of this cell-based system for drug discovery could produce new therapeutics.Together, this work makes A-body biology a new frontier for human health research.
Fluorescence microscopy, quantifications, and fluorescence recovery after photobleaching.Transfected cells were fixed with methanol (as described above), and probed with anti-CDC73 (Invitrogen, PA5-26189) antibody.Cells were then mounted on glass slides using Fluoromount G (ThermoFisher Scientific).Congo-red staining was carried out as described previously 27 .Briefly, following treatment cells were fixed in 4% formaldehyde, permeabilized with 0.5% Triton X-100, stained with a 1X (0.05%) Congo red solution, and mounted onto glass slides with 1% glycerol.Live-cells were stained with Amytracker 680 (Ebba Biotech) according to manufacturers instructions.Fluorescently-tagged and immunofluorescent, Congo red, or Amytracker stained cells were visualized on a Zeiss LSM880 laser scanning microscope with Airyscan and ZEN 2.3 software (Carl Zeiss Microscopy).For fluorescence quantification, images of 10 cells per biological replicate were captured.The percentage of cells with protein in the A-bodies was calculated by counting 100 cells and scoring each for the presence of A-body targeting.Relative A-body intensity was determined for treated samples to assess the effects of each compound on A-body recruitment as previously described 28 .Data was generated from at least 3 independent replicates.Fluorescence recovery after photobleaching (FRAP) experiments were carried out as described previously 27 .Briefly, cells were treated and visualized on a Zeiss LSM880 laser scanning microscope with Airyscan and ZEN 2.3 software (Carl Zeiss Microscopy).A-body sections, nuclear aggregates or sections of the nucleoplasm were bleached (100% argon laser at 488 nm) and monitored for the indicated time periods.Fluorescence intensity measurements were taken using ImageJ as described previously 67 .Data was generated from at least 10 cells per sample.
Thioflavin T aggregation assay and dot blot.Thioflavin T (Sigma Aldrich) stocks were prepared by dissolving powdered dye in double distilled water, with 0.2 µm filtration done to exclude clumps.The β-amyloid (1-42) peptide (UltraPure, HFIP treated) used in this assay was a recombinant variant (rPeptide) that was suspended in 1% ammonium hydroxide to a final concentration of 200 µM.Experiments were carried out in 96-well plates (black chimney wells, with µClear bottom plates) and sealed with optical film.Each well contained Thioflavin T (10 µM), β-amyloid (1-42) peptide (10 µM), and the indicated concentration of the chemical compounds, with the final volume of the well adjusted to 100 µL using PBS (pH = 7.4).Chemical compounds were dissolved in DMSO, which was used as a vehicle control for the experiments.The different components of the