Flow cytometry allows rapid detection of protein aggregates in cellular and zebrafish models of spinocerebellar ataxia 3

ABSTRACT Spinocerebellar ataxia 3 (SCA3, also known as Machado–Joseph disease) is a neurodegenerative disease caused by inheritance of a CAG repeat expansion within the ATXN3 gene, resulting in polyglutamine (polyQ) repeat expansion within the ataxin-3 protein. In this study, we have identified protein aggregates in both neuronal-like (SHSY5Y) cells and transgenic zebrafish expressing human ataxin-3 with expanded polyQ. We have adapted a previously reported flow cytometry methodology named flow cytometric analysis of inclusions and trafficking, allowing rapid quantification of detergent insoluble forms of ataxin-3 fused to a GFP in SHSY5Y cells and cells dissociated from the zebrafish larvae. Flow cytometric analysis revealed an increased number of detergent-insoluble ataxin-3 particles per nuclei in cells and in zebrafish expressing polyQ-expanded ataxin-3 compared to those expressing wild-type human ataxin-3. Treatment with compounds known to modulate autophagic activity altered the number of detergent-insoluble ataxin-3 particles in cells and zebrafish expressing mutant human ataxin-3. We conclude that flow cytometry can be harnessed to rapidly count ataxin-3 aggregates, both in vitro and in vivo, and can be used to compare potential therapies targeting protein aggregates. This article has an associated First Person interview with the first author of the paper.


Preparation of SCA3 cell cultures for FloIT
For flow cytometry experiments, SHSY5Y cells were seeded into 6-well plates and prepared for flow cytometry 24-hours post-transfection. Cultured SHSY5Y cells were imaged with 20x objective using an EVOS microscope (Invitrogen, catalogue number AMF4300) under brightfield and GFP LED light cube to determine transfection efficiency prior to harvesting. The total number of cells and total number of GFP-expressing cells were counted using automated cell counting functions within Fiji, allowing calculation of transfection efficiency. Cells were briefly washed in PBS and harvested using 0.5% trypsin-EDTA and pelleted. Pelleted cells were re-suspended in lysis buffer (phosphate buffered saline containing 0.5% Triton-X 100 and complete protease inhibitors [Roche]) and DAPI was added (final concentration 5 μM) to allow enumeration of the number of nuclei.
The use of DAPI to quantify nuclei aligns with experiments from Adriaenssens, Tedesco [26] but contrasts with many previous reports of FloIT [20][21][22]. Samples were incubated on ice and protected from light until analysis. All samples underwent flow cytometric analysis within 30 minutes of lysis with Triton-X 100.

Transgenic SCA3 Zebrafish
The present study utilised a previously described zebrafish model of SCA3 [32]. Within these studies we used embryos resulting from crossing zebrafish driver line mq15:

Imaging of aggregates in a transgenic zebrafish model of SCA3
At 6 days post-fertilisation (6 dpf), transgenic zebrafish were anaesthetised with 0.01% Tricaine (Sigma, catalogue number E10521) and embedded in 1% low-melting agarose (Sigma, catalogue number A4018). Confocal imaging was performed using an upright Leica SP5 confocal microscope (40x water submersible objective). An argon laser (39% laser power) was used to excite EGFP. Z-stacks spanning the depth of the spinal cord (~10 μm) were acquired and final images were obtained as maximum intensity projections using or Fiji software [34]. EGFP positive aggregates were manually counted and the aggregate size measured by a researcher blind to experimental group.

Western Blotting
SHSY5Y cells transiently expressing EGFP, EGFP-fused human ataxin-3 with 28 glutamines or EGFP-fused human ataxin-3 with 84 glutamines were harvested for protein extraction and lysed using ice-cold RIPA solution containing protease inhibitors (Complete ULTRA tablets, Roche) and phosphatase inhibitors (PHOSstop tablets, Roche). Cells were incubated in RIPA for 20 minutes (gently shaking) before manual scraping was used to harvest cells. To extract protein from 2 dpf zebrafish larvae, zebrafish larvae were euthanised and homogenised in RIPA solution (0.5 μL per larvae) containing protease inhibitors (Complete ULTRA Tablets, Roche), using a manual dounce. were then transferred to a PVDF membrane (0.45 μm pore size) for immunoblotting.
Immunoblots were probed with primary antibodies against human ataxin-3 (produced in rabbit, gift from H. Paulson) and GAPDH (produced in mouse, Proteintech). Immunoblots were probed with appropriate horseradish peroxidase secondary antibodies (Promega) and visualised by chemiluminescence (Supersignal detection kit, Pierce) using an ImageQuant LAS4000 imaging system.

Preparation of Dissociated of Zebrafish Larvae for Flow Cytometry
Dissociation of whole zebrafish was adapted from methods previously described by Acosta, Watchon [35]. In brief, whole zebrafish larvae (2 or 6 dpf) were euthanised and larvae bodies were manually transferred to a petri dish in a droplet of E3 media and manually dissected into pieces (<1 mm) using a scalpel blade. Larvae pieces were then transferred to an Eppendorf tube and centrifuged. The supernatant was removed and the pellet was enzymatically digested using 0.5% Trypsin-EDTA (500 μL for 2dpf larvae, 1000 μL for 6dpf larvae) for 30 minutes at 37 o C. Samples were vortexed frequently to aid digestion.
Trypsinisation was stopped by the addition of 1 mL of DMEM cell culture media containing 10% fetal bovine serum. Samples were pelleted and the DMEM/trypsin solution was removed. Samples were resuspended in 200 -500 μL of PBS containing 0.5% Triton-X and 1 x RedDot TM far red solution (Gene Target Solutions, catalogue number 40060) to identify nuclei.
In order to confirm that cells could survive the dissociation process, a subset of experiments were performed whereby dissociated cells were not used for flow cytometry and instead were stained with Hoechst 33342 live cell stain (Thermo Fisher, final concentration 400 μM) and imaged using an EVOS microscope (20x objective). The total number of cells and total number of Hoechst -positive cells were counted using Fiji automated cell counting functions, allowing calculation of cell survival post-dissociation (Supplementary Figure 1A).

Drug treatments in SCA3 cell cultures and transgenic SCA3 zebrafish
As a proof of principle, we examined the effect of prior treatment with compounds known to alter activity of the autophagy protein quality control pathway on the number of detergent-insoluble ataxin-3 particles detected using flow cytometry, in SCA3 cell cultures Treatment with DMSO alone acted as the vehicle control and the effect of all examined drug treatments were directly compared to vehicle treated controls.  Figure   1B).

Data analysis
Data analysis was performed using GraphPad Prism (Version 8) software. Group comparisons were made using one-way ANOVA tests, followed by a Tukey post-hoc to identify differences. In cases where only two groups were compared, comparisons were made using a student t-test. All graphs display group mean data ± standard error mean.

Expression of human ataxin-3 with an expanded polyglutamine region in neuronal-like SHSY5Y cells results in the formation of EGFP-positive aggregates
Microscopic analysis of SHSY5Y cells expressing EGFP fused-human ataxin-3 containing 28Q (EGFP-ataxin-3 28Q), EGFP-ataxin-3 84Q or EGFP alone, revealed the presence of more frequent EGFP-positive aggregates within the cells expressing human ataxin-3 containing 84Q than 28Q or EGFP-only vector control ( Figure 1A). Manual counting of the number of EGFP-positive protein aggregates and one-way analysis of variance revealed a significant difference in the number of EGFP-positive aggregates across genotypes (p < 0.0001). Post-hoc comparisons revealed that more aggregates were present in the SHSY5Y cells expressing EGFP-ataxin-3 84Q when compared to cells expressing EGFP alone (p < 0.0001) or EGFP-ataxin-3 28Q (p = 0.0002; Figure 1B). Further, analysis of manual counts revealed a difference in the number of cells affected with protein aggregates across Disease Models & Mechanisms • DMM • Accepted manuscript genotypes (p < 0.0001, Figure 1C). Multiple comparisons revealed that cells transiently expressing EGFP-fused ataxin-3 84Q displayed a significantly higher percentage of cells harbouring EGFP-positive aggregates than cells expressing EGFP only (p < 0.0001) and cells expressing EGFP-fused ataxin-3 with 28Q (p = 0.0033).
Next, we used western blotting to examine the relative expression of ataxin-3 across our three cell culture expression models, as the relative amount of protein expressed could influence ataxin-3 aggregation. Immunoblot analysis of SHSY5Y cells transiently expressing EGFP-fused human ataxin-3 or an EGFP only control revealed expression of endogenous ataxin-3 and overexpression of EGFP-fused human ataxin-3 ( Figure 1D, see Supplementary   Figure 2A for full immunoblot). The expression of EGFP-fused human ataxin-3 was found to differ across examined genotypes (p < 0.0001, Figure 1E), with expression human ataxin-3 found to be higher in cells expressing EGFP-fused ataxin-3 28Q and ataxin-3 84Q when compared to cells expressing EGFP only (p < 0.0001 and p = 0.0001, respectively).
Interestingly, the expression of EGFP-fused ataxin-3 28Q was found to be significantly increased when compared to the expression found in ataxin-3 84Q expressing cells (p < 0.0001), suggesting the increased presence of ataxin-3 aggregates in our MJD cells is not due to underlying differences in protein expression.
We next validated the rapid quantification of the number Triton-X 100 insoluble aggregates using the previously reported FloIT methodolgy [20]. Transfected cells underwent flow cytometric analysis to quantify the number and size of Triton-X insoluble particles ( Figure 2A). Firstly, fluorescent microscopy was used to confirm the number of transfected cells per sample (prior to harvesting). The calculated transfection efficiency of each sample was utilised to calculate the total number of detergent-insoluble particles per sample. Next, the total number of nuclei present within the lysed sample was determined by a DAPI stain for nuclei. The number of DAPI-positive particles was quantified using population gating based on intensity of UV fluorescence and relative size (forward scatter).
No statistically significant differences in the number of DAPI-positive nuclei were detected across examined groups (p = 0.996, Figure 2B). The number of Triton-X 100 insoluble EGFP-positive particles was then identified via gating of particle populations present within experimental replicates per group), with a greater number of detergent-insoluble particles per 100 cells found in EGFP-ataxin-3 84Q-expressing SH-SY5Y cells when compared to cells expressing EGFP alone (p = 0.0022) and EGFP-ataxin-3-28Q (p = 0.0118) ( Figure 2C). No significant differences were found between the number of detergent-insoluble particles expressed in EGFP alone or EGFP-ataxin-3-28Q (p = 0.4328). Additionally, use of non-fluorescent microspheres of known size allowed us to identify that EGFP-ataxin-3-84Q cells harbored detergent-insoluble particles that were similar in size to cells expressing EGFP-ataxin-3-28Q and EGFP alone ( Figure 2D).

Treatment of SCA3 cell cultures with compounds known to modulate autophagic activity altered the number of EGFP-fused ataxin-3 particles detected by FloIT
In order to validate the use of FloIT as a tool to screen the effect of compounds on detergent-insoluble protein species proteinopathy, we treated cells transiently transfected with EGFP-ataxin-3 84Q for 24 hours with 3MA, a known inhibitor of autophagy [36, 39], and calpeptin, a compound known to induce autophagy and reduce proteinopathy in polyQ disease models [32, 38, 40]. Indeed, we found that 24-hour treatment with modulators of the autophagy protein quality control pathway altered the number of EGFP particles detected by flow cytometry ( Figure 3A). We found that treatment with 3MA did not alter the number of DAPI-positive nuclei detected within each sample (p = 0.838, Figure 3B). In contrast, treatment with 3MA produced a 1.3-fold increase in the number detergent-insoluble EGFP particles present within cells expressing EGFP-ataxin-3 84Q when compared to vehicle treatment (p = 0.002, Figure 3C). Next, we examined the effect of increasing doses of calpeptin, a known autophagy inducer, on the number of Triton-X insoluble particles detected in cell culture samples transfected with EGFP ataxin-3 84Q ( Figure 3D). Calpeptin treatment did not alter the number of nuclei detected within each sample (p = 0.438, Figure   3E), however calpeptin was found to alter the number of detergent-insoluble EGFP particles detected (p = 0.028, Figure 3F). Post-hoc comparisons revealed that treatment with 1 μM calpeptin did not significantly alter the number of insoluble particles detected when compared to vehicle treatment (p = 0.331). In contrast, treatment with 2.5 μM or 5 μM calpeptin produced a statistically significant decrease in the number of detergent-insoluble particles, compared to vehicle treatment (p = 0.021 and p = 0.031, respectively.

Expression of expanded human ATXN3 in transgenic zebrafish results in the formation of ataxin-3-positive aggregates
We have previously described a transgenic zebrafish model of SCA3 that overexpresses EGFP-fused human ataxin-3 with expanded polyQ (84Q) tract, that develops impaired swimming behaviour and shortened lifespan [32]. Furthermore, histological analysis of the medulla from adult zebrafish (12 months of age) expressing human mutant ataxin-3 (84Q) evidenced a neuritic beading pattern, that is the presence of ataxin-3 protein aggregates in neuronal neurites, which was positive for both ataxin-3 and ubiquitin [32]. In this current study we performed confocal microscopy on transgenic zebrafish larvae expressing human ataxin-3 at 6 dpf to examine the rostral spinal cord ( Figure 4A). We observed similar levels of overall human ataxin-3 expression in zebrafish expressing ataxin-3 with 23 glutamine residues and ataxin-3 with 84 glutamine residues and observed EGFP-positive protein aggregates in neurons in both transgenic models. Manual counting of the EGFP-positive protein aggregates identified a statistically signficntly higher number of aggregates in larvae expressing EGFP-ataxin-3-84Q than in EGFP-ataxin-3-23Q (p = 0.0007; n = 5 zebrafish larvae imaged per ataxin-3 construct, Figure 4B). In addition to quantifying the number of aggregates present per zebrafish, we also quantified the percentage of neurons that harboured EGFP-positive protein aggregates. Transgenic MJD zebrafish were found to possess a significantly higher percentage of cells harbouring ataxin-3 aggregates when compared to zebrafish expressing EGFP ataxin-3 23Q (p = 0.0004, Figure 4C). Measurement of the size of observed protein aggregates within our maximum intensity microscopy images revealed that transgenic zebrafish expressing EGFP ataxin-3 23Q had significantly smaller aggregates (approximately 1 μm) compared to zebrafish expressing EGFP ataxin-3 84Q (containing aggregates greater than 2 μm in diameter) ( Figure 4D).
In order to confirm that the increase presence of ataxin-3 aggregates in transgenic MJD zebrafish was not due to underlying differences in protein expression, protein lysates from 2 dpf zebrafish were obtained and immunoblotted to examine ataxin-3 expression.

Disease Models & Mechanisms • DMM • Accepted manuscript
We also examined aggregate number and size at 6 dpf, an age at which the 84Q transgenic zebrafish present with a swimming phenotype characterised by reduced distance swum [32]. Again, we found similar numbers of nuclei were detected within 6 dpf samples (p = 0.861, Figure 5E). We also found that the number of detergent-insoluble ataxin-3 aggregates was consistent at both timepoints examined, with a significant difference across genotype groups still evident at 6 dpf (p = 0.0134). Post-hoc analysis revealed that significantly more detergent-insoluble EGFP-positive particles were present in dissociated zebrafish expressing EGFP ataxin-3-84Q compared to non-transgenic siblings (p = 0.0214) and EGFP ataxin-3-23Q-expressing transgenic fish (p = 0.0343, Figure 5F). The size of detergent-insoluble EGFP-positive particles was again found to be similar across analysed genotypes, with mean detergent-insoluble particle diameter around 3 microns in diameter by 6 dpf ( Figure 5G).
As proof of the utility of the FloIT approach to assess efficacy of drug treatments aimed at modifying protein aggregate formation and detergent-insoluble protein species, EGFP-ataxin-3 84Q larvae were treated with 3 mM chloroquine, an autophagy inhibitor for 24 hours. A 1.5-fold increase in detergent-insoluble EGFP particles was observed in chloroquine treated larvae compared to vehicle control larvae from the same clutch (p = 0.012, Figure 5H). FloIT also detected a difference in the number of detergent-insoluble EGFP positive particles present following 24-hour treatment with the autophagy inducer calpeptin (p = 0.006, Figure 5I). Post-hoc comparisons revealed that treatment with 25 μM calpeptin produced a statistically significant decrease in detergent-insoluble particles when compared to vehicle treatment (p = 0.038). Treatment with 50 and 100 μM doses appeared to produce a decrease in detergent-insoluble particles, however these comparisons were not statistically significant (p = 0.055 and p = 0.053, respectively).

Flow cytometry approach detects protein aggregates in a cell culture model of SCA3
Here we report that neuronal-like (neuroblastoma) SH-SY5Y cells expressing EGFP-ataxin-3 containing a polyQ expansion (84 polyQ repeats) develop EGFP-positive protein aggregates. These findings align with existing experimental evidence that suggests polyglutamine expanded ataxin-3 can form protein aggregates within cultured murine or

Disease Models & Mechanisms • DMM • Accepted manuscript
human neuroblastoma cells [11,41,43,44]. Interestingly, we observed ataxin-3 protein aggregates predominately in the cell cytoplasm, contrasting with reports of intranuclear ataxin-3 positive inclusions in SCA3 brain tissue [11,13,42,45] In addition to examining protein aggregation in our cell culture and transgenic zebrafish models of SCA3, we also utilised western blotting to examine the expression of ataxin-3 across our models. The purpose of our western blotting experiments was to confirm that the increased presence of EGFP ataxin-3 aggregates in our models expressing mutant expanded ataxin-3 was not due to underlying differences in protein expression. In our cell cultures, we found the expression of ataxin-3 to be significantly higher in cells transiently expressing EGFP-fused ataxin-3 with 28Q compared to cells transiently expressing EGFP-fused ataxin-3 with 84Q. In our transgenic zebrafish, expression of human ataxin-3 was found to be similar across larvae expressing EGFP ataxin-3 23Q and EGFP ataxin-3 84Q, however

Transgenic zebrafish expressing mutant ataxin-3 develop protein aggregates from two days of age
Here we provide the first adaption of FloIT for use with an in vivo model. We were able to successfully quantify the number of Triton-X insoluble EGFP-positive protein aggregates present within our transgenic zebrafish model of SCA3 that express EGFP fused to human ataxin-3 under a neuronal promotor (HuC, elavl3). We found that our transgenic zebrafish expressing EGFP-fused to human ataxin-3 containing a long polyQ stretch (84 glutamine residues) displayed significantly more detergent-insoluble EGFP-ataxin-3 inclusions than non-transgenic siblings or zebrafish expressing ataxin-3 with a short polyQ stretch fused to EGFP (23 glutamine residues). Interestingly, this phenotype was detectable at two different time points in SCA3 zebrafish development; two days post-fertilisation, prior to the onset of swimming deficits, and six days post-fertilisation, when a movement phenotype is detectable [32]. This suggests that formation of detergent-insoluble ataxin-3 aggregates may be an early disease phenotype that may contribute to the development of neurotoxicity, neurodegeneration and motor impairment. Further, we found that treatment with a compound that we have previously demonstrated to improve swimming of these zebrafish (calpeptin), does indeed also decrease the presence of this insoluble protein [32].

Disease Models & Mechanisms • DMM • Accepted manuscript
The ability to detect the presence of this protein aggregation phenotype at such an early timepoint, including the larval stages, has many advantages. Firstly, FloIT is a relatively inexpensive and efficient analysis tool that can be utilised to provide a rapid readout of treatment efficacy on protein aggregation in cells dissociated from large clutches of sibling zebrafish larvae. We suggest that using FloIT to examine proteinopathy phenotypes may be valuable within high-throughput drug screening pipeline in zebrafish models of proteinopathy and neurodegenerative diseases, either on its own, or together with tracking of locomotion behaviour (Figure 4). Zebrafish larvae can be assayed for improvements in swimming and then dissociated into single cells at the completion of the experiment, enabling identification of compounds that induce beneficial effects on animal movement and cellular phenotypes. At these early larval stages drugs and small compounds can easily be dissolved in the water that the larvae are incubated within, and are absorbed by the larvae, making dosing straight-forward [33]. Further, in comparison with more traditional methodologies for detecting proteinopathy, such as western blotting, live imaging confocal microscopy and immunostaining, this flow cytometry approach is much less time consuming and laborious. Whilst these other methodologies may provide critical insights relating to the relative expression and location of ataxin-3 positive inclusions, these approaches can also be incorporated into a possible drug testing workflow for greatest insight.
To validate the use of this FloIT approach to investigate potential treatments of SCA3, and related diseases, and to confirm that the detected particles are indeed protein aggregates, we examined the effect of treatment with autophagy modifying compounds on the number of particles detected. We hypothesised that treatment with autophagy inhibitors, 3MA in vitro or chloroquine in vivo, would result in an increase in the number of insoluble EGFP-ataxin-3 aggregates due to blockage of their removal. Further, we hypothesised that treatment with calpeptin, a compound known to inhibit activity of calpain proteases and induce autophagic activity, would decrease the number of detergent-insoluble EGFP-ataxin-3 aggregates present. Indeed, we found that treating SHSY5Y cells or zebrafish expressing EGFP-ataxin-3 Researchers should consider which approach may be most suited to their experimental aims before embarking on flow cytometric analysis of proteinopathy.
In conclusion, our findings highlight the utility of FloIT as a rapid approach to quantify protein aggregation, which can be utilised to screen novel compounds in vitro and in vivo.
We report this novel approach of applying FloIT to transgenic zebrafish samples that can be incorporated into a drug testing pipeline to aid identification of compounds that slow or prevent disease progression including ameliorating protein aggregation and swimming impairment.  (D) Detergent-insoluble particle size did not significantly differ across examined genotypes. * indicates p < 0.05 and ** indicates p < 0.01, *** indicates p < 0.001 and **** indicates p < 0.0001. Graphs depict mean ± standard error mean, n = 4-9 experimental replicates.