Development of a specific live-cell assay for native autophagic flux

Abstract

Autophagy is an evolutionarily conserved pathway mediating the breakdown of cellular proteins and organelles. Emphasizing its pivotal nature, autophagy dysfunction contributes to many diseases; nevertheless, development of effective autophagy modulating drugs is hampered by fundamental deficiencies in available methods for measuring autophagic activity or flux. To overcome these limitations, we introduced the photoconvertible protein Dendra2 into the MAP1LC3B locus of human cells via CRISPR/Cas9 genome editing, enabling accurate and sensitive assessments of autophagy in living cells by optical pulse labeling. We used this assay to perform high-throughput drug screens of four chemical libraries comprising over 30,000 diverse compounds, identifying several clinically relevant drugs and novel autophagy modulators. A select series of candidate compounds also modulated autophagy flux in human motor neurons modified by CRISPR/Cas9 to express GFP-labeled LC3. Using automated microscopy, we tested the therapeutic potential of autophagy induction in several distinct neuronal models of amyotrophic lateral sclerosis (ALS) and frontotemporal dementia (FTD). In doing so, we found that autophagy induction exhibited discordant effects, improving survival in disease models involving the RNA binding protein TDP-43, while exacerbating toxicity in neurons expressing mutant forms of UBQLN2 and C9ORF72 associated with familial ALS/FTD. These studies confirm the utility of the Dendra2-LC3 assay, while illustrating the contradictory effects of autophagy induction in different ALS/FTD subtypes.

Keywords: ALS; FTD; autophagy; high-content; high-throughput; neurodegeneration; photoconvertible; repurposing; screen; stem cell.

Figure 1

Creation of a stable cell line serving as a reporter for autophagic flux.A, illustration depicting the use of optical pulse labeling (OPL) to measure autophagic flux. 1, Dendra2-LC3 is a photoconvertible fusion protein that irreversibly shifts its fluorescence from green to red upon exposure to 405 nm light. 2, Dendra2-LC3 is an autophagy substrate that is incorporated into autophagosomes. Prior to degradation, red fluorescence is high. 3, autophagosomes mature into autophagolysosomes, where photoconverted Dendra2-LC3 is degraded over time, resulting in a drop in red fluorescence intensity. The time-dependent decay of red signal serves as an estimate of autophagic flux, independent of new (green) LC3-Dendra2 synthesis. B, schematic for tagging native LC3 using CRISPR/Cas9 genome editing. In HEK293T cells, the Dendra2 ORF was introduced into the MAP1LC3B locus upstream of exon 1 creating an N-terminal fusion protein upon translation. C, western blot confirming the successful labeling of LC3 with Dendra2. Dendra2-LC3 HEK293T cells were treated with 20 nM siRNA targeting LC3 or scrambled siRNA. Lysates were collected after 48h and immunoblotted with an LC3 antibody, demonstrating the Dendra2-LC3 fusion protein running at the expected MW of 43 kDa that disappears upon siRNA-mediated knockdown of LC3. GAPDH serves as a loading control. D, Dendra2-LC3 reporter line imaged in the GFP and bright-field channels 48h after application of siRNA. Scale bar = 100 μm. E, Dendra2-LC3 cells imaged 6h after treatment with vehicle, 1 μM Torin1, and 20 nM Bafilomycin-A1. Scale bar = 10 μm.

Figure 2

Time-dependent decay of Dendra2-LC3 serves as an accurate measure of autophagic flux.A, Dendra2-LC3 HEK293T cells were imaged prior to photoconversion to measure background RFP intensity. Immediately following photoconversion, cells were treated with DMSO, 1 μM Torin1, or 10 nM Bafilomycin-A1 and imaged at the indicated times. Images are pseudocolored to better highlight intensity differences. Scale bar = 50 μm. B–E, time-dependent changes in photoconverted Dendra2-LC3 fluorescence in the RFP (B and C) and GFP (D and E) channels. Intensity measurements were obtained prior to (dark gray) and following photoconversion (light gray) and normalized. For RFP measurements, the background intensity prior to conversion was set to 0 and the postconversion value to 1. GFP values are scaled to the preconversion intensity. Error bars represent SEM from three replicate experiments. B, treatment with 1 μM Torin1 accelerates Dendra2-LC3 decay, reflecting enhanced autophagic degradation of the reporter, while treatment with bafilomycin-A1 stabilizes reporter half-life. D, photoconversion results in a 40% drop in GFP intensity. As new Dendra2-LC3 is synthesized, GFP levels return to prephotoconversion levels over 13.5 h. Torin1 blocks the observed return in GFP fluorescence by accelerating flux. Genetic inhibition of autophagy via siRNA-mediated knockdown of ATG5 2 days prior attenuates Torin1’s effects in both the RFP (C) and GFP (E) channels. For (B–D), Two-Way ANOVA was performed, indicating significant effects of both time and treatment, as well as a significant interaction between time and treatment. These values are summarized in Table S1. ∗ denotes p < 0.05 using DMSO as reference group with Tukey’s multiple comparisons test; # indicates p < 0.05 with the scramble control for each drug treatment as the reference group (i.e., scramble siRNA 1 μM Torin1 versus ATG5 siRNA 1 μM Torin1). Superscript number indicates the first time point when significance was achieved.

Figure 3

Drug library screens in Dendra2-LC3 HEK293T cells confirm assay validity and identify new autophagy modulators.A, an unbiased screen of the Enzo autophagy compound library identified several known autophagy-modulating compounds, including enhancers (rapamycin, NVP-BEZ235, AKT inhibitor X) and inhibitors (bafilomycin-A1). All drugs were added at a final concentration of 10 μM via liquid handler, and autophagic flux estimated by their effects upon clearance of photoconverted (red) Dendra2-LC3 9 h after drug addition. B, representative images of red Dendra2-LC3 immediately postconversion and 9 h after drug addition, demonstrating relatively rapid clearance with application of enhancers and marked accumulation of Dendra2-LC3 with inhibitors. Both the 9 h/0 h RFP ratio and this value normalized to DMSO 9 h/0 h RFP intensity are listed below each set of images. Torin1 equivalents listed in Fig. 3-source data = [1 − (9 h/0 h RFP ratio)_drug]/[1 − (9 h/0 h RFP ratio)_Torin1]. Scale bar = 100 μm. C, the effect of autophagy enhancers on Dendra2-LC3 half-life is attenuated by siRNA-mediated knockdown of ATG5 2 days prior to drug application. Two-Way ANOVA indicated a significant interaction between time and treatment. F(126,300) = 15.22, p < 0.0001 and significant effects for treatment F(9,300) = 3427, p < 0.0001 and time F(14,300) = 957.3, p < 0.0001. D, similar effects are observed in the green channel. Two-Way ANOVA found a significant interaction between time and treatment F(126,300) = 20.71, p < 0.0001 as well as significant main effects for treatment F(9,300) = 721, p < 0.0001 and time F(14,300) = 320.8, p < 0.0001. In (C and D), error bars represent SEM from three biological replicates, # indicates p < 0.05 using Tukey’s multiple comparisons test with the scramble control for each drug treatment as the reference group (i.e., Scramble siRNA Torin1 versus ATG5 siRNA Torin1). Superscript number indicates the first time point when significance was achieved.

Figure 4

High-throughput screening identifies novel autophagy inhibitors.A, “90/10” experiment validating Dendra-LC3 intensity in the GFP channel as an indicator of autophagy flux. 320 wells of a 384w plate were treated with DMSO and 32 were treated with 1 μM Torin1. Plates were imaged immediately before, and 15 h after drug treatment. Z’=0.52 ± 0.04 in three replicates. B, schematic depicting the screening hierarchy used. C, the primary screen utilized the Maybridge 24K library, consisting of 24,000 chemically diverse compounds. For enhancers, changes in Dendra2-LC3 GFP intensity were normalized to Torin1’s effects using the formula [1 − (9 h/0 h RFP ratio)_drug]/[1 − (9 h/0 h RFP ratio)_Torin1]. Nontoxic compounds that passed the primary screen were filtered by retesting in the GFP channel, then evaluated in a secondary screen involving calculation of Dendra2-LC3 half-life in the RFP channel. Hits were retested in the secondary screen, followed by repeat evaluation in the RFP channel using fresh compound from a different distributor. The color and size of each dot denote the stage at which individual compounds were eliminated, in accordance with the key. Candidates that passed all filters are shown in black. D, representative images of the primary screen, secondary screen, and repeat secondary screen with fresh drug for five putative autophagy inhibitors, ranked based on the magnitude of inhibition measured in the secondary screen. Z-scores are reported for each screening phase. Scale bar = 100 μm. E, unmodified HEK293T cells were treated with vehicle or each compound at either 10 μM or 100 μM. Lysates were collected 9h after treatment and immunoblotted with an LC3 antibody. F, quantification of three biological replicates demonstrating significant inhibition autophagic flux by compounds 1 and 2 at 100 μM. For each group LC3-II was normalized to the loading control (GAPDH) and scaled to 100 nM bafilomycin-A1. Error bars represent standard error of the mean. One-way ANOVA showed significant differences between groups (F = 24.28, p < 0.0001), ∗p < 0.01 compared with DMSO, Dunnet’s multiple comparison test.

Figure 5

Evaluating the autophagy modulating effects of clinically relevant drugs.A, primary screen of the Prestwick drug library. Z-score is calculated as fold SD_DMSO greater or less than mean DMSO_15H/0H. Significant changes in flux, toxicity, or puncta for each drug are indicated by the color and size of representative dots, according to the key. The magnitude of effect is represented as % bafilomycin-A1 (left y-axis, for inhibitors) or % Torin1 (right y-axis, for enhancers). B, time-dependent decay in red (photoconverted) Dendra2-LC3 was used in a secondary screen of nontoxic candidates emerging from the primary screen. C, secondary screen of the SelleckChem Bioactive Compound-I library. Drugs with a normalized fluorescence intensity >1.0 were grouped together above the broken y-axis. D, stacked bar plot depicting enhancers (gray) identified from the Enzo, Prestwick, and SelleckChem drug libraries grouped by mechanism of action. Grey bars represent the compounds within each class with autophagy-enhancing activity. Red bars show all other compounds in the same class that had no observable effects. Drugs with distinct mechanisms of action, or those with incomplete or missing annotation, were labeled as “other.” Error bars in (B and C) represent SEM from six images (two images/well of three replicate wells).

Figure 6

Autophagy enhancers have mixed effects in human motor neurons.A, schematic describing the generation of human motor neurons differentiated from iPSCs with GFP knocked in to the MAP1LC3B locus. TALEN-mediated genome editing was used to introduce a DNA cassette containing the transcription factors Ngn2, Isl1, and Lhx3 downstream of a doxycycline responsive TRE3G regulatory element. Positive colonies were selected for using iRFP fluorescence. Two weeks of induction produce a homogeneous population of choline acetyl transferase (ChAT) and homeobox 9 (HB9) positive motor neurons (iMNs). Scale bars = 100 μm. B, day 14 iMNs imaged in the apical plane 5 h after treatment with the indicated compounds at 5 μM. Scale bar = 20 μm. C, percentage of neurons in the apical (red) and basal (gray) planes showing visible autophagosomes, as determined by a blinded observer. Black numbers above each column show the number of positive neurons over total neurons quantified, pooled from ≥3 replicates. D and E, representative western blots of Day 14 iMNs treated for 5h with 10 μM of each compound, with and without 10 mM ammonium chloride (NH4CL). F and G, quantification of western blots with intensity normalized to the mean of DMSO. Error bars represent standard error of the mean from at least three replicate experiments. ∗p < 0.0.05, Tukey’s posthoc test, when compared with DMSO. ^#p < 0.0.05, Tukey’s posthoc test, when compared with DMSO/NH4CL.

Figure 7

Autophagy modulation has distinct effects in ALS/FTD disease models.A, mixed rodent spinal neurons were transfected on DIV 4 with Dendra2-LC3, imaged 24h later (Day 1 pre), then pulsed with 405 nm light to photoconvert Dendra2-LC3 before imaging repeatedly and longitudinally over several days to track the time-dependent loss of red fluorescence and neuronal survival. Scale bars = 100 μm in each panel. B, experimental outline for determining the relationship between Dendra2-LC3 half-life and neuronal survival. After calculating Dendra2-LC3 half-life for individual neurons (Stage 1), each cell is prospectively tracked using automated microscopy to determine its time of death (Stage 2; red number and corresponding arrow). Scale bar = 100 μm. C, penalized spline Cox proportional hazards model depicting Dendra2-LC3 half-life (x-axis) versus relative risk of death (y-axis) for primary cortical (black) and spinal (red) neurons, demonstrating a strong proportional relationship for both populations (cortical: p = 3.4 × 10⁻⁹; spinal p = 1.1 × 10⁻⁶, linear Cox proportional hazards). Each hash mark represents an individual neuron, collected from three biological and eight technical replicates each. Gray dotted lines mark 95% confidence intervals. D, NVP-BEZ235 (25 nM NVP) treatment suppresses toxicity in primary cortical neurons cotransfected with WT-TDP43-GFP and scramble siRNA, but not those transfected with WT-TDP43-GFP and ATG5 siRNA. Table S2 summarizes the hazard ratio and statistical significance of each comparison as determined by Cox proportional hazards analysis. N for each group represents total neurons pooled from three biological replicates. ∗p < 0.05, Cox proportional hazards analysis. E, CCT128930 (1 μM) treatment increased toxicity in primary cortical neurons overexpressing iRFP-P497H-UBQLN2. Table S3 summarizes hazard ratio and statistical significance of each comparison. F, hazard ratios calculated in D and E. ∗p < 0.05, Cox proportional hazards analysis. Error bars mark 95% confidence intervals. G, a, schematic depicting the generation of WT and mutant C9ORF72 iPSC-derived neurons. b, day 14 neurons were treated with the indicated compound, imaged for an additional 10 days, and time of death recorded for each neuron (red numbers). Scale bar = 100 μm. H, NVP-BEZ235 (25 nM NVP) treatment increased toxicity in mutant C9ORF72 neurons, but not WT controls. Table S5 summarizes the hazard ratios and statistical significance. N for each group represents total neurons pooled from three replicate experiments. ∗p < 0.05, Cox proportional hazards analysis. I, hazard ratios from the experiments in (H). Error bars denote 95% confidence intervals.