TFEB ameliorates the impairment of the autophagy-lysosome pathway in neurons induced by doxorubicin

Doxorubicin promotes formation of pre-autophagosomal complexes

Doxorubicin causes problems in the heart by altering proteostasis systems, such as autophagy, in cardio-myocytes [31]. In neurons, autophagy has a protective role, and it is dysregulated in many neurological disorders [32]. Some autophagic pathways, however, or overstimulated autophagy are associated with cytotoxicity, which indicates that autophagy in neurons can be helpful, harmful, or even both [33,34]. We hypothesized that neurons upregulate autophagy in response to the treatment with doxorubicin. If so, we sought to determine if enhanced autophagy is a neuroprotective or neurotoxic mechanism.

Commonly used methods for monitoring autophagy are based on biochemical properties of an autophagy protein LC3. When an autophagosome forms, a proteolized form of LC3, LC3-I, is lipidated and can be detected as a mobility shift by western blotting. Increased LC3-II levels can also be a consequence of autophagic blockade at some step after the conversion of LC3-I to LC3-II. To determine if doxorubicin stimulates neuronal autophagy, we treated primary cortical neurons with doxorubicin or with doxorubicin and NH₄Cl, which neutralizes the lysosomal pH and inhibits the last step of autophagy-mediated degradation, or with NH₄Cl alone. LC3-II levels were higher in neurons incubated with the combination of doxorubicin and NH₄Cl than in neurons incubated only with doxorubicin, indicating that autophagy is indeed upregulated (Fig. 1A,B).

Since LC3-I lipidation might reflect the induction of a pathway not directly related to autophagy [35], we decided to confirm that doxorubicin indeed promotes the formation of the pre-autophagosomal structures. Beclin1 is a part of the pre-autophagosomal complex and is often used as an autophagy marker [36,37]. Beclin1-GFP and mApple, a marker of morphology, were expressed in two cohorts of primary cortical neurons. The first neuronal cohort was treated with a vehicle and the second cohort was treated with doxorubicin. As expected, doxorubicin induced the formation of beclin1-GFP-positive puncta (Fig. 1C,D), similar to what neuronal autophagy inducers do [37]. We recently showed that stimulating autophagy in neurons leads to the formation of beclin1-positive structures in the soma and neurites [37,38]. In these experiments, we also observed beclin1-positive structures in the soma and neurites in doxorubicin-treated neurons.

Levels of p62 increase in doxorubicin-treated neurons

The protein p62 (SQSTM1), which binds ubiquitin and LC3, is a selective substrate for autophagy [39,40]. When autophagy is enhanced, the levels of p62 are reduced; impeding autophagy is associated with higher p62 levels. In cardiomyocytes, doxorubicin leads to accumulation of p62, indicating that autophagy is inhibited [31]. We, therefore, tested if doxorubicin affects the degradation of p62 in primary neurons. p62-GFP and mApple were expressed in two cohorts of primary cortical neurons; the first neuronal cohort was treated with a vehicle, and the second was treated with doxorubicin. Transfected neurons were followed longitudinally with an automated microscope. Interestingly, neurons sometimes formed p62-GFP-positive aggresomes under basal conditions, but doxorubicin induced the formation of very large inclusion bodies formed by p62-GFP, suggesting that p62 accumulates and autophagy is ineffective (Fig. 2A,B).

Next, we tested if the levels of endogenous p62 change in neurons treated with doxorubicin. Cultured cortical neurons were treated with either a vehicle or doxorubicin, lysed and analyzed for p62 (Fig. 2C,D). As expected, p62 levels were higher in neurons exposed to doxorubicin. Our data suggest that doxorubicin-treated neurons exhibit upregulated autophagy and there is some flux though autophagy, but autophagy is impaired and inefficient.

Doxorubicin induces accumulation of vacuolar structures and organelles in neurons

Next, we examined whether the changes observed with the p62 reporter could be confirmed by electron microscopy. The first neuronal cohort was treated with a vehicle, and the second was treated with doxorubicin. Under control conditions, we found no autophagosomal structures (Fig. 3A); we have previously demonstrated the same result in untreated neurons [38]. In contrast, neurons treated with doxorubicin contained massive accumulations of vacuoles with often poorly identified content (Fig. 3B). Neurons also contained auto-phagosomes (Fig. 3B). Interestingly, we observed lipid droplets, which we never detected in control cells (Fig. 3B). Unexpectedly, more mitochondria were present in doxorubicin-treated cells (Fig. 3C). Some mitochondria were abnormally shaped or rounded (Fig. 3C). These data directly prove that doxorubicin impairs the de-gradative autophagic pathway, leading to accumulation of damaged organelles, autophagosomes, vacuoles, and lipid droplets.

Doxil induces accumulation of autophagosomes and lipid droplets in mouse brains

We were impressed by the changes we observed in doxorubicin-treated cultured neurons, and wondered if we could detect these pathologies in vivo. Mice underwent a course of chemotherapy with doxil, the liposomal doxorubicin frequently used in human patients. Mice were then perfused, and their brains were analyzed by electron microscopy. Remarkably, we observed autophagosomes, which often contained mitochondria. Some autophagosomes were empty, as in neurodegenerative Huntington’s disease [41] (Fig. 3D). We also detected the accumulation of lipid droplets and lipofuscin, a mixture of oxidized proteins and lipids, found in aged brains and in Batten disease (Fig. 3D) [42,43]. These data directly confirm that the treatment with doxorubicin leads to the accumulation of damaged organelles, autophagosomes, and lipid droplets.

Doxorubicin damages lysosomes in primary neurons

Doxorubicin is membrane-permeable in its basic form but impermeable when the molecule is protonated. Once inside the cell, doxorubicin enters acidic organelles such as lysosomes, in which doxorubicin is protonated and retained, leading to alkalinization of these organelles [44]. We hypothesized that doxorubicin impairs autophagy by affecting health of lysosomes in neurons. To test that, primary cortical neurons were treated with doxorubicin and stained with a green lysotracker dye, which stains acidic compartments in live cells. Remarkably, the green signal was significantly lower in cultures treated with doxorubicin than in controls (Fig. 4A, B). Cultures were then fixed and stained with MAP2c to confirm that visualized lysosomes belonged to neurons, and imaged again (Fig. 4A, B). Our data suggest that the drug affects lysosomal health in neurons, possibly by having an alkalinizing effect on the lysosomes, as it has in non-neuronal cells.

TFEB overexpression increases survival of doxorubicin-treated neurons and mitigates damage of the autophagy-lysosome pathway

We previously discovered that the survival of individual neurons can be predicted by the efficiency of their proteostasis system and that enhancing proteostasis capacity with Nrf2, a transcription factor that regulates expression of stress response genes, promotes neuronal longevity [45]. Therefore, we hypothesized that normalizing degradative machinery in neurons treated with doxorubicin would also be neuroprotective. Since Nrf2, an obvious choice, also strongly upregulates a number of anti-oxidant genes that are cytoprotective for cancer cells [46], we searched for another neuroprotective candidate.

TFEB, the bHLH-leucine zipper transcription factor EB, regulates lysosomal biogenesis and autophagy. Several studies showed potential involvement of TFEB in cancer, although that appears to be related to the cancer type. In non–small cell lung carcinomas, TFEB expression is associated with poor cancer prognosis, although the mechanism is not clear [47]. Conversely, GDC-0941, an inhibitor of phosphatidylinositol-3-kinase, upregulates TFEB, leading to death of glioblastoma cells [48].

We examined the survival of neurons that were treated with doxorubicin when TFEB was overexpressed. Four neuronal cohorts were transfected with mApple (a morphology and viability marker). The first and second neuronal cohorts were co-transfected with GFP; the third and fourth cohorts were co-transfected with TFEB-GFP. The first and the third cohorts were treated with a vehicle, and the second and the fourth were treated with doxorubicin. Neurons were followed longitudinally by automated microscopy, and the cumulative risk of neuronal death was calculated with longitudinal analysis (Fig. 5A,B). This technique allows us to track large cohorts of individual neurons over their lifetimes and to measure their survival with the statistical approaches used in clinical medicine [25,37,45,49]. By tracking neurons over their lifetimes, we can determine whether applied drugs or expressed proteins contribute positively, negatively or neutrally to neuronal fate. We found that the expression of TFEB promotes neuronal survival of doxorubicin-treated neurons, as expected (Fig. 5B).

2-Hydroxypropyl-β-cyclodextrin promotes survival of doxorubicin-treated neurons and mitigates neuronal damage

2-Hydroxypropyl-β-cyclodextrin (HPβCD) has been actively investigated as a treatment for Niemann-Pick type C disease, a cholesterol storage disorder characterized by the accumulation of cholesterol and glycolipids in lysosomes. HPβCD exerts its protective functions, at least in part, by activating TFEB and, therefore, enhancing the function of the lysosomes and autophagic capacity [43,55]. We therefore hypothesized that HPβCD mitigates autophagic damage in neurons associated with doxorubicin and improves their survival. In these experiments, neurons were transfected with mApple, and treated with doxorubicin alone or co-treated with doxorubicin and HPβCD. Interestingly, HPβCD indeed improved neuronal survival affected by doxorubicin, suggesting that HPβCD might be a co-therapy for treating cancer patients to reduce neuronal damage associated with the drug (Fig. 6A).

HPβCD, a cyclic molecule, may provide a hydrophobic environment and form a complex with doxorubicin, lowering its effective concentration. To determine if doxorubicin is still present in the culture media or, at least, if a potential complex of HPβCD/doxorubicin is weak, we investigated if HPβCD would prevent doxorubicin from binding to DNA. Indeed, we recently demonstrated that doxorubicin binds to nuclear DNA in cultured neurons [25]. Doxorubicin’s spectrum has an emission peak at 595 nm and a shoulder at 650 nm, and can be visualized with a regular red fluorescent protein (RFP) filter. Neurons were incubated with HPβCD/doxorubicin or with doxorubicin alone, fixed, and the red nuclear staining was analyzed (Fig. 6B). Expectedly, HPβCD did not affect doxorubicin’s binding to DNA.

Next, to confirm that HPβCD contributes to autophagy and is neuroprotective, we tested if the degradation of p62 is changed in the presence of HPβCD. Three neuronal cohorts were transfected with p62-GFP and mApple. The first cohort was treated with a vehicle; the second neuronal cohort was treated with doxorubicin; a third cohort was treated with doxorubicin and HPβCD. Indeed, co-treatment of doxorubicin and HPβCD lowered the levels of p62 in doxorubicin-treated neurons, suggesting that TFEB indeed enhanced the autophagic capacity in neurons (Fig. 6C).

Next, to confirm that HPβCD contributes to autophagy and is neuroprotective, we tested if the degradation of p62 is changed in the presence of HPβCD. Three neuronal cohorts were transfected with p62-GFP and mApple. The first cohort was treated with a vehicle; the second neuronal cohort was treated with doxorubicin; a third cohort was treated with doxorubicin and HPβCD. Indeed, co-treatment of doxorubicin and HPβCD lowered the levels of p62 in doxorubicin-treated neurons, suggesting that TFEB indeed enhanced the autophagic capacity in neurons (Fig. 6D).

Next, we investigated if the health of lysosomes can be improved by HPβCD. Neurons were pre-treated with vehicle or with doxorubicin or with a combination of doxorubicin and HPβCD and then stained with a green lysotracker dye. As expected, TFEB increased the intensity of the lysotracker dye in neurons, revealing that TFEB indeed improves the health of lysosomes in neurons treated with doxorubicin (Fig. 6E).

Would HPβCD improve the pathology that we observed by electron microscopy? Cultured neurons were pre-treated with HPβCD for several hours, and doxorubicin was added. Neurons were fixed and analyzed. Remarkably, we detected less pathology than in neurons treated with doxorubicin alone. The cytoplasm contained significantly fewer damaged organelles. No lipid droplets were detected (Fig. 6F). The neurons contained organelles that appeared to be young developing lysosomes. Given that HPβCD is in a clinical trial as a treatment for Niemann-Pick type C disease, our discovery was not unexpected, and is a promising candidate for rapid translation into the clinical setting. Would HPβCD improve the pathology and behavioral and cognitive impairments induced by doxorubicin? We plan to address these issues in a future manuscript.

Chemicals and plasmids

Doxorubicin was from Selleckchem (Houston, TX). HPβCD was from Cayman Chemical Company (Ann Arbor, MI). 10-(4′-(N-diethylamino)butyl)-2-chloro-phenoxazine hydrochloride (10-NCP) was from Calbiochem (San Diego, CA). NH4Cl was from Biomatik (Wilmington, DE). Lysotracker Green DND-26 was from Life Technologies (Carlsbad, CA). Antibodies against actin (#3700, 1:1.000) and p62 (#5114, 1:1.000) were from Cell Signaling (Danvers, MA). Antibody against tubulin (#ab7751, 1:1.000) was from Abcam (Cambridge, MA). Antibody against MAP2c (#sc-74421, 1:100) was from Santa Cruz Biotechnologies (Dallas, TX), and antibody against LC3 (#PD014, 1:1.000) was from MBL (Woburn, MA). Anti-mouse Alexa Fluor 488-labeled (#A11001, 1:200) was from Life Technologies. Hoechst dye was from Santa Cruz Biotechnology. pEGFP-N1-TFEB was from Addgene (#38119, deposited by Dr. Shawn Ferguson, Yale School of Medicine). pGW1-beclin1-GFP was described [37]. pGW1-GFP-p62 was cloned from pMXs-puro GFP-p62, which was from Addgene (#38277, deposited by Dr. Noboru Mizushima, the University of Tokyo).

Cell cultures

Cortices from rat embryos (E19) were dissected, digested with papain (Worthington Biomedical, Lakewood, NJ), resuspended in Optimem (Life Technologies) with 20 mM glucose (Biomatik, Wilmington, DE), and plated on 24-well tissue-culture plates (600.000 cells/well) coated with poly-D-lysine (EMD Millipore, Billeria, MA) as described [38,49]. Neurons were maintained in Neurobasal Medium (Life Technologies), supplemented with B-27 (Life Technologies), GlutaMAX (Thermo Scientific, Waltham, CA), and penicillin-streptomycin (Sigma, St. Louis, MO).

Immunocytochemistry

Cultured neurons were fixed with 4% paraformaldehyde (Santa Cruz Biotechnologies) for 15 min at room temperature. Cells were washed with PBS and permeabilized with 0.1% Triton X-100 (Santa Cruz Biotechnologies) in PBS for 10 min. Cells were washed before blocking overnight in PBS with 10% serum from the host species of a secondary antibody. Neurons were incubated overnight with antibodies against MAP2c (1:100) diluted in blocking buffer at 4ºC. Cells were washed three times and incubated with anti-mouse-Alexa488 (Life Technologies) diluted in blocking buffer for 1 h at room temperature. Nuclei were stained with the Hoechst dye in PBS.

Fluorescence microscopy and image analysis

Fixed neurons were imaged in automated fashion with the EVOS microscopy system (Life Technologies). The plate was placed on the EVOS microscope stage. The microscope automatically positions the 20x objective to the center of the first well of the 24-well tissue plate and collects fluorescence images with the RFP filter (mApple), the GFP filter (Beclin1; p62; lyso-tracker; TFEB), and the DAPI filter (Hoechst). Thereafter, the microscope moves the stage to each adjacent field in the well until a representative image is obtained from each well.

Puncta indexes reflect the redistribution of standard deviation of the fluorescence intensities of beclin1-GFP and GFP-p62 within soma and neurites before and after treatment [38]. Low puncta index represents diffuse localization, and high puncta index corresponds to punctate localization.

Neurons treated with vehicle or Dox were incubated with 400 nM Lysotracker Green for 10 min, imaged, and the green fluorescence intensity was quantified. Then cells were fixed and immunostained against MAP2c and stained with DAPI dye, as described above, to visualize morphology and nuclei of neurons.

Longitudinal fluorescent microscopy and survival analysis

To determine the role of TFEB in cell survival, primary cortical neurons were transfected with pGW1-mApple, as a morphology and viability marker, and pGW1-TFEB-EGFP. To elucidate the effect of HPβCD in cortical primary neurons, primary cortical neurons were transfected with pGW1-mApple, treated with a vehicle or with 50 nM Dox alone or in combination with different concentrations of HPβCD, and then imaged every 24 h for 6 days with the EVOS microscope system. For tracking the same neurons over time, an image of the fiduciary field with neurons on the plate was collected at the first time-point and used as a reference image. Each time the same plate was imaged thereafter, the fiduciary image was aligned with the reference image. Neurons that died during the imaging interval were assigned a survival time. These event times were used to obtain the exponential cumulative survival graphs and analyzed for statistical significance by Log-Rank test with JMP software (SAS Institute) as described [38,49]. One hundred to 150 neurons were analyzed from three independent experiments per condition.

Western blotting

Cultured neurons were collected in RIPA lysis buffer [150 mM NaCl, 1% Nonidet P40, 0.5% sodium deoxycholate, 0.1% SDS and 50 mM Tris/HCl (pH 8.0)] with a mix of protease and phosphatase inhibitors. Cellular suspensions were vortexed and centrifuged at 12.000 g for 10 min at 4ºC. Supernatants were collected, and protein concentration was determined using the BCA (bicinchoninic acid) Protein Assay Kit (Thermo Scientific), according to the manufacturer's instructions. Samples were analyzed by SDS/PAGE (4–12% and 16% gels, and proteins were transferred onto PVDF membranes by the iBlot2 system (Life Technologies). Membranes were blocked and incubated overnight with anti-tubulin, anti-actin, anti-p62 or anti-LC3. Membranes were washed with TBS [Tris-buffered saline; 10 mM Tris/HCl and 150 mM NaCl (pH 7.4)] and probed for 1 h with anti-rabbit or anti-mouse antibodies conjugated with horseradish peroxidase. Signals were detected using Lumi-Light Western Blotting Substrate (Roche Applied Science, Indianapolis, IN) on Medical X-Ray Film (Kodak, Rochester, NJ).

Mice

Female nude mice (Taconic Farms, Hudson, NY) were kept under pathogen-free conditions in facilities approved by the American Association for Laboratory Animal Care. Mice used in the experiments were 4–6 months of age. Mice were intraperitoneally injected with Doxil (Janssen Products, Beerse, Belgium; #NDC 59676-0960-01). Mice were injected weekly (4 mg/kg/week in 150 μl of sterile phosphate-buffered saline (PBS)) for 8 weeks. These conditions recapitulate a dose schedule used in human patients [62]. At 24 h after the last injection, two mice were anesthetized with isoflourane. The chest cavity was opened, and 20 ml of PBS were perfused into the left ventricle of the heart to purge the animal of blood. The animals were then perfused with 30 ml of 2.0% EM grade paraformaldehyde + 2.5% EM grade glutaraldehyde + 0.13 M sucrose in 0.1 M cacodylate buffer, pH 7.4.

Electron microscopy

The frontal cortices were dissected from the fixed mouse brains, minced into 1-mm cubes, and fixed for additional 2 days at 4° C. After primary fixation, the tissues were washed in 0.1 M cacodylate buffer and post-fixed with 1% cacodylate-buffered osmium tetroxide and 0.8% potassium ferricyanide for 90 minutes. The samples were washed several times in water, then dehydrated in increasing concentrations of ethanol, infiltrated, and embedded in Spurr’s Low Viscosity embedding medium. The samples were polymerized in a 62ºC oven for 3 days. Ultrathin sections were cut at 60 nm on a Leica U7 ultramicrotome (Leica, Deerfield, IL), stained with saturated aqueous uranyl acetate and counterstained with Reynold’s lead citrate. The sections were examined in a Hitachi H7500 transmission electron microscope using an accelerating voltage of 80 kV, and images were captured via an AMT XR-16 digital camera and AMT Image Capture, v602.600.51 software.

Ethics statement

Rats and mice were maintained in accordance with guidelines and regulations of the University of Texas, Houston, and the MD Anderson Cancer Center Institutional Animal Care and Use Committee (the protocol numbers #AWC-13-122, #AWC-16-0081, and #00001029-RN01). All experimental protocols were approved by the University of Texas, Houston, and by MD Anderson Cancer Center. The methods were carried out in accordance with the approved guidelines.