Cytoplasmic retention of polyglutamine-expanded androgen receptor ameliorates disease via autophagy in a mouse model of spinal and bulbar muscular atrophy

ARdNLS112Q transgenic mice express greater levels of AR than AR112Q mice

In order to understand the role of the nucleus in disease, we created transgenic mice bearing an AR with a mutation in the NLS. We previously created a transgenic mouse model of SBMA expressing full-length (human) polyQ-expanded AR (112Q) driven by the prion protein promoter (PrP) (22). The new transgenic mice (ARdNLS112Q and ARdNLS24Q) were created to express transgenic AR with a deletion of amino acids 628–640 within the bipartite NLS of the AR. A line of ARdNLS112Q was established which expresses over 2-fold more AR protein, in brain and spinal cord, than AR112Q mice (Fig. 1). In addition, ARdNLS24Q mice expressed equivalent levels of AR as ARdNLS112Q mice (Fig. 1), and both lines were used for behavioral analysis.

Motor deficits associated with SBMA are ameliorated in ARdNLS112Q male mice

We previously determined that the rotarod assay is a sensitive measure of motor function in SBMA mice (22). Therefore, we measured latency to fall from an accelerating rotarod, as well as grip strength, every 4 weeks beginning at 8 weeks of age. This behavioral analysis of a large age-matched cohort of ARdNLS112Q transgenic males (n = 18) and AR112Q males (n = 10) revealed delayed onset and reduction of motor deficits associated with SBMA, when compared with non-transgenic littermates (n = 18). While AR112Q males showed significant and progressive deficits in maintaining themselves on an accelerating rotarod at 8, 12 and 16 weeks of age, ARdNLS112Q males performed as well as non-transgenics (Fig. 2A). At 20, 24 and 28 weeks of age ARdNLS112Q males had significantly reduced rotarod function compared with non-transgenics, but still performed substantially better than AR112Q males (Fig. 2A). Female ARdNLS112Q (n = 10) performed as well as non-transgenic littermates (n = 15) until 24 weeks of age, when they had a slight reduction in rotarod performance (Fig. 2B). ARdNLS24Q (n = 15) males did not develop any rotarod deficits up to 28 weeks of age (Supplementary Material, Fig. S1).

Both forepaw and all paw grip strength was significantly reduced in AR112Q male mice beginning at 12 weeks of age, while both measures of grip strength of ARdNLS112Q mice were similar to non-transgenic males (Fig. 2C). At 16, 20 and 24 weeks of age (data not shown), grip strength results resembled those shown at 12 weeks of age. At 28 weeks of age, results of grip strength reflected those of rotarod analysis in that AR112Q males had significantly reduced grip strength compared with both ARdNLS112Q and non-transgenic males and ARdNLS112Q males were somewhat weaker than non-transgenic males. Female ARdNLS112Q mice showed grip strength similar to non-transgenic females up to 28 weeks of age (Fig. 2C).

AR112Q male mice fail to gain weight after 6 months of age (22). Analysis of the present cohort showed a failure of AR112Q male mice to gain weight after 28 weeks of age, while ARdNLS112Q males continued to gain weight in the same manner as non-transgenic male littermates (Supplementary Material, Fig. S2A). Female ARdNLS112Q mice had slightly greater weight gain over time compared with non-transgenic littermates (Supplementary Material, Fig. S2B). ARdNLS24Q male mice also showed no decrease in weight gain (Supplementary Material, Fig. S2C).

Additional tests of motor function, including footprint and rearing analysis, revealed similar results as rotarod and grip strength analysis (data not shown). From 8 to 16 weeks of age, ARdNLS112Q male mice demonstrated a normal gait, while AR112Q males exhibited a wider and shorter gait. From 20 to 28 weeks of age, ARdNLS112Q males had a slightly lower and wider stance compared with nTG males, but their gait was substantially better than that of AR112Q males. Analysis of vertical activity (during a 5-min period) was performed using a Versamax activity monitor (AccuScan Instruments, Columbus, OH). At ages when ARdNLS112Q males performed as well as nTG males on the rotarod, their vertical activity was also normal, while AR112Q males showed significant deficits. After 20 weeks of age, ARdNLS112Q males had decreased vertical activity compared with nTG males but increased vertical activity compared with AR112Q males. Female AR112Q mice had normal gait and vertical activity at all ages evaluated. As previously described, survival of the AR112Q mice was not substantially compromised; survival of ARdNLS112Q mice was also unchanged (data not shown).

ARdNLS112Q has delayed nuclear accumulation and forms oligomers later than AR112Q

Neuropathological analysis of AR112Q, ARdNLS112Q and non-transgenic male mice was performed at 8, 16 and 24 weeks of age. In AR112Q males at 8 weeks of age, AR112Q protein was localized primarily within nuclei of spinal motor neurons (Fig. 3A and Supplementary Material, Fig. S4) and these males had significant deficits in motor function (Fig. 2A). In contrast, at this age in ARdNLS112Q mice, transgenic AR protein was observed primarily within the cytoplasmic compartment, as observed both by immuofluorescence and fractionation (Fig. 3A and Supplementary Material, Fig. S4) and males had normal motor function (Fig. 2A). In addition, western blot analysis of brain and spinal cord revealed substantially more SDS-insoluble, high-molecular weight oligomeric species of AR112Q than ARdNLS112Q at this age (Fig. 3B). At 16 weeks of age oligomeric species of AR112Q were increased (Fig. 3B), although inclusions were not detected (Fig. 3A). Sixteen week-old male ARdNLS112Q revealed accumulated mutant AR within nuclei (Fig. 3A) and a small proportion of neurons contained small punctate intranuclear inclusions (Fig. 3A). Western analysis revealed oligomeric forms of ARdNLS112Q, although these were substantially less abundant than those from AR112Q mice (Fig. 3B), despite the abundance of AR protein. By 24 weeks of age, AR112Q male mice had a considerable proportion of neurons in the anterior horn of the spinal cord with large intranuclear inclusions (Fig. 3A); as previously shown, inclusions consisted of proteolyzed AR (data not shown) (22). At this age, ARdNLS112Q males also had a significant number of neurons with large intranuclear inclusions of mutant AR (Fig. 3A); intranuclear inclusions of ARdNLS112Q were also composed of fragmented AR, as they lacked the epitopes for antibodies AR441 and ARC-19 (data not shown). Similar results were observed in cortical neurons from these animals (Supplementary Material, Fig. S3A and B). In females, ARdNLS112Q was found largely within the cytoplasm at all ages, although by 37 weeks, a small number of neurons with large nuclear inclusions were observed (Fig. 3A). In males, ARdNLS24Q had also accumulated within neuronal nuclei by 37 weeks of age, but did not form intranuclear inclusions (Supplementary Material, Fig. S5). Inclusions of ARdNLS112Q were confirmed by confocal microscopy to be contained within nuclei (data not shown). In addition, our previous studies revealed decreased immunoreactivity of unphosphorylated neurofilament heavy chain (NF-H) in soma of both spinal motor neurons and neurons of the cerebral cortex (22). This alteration was absent from neurons of ARdNLS112Q mice (data not shown).

Polyglutamine-expanded ARdNLS fails to cause nuclear aggregation or toxicity in a cell model of SBMA

We previously created an inducible cell model of SBMA that expresses full-length human AR with 112 glutamines and forms intranuclear inclusions in response to DHT (27). Notably, as in patients’ tissue, nuclear inclusions in this model consist of proteolyzed N-terminal fragments of AR. In the present studies, we established a cell line expressing ARdNLS78Q to evaluate the metabolism of cytoplasmically retained polyQ-expanded AR in culture. In contrast to AR112Q-expressing cells, ARdNLS78Q-expressing cells showed a diffuse cytoplasmic distribution of AR in the presence of hormone (DHT), and failed to form intranuclear inclusions (Fig. 4A). Over time, these cells formed large cytoplasmic aggregates of full-length AR [detected with antibodies to the N-terminus (AR(N-20))], an internal epitope (AR441) (Fig. 4A) and the C-terminus [AR(C-19)] (data not shown). ARdNLS10Q cells also contained cytoplasmic AR in the presence of DHT and never formed nuclear or cytoplasmic aggregates (data not shown). To confirm that mutant ARdNLS78Q is capable of forming nuclear inclusions, we targeted ARdNLS to the nucleus with an exogenous NLS (NLSX3-ARdNLS63Q). This resulted in the hormone-dependent formation of nuclear inclusions of full-length AR (Fig. 4B and data not shown). In these cell lines, AR was expressed at comparable levels and was stabilized by DHT (Fig. 4C).

Hormone treatment of AR112Q-expressing cells resulted in toxicity (Fig. 4D). However, polyQ-expanded ARdNLS (ARdNLS78Q)-expressing cells (Fig. 4D) failed to die in response to hormone, indicating that nuclear localization is necessary, not only for AR aggregation, but for cell death as well. Targeting polyQ-expanded ARdNLS to the nucleus (NLSX3-ARdNLS63Q) resulted in DHT-dependent death (Fig. 4D), confirming that the deletion in the NLS does not affect the capacity of the polyQ-expanded AR to confer toxicity when localized to the nucleus. Thus, the possibility that deletion of the NLS alters AR conformation and prevents toxicity for reasons unrelated to its localization is unlikely, due to our finding that nuclear targeting confers on the mutant AR protein the capability of forming inclusions and causing toxicity.

Nuclear localization of polyglutamine-expanded AR is insufficient for aggregation and toxicity

Our results indicate a requirement for nuclear localization in both the nuclear aggregation and toxicity of polyQ-expanded AR. We next sought to determine whether nuclear localization is sufficient for these events. To accomplish this, we created PC12 inducible cell lines that express an AR targeted to the nucleus in the absence of hormone (NLSX3-AR). In the absence of DHT, NLSX3-AR76Q was localized within nuclei, while AR112Q was diffusely distributed within cytoplasm and nuclei (Fig. 5). No intranuclear inclusions of NLSX3-AR76Q were observed in the absence of DHT; inclusions consisting of N-terminal AR fragments were formed exclusively in response to DHT (Fig. 5). These data indicate that nuclear localization of the polyQ-expanded AR is insufficient for nuclear aggregation. Moreover, nuclear localization is insufficient for toxicity, as NLSX3-AR76Q PC12 cells failed to die in the absence of DHT, and only did so in the presence of DHT (Fig. 4C). It was also noted that NLSX3AR, containing an even shorter polyQ-tract (76Q) than non-NLS-tagged AR (112Q), induced a greater level of toxicity, despite equivalent protein levels (Fig. 4C), consistent with a role for the nucleus in mediating toxicity. NLSX3-AR35Q cells showed no aggregation or toxicity in response to DHT treatment (data not shown).

Primary motor neurons from ARdNLS112Q mice are protected from DHT-dependent toxicity by autophagy

In order to evaluate SBMA motor neuron toxicity in response to DHT, we initiated spinal cord cultures from non-transgenic, AR112Q and ARdNLS112Q transgenic mice. Monomeric levels of both AR112Q and ARdNLS112Q were increased (stabilized) in the presence of DHT; in addition, ARdNLS112Q was expressed at significantly higher levels than AR112Q (Fig. 6A). While DHT caused the loss of ∼40% of AR112Q-expressing motor neurons, it failed to cause the death of ARdNLS112Q-expressing motor neurons (Fig. 6B).

We next sought to determine the mechanism by which ARdNLS112Q motor neurons resist DHT-dependent death. Immunofluorescence staining revealed the presence of cytoplasmic puncta consisting of mutant AR in ARdNLS112Q motor neurons (Fig. 6C). With the knowledge that ARdNLS112Q enters the nucleus with reduced efficiency in the presence of hormone (Figs 3A and 4A), and that it forms cytoplasmic inclusions (Figs 3A and 6C), we considered autophagy to be a likely candidate. It is well established that activation of autophagy is neuroprotective in misfolded protein diseases (reviewed in 32). Therefore, we evaluated the essential autophagy marker LC3B (33) in primary motor neurons. Immunofluorescence analysis of LC3B in ARdNLS112Q motor neurons revealed punctate cytoplasmic staining of LC3B following treatment with DHT (Fig. 7A), indicating the activation of autophagy in these neurons. In addition, LC3B puncta were found to co-localize with AR (Fig. 7A). Punctate staining of LC3B was not detected in nTG or AR112Q motor neurons following DHT treatment (data not shown).

Given the suggestion that autophagy was activated in ARdNLS112Q motor neurons, we determined the role of autophagy in the resistance of ARdNLS112Q motor neurons to DHT-dependent death. To this end, we treated spinal cord cultures with 3-methyladenine (3-MA), a well-known inhibitor of autophagy (34). 3-MA failed to cause toxicity of non-transgenic motor neurons (data not shown); moreover, it did not enhance DHT-dependent toxicity of AR112Q-expressing motor neurons (Fig. 6B). In contrast, 3-MA induced DHT-dependent death of ARdNLS112Q motor neurons (Fig. 6B). Biochemical analysis of protein extracts from these cultures showed a large increase in the monomeric form of ARdNLS112Q following treatment with DHT and 3-MA (Fig. 7B), well above the stabilization of AR seen with DHT alone. In addition, the active form of LC3B (LC3B II) was decreased in the presence of 3-MA (Fig. 7B), validating the inhibitory effects of 3-MA on autophagy.

DHT-dependent death of motor neurons from AR112Q mice is prevented by activation of autophagy

The observation that endogenous autophagy can protect motor neurons from DHT-dependent death when polyQ-expanded AR is retained within the cytoplasm (ARdNLS112Q) confirms the importance of this degradation pathway in clearing misfolded cytoplasmic proteins. We next sought to determine whether pharmacologic activation of autophagy could rescue nuclear polyQ-expanded AR (AR112Q)-expressing motor neurons from DHT-dependent death. We used an AKT inhibitor (AKTi) to activate autophagy in spinal cord cultures from our SBMA (AR112Q) mice. Previous studies demonstrated the ability of the AKT inhibitor, phenoxazine, to activate autophagy in primary neurons expressing mutant huntingtin (Tsvetkov and Finkbeiner, unpublished results). Treatment of AR112Q motor neurons with AKTi for the last 3 days of a 7-day DHT treatment resulted in a substantial rescue from DHT-dependent death (Fig. 8A). As expected, AKTi-treated motor neurons contained cytoplasmic puncta of LC3B (Fig. 8B). Moreover, western analysis of AKTi-treated cultures revealed a significant increase in the active form of LC3B (LC3B II) (Fig. 8C). Non-transgenic motor neuron cultures also showed an increase in LC3B II following treatment with AKTi (data not shown). To confirm these findings, we evaluated another activator of autophagy, trehalose, which was previously shown to activate mTOR-independent autophagy (35) and relieve the neurotoxicity of polyQ-expanded huntingtin (36,37). Treatment of AR112Q spinal cord cultures with trehalose resulted in the formation of LC3B-positive cytoplasmic puncta (Fig. 8D), an increase in LC3B II (Fig. 8E) and rescue from DHT-dependent death (Fig. 8F). Non-transgenic cultures also showed increased LC3B II levels following trehalose treatment (data not shown). No effect on monomeric levels of AR112Q by either autophagy-inducing regimen was observed (Fig. 8C and E).

ARdNLS and NLSX3-AR inducible PC12 cell lines

Site-directed mutagenesis (Quick Change II XL, Stratagene) was performed on a pTRE plasmid (Clontech, Mountain View, CA), previously engineered to contain full-length human AR cDNA, bearing 10 or 112 CAGs, to delete the nucleic acids encoding amino acids 628–640 of the AR (within the NLS) (ARdNLS). Mutation and CAG repeat length were confirmed by sequence analysis.

NLSX3-AR was created as follows: The SV40 NLS in triplicate (NLSX3) was PCR-amplified from pShooter™ pEF/myc/nuc (Invitrogen, Carlsbad, CA) vector, and an EcoRI restriction digest site engineered on both the 5′ and 3′ ends. An NheI restriction was also engineered just upstream of the EcoRI site at the 5′ end. The PCR product was cloned into plasmid pCMVAR (16-CAG)ΔHA (9) (EcoRI site is just 5′ of the CTG start of the AR cDNA). The pCMV-NLSX3-AR (16-CAG)ΔHA was then digested with NheI and NarI, and pTRE-AR (112-CAG) was linearized with NheI and partially digested with NarI. The NLSX3-AR fragment from pCMVAR(16-CAG)ΔHA was ligated to pTRE-AR(112-CAG) (containing full length AR), resulting in pTRE-NLSX3-AR(112-CAG). ARdNLS was then cloned into this construct using NruI and BstBI. All constructs were sequenced to verify mutation and CAG length.

Stable transfections of Tet-On PC12 cells (Clontech) were performed using LipofectAMINE Plus (Invitrogen) with a plasmid conferring hygromycin resistance (pTK-hygromycin). Stable transformants were selected with 200 µg/ml hygromycin. Single colonies were isolated and expanded and screened for doxycycline-inducible AR protein expression by slot blot and western blot analysis using AR(N-20) antibody (Santa Cruz, Santa Cruz, CA). AR expression levels were adjusted with various doxycycline concentrations to achieve protein levels equivalent to AR112Q PC12 cells. Genomic DNA was extracted from each clone to verify mutation and CAG length via sequence analysis. Cells were maintained in normal growth media [Dulbecco’s modified Eagle’s medium with 10% heat-inactivated horse serum, 5% fetal bovine serum, 2 mm L-glutamine, 100 units/ml penicillin/streptomycin, 200 µg/ml hygromycin (Invitrogen) and 100 µg/ml G418 (Mediatech, Manassas, VA)] at 37°C, 10% CO₂.

Treatment of inducible PC12 cell lines

Stable Tet-On PC12 cell lines were treated with doxycycline to express AR for various times and with various concentrations of DHT in charcoal-stripped serum-containing cell-culture media.

PC12 cell toxicity assay

Stable Tet-On PC12 cell lines (AR10Q, AR112Q, ARdNLS78Q, NLSX3-AR76Q and NLSX3-ARdNLS63Q) were treated with doxycycline to express equivalent levels of AR, in the absence and presence of 10 nM DHT for 12 days. At the end of treatment, cells were harvested and stained with trypan blue. Two hundred cells were counted and the percentage of trypan blue-positive cells determined. Significance was determined with Student’s t-test.

ARdNLS PC12 cell and transgenic mouse constructs

The human AR gene, bearing either 24 CAGs (normal) or 112 CAGs (expanded), was previously cloned into the prion protein promoter (PrP) construct deleted of coding sequences (22). The deleted portion of the nuclear localization sequence of the AR (deleted of nucleic acids encoding amino acids 628–640) was cloned into the NruI and BstBI sites of both the 24 CAG and 112 CAG containing PrP-AR constructs. DNAs were linearized and the plasmid backbone (pBS) removed by digestion with NotI, gel-purified and injected into fertilized oocytes (C57Bl/6), by the Kimmel Cancer Center Transgenic Facility at Thomas Jefferson University. Both ARdNLS and AR112Q mice were maintained on a C57Bl/6 background (Charles River, Wilmington, MA). Founders were screened by genotyping tail clips and brain and spinal cord lysates from 5-week-old male mice were analyzed for ARdNLS protein expression compared with those from age-matched AR112Q SBMA male mice. Additionally, CAG repeat length was determined by sequence analysis of PCR products (Laragene, Inc., Los Angeles, CA).

Genotyping mice

DNA from mice was prepared from tail or ear biopsies using Red Extract-N-Amp Kit (Sigma). Transgenic animals were identified by PCR of the human AR: forward primer from the PrP promoter region (5′-ACTGAACCATTTCAACCGAGC-3′) coupled with a reverse primer from the AR sequence 5′ to the CAG repeat (5′AGGTGCTGCGCTCGCGGCCTCT-3′).

Western blot analysis

Freshly dissected tissue was flash-frozen in liquid nitrogen. Frozen sections were pulverized in a mortar and pestle on dry ice and homogenized in either 10 volumes of Triton-DOC buffer (1% sodium deoxycholate and 0.5% Triton X-100 in PBS with protease inhibitors) or RIPA buffer (50 mM Tris–HCl, pH 8.0, 0.15 M NaCl, 0.1% Nonidet P-40, 0.5% sodium deoxycholate, 0.1% SDS and protease inhibitors). PC12 cells and cells from primary spinal cord cultures were lysed with Triton-DOC buffer. All lysates were sonicated three times for 10 s using a Branson cup sonifier. A portion of tissue lysates in RIPA was centrifuged at 15 000g for 5 min at 4°C for detection of oligomeric species of AR (38). A DC protein assay (BioRad, Hercules, CA) was performed to determine protein concentration and lysates were electrophoresed by SDS–PAGE and transferred to 0.45 µm PVDF (Immobilon-P). Western hybridization was performed using the following antibodies: AR(N-20), GAPDH (1:1000) (Santa Cruz Biotechnology) and LC3B (1:500) (NB600-1384) (Novus Biologicals, Littleton, CO). Detection was performed with ECL (Amersham Biosciences, Arlington Heights, IL).

Behavioral analysis

Every 4 weeks, beginning at 8 weeks of age, an age-matched cohort of ARdNLS24Q males, AR112Q males, ARdNLS112Q males, ARdNLS112Q females, non-transgenic males and non-transgenic females was subject to various measures of motor function. Mice were tested during the light phase of a 12 h light/dark cycle for their latency to fall off a steadily accelerating rotarod (4–40 rpm over 10 min) (Ugo Basile, Comerio, VA-Italy). During the first week of testing, mice were tested four times per day for 3 consecutive days. The first 2 days constituted a learning period and data collected on the third day were used for analysis. For subsequent testing sessions, mice were only subjected to rotarod for 1 day, as statistical analysis revealed no loss of statistical power under this regimen (unpublished data). Mice were allowed a rest period of at least 15 min between testing sessions. Scores were analyzed by two-way repeated measures ANOVA with a Tukey post hoc analysis using SigmaStat 3.0 software (SPSS Inc., Chicago, IL). A grip strength meter (Columbus Instruments, Columbus, OH) was used to measure the force exerted by a mouse as it was pulled across a grid by its tail. Grip strength was measured for forepaws only or hindpaws and forepaws together. Six measures were taken for both measures of grip strength and the lowest and highest scores for each animal dropped. An average for each animal was used for statistical analysis. Significance was determined with two-way repeated measures ANOVA and a Tukey post hoc analysis (SigmaStat).

Primary motor neuron cultures

Dissociated spinal cord cultures were established according to Roy et al. (68). In brief, spinal cords were dissected on ice from 13.5-day-old embryos in dissection media (0.1% dextrose, 2% sucrose, 1.4 mm NaCl, 5.4 mm KCl, 0.17 mm Na₂HPO₄, 22 µM KH₂PO₄, 9.9 mm HEPES). Genotyping was performed from a tail biopsy. Transgenic AR112Q, ARdNLS112Q or non-transgenic spinal cords were pooled separately, plated for culture and incubated for 3 weeks in media conditioned by glial culture from 13.5-day-old non-transgenic brain [MEM, 35 mm NaHCO₃, 0.5% dextrose, 1% N3, 10 nM 2.5 S NGF (added after conditioning)]. During the development of this culture system, motor neurons were identified using antibodies to choline acetyltransferase, neuron-specific enolase and neurofilament heavy chain (SMI32). Motor neurons were identified to have much larger cell bodies relative to other spinal neurons and large, tapering, highly branched dendrites with a fibrillar appearance. In our experiments presented here, SMI32 immunoreactivity and morphology were used to identify motor neurons. Three weeks after initiation, cultures were treated with or without 10 µM DHT for 7 days. Additional reagents/drugs were administered for the last 3 days of the 7 day treatment period [5 mm 3-Methyladenine (3-MA), 100 mm Trehalose (Sigma), 2.5 µM AKT inhibitor X [10-(4′-N-diethylamino)butyl)-2-chlorophenoxazine] (AKTi) (Calbiochem, San Diego, CA)]. Three wells of each motor neuron culture line and treatment group were immunostained as described in what follows. Motor neurons were determined by SMI32 stain and morphology and counted. Significance was determined by a Student’s t-test.

Immunofluorescence

Dissected fresh whole brain and spinal cord were frozen in OCT, then sectioned with a cryostat (7 µm). Tissue sections, motor neuron cultures and PC12 cells were fixed with 4% paraformaldehyde for 10 min, washed in PBS, permeabilized with 0.3% Triton X-100 for 15 min (cells only), blocked in 1.5% goat serum (Jackson ImmunoResearch, West Grove, PA) for 20 min and incubated for 60 min in primary antibody diluted in 1.5% goat serum. Tissue sections or cells were washed in PBS and incubated for 30 min with secondary antibodies (FITC- or Texas Red-conjugated) (Jackson ImmunoResearch, West Grove, PA), washed in PBS, incubated for 10 min with Hoechst (2 µg/ml), washed in PBS and mounted in Vectashield (Vector Laboratories, Burlingame, CA). Fluorescence was visualized with a Leica (Leica Microsystems GmbH, Wetzlar, Germany) microscope; images were captured with a Leica camera and compiled with IP Lab software (BD Biosciences, Rockville, MD). Antibodies used include AR(N-20), ARH280, AR441 (1:100) (Santa Cruz), AR-318 (Vector Laboratories Burlingame, CA), SMI32 (1:1,000) (Sternberger Monoclonal, Baltimore, MD) and LC3B (NB600-1384) (1:200) (Novus Biologicals).