Guanabenz promotes neuronal survival via enhancement of ATF4 and parkin expression in models of Parkinson disease

Abstract

Reduced function of parkin appears to be a central pathogenic event in Parkinson disease (PD). Increasing parkin levels enhances survival in models of PD-related neuronal death and is a promising therapeutic objective. Previously, we demonstrated that the transcription factor ATF4 promotes survival in response to PD-mimetic stressors by maintaining parkin levels. ATF4 translation is up-regulated by phosphorylation of the translation initiation factor eIF2α. The small molecule guanabenz enhances eIF2α phosphorylation by blocking the function of GADD34, a regulatory protein that promotes eIF2α dephosphorylation. We tested the hypothesis that guanabenz, by inhibiting GADD34 and consequently increasing eIF2α phosphorylation and elevating ATF4, would improve survival in models of PD by up-regulating parkin.

We found that GADD34 is strongly induced by 6-OHDA, and that GADD34 localization is dramatically altered in dopaminergic substantia nigra neurons in PD cases. We further demonstrated that guanabenz attenuates 6-hydroxydopamine (6-OHDA) induced cell death of differentiated PC12 cells and primary ventral midbrain dopaminergic neurons in culture, and of dopaminergic neurons in the substantia nigra of mice. In culture models, guanabenz also increases eIF2α phosphorylation and ATF4 and parkin levels in response to 6-OHDA. Furthermore, if either ATF4 or parkin is silenced, then the protective effect of guanabenz is lost. We also found similar results in a distinct model of neuronal death: primary cultures of cortical neurons treated with the topoisomerase I inhibitor camptothecin, in which guanabenz limited camptothecin-induced neuronal death in an ATF4- and parkin-dependent manner. In summary, our data suggest that guanabenz and other GADD34 inhibitors could be used as therapeutic agents to boost parkin levels and thereby slow neurodegeneration in PD and other neurodegenerative conditions.

Introduction

Parkinson disease (PD) is a common and debilitating neurodegenerative disorder, with no effective treatments to slow its progression. Reduced function of parkin has emerged as a potential pathogenic mechanism in the development of PD. The Parkin gene was originally identified in families with an autosomal recessive, early-onset form of PD (Kitada et al., 1998). Parkin encodes an ubiquitin E3 ligase, and disease-linked mutations lead to loss of function (Henn et al., 2005). While Parkin mutations are relatively uncommon, parkin function is reduced in sporadic PD as well, via several mechanisms, including nitrosylation, oxidation, phosphorylation and aggregation (T. M. Dawson and V. L. Dawson, 2013). Furthermore, parkin plays a protective role in multiple neuronal death paradigms. Specifically, over-expression of parkin improves neuronal survival (Petrucelli et al., 2002; D. B. Wang et al., 2013; Yasuda et al., 2011), while reducing parkin levels favors cell death (Yang et al., 2007). Therefore, pathways that up-regulate parkin levels would favor neuronal survival, and could serve as targets for therapeutic intervention in PD.

Previously, our group (Sun et al., 2013) and others (Bouman et al., 2011) have identified the basic helix-loop-helix transcription factor ATF4 (activating transcription factor 4, or CREB2) as a positive regulator of parkin. ATF4 is up-regulated in response to several stressors, and can either promote or reduce neuronal survival, depending upon the context (Baleriola et al., 2014; Galehdar et al., 2010; Lange et al., 2008; Lewerenz et al., 2012; Wu et al., 2014). In cellular models, ATF4 contributes to neuronal survival in response to either 6-hydroxydopamine (6-OHDA) or 1-methyl-4-phenylpyridinium (MPP+), two toxins that model features of PD-related neurodegeneration. Importantly, ATF4 counteracts the toxin-induced loss of parkin protein in these cellular models. Furthermore, parkin is required for ATF4-mediated neuroprotection. Taken together, these findings suggest that, in PD-relevant models, ATF4 attenuates neuronal death by increasing parkin levels. Therefore, interventions that elevate ATF4 would in turn boost parkin levels and favor neuronal protection.

Guanabenz acetate (GA) was identified in a small molecule screen for suppressors of prion toxicity (Tribouillard-Tanvier et al., 2008). Subsequent studies have found that GA is protective in models of neurodegeneration based on mutant TDP-43 and mutant SOD1 over-expression (Vaccaro et al., 2013; L. Wang et al., 2014b). Guanabenz was originally developed as an α2 adrenergic agonist; however, its anti-prion effect is not related to α2 activity (Tribouillard-Tanvier et al., 2008). A subsequent study described an alternative activity for guanabenz: enhancement of eIF2α phosphorylation, via blockade of GADD34, a stress-induced regulator of the phosphatase PP1 that dephosphorylates eIF2α (Tsaytler et al., 2011). eIF2α is a translation initiation factor that is phosphorylated on Ser51 (P-eIF2α) in response to multiple cellular stressors. Phosphorylation of eIF2α leads to a global reduction in protein synthesis; however, the translation of ATF4 mRNA is paradoxically increased, due to short upstream open reading frames (uORFs) in its 5’UTR (Vattem and Wek, 2004). In this study, we tested the hypotheses that guanabenz would enhance ATF4 expression and therefore parkin levels by promoting eIF2α phosphorylation, and by this means, would lead to neuroprotection in both in vitro and in vivo models of PD.

Materials and methods

Materials and antibodies

Stock solutions of 6-hydroxydopamine (6-OHDA; Tocris), guanabenz acetate salt, clonidine, efaroxan (Sigma) were prepared in ddH2O; camptothecin (CPT; Tocris) was prepared in DMSO. The following antibodies were used: anti-ATF4 was commercially generated for our laboratory (Liu et al., 2014; Pasini et al., 2015); anti-GADD34 was from Proteintech; anti-tyrosine hydroxylase (TH) was from EMD Millipore; anti-ERK was from Santa Cruz Biotechnology; anti-parkin (PARK8), anti-phospho-eIF2α (S51) and total eIF-2α were from Cell Signaling Technology.

Plasmids and lentivirus preparation

All shRNA constructs for infection experiments were generated in pLVTHM (Addgene). The following are the target sequences for different silencing constructs: shCTR (a mutant version of shATF4, mutated bases are underlined); 5 ′ -GCCAGATTCAGCGGCCTACAT-3 ′); shATF4: 5 ′ -GCCTGACTCTGCTGCTTATAT-3’. shParkin: 5 ′ -ATCACCTGACAGTACAGAACT-3 ′. The specificity of these constructs has been demonstrated previously (Liu et al., 2014; Romaní-Aumedes et al., 2014; Sun et al., 2013). All plasmids were sequenced to confirm the correct insert.

Lentivirus was prepared by transient transfection of 293T cells with the target plasmid (pLVTHM) and second generation packaging plasmids (pMDLg/pRRE and CMV-VSVG, Addgene). The supernatant was collected, and virus particles were concentrated by Lenti-X concentrator (Clontech), resuspended in PBS, aliquoted and stored at −80°C. For lentiviral infection, cultures were transduced at day 3–4 of differentiation at an approximate multiplicity of infection (MOI) of 5, and experiments were performed at least 3 d later. Using these conditions, the infection rate was reliably >90%.

Cell culture and viability assays

PC12 cells were cultured as described previously (Greene and Tischler, 1976). For neuronal differentiation, cells were grown in RPMI-1640 media with 1% horse serum, penicillin/streptomycin, and 50 ng/ml recombinant human nerve growth factor for 6–12 d. Media was changed every other day. Primary cortical neuron cultures from E18 rats were prepared as described previously (Friedman et al., 1993) and maintained in Neurobasal media supplemented with B-27 and glutamine.

For survival experiments, cells were pretreated with guanabenz for 4 hours, then treated with 100 µM 6-OHDA or 10 µM CPT. Cell nuclei were counted (24 hours after toxin treatment unless otherwise stated) after adding counting cell lysis buffer as described previously (Rukenstein et al., 1991). Each condition was performed in triplicate, and each experiment was performed at least three times.

Primary ventral midbrain cultures from P0–P3 rat were prepared as described previously (Michael et al., 2007). After 6–7 days in culture, cells were pretreated with 2.5 µM guanabenz for 4 hours and then treated with 50 µM 6-OHDA for 24 h. The cells were then fixed with 4% paraformaldehyde and immunostained for TH. Cell survival was measured as the total count of TH cells in each culture. Three independent experiments were performed, with each condition performed in triplicate in each experiment.

Western immunoblot and real-time PCR

PC12 cells or cortical neurons were lysed in Cell Lysis buffer (Cell Signaling Technology) with protease inhibitor (cOmplete, Roche), supplemented with 2.5% β-mercaptoethanol, and prepared for SDS-PAGE with 4 X Laemmli sample buffer and boiled for 10 min at 100°C before running. Samples were separated by SDS-PAGE, then transferred to a nitrocellulose membrane (BioRad). The membrane was blocked with 5% milk, incubated with the indicated primary antibody and the appropriate secondary antibody, visualized with ECL reagent (Pierce), and developed on film. Images were scanned and quantified using image J.

Total cellular RNA was isolated using TRI Reagent (Ambion) following the manufacturer’s protocol. cDNA was synthesized using first-strand cDNA synthesis kit (Origene) with 1 µg of total RNA. Quantitative real-time PCR was performed using FastStart SYBR Green Master Mix (Roche) and an Eppendorf Realplex Mastercyler with the program: 95°C for 15 secs, 58C for 30secs, 72°C for 30secs. The primer pairs are: ATF4 forward 5′-CCTTCGACCAGTCGGGTTTG-3’ and reverse 5′-CTGTCCCGGAAAAGGCATCC-3′; Parkin Fwd 5’-CGGATGAGTGGAGAGTGC-3’ and reverse 5’-TGGCGGTGGTTACATTGG-3’; GADD34 forward 5′-AAGGCGTGTCCATGCTCTGG-3′, and reverse 5′-GTCCATTTCCTTGCTGTCTG-3′; α-tubulin forward 5′-TACACCATTGGCAAGGAGAT-3′ and reverse 5′-GGCTGGGTAAATGGAGAACT-3′; 18S rRNA forward 5′-TTGATTAAGTCCCTGCCCTTTGT-3′, and reverse 5′-CGATCCGAGGGCCTCACTA-3′. For quantification of relative expression, values were normalized to either α-tubulin or 18S levels. Each reaction was performed in triplicate, and each condition was performed in triplicate.

Immunohistochemistry in human brain sections

Postmortem brain samples from neuropathologically confirmed PD cases (n=5) and age- and gender-matched controls (n=7) were obtained from the New York Brain Bank at Columbia University (New York, NY). Midbrain sections (6 µm) were deparaffinized in xylene and rehydrated in an ethanol series. Sections were then cooked for antigen retrieval in citrate buffer for 45 min at 100°C. Sections were then blocked in goat serum for 20 min and incubated with anti-GADD34 (Proteintech) at 1:200 in blocking buffer overnight at 4°C. To test the specificity of the antibody, some sections were incubated with the antibody that was mixed with GADD34 fusion protein (Proteintech) Sections were then washed and incubated with biotinylated anti-rabbit secondary antibody for 1 h at room temperature, washed, and incubated in ready-to use ABC complex solution (Vectastain, Vector Laboratories) at room temperature, and then SG substrate (Vector Laboratories) was added and left on the sides for 15 min. Sections were counterstained with Nuclear Fast Red, dehydrated and mounted with coverslips. For quantification of GADD34 staining pattern, neuromelanin-positive neurons were assessed for the presence of GADD34-positive granules in a blinded manner. At least 30 neurons were assessed for each case.

In vivo 6-OHDA mouse experiments

Adult (8 week) male C57BL/6 mice (Charles River Laboratories) were divided into 3 groups, which were injected intraperitoneally (IP) with vehicle (normal saline + 5% DMSO), 1 mg/kg or 4 mg/kg guanabenz for 3 days. On day 0, 6-OHDA was injected as previously described (Chen et al., 2012). Briefly 6-OHDA was injected by microliter syringe at a rate of 0.5 µL/min by pump for a total dose of 15.0 µg/3 µl. Injection was performed into the left striatum at coordinates AP: +0.09 cm; ML: +0.22 cm; DV: −0.25 cm relative to bregma. After 2 minutes, the needle was slowly removed. On the following day, GA dosing was resumed, and continued by dosing 3 times per week thereafter for another 4 weeks.

Animals were sacrificed and analyzed for SN dopaminergic neuron count and striatal dopaminergic innervation using well-established protocols that have been previously described (Chen et al., 2012). Briefly, animals were perfused intracardially 4% paraformaldehyde in 0.1M phosphate buffer, pH 7.1. Brains were carefully removed, then post-fixed for 1 week, cryoprotected for 24 hours in 20% sucrose and flash frozen. A complete set of 30 micron thick sections through the SN were cut on a cryostat, and every fourth section was processed for TH immunostaining, in keeping with the fractionator sampling method. Sections were stained free-floating using rabbit anti-TH antibody (1:750, Calbiochem), followed by biotinylated protein A and ABC, and thionin counterstaining.

Dopaminergic SN counts were determined using stereological analysis with StereoInvestigator. All counts were performed in a blinded fashion on coded slides. The SN on each side was analyzed for each animal, and the entire SN was defined as the region of interest. TH-positive neurons were identified at 100×, focusing through the entire extent of each section. The program calculated the number of TH-positive neurons in each SN.

For striatal TH staining, forebrain sections containing the striatum were processed similarly to the midbrain sections, except that sections were post-fixed for 48 hours and frozen without cryoprotection. Optical density of striatal TH immunostaining was measured using an Imaging Research Analytical Imaging Station.

All animal procedures were approved by the Columbia University Animal Care and Use Committee.

Statistical analysis

All statistical analyses were performed with GraphPad Prism. Comparisons between two experimental groups were performed with a Student’s t-test. Multiple comparisons of more than two experimental groups were performed using one-way analysis of variance (ANOVA) and Tukey’s post-hoc test, or with two-way ANOVA and Bonferroni post-hoc test.

Results

Guanabenz suppresses neuronal death caused by 6-OHDA

We first tested the effect of guanabenz on 6-OHDA-induced toxicity in a well-established cellular system: neuronally differentiated (NGF-treated) PC12 cells. After exposure to NGF, PC12 cells assume a neuronal phenotype resembling sympathetic neurons, an affected cell population in PD. We observed a ~35% reduction in survival, as determined by counts of viable nuclei, after treatment with 100 µM 6-OHDA for 24 hours (Figure 1A). We then assessed the effect of pre-treatment with various concentrations of guanabenz on 6-OHDA-induced toxicity (Figure 1A–B). Guanabenz enhanced survival of 6-OHDA-treated PC12 cells, with a minimum effective concentration of 0.5 µM. Protection was maximal at concentrations of 1–5 µM guanabenz, with rescue of about 50% of cell loss (Figure 1A–B). Interestingly, guanabenz became less effective at higher concentrations, with no significant protection observed at 10–20 µM (Figure 1A–B).

Figure 1. Guanabenz acetate (GA) causes a dose-dependent improvement in survival after 6-OHDA.

(A) Differentiated PC12 cells were pretreated with different concentrations of GA as indicated for 4 hours, then treated with 100 µM 6-OHDA for 24 hours. Viable nuclei were counted. (B) The data in (A) expressed as % protection, defined as 100 – [(% cell loss with 6-OHDA + GA) / (% cell loss with 6-OHDA alone)]. (C) Cultured ventral midbrain neurons were pretreated with 2.5 µM GA for 4 hours, treated with 50 µM 6-OHDA for 24 hours, then fixed and immunostained for TH. The number of TH+ cells on each coverslip was counted. Data are mean +/− SEM (n=3–6 for A–B, n=3 for C); in each experiment, each condition was performed in triplicate. *p<0.05, **p<0.01 by 2-way ANOVA with Bonferroni posthoc test.

To extend our findings, we examined the effect of guanabenz on 6-OHDA-induced toxicity of dopaminergic neurons cultured from mouse ventral midbrain. Guanabenz (2.5 µM) significantly reduced the 6-OHDA-induced loss of dopaminergic neurons in this model (Figure 1C).

Guanabenz-mediated protection is independent of α2-adrenergic agonist activity

To assess the possible contribution of α2 activity to the protective effect of guanabenz, we tested clonidine, a distinct α2 agonist. Unlike guanabenz, clonidine has no anti-prion activity and does not interact with GADD34 (Tribouillard-Tanvier et al., 2008; Tsaytler et al., 2011). Clonidine had no effect on PC12 cell survival in response to 6-OHDA (Figure 2A), suggesting that α2 agonism is not sufficient to promote cell protection. Next, to assess whether α2 activation is necessary for protection, we tested the effect of the α2 antagonist efaroxan on the protective effect of guanabenz. Guanabenz reduced 6-OHDA-induced PC12 cell death equally well in the presence or absence of efaroxan (Figure 2B). Taken together, these results suggest that guanabenz promotes survival in an α2-independent manner.

Figure 2. Protection by guanabenz is independent of α2-adrenergic receptor activity.

(A) Differentiated PC12 cells were pretreated with clonidine, a distinct α2 agonist, or GA as indicated for 4 hours, then treated with 100 µM 6-OHDA for 24 hours. Viable nuclei were counted. (B) Differentiated PC12 cells were pretreated with GA and the α2 antagonist efaroxan as indicated for 4 hours, then treated with 100 µM 6-OHDA for 24 hours. Viable nuclei were counted. Data are mean +/− SEM from 3 independent experiments (except n=2 for 10 µM efaroxan alone); in each experiment, each condition was performed in triplicate. ***p<0.001 (vs. all other 6-OHDA conditions for A), by 2-way ANOVA with Bonferroni post-hoc test.

GADD34 is up-regulated by 6-OHDA in vitro

Guanabenz has been reported to bind directly to the regulatory protein GADD34, preventing its interaction with the phosphatase PP1. GADD34 is typically expressed at low levels in healthy cells, but is strongly induced by multiple stressors (Kojima et al., 2003). Therefore, we first assessed GADD34 expression in our cellular PD model. GADD34 levels are very low under basal conditions in differentiated PC12 cells (Figure 3A–D). In response to 6-OHDA, there is a dramatic dose-dependent increase in both GADD34 mRNA (Figure 3A) and protein (Figure 3B). This increase in protein is evident by 4 hours (Figure 3D), well before the onset of cell death. Furthermore, this initial increase in GADD34 protein (Figure 3D) preceded the increase in GADD34 mRNA, which occurred at later time points (Figure 3C). This suggests that both non-transcriptional and transcriptional mechanisms contribute to GADD34 induction in this system.

Figure 3. GADD34 up-regulation in response to 6-OHDA.

Differentiated PC12 cells were treated with the indicated concentration of 6-OHDA for 8 hours (A, B) or with 100 µM 6-OHDA for different lengths of time (C, D). In (A, C), total RNA was isolated, cDNA samples were prepared, and GADD34 mRNA (relative to 18S rRNA levels) was assessed by quantitative PCR. Results are mean +/− SD from a representative experiment; similar experiments (differing in exact timing and doses of 6-OHDA) were performed at least 3 times. In (B, D), parallel samples were used to prepare whole cell lysates and analyzed by Western blotting with the indicated antibodies.

Altered GADD34 localization in the substantia nigra of PD patients

Next, we examined GADD34 expression in dopaminergic neurons of the substantia nigra (SN) of age-matched PD and control brains by immunohistochemistry. GADD34 immunostaining was observed in both neuromelanin-positive dopaminergic neurons and neuromelanin-negative neurons (Figure 4A). GADD34-immunopositivity was detected in neuronal cell bodies and extending into proximal neuronal processes (Figure 4A), as well as in the surrounding neuropil. A pre-absorption control using the GADD34 antibody pre-incubated with a cognate fusion protein strongly diminished staining (Figure 4B). Both control and PD cases displayed GADD34 immunostaining in neuromelanin-positive neurons, but the pattern of immunoreactivity differed (Figure 4C–D). GADD34 staining was often found in discrete granules in PD cases (arrows in Figure 4D), while this pattern was rarely observed in controls. In PD cases, 89% of neuromelanin-positive cells displayed these GADD34-positive granules, compared to 8% in controls (Figure 4E). In addition, in PD cases, we occasionally observed beaded staining that appeared to be in neuritic processes (inset in Figure 4D); these structures were not seen in controls. Finally, GADD34 immunoreactivity in the neuropil appeared diminished in PD cases compared to controls. In sum, these data suggest alterations in GADD34 activity and/or distribution in both cellular models of PD and in dopaminergic neurons of PD patients.

Figure 4. Altered GADD34 localization in the substantia nigra in PD brain tissue.

(A–D) Fixed sections of midbrain from 7 controls and 5 PD patients were assessed by immunohistochemistry using GADD34 antibody (blue-gray). Panels (A) and (B) show sections from a control brain; (A) shows the staining with GADD34 antibody, and (B) shows staining when the GADD34 antibody was pre-incubated with its cognate fusion protein. Arrowheads show brown neuromelanin pigment-positive dopaminergic neurons. Diamond shows non-dopaminergic neuron. Arrows indicate a proximal process from a dopaminergic neuron. (C, D) show representative sections from a control and PD brain, respectively. In (D), arrows indicate GADD34-positive puncta in the soma of neuromelanin-positive neurons. The inset in (D) shows punctate GADD34 immunostaining in a neuritic process. (E) The percentage of neuromelanin-positive neurons with punctate GADD34 staining, as exemplified in (D), in each case was quantified (at least 30 neurons were assessed for each case), with values from individual cases, mean (bar) and SEM plotted. ***p<0.001 by t-test. Scale bar represents 150 µm in (AB), and 33 µm in (C–D).

Guanabenz enhances eIF2α phosphorylation, ATF4 and parkin levels in 6-OHDA-treated cells

Our rationale for testing guanabenz as a potential neuroprotectant in the context of PD arose from its potential to elevate parkin levels via inhibition of GADD34 function and consequent up-regulation of P-eIF2α and ATF4. Given the elevation of GADD34 in response to 6-OHDA, we next evaluated the effect of guanabenz on the potential downstream effectors P-eIF2α, ATF4 and parkin in 6-OHDA-treated PC12 cells.

As previously described (Sun et al., 2013), 6-OHDA caused a 50–60% loss of parkin protein after 10 hours, before the onset of cell death (Figure 5A, compare first two lanes; Figure 5B). This occurred despite a small, statistically insignificant increase in parkin mRNA (Figure 5C). These findings are consistent with our prior studies showing that parkin drops after 6-OHDA treatment as a result of increased protein degradation, rather than changes in parkin mRNA expression (Sun et al., 2013). Guanabenz treatment significantly preserved parkin levels after 6-OHDA exposure (Figure 5A–B). Importantly, guanabenz enhanced parkin levels only at a protective concentration (1 µM). That is, guanabenz had no effect on parkin at either lower (Figure 5B, 0.05 µM guanabenz) or higher (Figure 5B, 10 µM guanabenz) concentrations, both of which also failed to reduce 6-OHDA-induced toxicity (Figure 1B). Parkin mRNA levels were not significantly affected by guanabenz (Figure 5C).

Figure 5. Guanabenz increases eIF2α phosphorylation, ATF4 and parkin protein levels after 6-OHDA treatment.

Differentiated PC12 cells were pretreated with GA for 4 hours, then with 100 µM 6-OHDA for 10 hours. Parallel samples were used to prepare cell lysates for Western blotting with the indicated antibodies (A, B, D, F), or total RNA for quantification of indicated messages (normalized to alpha-tubulin levels) by quantitative real-time PCR (C, E). (A) shows a representative Western blot; (B, D, F) show densitometric quantification of the indicated proteins, normalized to ERK. Levels are plotted as fold change relative to control (0 GA) for parkin, and relative to 6-OHDA (0 GA) for ATF4 and P-eIF2α, as these proteins were undetectable in control cells in some experiments. Data are mean +/− SEM from 3–6 independent experiments. **p<0.01, ***p<0.001 compared to 6-OHDA alone by ANOVA with Tukey’s post-hoc test. # P<0.05 compared to control by ANOVA with Tukey’s post-hoc test (p ~0.14 compared to 6-OHDA alone).

Next, we examined the effects of 6-OHDA and guanabenz on ATF4 protein levels and eIF2α phosphorylation. We analyzed cultures 10 hours after 6-OHDA treatment since this is before the onset of cell death, but after an initial round of stress responses including activation of the eIF2α kinase PERK and elevation of genes such as ATF4 (Holtz and O'Malley, 2003; Ryu et al., 2002). As previously reported (Ryu et al., 2002; Sun et al., 2013), ATF4 protein and mRNA levels increased substantially in response to 6-OHDA (Figure 5A, compare lanes 1 and 2; Figures 5D–E). Guanabenz caused a near doubling of ATF4 protein levels above those induced by 6-OHDA alone (Figure 5A, D), but did not alter levels of ATF4 mRNA (Figure 5E), consistent with the predicted effect of guanabenz on P-eIF2 α and ATF4 translation. At the time-point analyzed (10 hours after 6-OHDA treatment alone), phosphorylation of eIF2α fell back to baseline levels (Figure 5A, F), consistent with induction of GADD34 and dephosphorylation of eIF2α. In contrast, guanabenz elevated P-eIF2α levels in 6-OHDA-treated cells at this time compared to control (Figure 5A, F). Total eIF2α levels did not change significantly under any of the treatment conditions (Figure 5a). Finally, in line with its effect on parkin, guanabenz increased ATF4 and P-eIF2α levels only at a protective concentration (Figure 5D, F, 1 µM guanabenz), but not at lower or higher concentrations. Taken together, these data demonstrate that guanabenz increases eIF2α phosphorylation, ATF4 and parkin levels after 6-OHDA exposure. Furthermore, there is a correlation between the guanabenz concentration that leads to neuroprotection and the elevation of these proteins.

Silencing either ATF4 or parkin abolishes the protective benefit of guanabenz

The previous experiments suggest that elevation of ATF4 and parkin play an important role in the pro-survival activity of guanabenz. To more directly test this notion, we examined the effect of reducing ATF4 or parkin levels on the protective effect of guanabenz. To reduce ATF4 levels, we used lentiviral-mediated delivery of a previously validated shRNA construct that targets ATF4. As in our prior studies (Romaní-Aumedes et al., 2014; Sun et al., 2013), we observed robust knockdown of ATF4 (Figure 6A, S1). Guanabenz had no effect on 6-OHDA-induced cell death when ATF4 was silenced, suggesting that ATF4 is necessary for guanabenz-mediated protection (Figure 6A).

Figure 6. ATF4 and parkin are necessary for guanabenz-mediated neuroprotection against 6-OHDA.

(A–C) Differentiated PC12 cells were infected with lentivirus carrying shRNA against ATF4 (shATF4, in A, or Parkin (shParkin, in C), or a control shRNA (shCTRL, a mutated version of the shATF4 sequence) for 3 days. In B, cells were treated with 20 nM ISRIB (I) at the same time as GA, where indicated. Whole cell lysates were analyzed by Western blotting with the indicated antibodies; for A and B, cells were treated with 100 µM 6-OHDA to induce robust ATF4 expression. In A, duplicate wells of cells infected with shATF4 were analyzed. For survival experiments, cells were pretreated with 2.5 µM GA (+/− 20 nM ISRIB in B) for 4 hours, then treated with 100 µM 6-OHDA for 24hrs. Viable nuclei were counted. Data are mean +/− SEM from 5 independent experiments; each condition was performed in triplicate in each experiment. (D) Cultured ventral midbrain neurons were infected with lentivirus carrying shRNA against Parkin or a scrambled control for 4 days. Then cultures were pretreated with 2.5 µM GA for 4 hours, treated with 50 µM 6-OHDA for 24 hours, and fixed and immunostained for GFP and TH. The number of TH/GFP+ cells on each coverslip was counted. Data are mean +/− SEM from 4 independent experiments (except for GA alone with either shRNA construct, which was performed in 2 experiments); each condition was performed either in duplicate or triplicate. **p<0.01, ***p<0.001 by ANOVA with Tukey’s post-hoc test.

As an alternative way of attenuating ATF4 protein induction, we used the small molecule ISRIB. ISRIB directly activates eIF2B, which negates the effect of eIF2α phosphorylation and prevents up-regulation of ATF4 translation (Sidrauski et al., 2015). Indeed, co-treatment of PC12 cells with GA and ISRIB prevented the 6-OHDA-induced up-regulation of ATF4 (Figure 6B). Similar to direct knockdown of ATF4, ISRIB also blocked the protective effect of GA on 6-OHDA-treated cells (Figure 6B), further supporting the notion that ATF4 is important for the beneficial effect of GA on survival.

Next, we performed similar experiments to assess the effect of parkin knockdown. Lentiviral delivery of parkin shRNA robustly lowered parkin levels and prevented guanabenz from improving survival after 6-OHDA exposure (Figure 6C). We repeated these experiments in ventral midbrain cultures and observed similar results: guanabenz was not protective against 6-OHDA-induced toxicity if parkin is silenced (Figure 6D). These results support the idea that ATF4 and parkin are required downstream mediators of the pro-survival activity of guanabenz.

Guanabenz attenuates DNA damage-induced cell death in cortical neurons in an ATF4- and parkin-dependent manner

We also tested the effect of guanabenz in a different neurodegenerative paradigm in which loss of parkin activity is important. Treatment of primary cultures of cortical neurons, a population that develops Lewy pathology in PD (Braak et al., 2003), with the topoisomerase I inhibitor camptothecin (CPT) leads to mitochondrial damage and subsequent neuronal death. This model is relevant for our studies because CPT kills cortical neurons, at least in part, via depletion of parkin (D. B. Wang et al., 2013). In addition, DNA damage is observed in PD brain tissue (Nakabeppu et al., 2007) and may contribute to neuronal death.

CPT (10 µM) led to approximately 60% reduction in the number of surviving primary cortical neurons after 24 hours (Figure 7A). Guanabenz partially prevented this CPT-mediated loss of cortical neurons (Figure 7A–B), producing approximately 30% protection at concentrations of 0.5–2.5 µM. Higher concentrations of guanabenz did not attenuate CPT-induced cell loss (Figure 7A–B). In the 6-OHDA model, ATF4 and parkin were required for guanabenz to improve cell survival. In cortical neuron cultures, silencing ATF4 abolished the protective activity of guanabenz against CPT-induced death (Figure 7C). Similarly, guanabenz failed to improve survival in CPT-treated cortical neurons when parkin was silenced (Figure 7D).

Figure 7. Guanabenz protects cortical neurons against DNA damage in an ATF4- and parkin-dependent manner.

(A) Primary cultures of rat E18 cortical neurons were pre-treated with GA for 4 hours, then treated with 10 µM CPT for 24 hours. Survival was assessed by counting viable nuclei. (B) The data in (A) expressed as % protection, defined as 100 – [(% cell loss with 6-OHDA + GA) / (% cell loss with 6-OHDA alone)]. (C–D) Neurons were first infected with lentivirus carrying shRNA against ATF4 (shATF4, in C) or Parkin (shParkin, in D), or a control shRNA (shCTRL, a mutated version of the shATF4 sequence) for 3 days. Survival was assessed by counting viable nuclei. Data are mean +/− SEM from 2–5 independent experiments for (A, B) and 4 independent experiments for (C, D); each condition was performed in triplicate in each experiment. *p<0.05, **p<0.01, ***p<0.001 by 2-way ANOVA with Bonferroni post-hoc test for (A, B) or 1-way ANOVA with Tukey’s post-hoc test for (C, D).

Treatment with CPT led to an initial increase in P-eIF2α levels, followed by a dramatic reduction at later time points (Figure 8A). Guanabenz led to a more sustained increase in P-eIF2α in response to CPT (Figure 8A–B; quantification in 8D). Furthermore, guanabenz led to partial preservation of ATF4 and parkin protein levels after CPT treatment (Figure 8C; quantification in 8E, 8G) without significantly changing the levels of either ATF4 or parkin mRNA (Figure 8F, H). As in 6-OHDA-treated PC12 cells, these biochemical changes occurred only with a protective concentration (1 µM) of guanabenz, but not with non-protective guanabenz concentrations (Figure 8C–E, G). In sum, guanabenz increases neuronal survival in CPT-treated primary cortical neurons by increasing P-eIF2α, ATF4 and parkin levels.

Figure 8. Guanabenz increases eIF2α phosphorylation, ATF4 and parkin protein levels after CPT treatment.

Primary cultures of rat E18 cortical neurons were pre-treated with GA (doses shown in µM) for 4 hours, then treated with 10 µM CPT for increasing times (A) or 16 hours (B–H). Parallel samples were used to prepare cell lysates for Western blotting with the indicated antibodies (A–E, G), or total RNA for quantification of indicated messages (normalized to alpha-tubulin levels) by quantitative real-time PCR (F, H). (B–C) show representative Western blots. (D, E, G) show densitometric quantification of the indicated proteins, normalized to ERK, and plotted as fold change relative to control (0 GA). Data are mean +/− SEM from 3–6 independent experiments. *p<0.05, **p<0.01 compared to CPT/0 GA by ANOVA with Tukey’s post-hoc test.

Guanabenz improves neuronal survival after intrastriatal injection of 6-OHDA

Finally, we sought to determine whether the protective effects of GA in vitro were also operative in vivo. We used the model of intrastriatal injection of 6-OHDA in adult mice, as an extension of the 6-OHDA models we studied in vitro. GA had been shown to readily cross the blood- brain barrier, and actually reaches higher levels in brain than in plasma (Meacham et al., 1980). Therefore, we felt confident that GA would access the brain. We tested two doses, 1 mg/kg and 4 mg/kg, given via IP injection three times per week. These doses were based upon prior in vivo studies using guanabenz (Ohri et al., 2014; Tribouillard-Tanvier et al., 2008; L. Wang et al., 2014a). In our preliminary injections in control animals, we observed significant, dose-dependent sedation. This effect was reduced after repeated administration, and was no longer apparent after daily dosing for 3 days (data not shown). Furthermore, GA is an α2 adrenergic agonist that lowers blood pressure, another effect, along with sedation, that we wanted to avoid at the time of 6-OHDA injection. To achieve this, we dosed with GA once daily for 3 days (days −3 to −1), performed the intrastriatal 6-OHDA lesion on day 0, resumed GA dosing on the following day (day 1), and continued dosing 3 times per week thereafter until the end of the study (Figure 9A). Animals were sacrificed 4 weeks after 6-OHDA lesion and analyzed for SN dopaminergic neuron count and nigrostriatal dopaminergic innervation. We found that GA at 1 mg/kg provided significant improvement in the number of TH+ neurons in the SN compared to saline control (Figure 9B–C). In contrast, the 4 mg/kg dose had no effect; this observation parallels our observation in vitro that high concentrations of GA were not protective. For striatal TH immunostaining, we found a trend towards improvement in the 1 mg/kg dose (Figure 9D; P=0.16 for GA1 vs vehicle) and no effect at 4 mg/kg. In sum, these in vivo experiments identify a dose of GA that provides protection against 6-OHDA induced degeneration of SN dopaminergic neurons.

Figure 9. GA improves dopaminergic neuron survival and striatal innervation after 6-OHDA lesioning.

8-week old mice were given an intrastriatal 6-OHDA injection and sacrificed after 4 weeks. GA was administered via IP injection for 3 days prior to injection, and then three times per week until sacrifice, at either 1 mg/kg (GA1) or 4 mg/kg (GA4), as shown schematically in (A). Sections were processed for TH immunostaining. Representative low power images of sections through the SN are shown in (B). Stereological counts of TH+ neurons in SN (C) and optical density of striatal TH immunostaining (D) were performed. Results displayed are means +/− SEM. * p<0.05 by 2-way ANOVA with Tukey post-hoc test.

Discussion

Guanabenz was originally developed as an α2-adrenergic agonist; however, our findings show that α2 activity is neither necessary nor sufficient for protection against 6-OHDA-induced toxicity. Rather, we provide several lines of evidence in support of an alternative mechanism: Guanabenz inhibits the function of GADD34, a PP1 regulatory subunit that targets eIF2α for dephosphorylation (Tsaytler et al., 2011). Our findings indicate that GADD34 is up-regulated in response to 6-OHDA, and that its distribution is markedly altered in dopaminergic SN neurons of PD patients compared to controls. In cells exposed to 6-OHDA, guanabenz promotes eIF2α phosphorylation, with consequent elevation of ATF4 and parkin levels. This elevation in ATF4 and parkin is essential for neuroprotection, since knockdown of either one abolishes the protective activity of guanabenz against 6-OHDA. We demonstrated parallel findings in another cellular model of neurodegeneration in which loss of parkin plays a critical role: CPT treatment of primary cortical neurons. In this model, guanabenz improved neuronal survival by increasing eIF2α phosphorylation, ATF4 and parkin, mirroring the findings with 6-OHDA treatment.

Taken together, our studies support a mechanism in which guanabenz promotes neuronal survival in several PD-related cellular models via serial up-regulation of eIF2α phosphorylation, ATF4, and parkin. Nevertheless, we cannot exclude the possibility that guanabenz may have additional effects that promote neuronal survival. For example, GADD34 can direct the dephosphorylation of other proteins, such as TSC1/2 and Akt, that can impact cell survival (Farook et al., 2013; Watanabe et al., 2007). In addition, guanabenz may have effects that are independent of GADD34. For example, guanabenz inhibits the protein folding activity of the ribosome (PFAR). PFAR inhibition appears to play an important role in the anti-prion activity of guanabenz (Nguyen et al., 2014; Tribouillard-Tanvier et al., 2008). The contribution of PFAR inhibition, if any, to the protection observed in our models remains to be determined.

We observed a therapeutic window for guanabenz-mediated neuroprotection. The protective window in cell culture is a concentration range of approximately 5–10-fold, and either higher or lower concentrations of guanabenz failed to both promote survival and augment P-eIF2α, ATF4 or parkin levels. Importantly, our in vivo experiments provide evidence of a similar phenomenon: the lower, 1 mg/kg, dose was protective, while a higher dose of 4 mg/kg was not. Why are higher concentrations of guanabenz no longer protective? As eIF2α phosphorylation is tightly regulated by multiple kinases and phosphatases, it is possible that higher doses of guanabenz trigger compensatory pathways that return eIF2α phosphorylation to a lower level. Alternatively, guanabenz might have “off-target” activities that promote cell death and become dominant at higher concentrations.

Guanabenz has also been studied in paradigms of other neurodegenerative diseases, with apparently conflicting results. TDP-43 aggregates and forms inclusion bodies in both amyotrophic lateral sclerosis (ALS) and fronto-temporal dementia. In both C. elegans and D. rerio models that over-express mutant TDP-43, guanabenz improves neuronal survival and motor function (Vaccaro et al., 2013). However, there are contradictory data regarding the effect of guanabenz on disease onset and progression in a mouse model of ALS based on overexpression of mutant SOD1 (Vieira et al., 2015; L. Wang et al., 2014b). Similarly, in models of prion disease, guanabenz (or compounds with similar effects on eIF2α signaling) have been found to have conflicting effects on survival and disease progression in different studies (Moreno et al., 2012; Tribouillard-Tanvier et al., 2008). Even in models of non-degenerative neurological conditions, guanabenz has been found to have opposing effects, e.g. slowing disease progression in a model of multiple sclerosis (Way et al., 2015), but having no effect in a model of spinal cord injury (Ohri et al., 2014). While differences in pathogenic mechanisms may account for some of these apparently conflicting findings, our results hint at another potential explanation: the doses of guanabenz that were employed exceeded those with protective activity. Furthermore, the sedating and hypotensive effects of GA might have negative effects, particularly in models involving acute injuries or surgeries. Careful consideration of guanabenz dosing and side effects therefore seems warranted in planning future animal studies. It is also notable that the role of parkin in any of the above disease models was not considered.

Guanabenz directly binds to and inhibits GADD34. We found that GADD34 was strongly up-regulated by 6-OHDA. Furthermore, we observed a dramatic difference between PD and control cases in the distribution of GADD34 in neuromelanin-positive neurons of the SN. In PD cases, most dopaminergic neuron cell bodies displayed GADD34-labelled puncta. Occasionally, similar GADD34-positive puncta were observed in neuritic processes. The nature of these GADD34-positive structures is unclear. GADD34 is present in cytosolic and membrane fractions in vitro, and GADD34 co-localizes with both endoplasmic reticulum (ER) and mitochondrial markers (Zhou et al., 2011). Furthermore, GADD34 overexpression causes a disrupted, dilated ER morphology (Zhou et al., 2011). Therefore, the GADD34-labelled puncta in dopaminergic neurons may be an abnormal membranous organelle, such as the ER. Alternatively, the structures might be insoluble aggregates that contain GADD34.

A critical downstream effector of guanabenz-mediated neuroprotection is the maintenance of parkin levels after cellular stress. Guanabenz partially prevents the loss of parkin that occurs in response to either 6-OHDA or CPT treatment, and parkin is required for the protective effect of guanabenz. Similarly, it was reported that in undifferentiated neuroblastoma cells exposed to rotenone, knockdown of parkin (or of ATF4) attenuates the protective effect of salubrinal, a broader inhibitor of eIF2α dephosphorylation (Wu et al., 2014). The mechanisms by which loss of parkin leads to neuron death are still unclear. While many studies suggest that parkin plays an important role in mitochondrial function (Narendra et al., 2008; Rakovic et al., 2013; X. Wang et al., 2011), parkin has been implicated in different facets of mitochondrial biology, and some studies even posit non-mitochondrial activities (Da Costa et al., 2009; Lim et al., 2007). Nevertheless, there is overwhelming evidence supporting the pro-survival activity of parkin in multiple paradigms (Petrucelli et al., 2002; Yasuda et al., 2011). Thus, even though parkin’s downstream effects are still being defined, elevating parkin levels with drugs such as guanabenz is a promising strategy to attenuate neurodegeneration.

Guanabenz has several properties that make it a promising candidate for the treatment of neurodegenerative disorders. Guanabenz readily crosses the blood-brain barrier (Meacham et al., 1980). Furthermore, guanabenz is an FDA-approved medication for the treatment of hypertension. This would facilitate clinical testing of the compound in humans. One drawback to guanabenz is its α2 adrenergic activity, which could lead to hypotension. However, guanabenz analogs have been developed that lack α2 adrenergic activity and that retain protective activity (Das et al., 2015; Nguyen et al., 2014).

An additional feature of guanabenz was that it provided only partial neuroprotection in our models, although a more robust effect was consistently observed with cultured dopaminergic neurons. This lack of full protection could reflect a detrimental off-target activity of guanabenz or the simultaneous activation of pro-apoptotic ATF4 targets such as Trib3 (Aimé et al., 2015). In any case, it is feasible that additional guanabenz analogs or alternative GADD34 inhibitors may show improved efficacy. The modest protection achieved in vivo might also be improved with a different dosing regimen. Nevertheless, even a relatively modest level of neuroprotection would have the potential to provide considerable benefit to PD patients.

In conclusion, several lines of evidence suggest that diminished parkin function is an important contributor to PD pathogenesis, and that elevation of parkin levels is a promising target for neuroprotection. In this study, we evaluate the small molecule guanabenz acetate as a candidate to increase parkin levels via enhancement of eIF2α phosphorylation and ATF4 expression, thus improving neuronal survival. We find that guanabenz enhances survival of both differentiated PC12 cells and cultured dopaminergic midbrain neurons treated with 6-OHDA. We also provide evidence that this effect is operational in vivo, as GA promoted dopaminergic neuronal survival in the substantia nigra after intrastriatal 6-OHDA injection. Guanabenz treatment also led to parkin-dependent protection against DNA damage-induced cell death in primary cortical neurons, a distinct model of neuronal death. Taken together, these data provide a compelling rationale to further study guanabenz and related compounds as therapeutic candidates to enhance parkin levels for the treatment PD and related neurological disorders.

Supplementary Material

supplement

Supplementary Figure 1. ATF4 antibody recognizes transgenic ATF4 at the same molecular weight as endogenous ATF4. PC12 cells were infected with control lentivirus (LV-CTRL) and then treated with 6-OHDA for 8 hours (to induce ATF4), or with lentivirus to over-express ATF4 (in duplicate). Whole cell lysates were prepared and immunoblotted for ATF4.

List of abbreviations

6-OHDA

6-hydroxydopamine

ATF4

activating transcription factor 4

CPT

camptothecin

GA

guanabenz acetate

MPP+

1-methyl-4-phenylpyridinium

PD

Parkinson’s disease

SN

substantia nigra

TH

tyrosine hydroxylase