Experimental Neurology 263 (2015) 177–189

Contents lists available at ScienceDirect

Experimental Neurology journal homepage: www.elsevier.com/locate/yexnr

Regular Article

GDNF signaling implemented by GM1 ganglioside; failure in Parkinson's disease and GM1-deficient murine model Piotr Hadaczek a,1, Gusheng Wu b,1, Nitasha Sharma a, Agnieszka Ciesielska a, Krystof Bankiewicz a, Amy L. Davidow c, Zi-Hua Lu b, John Forsayeth a,⁎, Robert W. Ledeen b,⁎⁎ a b c

Department of Neurological Surgery, University of California San Francisco, San Francisco, CA 94103-0555, USA Department of Neurology and Neurosciences MSB-H506, The State University of New Jersey, 185 South Orange Ave., Newark, NJ 07103, USA Department of Biostatistics/Epidemiology, New Jersey Medical School, Rutgers, The State University of New Jersey, 185 South Orange Ave., Newark, NJ 07103, USA

a r t i c l e

i n f o

Article history: Received 15 July 2014 Revised 16 September 2014 Accepted 16 October 2014 Available online 23 October 2014 Keywords: Parkinson's disease GM1 ganglioside GM1 analogs GDNF LIGA20 RET signaling B4galnt1 Neuroprotection

a b s t r a c t GDNF is indispensible for adult catecholaminergic neuron survival, and failure of GDNF signaling has been linked to loss of dopaminergic neurons in Parkinson's disease (PD). This study demonstrates attenuated GDNF signaling in neurons deficient in ganglio-series gangliosides, and restoration of such signaling with LIGA20, a membrane permeable analog of GM1. GM1 is shown to associate in situ with GFRα1 and RET, the protein components of the GDNF receptor, this being necessary for assembly of the tripartite receptor complex. Mice wholly or partially deficient in GM1 due to disruption of the B4galnt1 gene developed PD symptoms based on behavioral and neuropathological criteria which were largely ameliorated by gene therapy with AAV2-GDNF and also with LIGA20 treatment. The nigral neurons of PD subjects that were severely deficient in GM1 showed subnormal levels of tyrosine phosphorylated RET. Also in PD brain, GM1 levels in the occipital cortex, a region of limited PD pathology, were significantly below age-matched controls, suggesting the possibility of systemic GM1 deficiency as a risk factor in PD. This would accord with our finding that mice with partial GM1 deficiency represent a faithful recapitulation of the human disease. Together with the previously demonstrated age-related decline of GM1 in human brain, this points to gradual development of subthreshold levels of GM1 in the brain of PD subjects below that required for effective GDNF signaling. This hypothesis offers a dramatically different explanation for the etiology of sporadic PD as a manifestation of acquired resistance to GDNF. © 2014 Elsevier Inc. All rights reserved.

Introduction Parkinson's disease (PD) is characterized by gradual loss of dopaminergic (DA) and other catecholaminergic neurons, progressing from peripheral to central neuronal systems over periods of many years (Braak et al., 2003). Unlike familial forms, sporadic PD has no cogent theory that reconciles the disparate data into a compelling narrative regarding its etiology. The vulnerability of these neurons has drawn Abbreviations: aSyn, alpha synuclein; BDNF, brain-derived neurotrophic factor; BSS, basic salt solution; CtxB, B subunit of cholera toxin; DA, dopamine or dopaminergic; GDNF, glial cell line-derived neurotrophic factor; HPTLC, high performance thin-layer chromatography; HT, heterozygote; KO, knockout; LIGA20, membrane permeable analog of GM1; NGF, nerve growth factor; PBST, PBS with Tween-20; PD, Parkinson's disease; SNpc, substantia nigra pars compacta; TH, tyrosine hydroxylase; WT, wild type. ⁎ Correspondence to: J. Forsayeth, Department of Neurological Surgery, University of California San Francisco, San Francisco, CA 94103-0555, USA. ⁎⁎ Correspondence to: R. W. Ledeen, Department of Neurology & Neurosciences, MSBH506, New Jersey Medical School, Rutgers, The State University of New Jersey, 185 South Orange Ave., Newark, NJ 07103, USA. E-mail addresses: [email protected] (J. Forsayeth), [email protected] (R.W. Ledeen). 1 These authors contributed equally.

http://dx.doi.org/10.1016/j.expneurol.2014.10.010 0014-4886/© 2014 Elsevier Inc. All rights reserved.

attention to glial cell line-derived neurotrophic factor (GDNF), considered indispensable for adult catecholaminergic neuron survival. It belongs to a family of neurotrophic factor ligands related to TGFβ that includes neurturin, persephin, and artemin, all of which signal via multicomponent receptors consisting of the RET receptor tyrosine kinase plus a GPI-linked co-receptor (GFRα) that gives binding specificity. GDNF acts molecularly by high affinity binding to GFRα1, causing redistribution of RET into raft microdomains (Tansey et al., 2000). Interaction of these GDNF receptor components transmits survival signals to the cell commencing with homodimerization and autophosphorylation of RET (pRET). Successful results with GDNF-based therapies in animal studies led to clinical trials in which some PD patients experienced therapeutic benefit after putaminal infusions of GDNF (Gill et al., 2003; Slevin et al., 2005; Patel et al., 2005). However, a randomized, placebo-controlled trial that showed limited benefit did not achieve efficacy end-points (Lang et al., 2006) for reasons that may relate to adequate distribution of infused GDNF within the target putamen (Gimenez et al., 2011). In light of delivery-related problems, use of gene therapy for delivery of GDNF to the nigrostriatal system is now under consideration (Johnston et al., 2009; Kells et al., 2012). Another possible explanation for the variable results is inadequate GM1 in the affected

178

P. Hadaczek et al. / Experimental Neurology 263 (2015) 177–189

neurons. Association of this ganglioside with the RET/GFRα1 receptor is required for optimal GDNF neuroprotection, as revealed in the present study. This is consonant with previous studies revealing Parkinsonian degeneration in genetically altered mice deficient in GM1 (Wu et al., 2011a). Interestingly, heterozygous (HT) mice with only partial deficiency of GM1 showed a similar pattern of neurodegeneration as the homozygous (knockout, KO) mice, albeit less severely progressive (Wu et al., 2012). These PD manifestations were alleviated by peripheral administration of LIGA20, a membrane-permeable analog of GM1 that penetrates the blood brain barrier (Wu et al., 2005) and accesses both membrane and intracellular loci of neurons (Wu et al., 2001a, 2001b) where it mediates the multiple functions of GM1. GM1 itself, despite limited brain penetration, was nevertheless able at high doses and frequent application to improve motor symptoms and reduce the rate of symptom progression in PD subjects (Schneider et al., 2013). In view of the demonstrated requirement of GM1 for functional efficacy of other neurotrophic factors, such as NGF (Mutoh et al., 1995) and BDNF (Pitto et al., 1998), we considered the possibility that deficiency of this ganglioside might cause impaired GDNF signaling. The present study supports that hypothesis, based on evidence for deficient GDNF receptor activation in the above GM1-deficient mice as well as PD subjects. We have also performed in vitro experiments in cells that express RET and GFRα1 as well as GM1 and its metabolic precursor, GD1a. In each of these systems, reduced expression of GM1 (and GD1a) resulted in attenuated formation of phosphorylated RET (pRET), the first step in GDNF-induced signaling, and in downstream formation of pMAPK. This signaling deficit was largely alleviated with LIGA20. Coordinate activity of GM1 and GDNF was further indicated in the virtual elimination of alpha-synuclein (aSyn) aggregation both by LIGA20 (intraperitoneal) and GDNF application via AAV2 vector transmission into the striatum. The latter treatment also improved motor performance with a pole test and restored lost tyrosine hydroxylase (TH) immunoreactivity, similar to that previously shown for LIGA20 with the KO and HT mice (Wu et al., 2011a, 2012). These results demonstrate the ability of upregulated GDNF to override the GDNF resistance acquired with depressed GM1. That a similar mechanism based on GM1 insufficiency may be operating in PD itself was suggested by the earlier observation of depressed levels of GM1 in DA neurons of the substantia nigra pars compacta (SNpc) (Wu et al., 2012). We now report significantly diminished pRET formation in similar neurons of PD subjects, correlated with diminished GM1. Our concurrent finding of significant GM1 deficit in the occipital cortices of PD subjects further suggests the possibility of systemically low levels of GM1 in PD subjects which, with the progressive diminution of GM1 known to occur with aging, eventually falls below the level needed to maintain effective GDNF signaling. In this respect, our hypothesis offers a dramatically different explanation for the etiology of sporadic PD as a manifestation of acquired resistance to GDNF in a manner somewhat reminiscent of the insulin resistance seen in Type 2 diabetes mellitus. Materials and methods Generation of Neuro2a (α1-GA) cells with knockdown of B4galnt1 (GM2/ GD2 synthase) A Neuro2a cell line, previously engineered to over-express human GFRα1 (Crowder et al., 2004), (α1), was received from Dr. Jeffrey Milbrandt (Washington University, St. Louis, MO). We used a commercially available lentivirus vector encoding shRNA (3 to 5 expressing constructs) specific for mouse B4galnt1 mRNA (Santa Cruz Biotechnology, Inc.; sc-77390-V) to suppress GM2/GD2 synthase expression. To generate cells with a stable suppression of the enzyme (α1-GA), we followed a detailed protocol available on manufacturer's website (http://www.scbt. com/protocols.html?protocol=shrna_lentiviral_particles_transduction). The concentration of puromycin used for selection of clones was 2 μg/ml (puromycin death curve was prepared prior to lentiviral infection). Control

shRNA lentivirus (a scrambled shRNA sequence) was used as a negative control (sc-108080). The effect of lenti-shRNA transfection on GM2/GD2 synthase expression was analyzed by quantitative RT-PCR. RNA was isolated with QIAzol Lysis Reagent and RNAeasy MinElute columns from Qiagen. One microgram of RNA was used for the RT reaction (iScript™ cDNA synthesis Kit from Bio-Rad). One microliter of cDNA (1:10 dilution) was subsequently used for real-time PCR (TaqMan® Gene Expression Assay for mouse B4galnt1; cat. no. 4351372 and mouse Hprt1 used as a house-keeping control cat. no. 4453320, Applied Biosystems). The reaction was performed and analyzed in ViiA™ 7 Real-Time PCR System (Applied Biosystems). Threshold Ct was set to 0.1. Relative change of GM2/GD2 synthase expression was determined by subtraction of the ΔCt value for the control cell line from the ΔCt value for the B4gallnt1 shRNA-transfected cells (ΔΔCt value). Fold change was subsequently calculated with the formula 2−ΔΔCt, defined as the expression of GM2/ GD2 synthase relative to HPRT mRNA. Measurement of MAPK activation by GDNF in Neuro2a (α1) and Neuro2a (α1-GA) cells; dose response study (Fig. 1B) To measure the activation of MAPK1/2 in α1 and α1-GA cells by GDNF, a dose–response curve was generated by means of a custom incell ELISA. Cells were seeded at a density of 30,000 cells/well in a 96-well plate, Optilux™, black/clear bottom (Fisher Scientific). After overnight incubation at 37 °C, the RPMI-1640 medium with 10% FBS penicillin/streptomycin was removed and cells were starved in serumfree medium for 6 h. Following starvation, serum-free medium with various concentrations of GDNF (range: 0.78–100 ng/ml) was added to cells (6 replicates per concentration) and incubated for 10 min at 37 °C. Later, the medium was removed and cells were fixed immediately with cold methanol (200 μl/well) for 20 min. After 2 washes in PBS with 0.1% Tween-20 (PBST), the cells were blocked with 5% chicken serum in PBST for 1 h at room temperature. A rabbit monoclonal antibody against phospho-p44/p42 MAPK (Cell Signaling Technology) was then added at a dilution of 1:200 in PBST with 2% chicken serum and incubated at 4 °C overnight. After 3 rinses in PBST, a chicken anti-rabbit IgG (1:500) labeled with Alexa488 (Invitrogen) was added to each well and incubated at room temperature for 1 h. After 2 rinses in PBST containing 0.2 μg/ml DAPI (4′,6-diamidino-2-phenylindole), cells were rinsed in PBS and the plate was read in BioTek FL-800 fluorescence microplate reader. Fluorescence was determined via two pairs of filters: 488/515 green channel (for phospho-p44/p42 MAPK) and 360/420 blue channel (DAPI for nuclear staining). The Alexa 488 fluorescence was then divided by the DAPI fluorescence to normalize for cell count per well. The curve of MAPK1/2 activation (RFU, relative fluorescence units, RAlexa488/DAPI) was plotted against GDNF concentration. The difference between the two curves was obtained by regression analysis of RFU values with each dose treated as either a categorical variable or continuous variable; in both models the effect of the cell group was statistically significant (p b 0.0001). Individual GDNF doses were analyzed by two-way ANOVA with Bonferroni post-hoc test. Measurement of RET and MAPK activation by GDNF in SH-SY5Y, Neuro2a (α1), and Neuro2a (α1-GA) cells; restoration of lost signaling by LIGA20 (Figs. 1C, 2C) These cells were cultured in DMEM-10% FBS in 96-well Nunclon Delta black microwell plates (Nunc 137101) containing 30,000 cells/ 100 μL/well. Some SH-SY5Y cells were differentiated with retinoic acid (40 μM) for 6 days and some of these were treated with d-threoPDMP (10 μM; Matreya, Pleasant Gap, PA) during the last 24 h. Portions of the PDMP-treated SH-SY5Y cells and the above Neuro2a (α1-GA) cells were treated with LIGA20 (gift of Fidia Pharmaceuticals, Abano Terme, Italy) (10 μM) for 2 h. For signaling measurements, all cell types were replaced to DMEM without serum for 4 h, then stimulated

P. Hadaczek et al. / Experimental Neurology 263 (2015) 177–189

179

Fig. 1. GM1 regulation of GDNF signaling in modified Neuro2a cells. (A) The cell types employed were Neuro2a with GFRα1 over-expression (α1), and α1 cells with shRNA-depressed expression of GM2/GD2 synthase (B4galnt1) to reduce GM1 (α1-GA). The cells were stained with anti-GFRα1 antibody (Texas red-linked 2nd Ab) and CtxB-FITC (for GM1), showing depletion of GM1 and no depletion of GFRα1 in α1-GA cells. Confocal images were taken at 60×. Scale bar = 10 μm. (B) The same two cell types, α1 and α-GA, were stimulated with different concentrations of GDNF. The level of phosphorylation was measured by in-house-developed in-cell ELISA (see Methods) showing diminished GDNF signaling in GM1-depleted cells. Asterisks indicate p values b0.05 measured by two-way ANOVA with Bonferroni post-hoc test for individual GDNF doses; the difference between the two curves was obtained by regression analysis of RFU values with dose treated as either a categorical variable or continuous variable (p b 0.0001 in both cases). (C) The above two cell types and parental Neuro2a cells were each stimulated or not with 1 ng GDNF/100 μL/well in the above assay. Robust GDNF signaling in α1 cells was ablated in GM1-deficient (α1-GA) cells and restored in the latter with LIGA20. Nuclei were indicated with DAPI. Analyzed with one-way ANOVA with Tukey's multiple comparison test. (For interpretation of the references to color in this figure legend, the reader is referred to the web version of this article.)

or not with GDNF (Sigma-Aldrich, St. Louis, PA): one ng/100 μL at 37 °C for 10 min. The medium was then removed and the cells fixed with cold (−20 °C) methanol for 30 min. The fixed cells were stained overnight with rabbit anti-tyr 1062 phosphorylated RET (pRET) antibody (Santa Cruz Biotec., Santa Cruz, CA, Cat. # sc-20252-R) or rabbit anti-pMAPK (Cell Signaling Tec., Cat. # 4370S) at 4 °C, followed by 2 h incubation of FITC linked donkey anti-rabbit IgG and DAPI (1 ng/100 μL) at room temp; the latter was used to stain nuclei. The fluorescent ratio (RFU) for RFITC/DAPI was used to determine relative pRET and pMAPK expression. The data are expressed as mean ± SEM of 3–4 experiments with 6–8 wells in each group for each experiment, and were analyzed statistically by one-way ANOVA with Tukey's Multiple Comparison Test. Plates were read in a PerkinElmer Victo 3 V microplate reader. The same two filters as above were employed.

Immunohistochemistry, immunoblot analysis, and ganglioside patterns of modified Neuro2a and SH-SY5Y cells Neuro2a cells overexpressing GFRα1 (α1), along with same cells subjected to shRNA knockdown of B4galnt1 (α1-GA) (see above), were grown on coverslips in DMEM with 10% FBS. SH-SY5Y cells were grown in the same medium and were differentiated with 40 μM retinoic acid in DMEM-10%FBS for 6 days. Some SH-SY5Y cells were treated with d-threo-PDMP (10 mM) for the final 24 h to inhibit GM1 biosynthesis. LIGA20 was applied where indicated at 10 μM for the last 2 h. Cells were replaced in DMEM without serum for 4 h, then stimulated with GDNF (10 ng/mL, 10 min) as above before cells were fixed in 2% cold paraformaldehyde. Cells were stained with Ctx B-FITC (for GM1), DAPI (for nucleus) and primary antibodies against GDNF, GFRα1, RET or tyr-1062 phosphorylated RET (pRET), all from Santa Cruz Biotec; Texas red-linked second antibody was employed. Confocal images were taken at 60× with a Nikon A1R confocal laser microscope. Immunoblot (IB) analysis of SH-SY5Y cells was carried out by applying lysates (50 μg protein) to SDS-PAGE on 4 –15% polyacrylamide gels followed by transfer to PVDF membranes and application of antibodies to GFRα1, RET, and pRET (Santa Cruz Biotec, Santa Cruz CA). Actin levels were similarly revealed on a parallel gel to which equal volumes of lysates containing equal amounts of protein were applied, followed by staining with anti-actin antibody (Santa Cruz Biotec). Gangliosides from these cells were extracted with chloroform–methanol and

separated by HPTLC on silica-coated plates (EMD Chemicals, Philadelphia, PA) as described (Wu et al., 2011b). Reduced pRET expression in GM1-deficient mice and DA neurons of PD subjects. Restoration of mouse pRET with LIGA20 Heterozygous mice with disrupted B4galnt1 gene (GM2/GD2 synthase) were bred as described (Wu et al., 2011a, 2012) at Rutgers University, New Jersey Medical School to provide the three genotypes. All operations conformed with regulations of the Institutional Animal Care and Use Committee at that institution. Four HT and KO mice (~200 days of age) were given 2.5 mg/kg of LIGA20 (in basic salt solution, BSS) via intraperitoneal injection thrice per week over a fiveweek period, then sacrificed one day after the final injection via cardiac perfusion. Control WT, HT, and KO mice (four each) injected with BSS only were perfused similarly. Frozen SNpc sections from those mice were processed for immunohistochemistry (IHC) by staining with anti-TH antibody (EMD Millipore, Billerica, MA) as described (Wu et al., 2011a, 2012). Concurrent staining for pRET was performed with the above anti-pRET antibody. Five stained sections from each mouse were counted for (a) double-stained cells (TH, pRET) and (b) total TH+ cells (500–1000 cells/mouse), and the % determined as a/b. Four mice in each group were averaged and statistical comparisons made with one-way ANOVA (n = 4). Additional analyses measured fluorescent intensity of pRET staining in TH+/pRET+ cells with a Nikon AIR confocal laser microscope. To maintain consistency of the scanned images for all sections, the settings, including laser intensity, confocal aperture, PMT voltage, electronic gain, scan speed, image size, filter and Z thickness, were standardized throughout the experiment. Images were taken with a 60× objective. The optical intensity of indicated fluorescence ranging from 0 to 4095 was measured with NIS-Elements program; zero (0) represented a maximum black image and 4095 a maximum bright image. Five sections (total 100–200 cells) from each mouse were counted for four mice from each group. Statistical significance was assessed with one-way ANOVA (n = 4). The same pRET and TH stains were applied to human brain samples obtained from the New York Brain Bank at Columbia University, consisting of formalin-fixed paraffin-embedded 4–7 μm tissue sections from the SNpc region of sporadic PD patients and non-PD agematched controls. These PD subjects and controls were 66–88 years of age and included both genders. Four sections from each patient and

180

P. Hadaczek et al. / Experimental Neurology 263 (2015) 177–189

Fig. 2. GM1 regulation of GDNF signaling in SH-SY5Y cells. SH-SY5Y cells in culture were induced to differentiate with retinoic acid and were then treated or not with PDMP. Confocal images were obtained after IHC staining that was carried out with CtxB-FITC (green) and antibodies to GFRα1, RET, and pRET (Texas red as 2nd Ab in all cases); DAPI (blue) was employed to stain nuclei. PDMP caused partial suppression of GM1 but not GFRα1 or RET. Cells with or without PDMP treatment were subjected to GDNF stimulation followed by IHC to reveal suppression of pRET in cells with reduced GM1; pRET formation was restored by LIGA20. (A) Undifferented and (B) differentiated cells. Scale bar = 10 μm. (C) Quantification of pRET formation under these varying conditions employing cell ELISA. Analyzed via one way ANOVA with Tukey's multiple comparison test. Numbers preceded by b represent p values: *b0.05, **b0.01, and ***b0.001. Other designations are comparison to untreated cells (first bar). (For interpretation of the references to color in this figure legend, the reader is referred to the web version of this article.)

control were stained with anti-TH and anti-pRET antibodies as above and the results quantified as % of TH+ cells that also expressed pRET. Four sections with 1200–2100 cells for control and 500–1900 cells for PD samples were analyzed. The results for 11 PD subjects and 11 controls were analyzed by Mann–Whitney Rank Sum Test. To show correlation of pRET with GM1, four sections for each of the 11 PD subjects and controls were stained for TH and GM1 as described (Wu et al., 2011a) and the % of TH+ cells that also expressed pRET or GM1 were separately determined. The significance was determined using linear regression and correlation.

AAV2-GDNF injection and immunohistochemical staining Heterozygous mice with disrupted B4galnt1 gene (GM2/GD2 synthase) were bred as described (Wu et al., 2011a, 2012) to provide the three genotypes which received cerebral infusions of AAV2-GDNF. All surgical procedures were conducted in accordance with regulations of the Institutional Animal Care and Use Committee of the University of California San Francisco. The cDNA coding human GDNF was cloned into an AAV2 shuttle plasmid, and a recombinant AAV2 carrying GDNF under the control of the cytomegalovirus promoter was generated by

P. Hadaczek et al. / Experimental Neurology 263 (2015) 177–189

a triple transfection technique as previously described (Matsushita et al., 1998; Wright et al., 2003). The titer (vector genomes per ml) was determined by quantitative PCR. The final titer of AAV2-GDNF vector used in this study was 1.1 × 1013 vector genomes per ml. The mean age of GM2/GD2-deficient mice used for our experiments was 312 ± 19 days. Five mice each of each genotype (KO, HT, WT) received bilateral infusions of AAV2-GDNF (2 μl) per hemisphere into the striatum at stereotactic coordinates relative to bregma and dura: anteroposterior +0.2 mm, mediolateral +2.0 mm, dorsoventral −3.45 mm. Vector infusion was performed with the intra-cerebral delivery system previously described (Ciesielska et al., 2011). Control mice (5 of each genotype) were infused with PBS. Twenty weeks after AAV2-GDNF injections, mice were deeply anesthetized with sodium pentobarbital (90 mg/kg intraperitoneal) and perfused transcranially with ice-cold phosphate buffered saline (PBS) followed by 20 ml of 4% paraformaldehyde in PBS. The brains were removed, post-fixed for 24 h and processed for immunohistochemical evaluations. Single- and doublestaining protocols were carried out as described previously (Ciesielska, et al., 2011). For staining against tyrosine hydroxylase (TH), coronal sections (40 μm) were incubated with a mouse monoclonal anti-TH antibody from Millipore at a dilution of 1:5000 and a secondary, biotinylated anti-mouse IgG (1:200) followed by development with the Vectastain© ABC kit and DAB (Vector Laboratories) after blocking with 10% normal horse serum in PBS. Double-staining for TH/aSyn employed a cocktail of the same primary anti-TH antibody combined (1:500) with a rabbit polyclonal anti-aSyn (1:50) from Cell Signaling (Cell Signaling Technology, Inc.; Danvers, MA). The cocktail of fluorescent secondary antibodies (Alexa Fluor® 555 goat anti-mouse IgG and Alexa Fluor® 488 goat antirabbit IgG) was used for detection (both at 1:1000). No significant nonspecific signal was generated from the secondary antibodies, as determined by control experiments where the primary antibody was omitted from the IHC procedure. All sections were examined and digitally photographed on a Zeiss Axioskop microscope (Carl Zeiss, Thornwood, NY). Adobe Photoshop 6 software was used to establish co-localization of TH and aSyn staining at the level of substantia nigra. From each mouse, three consecutive sections, centered at the level of the medial terminal nucleus, were chosen for fluorescent staining. Images of TH (red channel) and aSyn (green channel) staining were superimposed and made visible by manipulation of each image's opacity to 50%. Motor performance and coordination Horizontal pole test Animals were trained to traverse the length of the 60-cm cylindrical (1.2 cm in diameter) pole. All mice received 2 days of training before each test (6 assisted trials 24 h prior to actual testing and 3 trials the following day just before the actual testing). The mean time to walk the 60-cm distance was calculated from 3 trials for each mouse at a given time point (pre-surgery, 2-, 3-, 4-, and 5 months after surgery). Results were expressed as a percentage change in the time needed to traverse the length of the pole compared to a pre-surgery baseline for each genotype separately: GM2/GD2 knockouts (KO); heterozygous GM2/GD2 knockouts (HT); and wild type controls (WT). The decremental percentage change confirms improvement (shorter times needed to complete the test), while the incremental percentage change attests to worsening (longer times needed to complete the test). Statistical analysis (two-way ANOVA with repeated measures and post-hoc Bonferroni's test) and graphs were performed with PRISM 6 software (GraphPad Software, Inc.) Gait analysis To measure gait, animals were trained to walk through a narrow, 120-cm alley leading into their home cage. Once trained, paper was placed along the alley floor, and each animal's hindlimbs were brushed

181

with nontoxic paint. Animals were then placed at the beginning of the alley. As they walked into their home cage, they left their paw-prints on the paper beneath. Stride-length was determined by measuring the distance between paw-prints. Only strides made while continuously walking (no stopping) were included in the analysis. Strides at the beginning and end of the alley were not counted. The results were expressed as change in stride-length compared to a pre-surgery baseline for each genotype. The incremental percentage change confirmed improvement (longer strides), whereas the decremental percentage change attested to worsening (shorter strides). Statistical analysis: two-way ANOVA with repeated measures and post-hoc Bonferroni's test and graphs were performed with PRISM 6 software (GraphPad Software, Inc.) Ganglioside quantification in PD occipital cortex Samples of occipital cortex of PD subjects and normal controls were obtained from the Brain and Body Donation Program, Banner Sun Health Research Institute, Sun City, Arizona (Arizona Parkinson's Disease Consortium), and the Brain and Tissue Bank, Univ. Maryland at Baltimore. Gray matter was carefully dissected free of white matter and total lipids extracted with chloroform–methanol (1:1, by vol). The protein residues were quantified by Lowry analysis following digestion of the pellet for several hours in NaOH/SDS (Lees and Paxman, 1972). The lipid extracts were reconstituted in smaller volumes of chloroform–methanol (1:1, by vol) and the gangliosides analyzed by HPTLC; applied volumes corresponded to equal protein. Varying amounts of bovine brain ganglioside mixture were applied on the same plate to serve as standards for densitometric quantification. Development of the plate, visualization of separated gangliosides, and quantification of each member of the ganglio-series were carried out as described (Wu et al., 2011b). Association of GM1 with GDNF receptor complex The substantia nigra was dissected from the brains of the three murine genotypes and homogenized in lysis buffer containing 1% Brij 98, 25 mM HEPES buffer (pH 7.5), 150 mM NaCl, 5 mM EDTA, and protease inhibitor cocktail (Roche, Indianapolis, IN). Lysate containing 100 μg protein was immunoprecipitated (IP) with anti-GFRα1 plus protein G-linked agarose beads (Santa Cruz Biotec, Santa Cruz, CA) according to manufacturer's instructions. The resulting IP products were subjected to SDS-PAGE on a 4–15% polyacrylamide gel and transferred to a PVDF membrane which was immunoblotted with anti-GFRα, RET, pRET (Santa Cruz Biotec, Santa Cruz, CA), and/or cholera toxin B subunit (CtxB-HRP for GM1; Invitrogen, Carlsbad, CA). Additional IP was carried out separately with anti-GFRα1 and anti-pRET and each precipitate extracted with chloroform–methanol followed by HPTLC as described (Wu et al., 2011b). Co-precipitated GM1 was detected on thin-layer plates with CtxB-HRP. To compare tissue levels of RET and GFRα1, lysates containing 50 μg protein from the three genotypes were directly subjected to SDS-PAGE and transferred to a PVDF membrane as above. Immuno-detection was carried out as above. Each IP and IB study was performed three times. Statistical analysis Data are presented as mean ± S.E.M. and were analyzed with GraphPad Prism software. These include Mann–Whitney rank sum test (Figs. 4D, 8B & C), one-way ANOVA with Tukey's multiple comparison test (Figs. 1C, 2C, 4B) or Bonferroni's post-hoc test (Fig. 7A & C), two-way ANOVA with repeated measures with Bonferroni's post-hoc test (Figs. 7B & D, S2) and linear regression and correlation (Fig. 4E). Fig. 1B was analyzed by regression analysis (difference between curves) and by two-way ANOVA with Bonferroni post hoc test (Individual GDNF doses). P b 0.05 was considered statistically significant.

182

P. Hadaczek et al. / Experimental Neurology 263 (2015) 177–189

Results Neuronal cell lines with diminished GM1 show impaired GDNF signaling We used a lentivirus vector encoding shRNA specific for mouse B4galnt1 mRNA to suppress GM2/GD2 synthase expression (Supplementary Fig. S1) in a Neuro2a cell line previously engineered to overexpress human GFRα1 (Crowder et al., 2004). Immunohistochemistry (IHC) with confocal images showed substantial depletion of GM1 in these shRNA-treated cells (α1-GA) (Fig. 1A), which was verified by high performance thin-layer chromatography (HPTLC; data not shown). Fig. 1A also demonstrates maintenance of GFRα1 levels despite GM1 reduction, and co-localization of that receptor component with GM1. The parental cell line evinced a brisk response to recombinant GDNF in a 96-well plate ELISA in which binding of anti-phosphoMAPK antibody to fixed cells was quantified with a fluorescent second antibody (Fig. 1B). The addition of GDNF increased MAPK phosphorylation up to 9-fold over basal with a maximum response at 25 ng GDNF/ ml. The derivative cell line, α1-GA, showed a significant shift to the right of the response curve that was particularly evident at low GDNF concentrations. The statistical difference between these curves was significant, as were the relative fluorescent units (RFUs) for individual doses (except at 25 ng GDNF/ml). At the lowest concentrations of GDNF, depletion of GM1 resulted in virtual ablation of GDNF response, whereas higher concentrations of GDNF could substantially overcome the GM1-dependent resistance. It is important to note that endogenous GDNF in primates occurs at extremely low levels (b 35 pg/mg protein) (Eslamboli et al., 2005). That GM1 was a crucial element in these changes was further demonstrated by restoration of elevated pRET and pMAPK after treatment of the GM1-deficient α1-GA cells with LIGA20; we observed no activation of RET by LIGA20 in the absence of GDNF (Fig. 1C). Similar studies were carried out with SH-SY5Y cells, previously shown to upregulate RET upon retinoic acid-induced differentiation (Yamada et al., 2007). We found that GFRα1 is also expressed both before (Fig. 2A) and after (Fig. 2B) differentiation with retinoic acid and that both pRET and pMAPK were significantly elevated with GDNF (Fig. 2B & C). Prior treatment with d-threo-PDMP, an inhibitor of glucosylceramide synthesis (Radin et al., 1993) that partially reduces GM1, caused reduction of pRET and pMAPK, thereby supporting the hypothesis that even partial GM1 reduction can block GDNF signaling.

Fig. 2C shows that PDMP treatment reduced pRET and pMAPK to basal levels with significant restoration by LIGA20. The IHC results in Fig. 2B suggested that neither GFRα1 nor RET was reduced due to GM1 reduction, an idea supported by IB analysis that shows retention of GFRα1 and RET levels coincident with reduced pRET upon GM1 reduction (Fig. 3A; housekeeping actin assay was carried out on a parallel IB). This reduced pRET was restored to control level with LIGA20. Fig. 3B indicates via HPTLC the degree of a-series ganglioside (GM1, GD1a) depletion by PDMP. To test for the possibility that reduced pRET may result from impaired GDNF binding, differentiated SH-SY5Y cells with and without prior PDMP exposure were treated with GDNF followed by washing and application of fluorescent anti-GDNF antibody (Fig. 3C); no observable difference in staining intensity was apparent, thus indicating no direct role of GM1 in binding of GDNF to GFRα1. Results below provide further insight into the inhibition mechanism. GM1 association with GDNF receptor; failed GDNF signaling in GM1-deficient mice and PD subjects Further evidence for the promotion of GDNF signaling by GM1 was obtained with brain tissues from the above mice (~200 days of age) either wholly (KO, B4galnt1−/−) or partially (HT, B4galnt1+/−) deficient in GM1; both of these were previously shown to manifest behavioral and neuropathological symptoms characteristic of PD (Wu et al., 2011a, 2012). Frozen sections of the substantia nigra (SN) showed diminished staining of pRET for HT mice compared to wild type (WT), and more substantial reduction for KO mice (Fig. 4A). Treatment of the mice with LIGA20 elevated pRET significantly for both genotypes (Fig. 4A, B), indicated in two ways: % of TH+ neurons with pRET and pRET fluointensity. This further indicated that LIGA20 functions as a replacement for deficient GM1. These sections were from the same mice depicted in Fig. 3 of our previous paper (Wu et al., 2012). Importantly, a similar reduction of pRET was found immunohistochemically in the SNpc of PD subjects (Fig. 4C), which was quantified by two means (Fig. 4D, E). These sections were from the same PD subjects and controls depicted in Fig. 7 of Wu et al. (2012) and the data corresponding to the X axis of Fig. 3 in that paper are drawn from the same figure. DA neurons of PD subjects were identified by positive staining for TH. All 11 of the sporadic PD subjects showed the same phenomenon, thus demonstrating severely impaired GDNF signaling in SNpc neurons characteristic of

Fig. 3. PDMP suppression of pRET formation without reduction of RET or GFRα1 in SH-SY5Y cells; no effect of PDMP on GDNF binding. (A) Cells were induced to differentiate with retinoic acid and were treated or not with PDMP. Some cells were then treated with LIGA20. Cell lysates were subjected to immunoblot analysis; actin, as housekeeping control, was determined on a parallel IB. PDMP, which reduced a-series gangliosides (GM1, GD1a), reduced pRET without changing RET or GFRα1; the pRET level was restored with LIGA20. (B) HPTLC of gangliosides extracted from cells before and after differentiation and after suppression of ganglioside synthesis in differentiated cells with PDMP. (C) GDNF was applied to differentiated cells that were treated or not with PDMP followed by washing and staining with CtxB-FITC (green) and anti-GDNF antibody (Texas red with 2nd antibody; DAPI (blue) for nuclei). Despite PDMP-induced reduction of GM1, GDNF staining remained unchanged in keeping with unchanged GFRα1. LIGA20 restored GM1-like fluorescence reduced by PDMP. (For interpretation of the references to color in this figure legend, the reader is referred to the web version of this article.)

P. Hadaczek et al. / Experimental Neurology 263 (2015) 177–189

183

Fig. 4. Failure of GDNF signaling in SNpc of GM1-deficient mice and PD subjects. (A) SNpc sections from heterozygous (HT) mice partially deficient and knockout (KO) mice wholly deficient in GM1 were subjected to IHC staining with anti-pRET and anti-tyrosine hydroxylase (TH) antibodies. Analysis with confocal imaging revealed that pRET expression was suppressed in HT mice and more strikingly in KO mice. Significant restoration of pRET occurred in both genotypes treated with LIGA20. (B) Quantification of above results: left panel, % of TH+ neurons with pRET; right panel, fluorescent intensity of pRET staining in TH+ neurons. Statistical analysis was with one-way ANOVA. *p b 0.05, **p b 0.01 vs BSS treated WT. (C) Similar IHC analysis with confocal imaging applied to SNpc sections from PD subjects and age-matched controls; shown are two representative samples out of 11 for each. pRET was markedly reduced in TH+ neurons of PD subjects. (D) and (E) Quantification of IHC results in (C) by Mann–Whitney Rank sum test (D) and linear regression and correlation (E). Scale bars = 100 μm.

PD. On the basis of these experiments and the previous finding of deficient GM1 in such neurons (Wu et al., 2012), we have concluded that effective GDNF signaling is dependent on the maintenance of an adequate level of GM1 in that brain region (Fig. 4E). To explore the nature of the GM1–GDNF interaction in vivo we carried out immunoprecipitation (IP) of GFRα1 from the SNpc of mouse

brains and found that, in addition to co-precipitation of RET and pRET, GM1 also co-precipitated with GFRα1 as shown in its migration with the SDS-PAGE solvent front (Fig. 5A). Significantly, RET and pRET were both diminished as co-precipitates in the HT SNpc and more so in KO SNpc whereas the total SNpc levels of RET and GFRα1 were not diminished in those mutants; this was in contrast to the reduction

184

P. Hadaczek et al. / Experimental Neurology 263 (2015) 177–189

Fig. 5. Association of GM1 with GFRα1 and pRET. (A) Substantia nigra tissues from WT, HT and KO mice were immunoprecipitated (IP) with anti-GFRα1 Ab and subjected to immunoblot (IB) analysis. GM1 migrates with the electrophoretic front. Note also the decrease of RET and pRET in HT mice, and more pronounced decrease in KO mice. (B) Direct immunoblot of GFRα1 and RET from substantia nigra tissues of the three genotypes. Note retention of both, despite diminution of GM1. (C) HPTLC of gangliosides extracted from the immunoprecipitates of both GFRα1 and pRET, indicating specific association of GM1 with the GDNF receptor complex. Each IP, IB, and HPTLC experiment was performed two additional times with results similar to those shown.

of SNpc pRET in HT and KO tissues (Fig. 5B). This agrees with the findings in Figs. 2 and 3 showing retention of GFRα1 and RET levels, despite reduced GM1, and suggests that reduced signaling by depleted GM1 is not based on reduced levels of the receptor proteins. Rather, we propose that a primary function of GM1 is to facilitate assembly of the GDNF receptor complex of which GM1 is an integral part. In addition to the SDS-PAGE evidence (Fig. 5A), this was demonstrated by co-precipitation of GM1 with both GFRα1 and pRET revealed by HPTLC of the lipid extracts (Fig. 5C). Over-expression of GDNF overcomes low GM1-induced parkinsonism We next explored the possibility that the parkinsonism seen in the GM1-deficient mice could be overcome by over-expression of GDNF. All three genotypes (312 ± 19 days of age) received bilateral striatal infusions of an adeno-associated virus encoding human GDNF (AAV2GDNF); employing the infusion methodology we previously showed enhances DA activity in the putamen of aged normal (Kells et al., 2012) or MPTP-lesioned nonhuman primates (Eberling et al., 2009). Post-mortem analysis of the infused mouse brains revealed abundant GDNF immunostaining throughout striatum in all treated animals (data not shown). These results showed that enhancing GDNF content in the basal ganglia of both KO and HT mice prevented Parkinsonian degeneration, in that treated animals showed substantial preservation of TH staining in contrast to untreated HT and KO mice (Fig. 6B). In accord with prior studies (Wu et al., 2011a, 2012), untreated KO and HT mice showed significant accumulation of aSyn; strikingly, GDNF-treated mice showed no evidence of this (Fig. 6A). In fact, GDNF treatment eliminated sporadic aSyn accumulation even in older WT animals (Fig. 6A). This finding forges a direct link between GDNF signaling and the formation of aSyn aggregates, a primary pathological feature of PD. Previously, we showed that aged nonhuman primates evinced significant upregulation in dopaminergic function and locomotor activity when given putaminal AAV2-GDNF, reflecting the ability of GDNF to reverse age-related nigro-striatal decline (Johnston et al., 2009). It takes

several months for GDNF to exert its full effect (increased density of dopaminergic terminals, increased number of TH+ neurons, and increased locomotion). Similar effects of GDNF were seen in our mice and this was evident in WT, HT and KO mice. In a horizontal pole test in which mice were timed in their transit of a wooden rod (Fig. 7A, B, and Supplementary Fig. S2A), WT and HT mice improved their times by N20% comparing pre-surgery to 5 months after surgery whether they received a striatal PBS injection or AAV2-GDNF. In contrast, PBS-injected KO mice displayed a relative decline in walk-time that was completely blocked by AAV2-GDNF treatment, giving the same improvement in walk-time as GDNF-treated WT and HT mice. Stride-length in WT, HT and KO mice did not change significantly when treated with PBS over the 5-month experiment (Fig. 7C, D, and Supplementary Fig. S2B), although the KO mice trended toward a shortening of stride. In all three lines, however, GDNF treatment increased stride-length by up to 20%. GM1 deficit in occipital cortex of PD subjects In view of these and previous results indicating that HT mice with partial deficit of GM1 show impaired GDNF signaling and multiple Parkinsonian symptoms, we considered the possibility that PD subjects might display systemically reduced levels of GM1. The very low levels of GM1 revealed in SNpc DA neurons of such subjects might well have reflected cellular pathology, but the moderate reductions evident in adjacent non-DA cells (glia as well as neurons) (Wu et al., 2012) suggested more widespread, possibly systemic deficiency. To further explore that possibility, we studied a less affected brain region, the occipital cortex of PD subjects. Carefully dissected gray matter samples from that region were subjected to ganglio-series analysis via HPTLC, revealing the pattern shown in Fig. 8A for two representative PD subjects and two controls. Also shown in this HPTLC are 3 different amounts of standard (bovine brain gangliosides) of known composition, used for densitometric quantification. Fig. 8B shows the results of such quantification, the GM1 average for PD subjects being significantly below that of agematched controls (n = 13). GD1a, a metabolic precursor to GM1

P. Hadaczek et al. / Experimental Neurology 263 (2015) 177–189

185

Fig. 6. Elimination of α-synuclein (aSyn) and restoration of TH expression in nigral neurons of GM1-deficient mice with AAV2-GDNF treatment. (A) Representative double immunostained brain sections from mice of the three genotypes, employing antibodies to aSyn and TH. Alpha-Syn expression in HT mice was enhanced compared to WT, and more so for KO mice. These elevated levels of aSyn were significantly reduced by treatment with AAV2-GDNF compared to phosphate-buffered saline (PBS). (B) Representative images showing IHC staining of TH+ neurons in mouse striatal sections. AAV2-GDNF treatment restored significant expression of TH in both HT and KO mice compared to PBS. Scale bar for (A) = 200 μm; for (B) = 2 mm.

(Miyagi et al., 1999), was also significantly reduced, showing similar deficit for the two members of the a-series of ganglio-type gangliosides. The b-series (GD1b, GT1b), on the other hand, although significantly different from controls, was less deficient than the a-series. Fig. 8C shows the combined results for each category of the ganglio-series. As with the HT mouse, such a-series deficiency might conceivably result over the years in incremental failure of sufficient GDNF signaling needed to maintain catecholaminergic neuron viability. Discussion The principal finding of this study is the essential role of GM1 ganglioside in regulating the neuroprotective signaling of GDNF, a function necessary for maintaining the viability of dopaminergic as well as other catecholaminergic neurons. This was exemplified in the reduced potency or complete loss of such signaling in cells or genetically altered mice deficient in GM1. Although such ganglioside changes included the entire ganglio-series, GM1 was accorded the primary regulatory role based on the restoration of GDNF signaling with LIGA20, a membrane permeable analog identical in carbohydrate and sphingosine structure to GM1. In-situ association of GM1 with the GDNF receptor complex was revealed by co-precipitation of GM1 with both pRET and GFRα1, as demonstrated by western blot analysis (Fig. 5A) and HPTLC (Fig. 5C). It was further indicative that immunoprecipitation of GFRα1 from the substantia nigra tissue of HT and KO mice resulted in progressively less co-precipitation of RET and pRET in correlation with reduced GM1 (Fig. 5A), despite no diminution of RET or GFRα1 in the mutant tissues (Fig. 5B). We view this as evidence for a requirement of GM1 for assembly of the GDNF receptor, now seen as a tripartite complex inclusive of GM1. GDNF signaling in the form of pRET and pMAPK formation was significant in Neuro2a cells overexpressing GFRα1, and in SH-SY5Y cells

induced to differentiate by retinoic acid with upregulation of GFRα1 and RET. Signaling was suppressed in both cell types upon partial GM1 reduction and, as tested in SH-SY5Y cells, this too was accompanied by no change in GFRα1 or RET levels (Fig. 3A). The application of GDNF to differentiated SH-SY5Y cells exposed to PDMP showed no reduction in GDNF binding (Fig. 3C), consistent with the conclusion that GM1 mediates GDNF receptor assembly rather than GDNF binding. Whereas the crystal structure of the GDNF binding site on GFRα1 has been reported (Leppänen et al., 2004), the binding sites of GM1 have yet to be determined. It was noteworthy that pRET was restored with LIGA20 in the presence of GDNF whereas LIGA20 alone induced no GDNF signaling (Figs. 1, 2). Further evidence for the coordinate function of GDNF and GM1 in vivo was provided by reduced pRET formation in nigrostriatal DA neurons of HT mice partially deficient in GM1, and more strikingly in KO mice completely devoid of GM1 (Fig. 4A, B). Both genotypes were previously shown to manifest multiple behavioral and neuropathological symptoms of PD including movement impairment, depletion of striatal DA, loss of DA neurons in the SNpc, and aggregation of aSyn. These symptoms were largely ameliorated by LIGA20 (Wu et al., 2011a, 2012), as were the pRET deficiencies in the present study (Fig. 4A, B). Importantly, pRET formation was also markedly reduced in SNpc DA neurons of PD subjects (Fig. 4C–E), the same tissues previously shown to express reduced GM1 and elevated aSyn aggregates (Wu et al., 2012). Of interest was the similar reduction of aSyn aggregates achieved by over-expression of GDNF via bilateral striatal infusion of AAV2-GDNF in both KO and HT mice (Fig. 6A). This treatment also restored TH staining to near-control levels in the striatal/midbrain region (Fig. 6B). The loss of TH-expressing neurons in the SNpc, another hallmark of PD, was previously demonstrated by unbiased stereology in both the HT and KO

186

P. Hadaczek et al. / Experimental Neurology 263 (2015) 177–189

Fig. 7. Motor performance and coordination of GM1-deficient mice treated with AAV2-GDNF. (A) Comparison of the three genotypes in terms of walk time on horizontal pole. KO mice showed significantly longer times than WT (one-way ANOVA with Bonferroni's post-hoc test). (B) Change in walk time at month 5 after treatment. The decrement (percentage change when compared with baseline) signified improvement (shorter walk time). The indicated p values are relative to PBS for same genotype: *b0.05, **b0.01 (two-way ANOVA with repeated measures and post-hoc Bonferroni's test). (C) Comparison of the three genotypes in terms of stride length. KO mice showed significantly shorter strides than WT. (D) Change in stride length after treatment at month 5. The incremental percentage change suggests improvement (longer strides).

mice, as was the LIGA20-induced restoration of TH staining in what were apparently dormant/dysfunctional DA neurons (Wu et al., 2012). It was demonstrated in PD patients that early loss of TH activity well precedes melanin-positive neuron loss (Kordower et al., 2013). Improved motor performance was also observed here by GDNF overexpression (Fig. 7). The fact that these improvements occurred in the GM1-negative KO mice suggests that excess GDNF is able to restore neuroprotective signaling even in the complete absence of GM1. The mechanism for this may relate to the demonstrated presence of RET-independent GDNF receptors (Trupp et al., 1997), including NCAM (Paratcha et al., 2003; Chao et al., 2003) and the heparin sulfate proteoglycan, syndecan-3 (Bespalov et al., 2011), in adult mammalian brain. An alternative or additional possibility is the substitution of other (non-ganglio-series) gangliosides for GM1, some of which show compensatory elevation in these genetically altered mice (Wu et al., 2001b); these, however, would be imperfect substitutes for GM1. Such mechanisms might also account for the remnant association of GFRα1 with RET and pRET observed in KO tissue despite complete absence of GM1 (Fig. 5A). Comparison of WT vs HT mice indicates that the constitutively low levels of GDNF in mature rodent brain (Laganiere et al., 2010) are sufficient to maintain effective neuroprotective signaling in the presence of normal GM1 but not in brain even partially deficient in this ganglioside. Robust pRET formation was restored with infused GDNF or membrane permeable analog of GM1. The pathognomonic role of aSyn has been well characterized (Braak et al., 2003) and its etiopathological role was recently highlighted in the demonstration that aSyn overexpression in DA neurons caused reduced expression of the transcription factor Nurr1 and its downstream target, RET (Decressac et al., 2012). As an additional consideration, since GM1

binds aSyn in a manner preventing fibrillation (Martinez et al., 2007; Bartels et al., 2014), overexpression of the latter could deplete the GM1 pool necessary to maintain all aSyn in helical non-fibrillating form and also mediate GDNF receptor cohesion. In that respect excessive sequestration of cellular GM1 by overexpressed aSyn may be functionally equivalent to underexpression of GM1. GDNF has been characterized as an indispensable neurotrophic factor for preserving the viability of catecholaminergic neurons of the adult mouse nervous system, as exemplified in the progressive death of such cells in a conditional GDNF-null mouse (Pascual et al., 2008). The pattern of cell death recapitulated many of the neuropathological hallmarks of PD, thereby suggesting a rationale for GDNF-based therapy. This was supported with rodent and primate models in which GDNF injected directly into brain afforded protection to DA nigrostriatal neurons as well as noradrenergic neurons in the locus coeruleus (Tomac et al., 1995; Gash et al., 1996; Choi-Lundberg et al., 1997; Rosenblad et al., 1998; Kordower et al., 2000; Akerud et al., 2001). Open-label clinical trials that followed gave promising results (Gill et al., 2003; Slevin et al., 2005; Patel et al., 2005), but a randomized, placebo-controlled study showed little efficacy and was halted partly because of safety concerns (Lang et al., 2006). Better results were obtained in clinical trials where greater attention was paid to GDNF distribution in the target putamen (Gill et al., 2003; Patel et al., 2013). The present study demonstrating the regulatory role of GM1 might account at least in part for such variability in terms of differing levels of intrinsic GM1, such levels possibly falling below that required for effective GDNF signaling in many if not most sporadic PD subjects. GM1 levels were shown to vary widely in normal human brain for persons of comparable age (Svennerholm et al., 1989, 1994). Our

P. Hadaczek et al. / Experimental Neurology 263 (2015) 177–189

187

Fig. 8. Comparison of ganglioside levels in occipital cortex of PD subjects vs controls. (A) HPTLC of representative samples of ganglio-series gangliosides extracted from the occipital cortex gray matter: two PD and two age-matched controls, plus three standards of bovine brain gangliosides of known composition for densitometric quantification. (B) Results for the individual 13 samples of PD and control; GM1 and GD1a were significantly lower in PD, whereas GD1b and GT1b showed less pronounced differences. (C) Results for combined a-series (GM1, GD1a) and b-series (GD1b, GT1b). Statistical calculations were made with Mann–Whitney rank sum test.

finding of subnormal levels of GM1 in the occipital cortices of PD subjects (Fig. 8), in conjunction with our previous observation of deficient GM1 in DA neurons as well as non-TH+ cells in the SNpc of PD subjects (Wu et al., 2012), suggests the intriguing possibility of systemic GM1 deficiency as a predisposing factor in PD. GM1 was shown to decline progressively with age in many regions of human brain and more so its metabolic precursor, GD1a (Svennerholm et al., 1989, 1994; Segler-Stahl et al., 1983; Kracun et al., 1992), thereby approximating over time the partial depletion engineered in the HT mouse (Wu et al., 2012). These findings accord with the idea of age as a primary risk factor for sporadic PD (Collier et al., 2011) and point to the HT mouse as a possibly unique model in reflecting the actual pathophysiology of sporadic PD. This proposal is reinforced by current findings that the HT mouse also shows gastrointestinal and cognitive dysfunctions characteristic of early and late PD, respectively (Wu, G. et al. in preparation). As with GDNF, GM1 has shown therapeutic potential in animal models and clinical trials. In a number of studies, GM1 partially restored depleted striatal DA while promoting neuron recovery in MPTP-treated mice (Hadjiconstantinou et al., 1986, 1989; Schneider et al., 1995) and primates (Schneider et al., 1992). LIGA20 proved superior to GM1 in murine models and was shown to have the important advantage of oral bioavailability (Schneider and DiStefano, 1995). This membranepermeable analog of GM1 proved effective in the present study in restoring GDNF signaling in GM1-deficient mice. A five-year openlabel study by Schneider et al. (Schneider et al., 2010) indicated that PD patients receiving GM1 acquired generally lower UPDRS motor scores and improved UPDRS Activities of Daily Living Scores than at baseline. This was followed by a randomized, controlled, delayed start trial that demonstrated GM1 superiority to placebo in reducing motor symptoms and slowing symptom progression over a two-year period

(Schneider et al., 2013). That even more benefit did not accrue in those trials may have been due to the limited ability of GM1 to enter the brain and, importantly, to enter the neurons themselves to mediate neuronal functions essential for long-term viability (see below). Precisely how GM1 and LIGA20 functioned in those animal and human studies is not well understood, though the present results suggest potentiation of GDNF through receptor association as a major factor. Among its several plasma membrane activities, GM1 is known to potentiate other neurotrophic factors, such as NGF (Mutoh et al., 1995) and BDNF (Pitto et al., 1998), through association with their tyrosine kinase receptors analogous to that shown here. GM1 also mediates certain critical intra-neuronal functions including regulation of nuclear calcium (Xie et al., 2002) and, of major significance to PD, maintenance of aSyn in a non-aggregating helical conformation (Martinez et al., 2007; Bartels et al., 2014). Gangliosides are essential for the formation of functional lipid rafts (Ohmi et al., 2012), and GM1 synthesis was shown necessary for the transport of TrkA to the cell surface (Mutoh et al., 2002). The present study revealing GM1 as necessary for association of GFRα1 and RET into a functional receptor indicate GM1 is itself a component of what we hypothesize is a tripartite GDNF receptor complex. LIGA20, owing to its membrane permeability, appears better able than extracellular GM1 to promote such assembly and to facilitate these additional intracellular functions. The above mentioned animal studies and clinical trials with GM1 were in keeping with reports that exogenous GM1 (and other gangliosides) exhibit multimodal neurotrophic activities in addition to the above mentioned neurotrophin receptor associations (Ledeen et al., 1998; Hadjiconstantinou and Neff, 1998; Mocchetti, 2005). These effects usually required relatively high concentrations of ganglioside and were often nonspecific in nature, as seen in the ability of five

188

P. Hadaczek et al. / Experimental Neurology 263 (2015) 177–189

other gangliosides besides GM1 to activate Trk when applied to intact striatal tissue (Duchemin et al., 2002). A recent report indicated that chronic GM1 application enhanced RET activity and phosphorylation in the striatum of MPTP-treated mice while promoting recovery of DA and DOPAC deficits (Newburn et al., 2014), consistent with the suggestion that applied GM1 may function as a GDNF mimetic. The experimental paradigm employing exogenous GM1, which may underlie the symptomatic improvements reported in PD clinical trials (Schneider et al., 2010, 2013), does not, in our opinion, reveal the functional role of GM1 in regard to normal GDNF signaling. The present study showing failure of such signaling and onset of Parkinsonism in animals with deficient GM1 points to the crucial nature of GM1 association with the GDNF receptor for effective neuroprotection. The ability of LIGA20 to restore a functional receptor complex is somewhat analogous to the restorative association of this analog with a nuclear Na+/Ca2 + exchanger in GM1-deficient mice (Wu et al., 2005), such association occurring intraneuronally as was also the case with the TrkA-GM1 association (Mutoh et al., 2002). The prolonged benefit of LIGA20 (Wu et al., 2012) compared to the transient effect of exogenous GM1 (Duchemin et al., 2002; Newburn et al., 2014) suggests that two fundamentally different mechanisms are at work. These findings suggest that a truly disease-modifying strategy for PD would necessarily be one that restores viability to dormantdysfunctional catecholaminergic neurons throughout the central and peripheral nervous systems. A direct route to that would be the revival of lost GDNF neuroprotective signaling, implemented by GM1 in accord with current results demonstrating receptor association. If, as appears to be the case, the GM1 deficiency we find in brains of PD subjects has the same parkinsonism-inducing effects as the GM1 deficiency in HT mice, the latter would appear to represent a true pathophysiological model of PD, the most faithful recapitulation of the human disease yet seen in an animal model. Moreover, treatments that rescued these mice, i.e. GDNF elevation, membrane-permeating analog of GM1 or combination thereof, might conceivably effect similar disease-modifying benefits in PD patients. Supplementary data to this article can be found online at http://dx. doi.org/10.1016/j.expneurol.2014.10.010. Acknowledgments This study was supported by NIH grant 2RO1 NS033912 to R.W.L. and a gift from from the Kinetics Foundation to K.B. References Akerud, P., Canals, J.M., Snyder, E.Y., Arenas, E., 2001. Neuroprotection through delivery of glial cell line-derived neurotrophic factor by neural stem cells in a mouse model of Parkinson's disease. J. Neurosci. 21, 8108–8118. Bartels, T., Kim, N.C., Luth, E.S., Selkoe, D.J., 2014. N−alpha−acetylation of alpha−synuclein increases its helical folding propensity, GM1 binding specificity and resistance to aggregation. PLOS ONE 9, e103727. Bespalov, M.M., Sidorova, Y.A., Tumova, S., Ahonen-Bishopp, A., Magalhaes, A.C., Kulesskiy, E., Paveliev, M., Rivera, C., Rauvala, H., Saarma, M., 2011. Heparan sulfate proteoglycan syndecan-3 is a novel receptor for GDNF, neurturin, and artemin. J. Cell Biol. 192, 153–169. Braak, H., Del Tredici, K., Rub, U., de Vos, R.A., Jansen Steur, E.N., Braak, E., 2003. Staging of brain pathology related to sporadic Parkinson's disease. Neurobiol. Aging 24, 197–211. Chao, C.C., Ma, Y.L., Chu, K.Y., Lee, E.H., 2003. Integrin αv and NCAM mediate the effects of GDNF on DA neuron survival, outgrowth, DA turnover and motor activity in rats. Neurobiol. Aging 24, 105–116. Choi-Lundberg, D.L., Lin, Q., Chang, Y.N., Chiang, Y.L., Hay, C.M., Mohajeri, H., Davidson, B.L., Bohn, M.C., 1997. Dopaminergic neurons protected from degeneration by GDNF gene therapy. Science 275, 838–841. Ciesielska, A., Mittermeyer, G., Hadaczek, P., Kells, A.P., Forsayeth, J., Bankiewicz, K.S., 2011. Anterograde axonal transport of AAV2-GDNF in rat basal ganglia. Mol. Ther. 19, 922–927. Collier, T.J., Kanaan, N.M., Kordower, J.H., 2011. Ageing as a primary risk factor for Parkinson's disease: evidence from studies of non-human primates. Nat. Rev. Neurosci. 12, 359–366. Crowder, R.J., Enomoto, H., Yang, M., Johnson Jr., E.M., Milbrandt, J., 2004. Dok-6, a Novel p62 Dok family member, promotes Ret-mediated neurite outgrowth. J. Biol. Chem. 279, 42072–42081.

Decressac, M., Kadkhodaei, B., Mattsson, B., Laguna, A., Perlmann, T., Björklund, A., 2012. α-Synuclein-induced down-regulation of Nurr1 disrupts GDNF signaling in nigral dopamine neurons. Sci. Transl. Med. 4, 163ra156. Duchemin, A.M., Ren, Q., Mo, L., Neff, N.H., Hadjiconstantinou, M., 2002. GM1 ganglioside induces phosphorylation and activation of Trk and Erk in brain. J. Neurochem. 81, 696–707. Eberling, J.L., Kells, A.P., Pivirotto, P., Beyer, J., Bringas, J., Federoff, H.J., Forsayeth, J., Bankiewicz, K.S., 2009. Functional effects of AAV2-GDNF on the dopaminergic nigrostriatal pathway in parkinsonian rhesus monkeys. Hum. Gene Ther. 20, 511–518. Eslamboli, A., Georgievska, B., Ridley, R.M., Baker, H.F., Muzyczka, N., Burger, C., Mandel, R.J., Annett, L., Kirik, D., 2005. Continuous low-level glial cell line-derived neurotrophic factor delivery using recombinant adeno-associated viral vectors provides neuroprotection and induces behavioral recovery in a primate model of Parkinson's disease. J. Neurosci. 25, 769–777. Gash, D.M., Zhang, Z., Ovadia, A., Cass, W.A., Yi, A., Simmerman, L., Russell, D., Martin, D., Lapchak, P.A., Collins, F., Hoffer, B.J., Gerhardt, G.A., 1996. Functional recovery in parkinsonian monkeys treated with GDNF. Nature 380, 252–255. Gill, S.S., Patel, N.K., Hotton, G.R., O'Sullivan, K., McCarter, R., Bunnage, M., Brooks, D.J., Svendsen, C.N., Heywood, P., 2003. Direct brain infusion of glial cell line-derived neurotrophic factor in Parkinson disease. Nat. Med. 9, 589–595. Gimenez, F., Krauze, M.T., Valles, F., Hadaczek, P., Bringas, J., Sharma, N., Forsayeth, J., Bankiewicz, K.S., 2011. Image-guided convection-enhanced delivery of GDNF protein into monkey putamen. Neuroimage 54, S189–S195. Hadjiconstantinou, M., Neff, N.H., 1998. GM1 ganglioside: in vivo and in vitro trophic actions on central neurotransmitter systems. J. Neurochem. 70, 1335–1345. Hadjiconstantinou, M., Rossetti, Z.L., Paxton, R.C., Neff, N.H., 1986. Administration of GM1 ganglioside restores the dopamine content in striatum after chronic treatment with MPTP. Neuropharmacology 25, 1075–1077. Hadjiconstantinou, M., Mariani, A.P., Neff, N.H., 1989. GM1 ganglioside-induced recovery of nigrostriatal dopaminergic neurons after MPTP: an immunohistochemical study. Brain Res. 484, 297–303. Johnston, L.C., Eberling, J., Pivirotto, P., Hadaczek, P., Federoff, H.J., Forsayeth, J., Bankiewicz, K.S., 2009. Clinically relevant effects of convection-enhanced delivery of AAV2-GDNF on the dopaminergic nigrostriatal pathway in aged rhesus monkeys. Hum. Gene Ther. 20, 497–510. Kells, A.P., Forsayeth, J., Bankiewicz, K.S., 2012. Glial-derived neurotrophic factor gene transfer for Parkinson's disease: anterograde distribution of AAV2 vectors in the primate brain. Neurobiol. Dis. 48, 228–235. Kordower, J.H., Emborg, M.E., Bloch, J., Ma, S.Y., Chu, Y., Leventhal, L., McBride, J., Chen, E.Y., Palfi, S., Roitberg, B.Z., Brown, W.D., Holden, J.E., Pyzalski, R., Taylor, M.D., Carvey, P., Ling, Z., Trono, D., Hantraye, P., Deglon, N., Aebischer, P., 2000. Neurodegeneration prevented by lentiviral vector delivery of GDNF in primate models of Parkinson's disease. Science 290, 767–773. Kordower, J.H., Olanow, C.W., Dodiya, H.B., Chu, Y., Beach, T.G., Adler, C.H., Halliday, G.M., Bartus, R.T., 2013. Disease duration and the integrity of the nigrostriatal system in Parkinson's disease. Brain 136, 2419–2431. Kracun, I., Rosner, H., Drnovsek, V., Vukelic, Z., Cosovic, C., Trbojevic-Cepe, M., Kubat, M., 1992. Gangliosides in the human brain development and aging. Neurochem. Int. 20, 421–431. Laganiere, J., Kells, A.P., Lai, J.T., Guschin, D., Paschon, D.E., Meng, X., Fong, L.K., Yu, Q., Rebar, E.J., Gregory, P.D., Bankiewicz, K.S., Forsayeth, J., Zhang, H.S., 2010. An engineered zinc finger protein activator of the endogenous glial cell linederived neurotrophic factor gene provides functional neuroprotection in a rat model of Parkinson's disease. J. Neurosci. 30, 16469–16474. Lang, A.E., Gill, S., Patel, N.K., Lozano, A., Nutt, J.G., Penn, R., Brooks, D.J., Hotton, G., Moro, E., Heywood, P., Brodsky, M.A., Burchiel, K., Kelly, P., Dalvi, A., Scott, B., Stacy, M., Turner, D., Wooten, V.G., Elias, W.J., Laws, E.R., Dhawan, V., Stoessl, A.J., Matcham, J., Coffey, R.J., Traub, M., 2006. Randomized controlled trial of intraputamenal glial cell line-derived neurotrophic factor infusion in Parkinson disease. Ann. Neurol. 59, 459–466. Ledeen, R.W., Wu, G., Lu, Z., Kozireski-Chubak, D.F., Fang, Y., 1998. The role of GM1 and other gangliosides in neuronal differentiation. Overview and new findings. Ann. N.Y. Acad. Sci. 845, 161–175. Lees, M.B., Paxman, S., 1972. Modification of the Lowry procedure for the analysis of proteolipid protein. Anal. Biochem. 47, 184–192. Leppänen, V.M., Bespalov, M.M., Runeberg-Roos, P., Puurand, U., Merits, A., Saarma, M., Goldman, A., 2004. The structure of GFRalpha1 domain 3 reveals new insights into GDNF binding and RET activation. EMBO J. 23, 1452–1462. Martinez, Z., Zhu, M., Han, S., Fink, A.L., 2007. GM1 specifically interacts with α-synuclein and inhibits fibrillation. Biochem. 46, 1868–1877. Matsushita, T., Elliger, S., Elliger, C., Podsakoff, G., Villarreal, L., Kurtzman, G.J., Iwaki, Y., Colosi, P., 1998. Adeno-associated virus vectors can be efficiently produced without helper virus. Gene Ther. 5, 938–945. Miyagi, T., Wada, T., Iwamatsu, A., Hata, K., Yoshikawa, Y., Tokuyama, S., Sawada, M., 1999. Molecular cloning and characterization of a plasma membrane-associated sialidase specific for gangliosides. J. Biol. Chem. 274, 5004–5011. Mocchetti, I., 2005. Exogenous gangliosides, neuronal plasticity and repair, and the neurotrophins. Cell. Mol. Life Sci. 62, 2283–2294. Mutoh, T., Tokuda, A., Miyadai, T., Hamaguchi, M., Fujiki, N., 1995. Ganglioside GM1 binds to the Trk protein and regulates receptor function. Proc. Natl. Acad. Sci. U. S. A. 92, 5087–5091. Mutoh, T., Hamano, T., Yano, S., Tokuda, A., Koga, M., Yamamoto, H., Furukawa, K., Ledeen, R.W., 2002. Stable transfection of GM1 synthase gene into GM1-deficient NG108-15 cells, CR-72 cells, rescues the responsiveness of Trk-neurotrophin receptor to its ligand, NGF. Neurochem. Res. 27, 801–806. Newburn, E.N., Duchemin, A.M., Neff, N.H., Hadjiconstantinou, M., 2014. GM1 ganglioside enhances Ret signaling in striatum. J. Neurochem. 130, 541–554.

P. Hadaczek et al. / Experimental Neurology 263 (2015) 177–189 Ohmi, Y., Ohkawa, Y., Yamauchi, Y., Tajima, O., Furukawa, K., Furukawa, K., 2012. Essential roles of gangliosides in the formation and maintenance of membrane microdomains in brain tissues. Neurochem. Res. 37, 1185–1191. Paratcha, G., Ledda, F., Ibanez, C.F., 2003. The neural cell adhesion molecule NCAM is an alternative signaling receptor for GDNF family ligands. Cell 113, 867–879. Pascual, A., Hidalgo-Figueroa, M., Piruat, J.I., Pintado, C.O., Gomez-Diaz, R., Lopez-Barneo, J., 2008. Absolute requirement of GDNF for adult catecholaminergic neuron survival. Nat. Neurosci. 11, 755–761. Patel, N.K., Bunnage, M., Plaha, P., Svendsen, C.N., Heywood, P., Gill, S.S., 2005. Intraputamenal infusion of glial cell line-derived neurotrophic factor in PD: a twoyear outcome study. Ann. Neurol. 57, 298–302. Patel, N.K., Pavese, N., Javed, S., Hotton, G.R., Brooks, D.J., Gill, S.S., 2013. Benefits of putaminal GDNF infusion in Parkinson disease are maintained after GDNF cessation. Neurology 81, 1176–1178. Pitto, M., Mutoh, T., Kuriyama, M., Ferraretto, A., Palestini, P., Masserini, M., 1998. Influence of endogenous GM1 ganglioside on TrkB activity, in cultured neurons. FEBS Lett. 439, 93–96. Radin, N.S., Shayman, J.A., Inokuchi, J.I., 1993. Metabolic effects of inhiting glucosylceramide synthesis with PDMP and other substances. Adv. Lipid Res. 26, 183–213. Rosenblad, C., Martinez-Serrano, A., Bjorklund, A., 1998. Intrastriatal glial cell line-derived neurotrophic factor promotes sprouting of spared nigrostriatal dopaminergic afferents and induces recovery of function in a rat model of Parkinson's disease. Neuroscience 82, 129–137. Schneider, J.S., DiStefano, L., 1995. Response of the damaged dopamine system to GM1 and semisynthetic gangliosides: effects of dose and extent of lesion. Neuropharmacology 34, 489–493. Schneider, J.S., Pope, A., Simpson, K., Taggart, J., Smith, M.G., DiStefano, L., 1992. Recovery from experimental Parkinsonism in primates with GM1 ganglioside treatment. Science 256, 843–846. Schneider, J.S., Kean, A., DiStefano, L., 1995. GM1 ganglioside rescues substantia nigra pars compacta neurons and increases dopamine synthesis in residual nigrostriatal dopaminergic neurons in MPTP-treated mice. J. Neurosci. Res. 42, 117–123. Schneider, J.S., Sendek, S., Daskalakis, C., Cambi, F., 2010. GM1 ganglioside in Parkinson's disease: results of a five year open study. J. Neurol. Sci. 292, 45–51. Schneider, J.S., Gollomp, S.M., Sendek, S., Colcher, A., Cambi, F., Du, W., 2013. A randomized, controlled, delayed start trial of GM1 ganglioside in treated Parkinson's disease patients. J. Neurol. Sci. 324, 140–148. Segler-Stahl, K., Webster, J.C., Brunngraber, E.G., 1983. Changes in the concentration and composition of human brain gangliosides with aging. Gerontology 29, 161–168. Slevin, J.T., Gerhardt, G.A., Smith, C.D., Gash, D.M., Kryscio, R., Young, B., 2005. Improvement of bilateral motor functions in patients with Parkinson disease through the unilateral intraputaminal infusion of glial cell line-derived neurotrophic factor. J. Neurosurg. 102, 216–222.

189

Svennerholm, L., Bostrom, K., Fredman, P., Mansson, J.E., Rosengren, B., Rynmark, B.M., 1989. Human brain gangliosides: developmental changes from early fetal stage to advanced age. Biochim. Biophys. Acta 1005, 109–117. Svennerholm, L., Bostrom, K., Jungbjer, B., Olsson, L., 1994. Membrane lipids of adult human brain: lipid composition of frontal and temporal lobe in subjects of age 20 to 100 years. J. Neurochem. 63, 1802–1811. Tansey, M.G., Baloh, R.H., Milbrandt, J., Johnson Jr., E.M., 2000. GFRalpha-mediated localization of RET to lipid rafts is required for effective downstream signaling, differentiation, and neuronal survival. Neuron 25, 611–623. Tomac, A., Lindqvist, E., Lin, L.F., Ogren, S.O., Young, D., Hoffer, B.J., Olson, L., 1995. Protection and repair of the nigrostriatal dopaminergic system by GDNF in vivo. Nature 373, 335–339. Trupp, M., Belluardo, N., Funakoshi, H., Ibanez, C.F., 1997. Complementary and overlapping expression of glial cell line-derived neurotrophic factor (GDNF), c-ret protooncogene, and GDNF receptor-alpha indicates multiple mechanisms of trophic actions in the adult rat CNS. J. Neurosci. 17, 3554–3567. Wright, J.F., Qu, G., Tang, C., Sommer, J.M., 2003. Recombinant adeno-associated virus: formulation challenges and strategies for a gene therapy vector. Curr. Opin. Drug Discovery Dev. 6, 174–178. Wu, G., Lu, Z., Xie, X., Li, L., Ledeen, R.W., 2001a. Mutant NG108-15 cells (NG-CR72) deficient in GM1 synthase respond aberrantly to axonogenic stimuli and are vulnerable to calcium-induced apoptosis: they are rescued with Liga-20. J. Neurochem. 76, 690–702. Wu, G., Xie, X., Lu, Z., Ledeen, R.W., 2001b. Cerebellar neurons lacking complex gangliosides degenerate in the presence of depolarizing levels of potassium. Proc. Natl. Acad. Sci. U. S. A. 98, 307–312. Wu, G., Lu, Z., Wang, J., Wang, Y., Xie, X., Meyenhofer, M.F., Ledeen, R.W., 2005. Enhanced susceptibility to kainate-induced seizures, neuronal apoptosis and death in mice lacking gangliotetraose gangliosides. Protection by LIGA 20, a permeant analog of GM1. J. Neurosci. 25, 11014–11022. Wu, G., Lu, Z.H., Kulkarni, N., Amin, R., Ledeen, R.W., 2011a. Mice lacking major brain gangliosides develop parkinsonism. Neurochem. Res. 36, 1707–1714. Wu, G., Lu, Z.H., Gabius, H.J., Ledeen, R.W., Bleich, D., 2011b. Ganglioside GM1 deficiency in effector T cells from NOD mice induces resistance to regulatory T-cell suppression. Diabetes 60, 2341–2349. Wu, G., Lu, Z.H., Kulkarni, N., Ledeen, R.W., 2012. Deficiency of ganglioside GM1 correlates with Parkinson's disease in mice and humans. J. Neurosci. Res. 90, 1997–2008. Xie, X., Wu, G., Lu, Z., Ledeen, R.W., 2002. Potentiation of a sodium–calcium exchanger in the nuclear envelope by nuclear GM1 ganglioside. J. Neurochem. 81, 1185–1195. Yamada, S., Nomura, T., Uebersax, L., Matsumoto, K., Fujita, S., Miyake, M., Miyake, J., 2007. Retinoic acid induces functional c-Ret tyrosine kinase in human neuroblastoma. Neuroreport 18, 359–363.