Serum antibodies against amyloid-β peptide (Aβ) in humans with or without diagnosis of Alzheimer's disease (AD) indicate the possibility of immune responses against brain antigens. In an unbiased screening for antibodies directed against brain proteins, we found in AD patients high serum levels of antibodies against the neuronal cytoskeletal protein ankyrin G (ankG); these correlated with slower rates of cognitive decline. Neuronal expression of ankG was higher in AD brains than in nondemented age-matched healthy control subjects. AnkG was present in exosomal vesicles, and it accumulated in β-amyloid plaques. Active immunization with ankG of arcAβ transgenic mice reduced brain β-amyloid pathology and increased brain levels of soluble Aβ42. AnkG immunization induced a reduction in β-amyloid pathology, also in Swedish transgenic mice. Anti-ankG monoclonal antibodies reduced Aβ-induced loss of dendritic spines in hippocampal ArcAβ organotypic cultures. Together, these data established a role for ankG in the human adaptive immune response against resident brain proteins, and they show that ankG immunization reduces brain β-amyloid and its related neuropathology.
Naturally occurring immune responses targeting brain resident proteins are associated with possible roles in modifying the disease courses in several neurodegenerative diseases including Alzheimer's disease (AD), Parkinson's disease, Huntington's disease and amyotrophic lateral sclerosis.1 The related immune responses are likely triggered by brain deposition of pathological protein aggregates with possible roles in neurodegeneration as indicated by naturally occurring antibodies against β-amyloid in both healthy human subjects and patients with AD.2, 3 Antibodies generated against pathological Aβ aggregates can be biologically active and participate in clearing β-amyloid deposits from affected brains, both in transgenic mouse models and in human patients.4 The clearance of amyloid plaques was followed by improved neuronal morphology and other signs of neuronal regeneration,5 suggesting that immunotherapy targeting pathological brain protein aggregates can be beneficial in the related neurodegenerative diseases. Ankyrin G (ankG) is part of the actin–spectrin cytoskeleton of neurons, forming an intracellular scaffolding protein, which directs and anchors proteins such as L1 and voltage-gated sodium channels to the axon initial segment.6 Recently ankG has been shown to be required for neurogenesis.7 Interestingly, scaffolding proteins like ankG have recently been identified as important key players in regulating signaling pathways, which control the immune response.8 ANK3, the gene encoding for ankG, maps to chromosome 10 within a region that has been identified as a susceptibility locus for late-onset AD.9, 10 Evidences for a genetic association of an ANK3 polymorphism with late-onset AD were found.11
In an unbiased exploratory screen of antibodies directed against brain proteins, we found in AD patients high serum levels of antibodies against ankG. AnkG expression was high in both cortical neurons in AD patients and in β-amyloid plaques, and active vaccination against ankG reduced β-amyloid plaque pathology.
Materials and methods
AD patients and healthy control subjects (HCS) were recruited at the gerontopsychiatric department of the University of Zurich. AD cases used for sera analysis had an age at onset ⩾65 years and a diagnosis of definite or probable AD, according to the National Institute of Neurological and Communication Disorders and Stroke-Alzheimer's Disease and Related Disorders Association diagnostic criteria. HCS were matched for age and sex. Cognitive assessment was recorded using mini-mental score evaluation (MMSE) at baseline. All subjects were followed up for a period of up to 28 months with MMSE assessment at end point. The rate of cognitive decline for each AD group was based on the average slope of MMSE weighted for number of follow-up months. In each group there was one subject drop out. Human brains were obtained from the Kathleen Price Bryan Brain Bank from Duke University (USA). AD classification was assessed according to The Consortium to Establish a Registry for Alzheimer's Disease (CERAD) guidelines.
Tissue micro arrays
Immunohistochemical quantification of ankG expression was performed by tissue micro array analysis from 288 tissue samples, each from AD and HCS (n=24). Primary antibodies were visualized using either a standard diaminobenzidine horseradish-peroxidase method for Aβ (Histofine simple Stain Max PO, Nichirei Bioscience, Nichirei Inc., Tokyo, Japan) or alkaline phosphatase method for ankG staining (Histofine simple Stain Max AP, Nichirei). AnkG staining was assessed in seven grades determined by the sum of a score for the percentage of immunopositive cells (0=none, 1=1-10%, 2=10–50%, 3⩾50%), and the intensity of the staining (0=none, 1=weak, 2=moderate, 3=strong, 4=very strong).
ArcAβ mice were produced as described previously by Knobloch et al.18 Nontransgenic littermates and APP-deficient mice were used as controls when specified. All animals were housed 3–5 per cage and had access to food and water ad libitum, under a 12-h light and dark cycle. All in vivo experiments were performed with the approval of the Swiss veterinary council (BVET).
Anesthesia and perfusion
Mice were anesthetized using a cocktail of 1.25% ketamine and 0.25% xylazine. Blood was collected from the right ventricle using a syringe with EDTA and samples were centrifuged at 8 000 rpm for 8 min at 4 °C, and plasma extracted using a sterile pipette and stored at −80 °C for later batch analysis. After transcardial perfusion with ice-cold buffer PBS (10 mM Na2HPO4, 2.5 mM NaH2PO4, 150 mM NaCl, 3 mM KCl), one hemisphere was processed biochemically and the other hemisphere incubated in paraformaldehyde, 4% in PBS for 4 h at 4 °C followed by incubation in sucrose, 30% in PBS for 24–36 h for immunohistochemistry purposes. Paraformaldehyde-fixed and cryoprotected hemibrains were cut in 35 μm thick coronal slices at −80 °C using a microtome (Leica Jung HN40) and kept at −20 °C in an antifreeze solution (phosphate buffer 0.50 M in MilliQ water: ethyleneglycol: glycerol=1.3:1:1) until staining was performed.
Active ankG immunization
ArcAβ mice or Tg2576 mice and their nontransgenic littermates were actively immunized with either 10 μg ankG protein in PBS+Freund's adjuvant (FA) complete or PBS+FA (control-immunized). Blood was drawn from the tail of each mouse before and during (every 4 weeks) immunization to obtain plasma for antibody titer analysis by 8 000 rpm centrifugation for 8 min at 4 °C. Immunization cocktails were injected subcutaneously between the scapulae. Three 4-weekly immunization re-boosts were performed with FA incomplete instead of FA. At the end of the 3-month immunization period the mice were perfused. One hemisphere of the brain was fixed in paraformaldehyde solution for 4 h at 4 °C and subsequently put in a solution of 30% sucrose at 4 °C for cryoprotection for 24–36 h for immunohistochemistry. The other hemisphere of the brain was processed for biochemical analysis.
A battery of well-validated and carefully controlled tests was used to behaviorally assess mice for motoric and cognitive performance.18 At the time of testing, mice were weighed and examined for general health measures to ensure that the mice were physically able to conduct the cognitive-behavioral test and to rule out any adverse side effects due to the ankG immunization.
Spatial working memory was assessed in mice using the Y-maze (Y-shaped plastic maze, with 40 × 20 × 10 cm arm sizes). During a 5 min trial, the sequence of arm entries was recorded using the ANY-maze Video Tracking System (Stoelting, Wood Dale, IL, USA). The percentage alternation was calculated as the ratio of actual to possible alternations (defined as the total number of arm entries −2) × 100%. Data were analyzed using Prism statistical program. The percentage of alternations was analyzed using repeated measures analysis of variances (ANOVAs).
Mini neurological examination was performed before, during and after each immunization. In brief, all animals had normal fur appearance, no secretory signs, normal body postures and normal basic reflex, including the eye blink, pupillary, flexion and righting reflexes. All animals had age-appropriate body weight and no difference in muscular strength (grip strength) as measured with a spring scale.
Human embryonic kidney cells (HEK 293, DSMZ, ACC 305) were cultured in Dulbecco's modified eagle medium (DMEM, Invitrogen #52100039; Basel, Switzerland) supplemented with 10% fetal calf serum (FCS) and penicillin/streptomycin (PenStrep, Invitrogen #10378–016) at 37 °C, 5% CO2, 95% humidity. SHY-5Y neuroblastoma cells were grown in DMEM/F12 supplemented with 20% fetal bovine serum. For differentiation, cells were sequentially treated for 5 days with 5 μM retinoic acid (Sigma, Buchs, Switzerland) and for 5 days with 20 ng ml−1 BDNF (Peprotech, London, UK), with serum reduced to 2%, after which they were transfected identically to HEK293 cells. HeLa cells stably expressing Swedish APP were a kind gift from Professor Gang Yu, Dallas, USA.
Silencing of the ankG gene (Validated siRNA cat. n. SI02780204, Qiagen AG, Switzerland), ankB gene (Validated siRNA cat. n. SI03031238, Qiagen AG) and negative control small interfering RNA (siRNA) Alexa Fluor 488 (cat. n. 1022563), Qiagen AG, Hombrechtikon, Switzerland) in differentiated SHY-5Y cell, HEK293, Citrine-HEK293, Swedish APP HEK293 and Swedish APP Hela HeLa cells was carried out using the RNAi kit and RNAi nucleotides as per the manufacturer's instructions (Qiagen). Silencing was repeated every 24 hours for three days, followed by immunoblotting analysis. To inhibit γ-secretase activity, cells were cultured in the presence of 1 nM DAPT (Sigma–Aldrich, Switzerland).
Protein antigen arrays screening
The protein antigen arrays screening was custom made by imaGenes GmbH, Berlin, Germany, containing 37 000 clones from a human fetal brain complementary DNA (cDNA) expression library as described by Cepok et al.12 Briefly, the protein arrays were utilized for high-throughput antibody screening and each clone was spotted in duplicates onto PVDF filter membranes (two parts, each 26 × 26 cm in size). For screening with sera antibodies, the protein array filters were first incubated with blocking buffer (3% non-fat milk in Tris-buffered saline (TBS)) for 2 h and then incubated with the patient's sera (dilution 1:1000) for 14 h. After extensive washing with TBS containing 0.05% Tween-20 and 0.5% Triton X-100 (TBS-T), filters were incubated with HRP-conjugated anti-human IgG (1:2000 diluted in blocking buffer) for 1 h. After several washings, the filters were developed by stabilized tetramethylbenzidine-blotting (TMB-blotting) substrate (Pierce, Rockford, IL, USA) for 5 min. Positive spots in duplicates indicated specific binding of sera antibodies to recombinant proteins. The corresponding expression clones were obtained from the RZPD and were cultured in LB Broth Base medium (Invitrogen) supplemented with ampicillin. Their plasmids were isolated, and the cDNA inserts were sequenced for identification of the proteins. Immunoreactivity was assigned with the following score: (0=none, 1=weak, 2=moderate, 3=strong, 4=very strong). The strongest immunoreactive clones encoding for proteins expressed in the correct reading frame and that were immunopositive in at least two different sera were considered as positive hits.
Immunohistochemical images were acquired on a Leica DM4000B microscope using an Olympus DP71 colour digital camera and newCAST software (Visiopharm, Copenhagen, Denmark). Image analysis was carried out with ImageJ image analysis software. Quantification of the number of β-amyloid plaques and Iba-1 positive microglia was performed using the area measurement tool. For measurement of plaque diameter, the particle count tool was used. The measurements were done on 10 sequential sections per mouse.
Western blot (WB) analysis
Densitometric analyses of the WBs were done using ImageJ (Scion Corporation, Frederick, National Institutes of Health, Bethesda, MD, USA) or Tina Image analysis software (University of Manchester, UK). Mean optical intensities were plotted after standardization of intensities to loading controls.
Data analysis, statistical evaluation and graphical representation of data were done using GraphPad Prism software. ANOVAs, t-test or χ2-test was used to determine significance between the groups.
Biochemical enrichment of Aβ
Frozen brain samples were homogenized using a sequential extraction protocol in which tissue was homogenized using a glass-Teflon homogenizer (20 strokes, 315 rpm) in buffer containing 100 mM Tris-HCl, 150 mM NaCl, 2% SDS and complete protease inhibitor (Roche, Indianapolis, IN, USA), centrifuged at 100 000 g at 8 °C for 1 h, and supernatant extracted and stored at −80 °C for later biochemical batch analysis. The remaining pellet was resuspended in 70% formic acid, sonicated once for 30 s at 30% power, ultracentrifuged (100 000 g at 4 °C for 30 min), supernatant extracted, lyophilized, reconstituted in buffer containing 100 mM Tris-HCl, 150 mM NaCl and complete protease inhibitor (Roche), stored at −80 °C for later batch analysis. The protein concentration of each fraction was quantified by BCA assay kit (Pierce) and subjected to immunoblotting and Aβ ELISA. All buffers were used at 4 °C with protease inhibitors (Roche Applied Science, Indianapolis, IN, USA).
Exosomes were isolated from HEK 293 and N2a cells. Briefly, cells from 3–6 T175 flasks were cultured in DMEM with 10% fetal calf serum. One day before the exosome preparation, culture medium was replaced with AIM-V medium. Culture supernatants of cells grown for 24 h in AIM-V medium were collected and spun at 300 g for 10 min to remove cells. The supernatants were then sequentially centrifuged at 1200 g for 20 min, 10 000 g for 30 min, and 100 000 g for 1 h. The 100 000 g pellet was washed with phosphate buffer saline (PBS), and again spun at 100 000 g for 1 h. The second 100 000 g pellet (exosomal pellet) was resuspended in PBS. The pelleted fractions were then used for immunoblotting.
Equal amounts of total protein were subjected to separation on 10-20% Tricine gels (Invitrogen), blotted on nitrocellulose membranes (0.45 μm; Biorad, Reinach, Switzerland) or Protran BA 79 cellulose-nitrate membranes (0.1 μm; Schleicher & Schuell, Dassel, Germany). The immunoblot was then incubated with primary antibodies (Supplementary Table S3) followed by incubation with HRP-tagged secondary antibodies. Detection was performed using chemiluminescence, visualized using ECL WB reagents (Amersham Pharmacia, GE, Germany) or SuperSignal West Dura Extended Duration reagents (Pierce) on BIOMAX films (GE, Germany).
Brain homogenates were lysed for 1 h with lysis buffer, pH 7.5, containing 50 mM Tris-HCl, 150 mM NaCl, 1 % Nonidet P-40, 1 mM Na2P2O7, 1 mM NaF, 2 mM Na3VO4, 0.1 mM PMSF, 2 mM EDTA, and EDTA-free protease inhibitor cocktail (Roche Diagnostics). Lysates were precleared with non-specific IgG antibodies followed by protein A/G coated magnabeads (Invitrogen, Basel, Switzerland) for 1 h each. The precleared lysate was then incubated with indicated primary antibodies or non-specific IgG for 1 h. Antibody complexes were then precipitated with beads applied for 1 h. Beads were washed five times with lysis buffer and once with PBS, and used for WB immunoblot analysis. All steps were carried out at 4 °C. For every 1 mg ml−1 of total protein content of homogenate 3 μg of antibodies were used.
Serum antibody titers
Blood samples were spun down at 3000 g for 10 min at 400 °C to obtain sera for titer analysis. Anti-ankG antibodies in the serum of immunized mice were determined by ELISA techniques. Briefly, 0.5 μg ml−1 rat recombinant ankG (or 2 μg ml−1 BSA) in PBS was immobilized on the polyvinylchloride surface overnight at 4 °C. Wells were washed five times for 5 min at room temperature (RT), with PBS containing 0.05% Tween-20 (PBS-T) followed by blocking for 1 h at RT with 2% BSA in PBS, and subsequently incubating with immunized sera diluted 1:500 in PBS-T containing 3 % BSA for 2 h at RT. Wells were washed and bound-serum antibodies were detected with HRP-conjugated anti-mouse antibodies. The reaction was developed with 0.1% ABTS (Roche) in 100 mM acetate buffer, pH 5.0. The reaction was stopped with 100 mM sodium fluoride. Optical density (OD) was measured at 450 nm.
Protein ligand binding assay
APP CTF-50 peptide (Calbiochem, Billerica, MA, USA), Aβ40 and Aβ42 (Bachem AG, Weil am Rhein, Germany) were coated on 96-well polyvinyl chloride plates at 0.5 μg ml−1 in PBS at 4 °C overnight. Wells were then blocked for 1.5 h with PBS containing 2% BSA and incubated for 2 h at RT with rat recombinant ankG (0-2.5 μg ml−1) diluted in PBS-T. Washing steps were carried out between all incubations using PBS-T. Plates were incubated for 1.5 h at RT with goat antibodies against ankG in PBS-T containing 2% BSA. Detection was as described before for serum titer analysis.
ELISA quantification of Aβ levels
Concentrations of Aβ40 and Aβ42 from mice serum and SDS insoluble brain fractions were quantified using commercially available ELISA kits (The Genetics Company, Schlieren, Switzerland) as per manufacturer's instructions. Sera were diluted 1:100 in PBS before use. Two to four micrograms at 25 μg ml−1 of total protein of SDS insoluble fractions in PBS were used per well.
Fixed and cryoprotected hemibrains were cut in 30 μm thick slices at ∼−80 °C using a microtome (Leica Jung HN40) and kept at −20 °C in an antifreeze solution (phosphate buffer 0.50 M in MilliQ water: ethyleneglycol: glycerol=1.3:1:1) until staining was performed. All immunohistochemical stainings were executed using the free-floating method. Washing steps were carried out between all incubations using washing buffer (TBS pH 7.4 containing 0.2% Triton X-100) at RT. Antigen retrieval was performed when required (Supplementary Table S3), by incubating the slices in 1 M HCl at 65 °C for 30 min. Slices were blocked for 1 h at RT using blocking buffer (5.0% goat serum and 5.0% donkey serum in washing buffer). Blocked slices were incubated overnight at 4 °C with slight agitation in primary antibody incubation buffer (2.5% goat serum and 2.5% donkey serum in washing buffer). This was followed by secondary antibody incubations carried out for 2 h at RT. Slices were washed in washing buffer, mounted on chrom-gelatin-coated microscopy slides (Super-frost-plus, Menzel, Braunschweig, Germany) and glass-covered using Hydromount® (National Diagnostics, Hull, UK).
Organotypic hippocampal slice cultures and Sindbis virus infection
Organotypic hippocampal slice cultures were prepared and cultured according to standard protocols. Briefly, 6- to 7-day-old arcAβ transgenic and nontransgenic mice were decapitated, brains were removed, both hippocampi isolated and cut into 400 μm thick slices using tissue chopper. Slices were cultured on Millicell culture plate inserts (0.4 μm, Millipore, Bedford, MA, USA) in 6-well plates containing 1 ml culture medium (46% minimum essential medium eagle with HEPES modification, 25% basal medium with earls modification, 25% heat-inactivated horse serum, 2 mM glutamine, 0.6% glucose, pH7.2). Culture plates were kept at 37 °C in a humidified atmosphere containing 5% CO2. Slices were kept in culture for 12 days before start of the experiments. Culture medium was exchanged every second to third day. On day 11, culture medium was replaced by low-serum Nb–N1 medium (94.5% Neurobasal medium, 0.5% heat-inactivated horse serum, 2 mM glutamine, 0.6% glucose, 1 × N1 supplement, pH 7.2). On day 12, in vitro slice cultures were infected with Sindbis virus expressing EGFP using droplet method. For spine analysis, cultures were fixed at day 3 postinfection within 6-well plates. Slices were left attached to the culture plate membrane to preserve hippocampal structure and rinsed with PBS. Slices were then fixed with 4% paraformaldehyde in PBS containing 4% sucrose for 2 h at 4 °C. After washing with PBS, cultures were mounted and coverslipped. For analysis of effects of anti-ankG antibodies, cultures were treated with 10μl antibody per 1 ml culture medium for 1 week.
Confocal imaging of fixed hippocampal slice cultures and analysis of spine density
Confocal high-resolution imaging of spines was performed using TCS/SP2 Leica confocal laser scanning microscope with × 63 objective (oil, numerical aperture: 1.4). Fragments of apical and basal dendrites of hippocampal CA1 and CA3 pyramidal neurons were imaged with voxel size of 0.058 × 0.058 × 0.25 μm in x–y–z direction. To determine spine density, the length of the dendrite was measured and spines were counted as protrusions in x- and y- axes using NIH ImageJ software.
Immune response against the cytoskeletal adaptor protein ankG in patients with AD correlates with slower cognitive decline
To explore the existence of an endogenous immune response against brain proteins in AD, we investigated the spectrum of antibodies directed against brain antigens in sera of AD patients. To this aim, we used unbiased protein arrays generated from a human fetal brain cDNA expression library comprising 37 000 clones.12, 13, 14 In addition to ubiquitously expressed genes, this library contained a subset of proteins that are predominantly expressed in brain tissue. Sera from 15 AD patients and 10 nondemented age-matched HCS were applied to separate protein arrays (Figure 1a). A maximum of 15 immunoreactive cDNA expression products were identified with serum samples of each subject. Sequencing of immunopositive clones identified ankG, a key protein in neuronal cytoskeletal organization that is encoded by a gene mapping to chromosome 10 within a susceptibility locus for AD.10, 15 WB analyses using purified rat ankG were performed with an independent set of sera from 31 AD patients and 33 HCS (Supplementary information, Table S1) to test the hypothesis of AD-related immunoreactivity against ankG (Figure 1b). AnkG-reactive IgGs were found in 54.84 % of the AD sera and 24.24% of the HCS sera (P=0.01, χ2-test). No IgMs against ankG were found in the same sera. The ankG-reactive IgGs were subtype-selective, indicated by the absence of cross-reactivity with the ankG homolog protein ankyrinB,16 (data not shown). AD patients, immunopositive for ankG IgGs, displayed stable or even improved cognitive functions over a time period of 2 years, as measured with MMSE, when compared with the immunonegative AD patients individually (Figure 1d, e and h). The number of stable or improved patients was significantly higher in the immunopositive group than in the immunonegative group (P=0,03, χ2-test). We found that the change in MMSE score (from baseline, t0, to follow-up, tf) was significantly different between immunopositive and immunonegative AD patients (Figure 1f P=0,0075, Student's t-test); importantly, this was not owing to a difference in observation times, as the months at follow-up did not significantly differ (P=0,6, Figure 1h). The analysis of MMSE decline weighed for the time of follow-up confirmed that the presence of ankG-reactive IgGs was accompanied by a reduced rate of cognitive decline (Figure 1g and h, P=0,0075, Student's t-test). Taken together, our results suggest that the presence of ankG antibodies in sera from AD patients might represent a protective mechanism that slows down the progression of the disease.
AnkG is overexpressed in AD brains and is redistributed from intraneuronal compartments to extracellular β-amyloid plaques
We first applied immunohistochemical methods to study the distribution of ankG in AD compared with HCS. As expected, ankG was found to be present in neuronal cell bodies and axons in brains obtained both from AD patients and HCS.17 In AD brains, however, ankG was specifically associated with β-amyloid plaques, and was localized on structures reminiscent of dystrophic neurites (Figure 2a, arrows). Quantitative immunohistochemistry showed overall higher cortical levels of ankG in AD as compared with HCS (Figure 2b, P<0.01, Student's t test, n=24).
A similar pattern of ankG immunostaining was present in arcAβ transgenic mice expressing human APP containing both the familial AD-causing ‘Swedish’ and ‘Arctic’ mutations18 (Figure 2c). AnkG localized with β-amyloid deposits in arcAβ mice at 12 months (Figure 2d, lower panels), whereas very little immunoreactivity was observed in 6-month-old animals, when the brains are barely affected by β-amyloid pathology (Figure 2c, upper panel). The redistribution of ankG toward the neuropil and the extracellular plaques, therefore accompanied the progression of the disease. The staining patterns of the neuronal marker neurofilament showed intact gross hippocampal architecture in these mice (Supplementary Figure 1a), suggesting that the redistribution of ankG from the neuronal cytoplasm to β-amyloid deposits is specifically related to β-amyloid pathology rather than to more general effects of either neurodegeneration or aging on neuronal architecture.
Similarly, WB of brain protein extracts prepared from human cortical samples indicated higher levels of both ankG and Aβ in AD as compared with HCS (Figures 2d and e), confirming our quantitative immunohistochemical data. In AD samples, ankG was detected together with insoluble Aβ in SDS-insoluble fractions (Figure 2e). In order to investigate how a cytosolic protein such as ankG could be released and presented to the immune system, we investigated whether exosomes could be involved in the mechanism of ankG release. Exosomes are small membrane vesicles involved in trafficking, cell-to-cell communication and immune response activation.19, 20, 21 Several cytosolic proteins, including heat-shock proteins, α-synuclein and caspase-1, have been shown to be released from cells through exosomes. Exosomes purified from media of HEK293 cells contained ankG, as indicated by WB (Figure 2f, left panel). Similarly, we found ankG enrichment in exosomal preparations from neuroblastoma cell line N2a (Figure 2f, right panel). Interestingly, the neuron-like N2a cells displayed an enrichment of a higher isoform of ankG in exosomes (around 210 kDa), which is the most abundant isoform observed in brains (Figure 2d,e). This finding suggested a mechanism through which the cytosolic ankG could be released from neurons, accumulate in β-amyloid plaques in the neuropil, and possibly be transferred to antigen-presenting cells of the immune system. Together these results show that brain levels of ankG increase with the development of brain β-amyloid pathology, redistribute to extracellular space through exosomal release and elicit a humoral immune response.
Active immunization with ankG reduces β-amyloid pathology in vivo
To determine the role of an immune response against ankG in β-amyloid brain pathology, we actively vaccinated 14-week-old arcAβ mice and their wild-type littermates with ankG or control. The mice were vaccinated monthly for 3 months with 10 μg recombinant ankG protein in FA or PBS in FA as control. ELISA of plasma from ankG-immunized mice showed monthly increasing titers of ankG-reactive IgGs, which was not found in control mice (Supplementary Figure 1b). One month after the last immunization mice were sacrificed. Immunostaining of ankG-immunized arcAβ mice showed IgGs binding within β-amyloid plaques, with no such staining in control mice, suggesting that anti-ankG antibodies migrated from the blood across the blood-brain barrier and bound to ankG localized on β-amyloid plaques (Supplementary Figure 1c). Quantitative histological analyses showed that active ankG immunization significantly reduced both the size and the number of β-amyloid plaques in arcAβ mice (Figure 3a-c).
Both WB and ELISA assays confirmed lower concentrations of SDS insoluble Aβ levels in ankG-immunized arcAβ mice as compared with controls (Figure 3d,e and Figure 4a). In contrast to the reduced SDS-insoluble Aβ pool, the SDS-soluble Aβ42 fraction was significantly higher after ankG immunization (Figure 4b). These findings suggest that ankG immunization induce disaggregation of SDS-insoluble β-amyloid material with decreased insoluble Aβ40 and Aβ42 in the formic acid fraction but followed by increased Aβ42 released into the soluble fraction.
In contrast to the strong increase in ankG antibody titers, ankG-immunized mice produced no antibodies against Aβ, excluding the possibility that a cross-immune reaction against Aβ was related to the Aβ clearance from arcAβ-mice brains (Supplementary Figure 2a).
To explore the role of microglia in the clearance of β-amyloid plaques by phagocytosis,22 we determined whether ankG immunization was associated with the uptake of ankG and Aβ by microglia (Figure 3a). Triple immunostainings for Iba1, a marker for microglia, Aβ and ankG indicated the presence of ankG within microglia together with Aβ-positive material in both ankG-immunized and control-treated arcAβ mice, suggestive of phagocytosis of both ankG and Aβ in the presence and the absence of ankG-reactive antibodies (Figure 3a). In ankG-immunized mice, however, the density of microglia was significantly higher within the β-amyloid plaques as compared with controls, suggesting that ankG-reactive antibodies bound to β-amyloid plaques guided microglia to ankG-containing β-amyloid plaques, presumably via interaction of ankG IgGs with FcRs known to be present at the surface of microglia (Figure 3a, Supplementary Figure 2b).23 Together, our results suggest that ankG immunization promotes the interaction of microglia with brain β-amyloid, possibly resulting in accelerated clearance of β-amyloid plaques.
To confirm the efficacy of ankG immunization in clearing brain β-amyloid in a second mouse model, and to exclude that the effect of ankG immunization was related to the ‘Arctic’ mutation at position 22 within the Aβ domain, we immunized 4-month-old Tg2576 mice expressing human APP with the ‘Swedish’ mutation24 with the identical immunization protocol. The mice were sacrificed at 7 months of age, an age at which only intraneuronal Aβ is usually seen in this mouse model.18 ELISA of plasma of ankG-immunized mice showed monthly increasing titers of ankG-reactive IgG only in 4 out of 10 immunized mice (data not shown). AnkG immunization reduced brain levels of insoluble Aβ42 (Supplementary Figure 1d) with no significant reduction of Aβ40 (Supplementary Figure 1e). In these mice, ankG immunization was associated with a plasma spike of Aβ42, as previously observed with some forms of Aβ immunotherapy that cleared soluble Aβ from the brain into the plasma via a so-called peripheral ‘sink’ mechanism25 (Supplementary Figure 1f).
Finally, we addressed whether ankG immunization influences memory and spatial reference learning in arcAβ mice. Mice were subjected to Y-maze testing to measure hippocampal-dependent memory; no significant difference in performance between ankG-immunized and nonimmunized mice was found. Likewise, there was no difference in behaviour in ankG-immunized wild-type mice (Supplementary Figure 3a,b), suggesting that ankG immunization was neither neurotoxic nor interfering with the known physiological functions of ankG in neurotransmission. Taken together with the absence of ankG antibodies in neuronal cell bodies and axons (Supplementary Figure 1c), these findings support the idea that ankG-reactive antibodies preferentially target extracellular ankG including ankG bound to β-amyloid plaques rather than intraneuronal ankG.
Antibodies against ankG reduce insoluble Aβ and attenuate Aβ-induced spine loss in arcAβ organotypic hippocampal slices
To directly demonstrate the activities of ankG-reactive antibodies in the rescue of neuronal morphology, we produced mouse monoclonal antibodies against ankG by using hybridoma cell fusions.26 Monoclonal antibody A (mAbA), with the strongest combined immunoreactivity against ankG on WB (Supplementary Figure 2c), was added for 7 days to organotypic hippocampal slice cultures from arcAβ mice. ELISA of medium showed that mAbA significantly reduced both Aβ40 and Aβ42 levels in the hippocampal slice cultures to <50% of untreated slices (Figure 5a,b). A control mouse monoclonal antibody against a bovine-herpes-virus-1 that is not present in mouse brain had no effect on Aβ levels, indicating the absence of non-specific IgG activity on Aβ levels in the slices (data not shown). To determine the role of ankG antibodies on the known Aβ-induced loss of dendritic spines in hippocampal neurons, we expressed EGFP in the organotypic hippocampal slice cultures by using Sindbis virus-mediated neuronal expression to allow for the detailed visualization of neuronal morphology including dendritic spines.27 As expected, hippocampal cultures prepared from arcAβ mice showed significantly reduced spine density as compared with hippocampal slices prepared from wild-type, nontransgenic littermates (Figure 5c,d). Treatment with mAbA almost completely rescued spine loss in the arcAβ hippocampal cultures (Figure 5c,d). This activity was unrelated to direct binding of mAbA to Aβ40 and Aβ42, which was shown to be absent by ELISA. mAbA had no effect on spine density in slices prepared from nontransgenic littermates, supporting the conclusion that the rescue effect on spine density was related to transgene-induced, Aβ-related neurotoxicity (Figure 5c,d).
AnkG is a cytoplasmic adaptor protein involved in the anchoring and assembly of voltage-gated ion channels in the axonal initial segment of neurons throughout the brain.28, 29 To the best of our knowledge, this is the first report documenting the involvement of ankG and ankG antibodies in AD. The results of our study establish that ankG can function in human subjects as a self-antigen triggering a humoral immune response with the generation of ankG-reactive IgG antibodies, and they point to a novel role of ankG in the neuropathology of AD.
The frequency of ankG immunoreactive sera as well as the ankG expression in frontal cortex was higher in patients with AD as compared with HCS, indicating that the ankG immunoresponse is specifically linked to the AD neuropathology (Figure 1). However, the rate of cognitive decline in AD was faster in the absence of ankG-reactive antibodies, suggesting that the presence of sera ankG antibodies can have a protective effect on the mental function of AD patients. AD patients immunonegative for ankG lost an average of five points in the MMSE evaluation over a period of two years, as expected and in line with previous observations.30 In contrast, ankG immunopositive patients were mostly cognitively stable over the same period, and some even improved their MMSE score (Figure 1d-h). Although these results warrant further confirmation, they suggest that endogenous ankG-reactive antibodies can slow the disease course of AD or even modify it.
AnkG is known to reside in the cytoplasm in physiological conditions, hence in a compartment hardly reachable by extracellular antibodies. It was therefore somehow counterintuitive that AD patients mounted an immune response directed toward an intracellular protein. We therefore studied the localization of ankG in AD brains. This analysis allowed us to discover that cortical levels of ankG are significantly higher in AD than in HCS, accompanied by a redistribution of this protein to extracellular β amyloid plaques in AD brains and in brains from APP Tg mouse models (Figure 2a-c). The partial re-localization of ankG from the neuronal cytoplasm to β-amyloid plaques in the neuropil of AD brains is possibly mediated through exosomal release, as suggested by the presence of ankG in exosomes. Interestingly, we were able to show ankG enrichment in exosomal preparations from HEK293 cells as well as from neuroblastoma, neuron-like cells. The latter appeared to specifically accumulate a high molecular weight ankG isoform (ca. 200 kDa) in the exosomes. As this is the isoform that is most abundantly found in brain homogenates (compare with Figure 2d,e), these results suggest that exosomal ankG transport might be a neuronal-specific event. Exosomes are small membrane-bound vesicles (50–90 nm) involved in intercellular communication. Although their function remains still largely enigmatic, their involvement in the immune response (in particular in antigen presentation to T cells) is well-established.21 Exosomal ankG, in particular when associated with β-amyloid plaques, could serve as an immunogen triggering the observed humoral immune responses in AD patients.
Our active ankG-immunization protocol reduced brain levels of insoluble Aβ in two independent transgenic mouse models (Figures 3,4 and Supplementary Figure 1d–f). Reduced β-amyloid plaque load and brain concentrations of insoluble aggregates of Aβ40 and Aβ42 in formic acid fractions of arcAβ mice was accompanied by increased levels of soluble Aβ42 in SDS-soluble fractions (Figures 3d,e and 4), compatible with immunization-induced disaggregation of insoluble material and resulting in the release of Aβ42 into the soluble pool. Aβ42 has synaptotoxic activities mediated through interactions with nicotinic receptors and NMDA receptors,31 and transient increases may occur during disaggregation of fibrillar β-amyloid followed by microglial phagocytosis and clearance. It is possible that the initial disaggregation phase of β-amyloid clearance is not accompanied by immediate beneficial effects on neuronal function and behavior, which may be expected to follow clearance of the toxic peptides and regeneration of neuronal structures.5 We observed no changes in behavioral tests during the time interval of our ankG-immunization experiments, neither in transgenic nor in wild-type mice (Supplementary Figure 3). This finding could be interpreted as resembling the observation of progressive dementia in AD patients during Aβ42 immunization, which clearly lowered brain β-amyloid plaque load.5, 32
To address more directly the role of ankG-reactive antibodies on the morphology of neurons subjected to Aβ-related toxicity, we analyzed dendritic spine morphology in organotypic brain slices prepared from transgenic mice expressing human mutant APP (Figure 5). The results of these experiments showed that ankG antibodies almost completely rescued the Aβ-related spine loss in brain slices from transgenic as compared with nontransgenic wild-type littermates. This was associated with decreased levels of Aβ within the organotypic slice preparation. Because dendritic spines are known to react very sensitively to Aβ, these data suggest Aβ reduction as a potential mechanism for the beneficial effects of ankG antibodies on dendritic spines in arcAβ mouse brains slices.
Our data showed that microglial cells are involved in phagocytosis of ankG present within β-amyloid plaques, possibly via Fc receptor-mediated phagocytosis of IgG reacting with ankG bound to β-amyloid. Importantly, active immunization with ankG in mice did not lead to any obvious side effects. Although ankG immunization showed the potential to reduce and clear β-amyloid to protect neurons from Aβ-related damage, it is necessary to determine whether these potentially beneficial effects can be achieved without disrupting ankG's physiological functions in anchoring voltage-gated ion channels and adhesion molecules to the axonal cytoskeleton. This would require, however, specific neutralization of ankG in its physiological cytoplasmic localization, a compartment that is unlikely to be targeted by extracellular antibodies. Even if the antibodies may be internalized into luminal endosomal compartments, they would have limited capabilities to escape into the cytoplasm.
Previous studies describe autoantibodies in AD patients against spectrin, glial fibrillary acidic protein, myelin basic protein and aldolase.33, 34, 35, 36 Our data of reduced cognitive decline in AD patients with positive serum titers of ankG-reactive antibodies provide evidence for a protective potential of such antibodies in AD. Should similar antibodies be present in IVIG preparations consisting of pooled IgG fractions derived from a large number of human donors, these may contribute to some of the beneficial effects of IVIG observed in clinical trials. Although currently active and passive Aβ immunotherapies are clearly the most actively pursued immunotherapeutic approaches,37 other immunotherapeutic candidate targets for the treatment of neurodegeneration are being explored. Recent findings showed that immunization with tau in a tau-transgenic mouse model reduces aggregated tau in brains and slows progression of the tangle-related phenotype,38 and α-synuclein immunization reduced the related pathology in transgenic mouse models.39
ANK3, the gene encoding AnkG, is one of the 23 functional candidate genes associated with late onset AD,11, 40 and our results suggest extracellular ankG as a novel target for AD immunotherapy. In addition to AD, ANK3 has been shown to be associated with bipolar disorder,41 but immunogenicity of ankG in this condition has not been investigated. It is likely that the presence of anti-ankG antibodies may also influence the clinical course of bipolar disorder. Further studies addressing this possibility in this cerebral condition are required.
In conclusion, we demonstrate that ankG may constitute a β-amyloid-related antigen, and that both nondemented aged subjects and AD patients can generate a humoral immune response against it. Slowed disease course suggests a beneficial role for ankG antibodies in AD, supported by reductions in β-amyloid pathology and improved dendritic spine morphology in transgenic mouse models after ankG immunization.
Diamond B, Huerta PT, Mina-Osorio P, Kowal C, Volpe BT . Losing your nerves? Maybe it's the antibodies. Nat Rev Immunol 2009; 9: 449–456.
Togo T, Akiyama H, Iseki E, Kondo H, Ikeda K, Kato M et al. Occurrence of T cells in the brain of Alzheimer's disease and other neurological diseases. J Neuroimmunol 2002; 124: 83–92.
Akiyama H, Barger S, Barnum S, Bradt B, Bauer J, Cole GM et al. Inflammation and Alzheimer's disease. Neurobiol Aging 2000; 21: 383–421.
Nitsch RM, Hock C . Targeting beta-amyloid pathology in Alzheimer's disease with Abeta immunotherapy. Neurotherapeutics 2008; 5: 415–420.
Serrano Pozo A, William C, Ferrer I, Delisle M-B, Maurage C-A, Nitsch RM et al. Beneficial effect of human anti-Aβ active immunization (AN1792) on amyloid deposition and neurite morphology. Program No 8255 2009 Neuroscience, Meeting Planner 2009.
Bennett V, Healy J . Organizing the fluid membrane bilayer: diseases linked to spectrin and ankyrin. Trends Mol Med 2008; 14: 28–36.
Paez-Gonzalez P, Abdi K, Luciano D, Liu Y, Soriano-Navarro M, Rawlins E et al. Ank3-dependent SVZ niche assembly is required for the continued production of new neurons. Neuron 2011; 71: 61–75.
Shaw AS, Filbert EL . Scaffold proteins and immune-cell signalling. Nat Rev Immunol 2009; 9: 47–56.
Bertram LBD, Mullin K, Keeney D, Jones J, Basu S, Yhu S et al. Evidence for genetic linkage of Alzheimer's disease to chromosome 10q. Science 2000; 290: 2302–2303.
Myers A, Holmans P, Marshall H, Kwon J, Meyer D, Ramic D et al. Susceptibility locus for Alzheimer's disease on chromosome 10. Science 2000; 290: 2304–2305.
Morgan AR, Turic D, Jehu L, Hamilton G, Hollingworth P, Moskvina V et al. Association studies of 23 positional/functional candidate genes on chromosome 10 in late-onset Alzheimer's disease. Am J Med Genet B Neuropsychiatr Genet 2007; 144B: 762–770.
Cepok S, Zhou D, Srivastava R, Nessler S, Stei S, Bussow K et al. Identification of Epstein-Barr virus proteins as putative targets of the immune response in multiple sclerosis. J Clin Invest 2005; 115: 1352–1360.
Lalive PH, Menge T, Barman I, Cree BA, Genain CP . Identification of new serum autoantibodies in neuromyelitis optica using protein microarrays. Neurology 2006; 67: 176–177.
Horn S, Lueking A, Murphy D, Staudt A, Gutjahr C, Schulte K et al. Profiling humoral autoimmune repertoire of dilated cardiomyopathy (DCM) patients and development of a disease-associated protein chip. Proteomics 2006; 6: 605–613.
Morgan AR, Hamilton G, Turic D, Jehu L, Harold D, Abraham R et al. Association analysis of 528 intra-genic SNPs in a region of chromosome 10 linked to late onset Alzheimer's disease. Am J Med Genet B Neuropsychiatr Genet 2008; 147B: 727–731.
Mohler PJ . Ankyrins and human disease: what the electrophysiologist should know. J Cardiovasc Electrophysiol 2006; 17: 1153–1159.
Zhang X, Bennett V . Restriction of 480/270-kD ankyrin G to axon proximal segments requires multiple ankyrin G-specific domains. J Cell Biol 1998; 142: 1571–1581.
Knobloch M, Konietzko U, Krebs DC, Nitsch RM . Intracellular Abeta and cognitive deficits precede beta-amyloid deposition in transgenic arcAbeta mice. Neurobiol Aging 2007; 28: 1297–1306.
Emmanouilidou E, Melachroinou K, Roumeliotis T, Garbis SD, Ntzouni M, Margaritis LH et al. Cell-produced alpha-synuclein is secreted in a calcium-dependent manner by exosomes and impacts neuronal survival. J Neurosci 2010; 30: 6838–6851.
Aguzzi A, Rajendran L . The transcellular spread of cytosolic amyloids, prions, and prionoids. Neuron 2009; 64: 783–790.
Qazi KR, Gehrmann U, Domange Jordo E, Karlsson MC, Gabrielsson S . Antigen-loaded exosomes alone induce Th1-type memory through a B-cell-dependent mechanism. Blood 2009; 113: 2673–2683.
Morgan D . The role of microglia in antibody-mediated clearance of amyloid-beta from the brain. CNS Neurol Disord Drug Targets 2009; 8: 7–15.
Koenigsknecht-Talboo J, Meyer-Luehmann M, Parsadanian M, Garcia-Alloza M, Finn MB, Hyman BT et al. Rapid microglial response around amyloid pathology after systemic anti-Abeta antibody administration in PDAPP mice. J Neurosci 2008; 28: 14156–14164.
Hsiao K, Chapman P, Nilsen S, Eckman C, Harigaya Y, Younkin S et al. Correlative memory deficits, Abeta elevation, and amyloid plaques in transgenic mice. Science 1996; 274: 99–102.
Zlokovic BV, Deane R, Sagare AP, Bell RD, Winkler EA . Low-density lipoprotein receptor-related protein-1: a serial clearance homeostatic mechanism controlling Alzheimer's amyloid beta-peptide elimination from the brain. J Neurochem 2010; 115: 1077–1089.
Kohler G, Milstein C . Continuous cultures of fused cells secreting antibody of predefined specificity. Nature 1975; 256: 495–497.
Tackenberg C, Brandt R . Divergent pathways mediate spine alterations and cell death induced by amyloid-beta, wild-type tau, and R406W tau. J Neurosci 2009; 29: 14439–14450.
Dzhashiashvili Y, Zhang Y, Galinska J, Lam I, Grumet M, Salzer JL . Nodes of Ranvier and axon initial segments are ankyrin G-dependent domains that assemble by distinct mechanisms. J Cell Biol 2007; 177: 857–870.
Komada M, Soriano P . Beta]IV-spectrin regulates sodium channel clustering through ankyrin-G at axon initial segments and nodes of Ranvier. J Cell Biol 2002; 156: 337–348.
Han L, Cole M, Bellavance F, McCusker J, Primeau F . Tracking cognitive decline in Alzheimer's disease using the mini-mental state examination: a meta-analysis. Int Psychogeriatr 2000; 12: 231–247.
Bordji K, Becerril-Ortega J, Buisson A . Synapses, NMDA receptor activity and neuronal Abeta production in Alzheimer's disease. Rev Neurosci 2011; 22: 285–294.
Holmes C, Boche D, Wilkinson D, Yadegarfar G, Hopkins V, Bayer A et al. Long-term effects of Abeta42 immunisation in Alzheimer's disease: follow-up of a randomised, placebo-controlled phase I trial. Lancet 2008; 372: 216–223, .
Singh VK, Yang YY, Singh EA . Immunoblot detection of antibodies to myelin basic protein in Alzheimer's disease patients. Neurosci Lett 1992; 147: 25–28.
Kellner A, Matschke J, Bernreuther C, Moch H, Ferrer I, Glatzel M . Autoantibodies against beta-amyloid are common in Alzheimer's disease and help control plaque burden. Ann Neurol 2009; 65: 24–31.
Mor F, Izak M, Cohen IR . Identification of aldolase as a target antigen in Alzheimer's disease. J Immunol 2005; 175: 3439–3445.
Mecocci P, Parnetti L, Romano G, Scarelli A, Chionne F, Cecchetti R et al. Serum anti-GFAP and anti-S100 autoantibodies in brain aging, Alzheimer's disease and vascular dementia. J Neuroimmunol 1995; 57: 165–170.
Orgogozo JM, Gilman S, Dartigues JF, Laurent B, Puel M, Kirby LC et al. Subacute meningoencephalitis in a subset of patients with AD after Abeta42 immunization. Neurology 2003; 61: 46–54.
Asuni AA, Boutajangout A, Quartermain D, Sigurdsson EM . Immunotherapy targeting pathological tau conformers in a tangle mouse model reduces brain pathology with associated functional improvements. J Neurosci 2007; 27: 9115–9129.
Masliah E, Rockenstein E, Mante M, Crews L, Spencer B, Adame A et al. Passive immunization reduces behavioral and neuropathological deficits in an alpha-synuclein transgenic model of Lewy body disease. PLoS One 2011; 6: e19338.
Bertram L, Blacker D, Mullin K, Keeney D, Jones J, Basu S et al. Evidence for genetic linkage of Alzheimer's disease to chromosome 10q. Science 2000; 290: 2302–2303.
Ferreira MA, O'Donovan MC, Meng YA, Jones IR, Ruderfer DM, Jones L et al. Collaborative genome-wide association analysis supports a role for ANK3 and CACNA1C in bipolar disorder. Nat Genet 2008; 40: 1056–1058.
We thank Dr Annamaria Calella and Dr Sarah Hoey for helpful discussions (University of Zurich); Dr Burkhardt Seifert for help with statistical analysis; Dr Vann Bennett (Duke University Medical Center) for providing us with ankG and ankB recombinant proteins; Dr Uwe Konietzko (University of Zurich) for APP-citrine HEK293 cells; Dr Gang Yu (University of Texas) for HeLa cells expressing the Swedish mutation in APP; Esmeralda Gruber and Dr Anton Gielt for patient care and sampling (University of Zurich); Diana Bundschuh for human DNA genotypig and Sebastian Zurbriggen for help with cell culture (University of Zurich). Rebecca Derungs for help with himmunohistochemistry (University of Zurich). We thank Daniel Schuppli for animal care, Manuela Hitz and Bjorn Henz for technical help in behavioral experiments (University of Zurich). ACS was supported by the Novartis Foundation. AS was supported from the Swiss Commission for Research Partnerships with Developing Countries. LR acknowledges funding support from the Velux foundation, Bangerter Foundation and the Swiss National Science Foundation. MM and RMN were funded by the Swiss National Science Foundation SNSF 3200B0-112616/1, and the National Centre of Competence in Research NCCR ‘Neural Plasticity and Repair’. CH was supported by SNF 320030_125378 and SNF 33CM30_124111. CT was supported from DFG, Ta762/1-1. CB and MG were supported by DFG, (FOR885, GRK1459).
ACS, RMN and MM conceived the study. ACS, RMN, LR, MG, JG, and CH raised funds for the study. ACS, MM, AS, RMN, JB, LR, CT, CB, LK, MTF designed and performed the experiments. TW assisted with the experiments. ACS, RMN, MM, LR, CB, MG, ACS, RMN, MM, supervised the experiments and analyzed the data. ACS, MTF, MM, AS, RMN wrote the drafts of the manuscript. All authors edited the manuscript.
A patent application that includes data from this manuscript was filed by the University of Zurich with ACS, MM, JG, CH and RMN listed as inventors.
Supplementary Information accompanies the paper on the Molecular Psychiatry website
About this article
Cite this article
Santuccione, A., Merlini, M., Shetty, A. et al. Active vaccination with ankyrin G reduces β-amyloid pathology in APP transgenic mice. Mol Psychiatry 18, 358–368 (2013). https://doi.org/10.1038/mp.2012.70
- Alzheimer's disease
- ankyrin G
- immune system
AnkG hemizygous mice present cognitive impairment and elevated anxiety/depressive-like traits associated with decreased expression of GABA receptors and postsynaptic density protein
Experimental Brain Research (2017)
Statins induce insulin-degrading enzyme secretion from astrocytes via an autophagy-based unconventional secretory pathway
Molecular Neurodegeneration (2015)