Alzheimer’s disease (AD) is the most common form of dementia in the elderly. It is a progressive neurodegenerative disorder that is characterized by the abundant presence of cerebral β-amyloid (Aβ) plaques and neurofibrillary Tau tangles. The rare, early-onset AD is caused by mutations mainly within either the amyloid precursor protein (APP) or Presenilins 1 or 2 (PS1 or PS2), the catalytic subunits of the γ-secretase complex. These mutations either increase overall Aβ production or specifically alter γ-secretase mediated processing towards an increased production ratio of Aβ42:Aβ40. For late-onset AD, however, which accounts for the vast majority of AD cases, the exact mechanisms by which the disease is caused are not known. While genome-wide association studies (GWAS) have identified certain genetic risk factors associated with late-onset AD, the mechanisms through which they contribute to the pathogenesis are still elusive. Previously, using a HeLa cell model of Aβ production, it was shown that, in contrast to the early-onset AD causing mutations, knockdown of late-onset AD susceptibility genes did not specifically affect the Aβ42:Aβ40 ratio^[1]. To validate these findings in a neuronal setting without any overexpression of APP, here we re-addressed the role of 6 late-onset AD risk genes (APOE, BIN1, PICALM, CLU, PRNP and CST3) in the regulation of γ-secretase mediated APP processing in wild-type mouse primary neurons by analyzing Aβx-40 and Aβx-42. In addition, we extended the analysis by also including measurements of total Tau protein and phosphorylation of Tau at Threonine 231, a non-physiological phosphorylation site that is associated with AD. The siRNA-mediated knockdown of the studied LOAD risk genes neither affected the Aβx-42:Aβx-40 ratio nor altered the levels of total Tau or phospho(Thr231)-Tau. Our results thus show that acute downregulation of these genes in wild-type mouse primary neurons does not significantly impact on γ-secretase mediated APP processing nor Tau homeostasis or Tau phosphorylation.

The brains of Alzheimer’s disease (AD) patients show widespread formation of extracellular senile plaques composed of aggregated β-amyloid (Aβ) peptides as well as intraneuronal aggregates of misfolded Tau protein, so-called neurofibrillary tangles. According to the onset of disease one can distinguish two types of AD: a rare, early-onset AD (EOAD) with an onset before 65 years of age, and the common, late-onset AD (LOAD) with an onset after 65 years of age. EOAD is caused by a specific set of highly penetrant mutations, which affect almost exclusively either the amyloid precursor protein (APP) or the catalytically active Presenilin subunit of the Aβ-producing γ-secretase complex. These mutations enhance the overall production of Aβ peptides and/or increase the ratio of the aggregation prone and neurotoxic Aβ42 over the common, shorter Aβ40. The genetic contribution to LOAD, on the other hand, is not very well understood. Inheritance patterns within families point to multifactorial inheritance of LOAD, involving both genetic and environmental factors. Initial linkage analysis studies and GWAS analyses have identified a number of susceptibility genes for LOAD (Alzgenes), the majority of which having only small effect sizes on disease risk^[2]. The only exception is the APOE ɛ4 locus which increases LOAD risk 2- to 4-fold in individuals with one copy of the allele and 8- to 15-fold in individuals carrying two copies^{[3][4][5][6][7]}.

While an increased production of Aβ and/or an increased Aβ 42:40 ratio have been acknowledged as the driving pathogenic mechanism in EOAD, the etiology of LOAD appears to be much more complex and it is still not absolutely clear if Aβ accumulation plays an active causative role in the initial pathogenesis of the disease. Also the mechanisms that drive Aβ accumulation appear to be different in LOAD and EOAD. While some studies showed effects of LOAD risk genes on Aβ production, most of the data suggest that Aβ accumulation in the brains of LOAD patients results from impaired Aβ clearance rather than an increased production as seen in EOAD^[8][9]. For example APOE, which is the strongest genetic risk factor for LOAD, was shown to bind to Aβ and mediate its clearance across the blood brain barrier (BBB) or its endocytosis into brain cells for lysosomal degradation^[10][11]. Similarly, also CLU and PICALM had been shown to enhance the transport of Aβ across the blood-brain barrier and the blood-cerebrospinal fluid barrier^{[12][13][14][15]}. In line with the notion that LOAD genes do not specifically affect Aβ production, a recent study did not detect specific effects of the knockdowns of several LOAD susceptibility genes on Aβ levels or the 42:40 ratio in a HeLa cell APP overexpression system^[1].

We wanted to study the effect of acute downregulation of selected LOAD risk genes on γ-secretase mediated processing of endogenous APP and on levels of endogenous total Tau and phospho-Tau (pThr231) in wild-type neurons. The selected risk genes comprised four high ranking Alzgenes (Apolipoprotein E [APOE], Bridging integrator 1 [BIN1], Phosphatidylinositol binding clathrin assembly protein [PICALM], Clusterin [CLU] and two LOAD risk genes with weaker association (Prion protein [PRNP] and Cystatin C [CST3]).

In order to study the role of selected LOAD risk genes on γ-secretase mediated processing of endogenous APP and on levels of endogenous total Tau and phospho-Tau in neurons, we performed individual siRNA-mediated gene knockdowns in primary neuronal cultures prepared from embryonic wildtype mice. We chose a set of 4 high ranking Alzgenes (APOE, BIN1, CLU, PICALM) and 2 risk genes with weaker association (PRNP and CST3) which were readily detectable at mRNA level in the primary cultures and amenable to gene knockdown by RNAi. As negative control, we transfected a pool of non-targeting siRNAs. In addition, we chose four non-AD control genes (TARDBP: TAR-DNA-binding protein 43, RELN: Reelin, KAT5: Lysine acetyltransferase 5 and SNCA: α-synuclein) that are implicated in other neurodegenerative diseases and neurological conditions including amyotrophic lateral sclerosis (ALS), frontotemporal lobar degeneration (FTD), schizophrenia, autism, multiple sclerosis (MS) and Parkinson’s disease (PD). As positive controls, we transfected siRNAs that target either APP or MAPT. Knockdown-efficiencies of all candidate and control genes were assessed at the mRNA level by real-time PCR in each experimental round. In addition, knockdown of candidate genes at the protein level were confirmed by Western Blot analysis. For the simultaneous detection of both Aβx-40 and Aβx-42 or phospho(Thr231)-Tau and total Tau, we used specific multiplex electrochemiluminescence assays for analysis of the conditioned neuronal culture medium or the recovered protein fraction of the whole cell lysate respectively. Knockdown of the positive control APP led to the expected reduction of both Aβx-40 and Aβx-42 without altering Tau levels (Fig. 1A) while transfection of MAPT siRNA significantly lowered the signal of total and phospho (Thr231)-Tau without affecting Aβ (Fig. 1B), thus proving the functionality of the RNAi approach and the validity of the assays.

Silencing any of the tested Alzgenes did not affect secreted Aβx-40 and Aβx-42 or the Aβx-42:Aβx-40 ratio (Fig. 1A) nor did it alter total Tau and phospho(Thr231)-Tau levels (Fig.1B). Our results suggest that, in wildtype mouse primary cortical/hippocampal neurons, acute knockdown of the studied LOAD risk genes does not crucially impact on γ-secretase mediated APP processing nor on total Tau levels or Tau(Thr231)-phosphorylation. Susceptibility to AD through these genes might involve effects on APP processing or Tau that build up only over a longer time or only at a later stage, when neurons have already undergone substantial aging. Alternatively, these genes might play a role in mediating Aβ toxicity to neurons, or alter Aβ clearance pathways by non-neuronal cell types, or modulate Aβ aggregation. Indeed, experimental evidence for multiple of these possible mechanisms have been provided by previous studies for several of the herein studied susceptibility genes. For example, in addition to the aforementioned roles of APOE, CLU and PICALM in facilitating Aβ clearance, APOE lipoproteins were shown to regulate the association of oligomeric Aβ with synapses^[16], CLU and CST3 were shown to inhibit Aβ aggregation^{[17][18][19][20]} and PRNP was shown to play a role in mediating the deleterious effects of Aβ on synaptic transmission and learning and memory^{[21][22][23][24][25][26][27]}, though this finding was challenged by at least 3 other groups^[28][29][30]. Importantly, while in our study the knockdown of the tested genes in mouse primary neurons altered neither Aβx-40 and Aβx-42 nor total Tau and phospho(Thr231)-Tau levels, a number of previous studies using different systems reported effects on Aβ or Tau pathology for some of the tested genes. APOE, for example, was shown to regulate neuronal Aβ production by modulating APP recycling^[31] or APP transcription^[32] when applied extracellularly. It is conceivable that the reason why we did not observe any effects of APOE knockdown on Aβx-40 and Aβx-42 is because it is expressed mainly by astrocytes and produced only at low amounts by primary neuronal cultures as used for our study. For BIN1, conflicting results have been reported on its role in Aβ regulation: while depletion of BIN1 in neurons resulted in a slight decrease in secreted Aβ40 with a concomitant increase in intracellular Aβ40 and Aβ42 in one study^[33], increased levels of secreted Aβ were observed by another group^[34] and no effects were detected in neuroblastoma cells in a third study^[35]. Similarly, also for PICALM the observed effects on Aβ are diverging: in one study, PICALM was reported to promote Aβ generation in cells and to accelerate Aβ pathology in APP transgenic mice^[36] whereas another study observed no effect on cellular Aβ production^[37]. As for PRNP, a direct role was proposed in the regulation of Aβ production through inhibition of β-secretase-mediated cleavage of APP^[38]. In our study, neither Aβx-40 nor Aβx-42 levels changed upon knockdown of BIN1, PICALM or PRNP in mouse primary neurons. Our results thus suggest that the overall APP processing by γ-secretase is not altered by downregulation of either of these genes in young cultured wild-type neurons. Since the assay used in our analysis was chosen so as to detect all Aβx-40 and Aβx-42 species produced by γ-secretase, comprising full-length Aβ and the shorter Aβ' and p3 peptides^[39], our data do not necessarily challenge earlier findings on altered full length Aβ levels. Regarding Tau pathology, only for BIN1 a link has been reported previously: BIN1 expression was shown to correlate with the amount of neurofibrillary tangles (NFT) or total-Tau/phospho-Tau in AD patients in 2 studies^[40][41] and a direct interaction between BIN1 and Tau was suggested by another^[42]. However, no significant overlap between BIN1 and neurofibrillary tangles was seen in a more recent study^[43]. Previously reported effects of experimentally altered BIN1 expression on Tau pathology have been controversial too. While loss of BIN1 was shown to reduce Tau mediated neurodegeneration in Drosophila^[42], it reduced the propagation of Tau pathology in an in vitro system^[44]. We did not observe a significant effect of BIN1 downregulation on wildtype total Tau or pTau(Thr231) levels in our study. Effects on Tau pathology progression as seen in the earlier studies might possibly require a disease context as provided by the presence of Tau mutations in these studies.

Our results show that acute downregulation of the studied LOAD risk genes in mouse primary neurons does not significantly alter γ-secretase mediated APP processing, Tau levels or Tau(Thr231)-phosphorylation. Susceptibility to AD through these genes might be conferred through other mechanisms, for example modulation of Aβ clearance or Aβ aggregation/toxicity as it has been proposed by some of the studies discussed above and/or Aβ and Tau-independent signaling pathways that directly affect neuronal function.

There are 3 important limitations to our study:

1. The in vitro character of the study: The simple intercellular connections that are formed in a primary neuronal culture do not allow to model the complex interactions that occur in an intact brain. Any effects that depend on the network activity within or between certain brain areas can therefore not be assessed by our model.

2. The limited time frame during which we assessed the role of the studied Alzgenes: Knockdowns were performed for about 72 h. Any effects on APP processing or Tau that build up only over a longer period, as it might be the case in AD, would not be detected in our system.

3. The age of the primary neurons: Primary neuronal cultures were prepared from embryonic neurons, thus neurons in their very early developmental stage. LOAD, however, is a disease of the aged brain. The expression profiles of the tested Alzgenes and their interaction partners might be very different in young and old neurons.

We are now planning to study the role of the respective risk gene variants in neurons derived from human induced pluripotent stem cells (iPSCs) which have been modified using CRISPR/Cas9 technology for targeted genome editing to introduce the respective LOAD associated genetic polymorphisms. This will not only allow us to assess long-term effects of the respective risk variants on APP processing and Tau homeostasis but also to test whether these genes would have any roles in neuronal physiology and function in an Aβ and Tau independent manner.

Mouse primary neuronal cultures

Mixed cortical/hippocampal primary neuronal cultures were prepared from E16 ICR (CD-1®) outbred mice (Harlan Laboratories, Horst, Netherlands). Cortices with adjacent hippocampi were dissected in ice-cold HBSS, incubated in 7 ml TrypLe Express for 10 min at 37°C and triturated in Dulbecco’s modified Eagle’s medium (DMEM) (1 g/l glucose) containing 5% FCS (all from Life Technologies) through repeated pipetting with a 5 ml serological pipette and then passed through a 70 µM cell strainer. Neurons were cultured in Neurobasal medium supplemented with B27 (1:50) and 1 mM GlutaMax (all from Life Technologies, Zug, Switzerland) on 96-well plates coated with poly-D-lysine (Sigma Aldrich, St. Gallen, Switzerland) in a humidified incubator at 37°C in a 5% (vol/vol) CO2 atmosphere.

siRNA transfection

4DIV primary neurons were transfected with 50 nM of siRNA (stealth siRNA, Life Technologies, Zug, Switzerland) using Lipofectamine RNAiMax (Life Technologies, Zug, Switzerland) as transfection reagent. For each gene, a pool of 4 different siRNA duplexes was used. A pool of 3 different non-targeting siRNAs served as negative control. For each well of a 96-well plate 0.4 µl of RNAiMax were mixed with siRNA and Neurobasal (Life Technologies, Zug, Switzerland) in a final volume of 20 µl. The mix was incubated for 25 min at RT and then further diluted with Neurobasal medium to 100 µl. Cells were incubated with the transfection mix for 6 h, after which the medium on the cells was again replaced with Neurobasal medium supplemented with B27 (1:50) and 1 mM GlutaMax (all from Life Technologies, Zug, Switzerland).

siRNAs

negative controls

medGC duplex #1 Life Technologies #12935-111

medGC duplex #2 Life Technologies #12935-112

medGC duplex #3 Life Technologies #12935-113

APP

Sense sequence 1 GCGGAUGGAUGUUUGUGAGACCCAU

Sense sequence 2 UCAGGAUUUGAAGUCCGCCAUCAAA

Sense sequence 3 GACCAGGUUCUGGGCUGACAAACAU

Sense sequence 4 CACACACCCACAUCGUGAUUCCUUA

APOE

Sense sequence 1 GGUUCGAGCCAAUAGUGGAAGACAU

Sense sequence 2 GCAGAGCUCCCAAGUCACACAAGAA

Sense sequence 3 GAUGGAGGAACAGACCCAGCAAAUA

Sense sequence 4 GAGAAUCAAUGAGUAUCCUUCUCCU

BIN1

Sense sequence 1 CCUGGCAGGGAUGAAGCAAACAAGA

Sense sequence 2 UCGGACCUAUCUGGCUUCUGUUAAA

Sense sequence 3 GAUGACGCAUUUGUCCCUGAGAUCA

Sense sequence 4 AAGAGAUGAGUAAGCUCAAUCAGAA

CLU

Sense sequence 1 UCUCUGACAAUGAGCUCCAAGAACU

Sense sequence 2 UGUACUUGAGCAGAGCGCUAUAAAU

Sense sequence 3 ACGCCAUGAAGAUUCUCCUGCUGUG

Sense sequence 4 CCACUCAAGGGAGUAGGUAUAUUAA

PICALM

Sense sequence 1 GGGAGAUCCUUUCUCUGCUACUCUA

Sense sequence 2 GCUUGACUUGCAGCAGCCAACCUUU

Sense sequence 3 UGGCUCCGCGGUAUCUAAGACAGUA

Sense sequence 4 CAGCAGUCUUCUUGAUGCUUUAGAA

PRNP

Sense sequence 1 GGGACAACCUCAUGGUGGUAGUUGG

Sense sequence 2 CCAGUGGAUCAGUACAGCAACCAGA

Sense sequence 3 UGGAGCAGAUGUGCGUCACCCAGUA

Sense sequence 4 CACGACUGCGUCAAUAUCACCAUCA

CST3

Sense sequence 1 CCAGACAAAUUUGACUGACUGUCCU

Sense sequence 2 AGGCACUCUGCUCCUUCCAGAUCUA

Sense sequence 3 GACUUCGCUGUGAGCGAGUACAACA

Sense sequence 4 CAGCUCGUGGCUGGAGUGAACUAUU

MAPT

Sense sequence 1 CAGGAGGUGGCCAGGUGGAAGUAAA

Sense sequence 2 CAGGAGGUGGCAAGGUGCAGAUAAU

Sense sequence 3 CAGUCGAAGAUUGGCUCCUUGGAUA

Sense sequence 4 CAAGACAGACCAUGGAGCAGAAAUU

TARDBP

Sense sequence 1 GCAAUCUGGUAUAUGUUGUCAACUA

Sense sequence 2 CGAAAGGGUUUGGCUUUGUUCGAUU

Sense sequence 3 GAAAUACCAUCAGAAGACGAUGGGA

Sense sequence 4 CCUCCCUGUUGAGUGAGGCUAUUUA

RELN

Sense sequence 1 GCUCUCAAACUGGAUUUCAAGAUAA

Sense sequence 2 CCUGGGUGAUCGACCAGAUUCUUAU

Sense sequence 3 GAGAGCUCAUUAUACAGCCAGGAUA

Sense sequence 4 CAGUUCCAUGAAGCCACCAUUUAUA

KAT5

Sense sequence 1 AGCCUGGACGGAAGCGGAAAUCUAA

Sense sequence 2 CGGCACCCUCCAGGCAAUGAAAUUU

Sense sequence 3 CGUAAUGACGGAGUAUGACUGCAAA

Sense sequence 4 CACACUGCAGUAUCUCAACCUCAU

SNCA

Sense sequence 1 GCAAGUGACAAAUGUUGGAGGAGCA

Sense sequence 2 GGGAGUCCUCUAUGUAGGUUCCAAA

Sense sequence 3 CCAAGACUAUGAGCCUGAAGCCUAA

Sense sequence 4 CACAGGAAGGAAUCCUGGAAGACAU

Alamar Blue assay

Alamar Blue metabolisation was used as an indirect measure of cell number and viability of primary neurons. Alamar BlueTM assay reagent (AbD Serotec Ltd, Bio-Rad, Cressier, Switzerland) was added to the cells at a final concentration of 10% for the final 3 h before termination of the medium collection period. Levels of the metabolite resofurin were assessed by 544EX nm/590EM nm fluorescence measurements with the Spectra MAX-GEMINI-XS spectrofluorometer (Molecular-Devices, Sunnyvale, CA, USA).

RNA isolation

For RNA extraction, adherent neurons were briefly washed with PBS (containing CaCl2 and MgCl2) and lysed in 100 µl TRI-Reagent (Sigma-Aldrich, Buchs, Switzerland) per well. Lysates of technical replicates of each experimental condition were pooled and isolation of RNA was performed according to manufacturer's instructions. All RNA samples were subjected to DNase treatment using 0.5 units of DNaseI (Fermentas; Life Technologies, Zug, Switzerland) for 30 min at 37°C. DNase was heat inactivated by a 10 min incubation at 65°C in the presence of 5 mM EDTA. RNA concentrations were measured with a Nanodrop 2000 UV-Vis spectrophotometer (Thermo Scientific, Waltham, MA, USA).

Protein isolation

Protein was recovered from the organic/phenol phase that was obtained during RNA isolation with TRI-Reagent (Sigma-Aldrich, Buchs, Switzerland) following the manufacturer’s protocol for protein precipitation with acetone. The obtained protein pellet was resuspended in 9.5 M Urea (pH 9.0), 2% CHAPS. For analysis by electrochemiluminescence multiplex assay, samples were diluted 1:100 with TBS (pH 7.5) containing 0.05% Tween and 1% BlockerA (Meso Scale Discovery).

cDNA synthesis

For each sample 500 ng of total RNA was reverse-transcribed using the iScript cDNA-Synthesis-kit (Bio-Rad, Cressier, Switzerland) according to manufacturer’s instructions.

Real-time PCR

Real-time PCR for relative quantification of cDNA levels was performed with the 7900HT Real-Time PCR System (Life-Technologies, Zug, Switzerland), using the iTaq-SybrGreen Supermix with ROX (Bio-Rad, Cressier, Switzerland). Relative gene expression levels were calculated with the ΔΔCt-method using GAPDH for normalization.

Real-time PCR primer

GAPDH

Forward primer ATCACTGCCACCCAGAAGAC

Reverse primer GGATGCAGGGATGATGTTCT

APP

Forward primer ACCGTTGCCTAGTTGGTGAGT

Reverse primer CGGTGTGCCAGTGAAGATG

APOE

Forward primer CTGAACCGCTTCTGGGATTAC

Reverse primer CCATCAGTGCCGTCAGTTCT

BIN1

Forward primer GAAGATCGCCAGCAACGTAC

Reverse primer TGCTCAAACTGCTCGTCCTT

CLU

Forward primer CGTCCAGGGAGTGAAGCA

Reverse primer AATCCCTAGTGTCCTCCAGAGC

PICALM

Forward primer AAGGTTGCACCAACAACTGC

Reverse primer CTATCATGCCCGTTGGTGTAGT

PRNP

Forward primer TCCATTTTGGCAACGACTG

Reverse primer TCGTGCACGAAGTTGTTCTG

CST3

Forward primer TACAACAAGGGCAGCAACG

Reverse primer TAGTTCGGCCCATCTCCAC

MAPT

Forward primer TCGCCAGGAGTTTGACACA

Reverse primer GTCTCCGATGCCTGCTTCT

TARDBP

Forward primer GAAGACGATGGGACGGTGT

Reverse primer TCCACCAGTCGGACTCCTC

RELN

Forward primer CAAGGTGACGACTGCTCTGTC

Reverse primer ACTCCACCCTGGATGGTTTC

KAT5

Forward primer CCTGTGTCTTCTGGCCAAGT

Reverse primer CCCACGATGTGGAAACCTT

SNCA

Forward primer ATGTTGGAGGAGCAGTGGTG

Reverse primer GCCCATCTGGTCCTTCT

Electrochemiluminescence assay

Aβx-40 and Aβx-42 levels in 24 h conditioned medium of 8DIV mouse primary neurons were analyzed with the Aβ-Panel1 (4G8) Kit (Meso Scale Discovery, Maryland, USA) following the manufacturer’s instructions. To each well of a 96-well assay plate 25 µl of 1:50 diluted Sulfo-Tag 4G8 mAb detection antibody and 25 µl of conditioned medium were added and incubated overnight at 4°C. Total Tau and phosphoThr231-Tau from 8DIV mouse primary neuronal cultures were assayed in the recovered protein fraction with the Phospho(Thr231)/Total Tau Kit (Meso Scale Discovery, Maryland, USA) following the manufacturer’s instructions. To each well of a 96-well assay plate 25 µl of sample were added and incubated for 2 h at RT on a shaker at 750 rpm. Incubation with the detection antibody was for 1 h at RT on a shaker at 750 rpm using 25 µl of Sulfo-Tag anti-Tau antibody per well at a concentration of 1 µg/ml. Measurements were taken on a Sector-Imager-6000 (Meso-Scale-Discovery, Maryland, USA). Electrochemiluminescence values were normalized to the corresponding Alamar-Blue assay values.

SDS-PAGE and Western blot

Proteins were separated on 4–12% Bis-Tris gels (Life Technologies, Zug, Switzerland) and blotted on 0.2µm Nitrocellulose membranes (Life Technologies, Zug, Switzerland). Unspecific binding was blocked by preincubation of membranes with TBS-Tween(0.05%) containing 5% w/v nonfat milk powder. Incubation with primary antibodies was performed overnight at 4°C, incubation with secondary antibodies for 1 h at room temperature. Infrared signal at 700 nm and 800 nm were acquired with an Odyssey CLx Imaging System (Li-COR Biosciences, Bad Homburg, Germany).

Primary antibodies

goat anti-ApoE (Santa Cruz # sc-6384) 1:100

rabbit anti-Bin1 (Proteintech # 14647-1-AP) 1:1000

goat anti-ApoJ/Clu (Abcam # ab79280) 1:500

rabbit anti-Picalm (Abcam # ab106409) 1:500

mouse anti-PrnP (Abcam # ab61409) 1:500

rabbit anti-Cst3 (Proteintech # 12245-1-AP) 1:1000

mouse anti-β-Actin (Abcam # ab6276) 1:10000

mouse anti-Gapdh (Life Technologies # AM4300) 1:5000

Secondary antibodies

donkey anti-mouse IRDye 800CW (Li-COR Biosciences # 926-32212) 1:5000

donkey anti-rabbit IRDye 800CW (Li-COR Biosciences # 926-32213) 1:5000

donkey anti-goat IRDye 800CW (Li-COR Biosciences # 926-32214) 1:5000

donkey anti-mouse IRDye 680RD (Li-COR Biosciences # 926-68072) 1:5000

L. R. was supported by the Swiss National Science Foundation grants (Sinergia and Core Interdisciplinary grants), by the Velux Foundation, the Novartis Foundation grant, the Cure Alzheimer Foundation, the Hurka Stiftung, the Bangerter Stiftung, the Baugarten Stiftung and the Synapsis Foundation. G. S. was supported by an EMBO long-term fellowship (ALTF 668-2011) and the University of Zurich’s Forschungskredit (K-82033-02-01).

We thank the lab members for their stimulating discussions.

The authors declare no conflicts of interest.

All animal experiments were done according to the guidelines of and approved by the veterinary office of the Canton of Zürich, Switzerland.

No fraudulence is committed in performing these experiments or during processing of the data. We understand that in the case of fraudulence, the study can be retracted by ScienceMatters.