German Center of Neurodegenerative Diseases, University of Tübingen, Tübingen, Germany
Laboratory of Functional Neurogenetics, Department of Neurodegeneration, Faculty of Medicine, Hertie Institute for Clinical Brain Research, University of Tübingen, Tübingen, Germany
Address correspondence and reprint requests to Philipp Kahle, Laboratory of Functional Neurogenetics, Department of Neurodegeneration, German Center of Neurodegenerative Diseases and Hertie Institute for Clinical Brain Research, University of Tübingen, Otfried Müller Str. 27, 72076 Tübingen, Germany. E-mail: email@example.com
Parkinson's disease (PD) and diabetes belong to the most common neurodegenerative and metabolic syndromes, respectively. Epidemiological links between these two frequent disorders are controversial. The neuropathological hallmarks of PD are protein aggregates composed of amyloid-like fibrillar and serine-129 phosphorylated (pS129) α-synuclein (AS). To study if diet-induced obesity could be an environmental risk factor for PD-related α-synucleinopathy, transgenic (TG) mice, expressing the human mutant A30P AS in brain neurons, were subjected after weaning to a lifelong high fat diet (HFD). The TG mice became obese and glucose-intolerant, as did the wild-type controls. Upon aging, HFD significantly accelerated the onset of the lethal locomotor phenotype. Coinciding with the premature movement phenotype and death, HFD accelerated the age of onset of brainstem α-synucleinopathy as detected by immunostaining with antibodies against pathology-associated pS129. Amyloid-like neuropathology was confirmed by thioflavin S staining. Accelerated onset of neurodegeneration was indicated by Gallyas silver-positive neuronal dystrophy as well as astrogliosis. Phosphorylation of the activation sites of the pro-survival signaling intermediate Akt was reduced in younger TG mice after HFD. Thus, diet-induced obesity may be an environmental risk factor for the development of α-synucleinopathies. The molecular and cellular mechanisms remain to be further elucidated.
Life-long high fat diet (HFD) induces obesity and glucose intolerance in a transgenic mouse model for α-synucleinopathy and thereby leads to decreased life span as well as accelerated age of onset of the terminal phenotype. This is accompanied by increased neuroinflammation and premature α-synuclein pathology in the brainstems of the HFD-fed mice.
Amyloid-like fibrillar aggregates (Lewy bodies and Lewy neurites) composed of α-synuclein (AS) are the neuropathologically diagnostic hallmark aggregates of Parkinson's disease (PD) and related disorders (Spillantini et al. 1997). The α-synuclein gene (SNCA) encoding AS was discovered as a cause of rare familial forms of PD (Polymeropoulos et al. 1997; Krüger et al. 1998; Chartier-Harlin et al. 2004) and SNCA is one of the most significant genetic risk factors for PD (Satake et al. 2009; Simón-Sánchez et al. 2009). Aside from genetic predispositions, aging is the most established PD risk factor. Compared to several environmental or intrinsic risk factors, diabetes mellitus was insignificant (Noyce et al. 2012). Nevertheless, diabetes did increase PD risk in prospective studies (Hu et al. 2007; Driver et al. 2008; Xu et al. 2011) and was associated with PD severity (Cereda et al. 2012). Moreover, dietary uptake of animal fat has been reported to increase PD risk (Logroscino et al. 1996; Anderson et al. 1999; Johnson et al. 1999). The hypothesis is emerging that PD and diabetes might share common dysregulated pathways related to mitochondrial dysfunction, protein degradation, (neuro)inflammation, and neurotrophic insulin signaling (Santiago and Potashkin 2013).
Beyond the general risk effects of insulin resistance on neurodegeneration (Aviles-Olmos et al. 2013), there are specific links between AS and diabetic processes. Though most abundant in brain synapses, AS is expressed also in pancreatic islet β-cells, where it tones insulin secretion (Geng et al. 2011; Steneberg et al. 2013). Moreover, AS is present in surprisingly large amounts in blood (Barbour et al. 2008), and acute administration of AS protein was recently reported to regulate glucose uptake in adipocytes (Rodriguez-Araujo et al. 2013). The AS protein not only binds to phospholipid vesicles (Davidson et al. 1998; Nuscher et al. 2004), but it also interacts with free fatty acids (Sharon et al. 2001), which are elevated in obesity and diabetes. Polyunsaturated fatty acids induce oligomerization of AS protein (Perrin et al. 2001; Sharon et al. 2003a; Assayag et al. 2007). Fatty acid composition is altered in neurons over-expressing AS and in brain samples from α-synucleinopathy patients (Sharon et al. 2003b), and studies on Snca−/− mice showed specific roles of AS in brain fatty acid uptake and metabolism (Golovko et al. 2009). Finally, glucose stress was detected in Snca−/− mice, suggesting an influence of AS on brain glucose metabolism as well (Kurz et al. 2011).
To assess a possible interplay between dietary risk factors and genetic predisposition to α-synucleinopathy, we induced by high fat diet (HFD) obesity and glucose intolerance in transgenic (TG) mice expressing the human mutant [A30P]AS under control of the brain neuron-specific Thy1 promotor (Kahle et al. 2000b). These animals develop age-dependent AS pathology similar to human patients (Kahle et al. 2001; Neumann et al. 2002). (Thy1)-h[A30P]AS mice develop behavioral impairments during aging and terminate with a lethal locomotor phenotype within the second year of life (Neumann et al. 2002; Freichel et al. 2007; Schell et al. 2012). Aggregation of AS is associated with pathological protein modifications of which especially phosphorylation at serine-129 (pS129) (Okochi et al. 2000; Pronin et al. 2000) is thought to be disease-relevant (Fujiwara et al. 2002; Anderson et al. 2006). Several kinases can phosphorylate AS at this residue, such as casein kinases (Okochi et al. 2000), G protein-coupled receptor kinases (GRK2 and GRK5) (Pronin et al. 2000), and polo-like kinases 2 and 3 (Mbefo et al. 2010). Here we subjected (Thy1)-h[A30P]AS TG mice after weaning (5 weeks of age) to a lifelong HFD and investigated the effect on lifespan, terminal locomotor phenotype, neuropathology and neurodegeneration.
Materials and methods
Animals and diets
Transgenic mice of C57BL/6 background expressing human mutant [A30P]AS under the control of the CNS neuronp-specific Thy1 promotor (Kahle et al. 2000a; Neumann et al. 2002) were maintained as a homozygous colony. Wild-type (WT) controls were derived from the same TG outcross with C57BL/6 mice and maintained as a parallel colony. Only male mice were used in this study. From the age of 5 weeks onward, homozygous (Thy1)-h[A30P]AS mice (n = 56) were either kept on standard chow diet (SD, n = 56) (3.8% total fat, 3.1 kcal/g, sniff R/M H Extrudat; ssniff Spezialitäten GmbH, Soest, Germany) or on HFD (22.8% total fat, 4.6 kcal/g, TD.06415 Adjusted Calories Diet 45/Fat; Teklad Custom Research Diets, Harlan Laboratories, Boxmeer, the Netherlands) and followed throughout their life span. WT controls (n = 52) were kept on HFD. All animal procedures were approved by local government authorities for animal research (file references N2/10 and N9/12) according to the guidelines of laboratory animal care.
Glucose tolerance test
Mice were fasted overnight (12 h) and subsequently 2 g glucose per kg body weight (Glukosteril 50%; Fresenius Kabi GmbH, Bad Homburg, Germany) was injected intraperitoneally. Blood glucose was measured from tail bleeds right before injection as well as 15, 30, 60, 90 and 120 min afterward using the Accu-Check Sensor (Roche, Mannheim, Germany).
CatWalk locomotion assays
The CatWalk system (CatWalk XT; Noldus Information Technology, Wageningen, the Netherlands) was used to analyze locomotor phenotype. The camera was fixed 22 cm under the glass plate and a camera gain was fixed on 37, intensity threshold on 0.27. Data were processed and analyzed using CatWalk XT 9.0 (Noldus Information Technology). Mice were trained to pass the CatWalk at the age of 23 weeks and subsequently tested every second week. Each trial consisted of three single runs, which were only compliant when shorter than 10 s and with a maximum variance lower than 60%.
Histological analyses were performed as described before (Schell et al. 2009, 2012) with brain slices from (Thy1)-[A30P]AS mice and WT mice at different ages. Fresh mouse brains were fixed in 4% paraformaldehyde in a phosphate buffer (pH 7.4) and embedded in paraffin. Sections (4 μm) were incubated at 60°C for at least 3 h and deparaffinized and hydrated with xylene and a graded ethanol series. Peroxidase activity was eliminated by 1% H2O2 treatment for 30 min. Antigen retrieval was performed by heating sections to 90°C for 30 min in citrate buffer (pH 6.0). Sections were blocked with 5% normal goat serum in Tris-buffered saline (TBS) for 30 min and incubated overnight with rat monoclonal anti-AS (15G7, 1 : 5) (Kahle et al. 2000b), rabbit monoclonal anti-pS129 [clone EP1536Y (ab51253), 1 : 500; Abcam, Cambridge, MA, USA], rabbit polyclonal anti-GRK5 (ab64943, 1 : 250; Abcam), or rabbit polyclonal antibody against glial fibrillary acidic protein (GFAP ab7260, 1 : 5000; Abcam). Subsequently, sections were incubated with matching biotinylated secondary antibodies, followed by incubation with an avidin–biotin complex (all components Vectastain ABC Kit; Vector Laboratories, Burlingame, CA, USA). Signal was visualized using Vector SG-Blue (Vector Laboratories). Nuclear counterstaining was done with Nuclear Fast Red (Vector Laboratories). Sections were dehydrated and mounted with Pertex (Medite GmbH, Burgdorf, Germany).
For thioflavin S (TS) staining, sections were deparaffinized and hydrated as described above. Sections were incubated for 3 min in 0.1% TS (Sigma, Steinheim, Germany) in TBS and subsequently washed 2 × 2 min in 70% ethanol, followed by washing steps in water and TBS. Sections were then mounted using Mowiol 4-88 (Roth, Karlsruhe, Germany).
For Gallyas silver staining sections were deparaffinized and hydrated as above. Subsequently sections were oxidized in 5% periodic acid, incubated for 1 min in alkaline silver iodide and for 10 min in 0.5% acetic acid. Sections were then developed for 20 min in freshly prepared developing reagent (3 volumes reagent A: 0.2% ammonium nitrate, 0.2% silver nitrate, 1% tungstosilicic acid; 10 volumes reagent B: 5% sodium carbonate; 7 volumes reagent C: 37% formaldehyde), followed by 3 min incubation in 0.5% acetic acid, 5 min in 0.1% gold chloride and 5 min in 1% sodium thiosulfate. Nuclear staining and mounting was performed as described above.
Microscopy was performed with an Axioplan 2 Imaging microscope (Carl Zeiss, Jena, Germany) and processed with AxioVision 4.3 imaging software. Immunohistochemistry pictures were taken with an Axiocam HR, TS fluorescence pictures with an Axiocam MR. For quantification, one representative 610 × 610 μm square was taken per mouse and staining condition from brainstem microscopy pictures at 20× magnification. Staining intensity was measured using ImageJ software (National Institutes of Health, Bethesda, MD, USA) by determining first the total area and secondly the stained area. Percentage of stained area was calculated and the mean was determined for all groups.
Western blot analysis
Brainstems were dissected, homogenized and lysed at 4°C in radioimmunoprecipitation assay buffer (50 mM Tris–HCl pH 7.4, 150 mM sodium chloride, 1% NP-40, 1% sodium deoxycholic acid, 0.1% sodium dodecylsulfate, 2.5 mM tetrasodium pyrophosphate) with freshly added protease inhibitors (Complete Protease Inhibitor Cocktail Tablets; Roche). Homogenates were allowed to solubilize for 15 min on ice and subsequently centrifuged for 15 min at 30 000 g. Supernatant was removed and protein content was measured using BCA Protein Assay (Thermo Fisher Scientific, Waltham, MA, USA). Protein (20 μg) was subjected to denaturing polyacrylamide gel electrophoresis and transferred on a polyvinylidene difluoride membrane (Immobilon-P; Millipore, Billerica, MA, USA). Membranes were probed with anti-AS, anti-GRK5 or anti-GFAP (see above) and reprobed for glyceraldehyde-3-phosphate dehydrogenase (GAPDH H86504M; Meridian Life Science, Memphis, TN, USA); or probed with pT308-Akt (#4056) or anti-pS473-Akt (#4060), stripped and reprobed for total AKT (#9276) (all Akt antibodies from Cell Signaling Technology, Beverly, MA, USA). Horseradish peroxidase conjugated anti-rabbit or anti-mouse (Jackson ImmunoResearch, Newmarket, UK) was used as a secondary antibody. Detection was performed with Immobilon Western horseradish peroxidase substrate (Millipore). Signals were visualized on Amersham Hyperfilm for enhanced chemiluminescence (GE Healthcare, Garching, Germany). Bands were quantified by optical densitometry using ImageJ.
Data are expressed as mean ± SD unless indicated differently. All data were analyzed using Origin Pro 8G (OriginLab, Northampton, MA, USA). Significance was tested using two-sided Student's t-test or the Kaplan–Meier estimator for survival curves.
HFD-induced obesity and glucose intolerance
Within the first 2 months of life, all mice rapidly gained weight. After this juvenile phase, body weight gain leveled to a slow constant increase on SD. On HFD, adult mice gained more body weight (Figure S1a). At 1 year of age, mice on HFD weighed ≈ 50 g instead of the normal ≈ 35 g on a standard diet. There was no significant difference in the additional weight gain between TG and WT mice (Figure S1a).
To monitor the obesity-induced pre-diabetic state, glucose tolerance tests were performed with mice aged 13 weeks (Figure S1b) and 43 weeks (Figure S1c). As expected, mice on SD normalized blood glucose levels significantly faster than those on HFD at both time points (Figure S1b and c). Consistent with the same elevated body weights of fat WT and TG mice, there was no difference in glucose tolerance between WT and TG mice.
HFD reduces life span of TG mice
(Thy1)-h[A30P]AS mice die prematurely when reaching the locomotor end stage disabling them to move and feed themselves. At that point, mice are euthanized. On SD, first (Thy1)-h[A30P]AS mice reached end stage after 15 months. By 24 months, all mice in the cohort were dead. The average life span of the SD cohort was 20.1 ± 1.74 months (Fig. 1a). In contrast, first TG mice on HFD reached end stage already after 9 months. The average life span of (Thy1)-h[A30P]AS mice on HFD was 16.5 ± 3.46 months, significantly shorter than on SD (Fig. 1a). It should be noted that several mice on HFD faded away (precipitous weight loss likely due to diabetic problems) without developing the locomotor end stage phenotype characteristic for (Thy1)-h[A30P]AS mice. However, even when excluding these dropouts, the HFD TG mice that reached the end point were on average significantly younger than those on SD (Fig. 1b).
As the terminal point of (Thy1)-h[A30P]AS mice is due to inability to move, locomotor functions were measured in a subcohort of animals. On SD, TG mice on average behaved relatively normally on the CatWalk up to ≈ 20 months. Then they generally lost locomotion, shortly preceding death (see above). By comparison, TG mice on HFD failed on the CatWalk significantly sooner (Fig. 1c). Detailed information for the CatWalk parameters measured is given in Figure S2. Thus, HFD-induced obesity accelerates the onset of terminal locomotor phenotype in (Thy1)-h[A30P]AS TG mice.
HFD accelerates the onset of terminal brainstem α-synucleinopathy in TG mice
To correlate the accelerated terminal phenotype of AS TG mice with markers of α-synucleinopathy, we sacrificed mice and performed immunostaining for human TG AS and its pathology-associated form pS129 in pontine brainstem sections. Qualitatively, the AS staining patterns were the same regardless of diet. Immunostaining of total human TG AS showed the normal synaptic distribution and in addition somal accumulations (Figure S3a–f). The pathological signal-to-noise ratio was more impressive with anti-pS129 (Fig. 2a–f). HFD-fed WT mice did not show any specific signals, even at oldest age (Figure S3g and h and Fig. 2g and h). Higher magnification confirmed the pathological morphologies of pS129 immunostaining patterns resembling Lewy neurites (Fig. 2c′), spheroid-like structures (Fig. 2c″), and Lewy body-like accumulations in neuronal cell bodies (Fig. 2d) in HFD-fed TG mice aged 16 months, when pS129 just appeared in TG mice on SD (Fig. 2a and b) and only later developed into similar pS129 immunostaining patterns (Fig. 2e and f). Densitometric measurements of pS129 staining intensities confirmed the significant increase in 16-month-old HFD-fed TG mice to a level that TG mice on SD reached on average only later at 20 months (Fig. 2i). Western blot analysis confirmed the trend of increased amounts of pS129 in brainstem lysates of 16-months-old TG mice after HFD compared to age-matched TG mice on SD, while total AS was only slightly elevated (Figure S3i and j).
To check if the observed enhancement of pS129 would correlate with increases of candidate kinases, we checked GRK5 (Arawaka et al. 2006) using antibodies that reacted reliably on sections and Western blots from brainstem lysates (Fig. 2j–t). In the oldest TG mice, strong increases in GRK5 signals were detected (Fig. 2s). However, in younger mice, no such GRK5 accumulations were detected, even after life-long HFD. Densitometric band quantifications confirmed the significant GRK5 increase only for end stage (Thy1)-[A30P]AS mice (Fig. 2t). Anti-GRK5 immunostainings confirmed the biochemical measurements (Fig. 2j–r). Interestingly, the enhanced GRK5 immunostaining in end stage (Thy1)-[A30P]AS mice was found on neuronal cell bodies and in the nucleus (Fig. 2n and o), consistent with the known localization of GRK5 at the plasma membrane and its nuclear translocation (Johnson et al. 2004; Bychkov et al. 2012).
The amyloid-like quality of α-synucleinopathy was detected with the dye TS. Similar to the apparently pathological AS staining patterns, (Thy1)-h[A30P]AS mice on SD showed no TS positive Lewy body-like profiles at earlier time points (Fig. 3a and b) when HFD-fed TG mice already showed the TS neuropathology (Fig. 3c and d) that developed later in life in TG mice on SD (Fig. 3e and f). WT controls did not show any specific TS staining (Fig. 3g and h). Densitometric measurements of TS staining intensities confirmed the significant increase in 16-month-old HFD-fed TG mice to a level that TG mice on SD reached on average only later at 20 months (Fig. 3i).
Dystrophic neurites as detected by silver staining were distributed alike in both groups. However, HFD-fed (Thy1)-h[A30P]AS mice showed silver-positive neuronal dystrophy at earlier time points (16 months), when silver staining hardly revealed any neuropathology in TG mice on SD (Fig. 3j–m). Only at later time points (20 months) did TG mice on SD develop silver-positive neuronal dystrophy (Fig. 3n and o). In contrast, WT mice on HFD did not show silver staining up to 20 months of age (Fig. 3p and q). Higher magnification confirmed the pathological morphologies of silver staining patterns mostly resembling Lewy neurites (Fig. 3l′ and m′) and some in neuronal cell bodies (Fig. 3l″ and m″) in HFD-fed TG mice aged 16 months. Only later developed similar silver staining patterns in TG mice on SD (Fig. 3n and o). Densitometric measurements of the silver staining intensity confirmed the significant increase in 16-month-old HFD-fed TG mice to a level that TG mice on SD reached on average only later at 20 months (Fig. 3r).
HFD accelerates the onset of astrogliosis in TG mice
Glial fibrillary acidic protein-positive astrogliosis often highlights neurodegenerative changes in TG mouse models, including the one used here (Neumann et al. 2002). (Thy1)-h[A30P]AS mice on HFD showed massive astrogliosis at earlier time points that were overall less conspicuous on SD at the age of 16 months (Fig. 4a–d). Astrogliosis became massive in TG mice on SD later in life (Fig. 4e and f), when the phenotype occurred in due course. WT controls showed much less GFAP signals, offering no evidence for strong astrogliosis in the brainstem by HFD alone (Fig. 4g and h). Higher magnifications revealed the enlarged and stellate morphology of reactive astrocytes in relatively young HFD-fed TG mice (Fig. 4c and d) and old SD-fed TG mice (Fig. 4e and f), compared to younger TG mice on SD (Fig. 4a and b) and controls on HFD (Fig. 4g and h). Densitometric measurements of GFAP staining intensities confirmed the significant increase in 16-month-old HFD-fed TG mice to a level that TG mice on SD reached on average only later at 20 months (Fig. 4i). Western blot analysis confirmed the increased amounts of GFAP in brain stem lysates of 17-month-old TG mice after HFD compared to the age-matched TG mice on SD (Fig. 4j). TG mice on SD showed higher amounts of GFAP later at the age of 20 months (Fig. 4j). In contrast, WT mice hardly showed any GFAP signals even at the oldest age. Thus, astrogliosis seems to coincide with the accelerated onset of neurodegenerative α-synucleinopathy in HFD-fed TG mice.
HFD reduces Akt phosphorylation in TG mice
In addition to neuroinflammatory mechanisms, HFD-induced obesity may also sensitize to neurodegeneration by insulin resistance (Kim and Feldman 2012). A key signaling hub for insulin and in general for neurotrophic factors signaling via receptor tyrosine kinases is Akt (Hemmings and Restuccia 2012). Activation of Akt involves phosphorylation at threonine-308 within the kinase domain as well as phosphorylation at serine-473 at the carboxy-terminus. Although the variability of Akt phosphorylation signals in whole brainstem lysates was fairly high, a reduction of pT308-Akt (Fig. 5a) and pS473-Akt (Fig. 5b) could be observed in 17-month-old HFD-fed TG mice compared to age-matched TG mice on SD.
Here, we show that HFD-induced obesity accelerates the onset of terminal locomotor phenotype in (Thy1)-[A30P]AS mice, accompanied by earlier α-synucleinopathy and astrogliosis. Protein aggregation and neuroinflammation are considered common dysregulated pathways in diabetes and PD (Santiago and Potashkin 2013). Additionally, glucose intolerance and insulin resistance may lead to reduced neurotrophic signaling directly in CNS neurons, or via cerebrovascular pathology. These mechanistic aspects remain to be further elucidated. As an initial approach, we have determined Akt phosphorylation states, as Akt is a signaling hub for insulin and neurotrophic factors (Hemmings and Restuccia 2012). Moreover, single nucleotide polymorphisms in the AKT1 gene might increase the risk for PD and diabetes (Xiromerisiou et al. 2008). Although there was considerable individual animal variability in total brain stem lysate phospho-Akt levels (Fig. 5), the significant reduction in 17-month-old TG mice fed HFD compared to SD could point to insulin resistance as a contributing factor for the observed accelerated onset of neurodegenerative phenotype in this mouse model of α-synucleinopathy.
Astrogliosis seems to follow neurodegeneration in TG brainstem, as we did not detect such an extent of GFAP staining at the age of 1 year regardless of diet or genotype (data not shown). HFD-induced obesity can stimulate astrogliosis in the mouse CNS, but in a very brain region-specific manner (Buckman et al. 2013). In our experiments, overt astrogliosis does not seem to precede α-synucleinopathy and neurodegeneration by months in the brainstem. However, we cannot rule out that astrogliosis precedes the phenotypic onset by weeks or days, potentially triggering a neuroinflammatory cascade, or at least driving a vicious cycle for α-synucleinopathy (Fellner et al. 2011).
The pS129 is a marker of α-synucleinopathy, but the causal relationship with neurotoxic AS aggregation is not fully understood (Sato et al. 2013). We found that GRK5 steady-state levels clearly increased in oldest (Thy1)-[A30P]AS mice. Thus, GRK5 is a candidate kinase for pS129 modification at end stage, regardless of diet. It remains to be determined if HFD-induced AS neuropathology correlates with kinase activities. AS candidate kinases include casein kinases (Okochi et al. 2000), GRK2 and GRK5 (Pronin et al. 2000), and the polo-like kinases 2 and 3 (Mbefo et al. 2010). Conversely, pS129 could also be regulated at the phosphatase level (Lee et al. 2011; Pérez-Revuelta et al. 2014).
Our data are in contrast to a very recent report on (Thy1)-[A53T]AS mice exposed to a high-calorie diet with additional sugar supplementation (Rothman et al. 2014). Unlike our (Thy1)-[A30P]AS mice fed with HFD with a comparatively mild caloric enrichment that gained weight like WT, the (Thy1)-[A53T]AS were practically resistant to obesity and metabolic effects in response to a high calorie/sugar diet. Thus, in addition to potential genetic background differences, the diet quality may have a dramatic influence on the outcome of AS effects. Although it generally emerges that nutritional factors can have a significant impact on α-synucleinopathy, clearly more work is necessary to establish the specific details and mechanisms by which HFD-induced obesity, diabetes and α-synucleinopathiy might be interconnected. Our study could provide a starting point, documenting that this HFD-induced obesity can accelerate the onset of neurodegenerative phenotypes in (Thy1)-[A30P]AS mice.
Acknowledgements and conflict-of-interest disclosure
This work was supported by the German Center for Neurodegenerative Diseases, the Helmholtz Alliance ‘Aging Brain’, and the Hertie Foundation. We thank Cindy Boden for technical assistance and Anita Hennige and Tina Sartorius for helpful discussions. The authors have no conflicts of interest to declare.
All experiments were conducted in compliance with the ARRIVE guidelines.