Phosphorylation of the amyloid β-peptide at Ser26 stabilizes oligomeric assembly and increases neurotoxicity

Synthetic non-phosphorylated Aβ1–40 (npAβ), phosphorylated Aβ1–40 variants (pSer8Aβ and pSer26Aβ) and other modified Aβ (Tyr10 nitrated, Glu3 pyroglutamate and truncated 3–42) peptides were purchased from Peptide Speciality Laboratory (PSL, Germany). Thioflavin T, 4′,6-diamidino-2′phenylindole dihydrochloride (DAPI), 3,3′-diamino-benzidine (DAB) and methanol were from Sigma-Aldrich (USA). Congo red was purchased from AppliChem GmbH (Germany). Precast 4–12 % NuPAGE Bis–Tris mini and midi gels, prestained protein molecular weight markers and PrestoBlue^® cell viability reagent were from Life technologies (Germany). Nitrocellulose membranes were from Schleicher and Schuell (Germany). ECL Western blotting detection reagents were from GE Healthcare (UK). Vectastain ABC kit and hematoxylin were from Vector laboratories (USA). Protease and phosphatase inhibitors were from Roche laboratories (Germany). BCA™ protein assay kit was from Thermo Scientific (USA). Monoclonal Aβ antibodies 6E10 and 4G8 were purchased from Covance Laboratories (USA), and 82E1 antibody was from IBL Corporation (Japan). Mouse monoclonal GFAP antibody was from Synaptic systems (Germany), and 22C11 antibody specific against amyloid precursor protein (APP) (a.a. 66–81 of APP at N-terminus) was from Merck Millipore (Germany). A Mouse monoclonal Phospho-PHF-tau specific AT8 antibody was purchased from Thermo scientific (USA). Rabbit polyclonal anti-CK1δ (antiserum 108) and anti-CK1ε (antiserum 712) were generously provided by Dr. Uwe Knippschild from University Hospital Ulm, Germany. The anti-mouse, anti-rabbit secondary antibodies conjugated to horseradish peroxidase were from Sigma Aldrich (Germany), Secondary fluorescent anti-mouse 594 DyLight, anti-rabbit 488 antibodies were from Thermo Scientific (USA), IRDye800CW and IRDye680RD were from LI-COR Biotechnology. Biotinylated secondary anti-mouse and anti-rabbit antibodies were from DAKO (Glostrup, Denmark). The dilutions of each antibody stock are mentioned for the respective methods or in figure legends.

The pSer26Aβ-specific polyclonal antibody SA6192 was generated in rabbits by injecting synthetic Aβ19–31 peptides with Ser26 in phosphorylated state (antigen sequence: FFAEDVG (p) SNKGAI) conjugated with keyhole limpet hemocyanin (KLH) (Eurogentec, Belgium). Phosphorylation state-specific antibodies were purified from the serum by double-affinity purification using pSer26Aβ and npAβ peptide. The specificity of the purified antibodies was characterized by enzyme-linked immunosorbent assay (ELISA) and Western blotting (WB). Further details are described in the Supplementary Information.

For biochemical analysis of pSer26Aβ, whole brain homogenates from APP/PS1KI were prepared as described previously [26, 29]. Immunohistochemistry was performed on 4 µm sagittal paraffin sections as described previously [58]. Further details of Aβ extraction and immunohistochemistry of transgenic mouse brains are described in the Supplementary Information.

Human autopsy brains were received from University Hospital Bonn (Germany) and from University Hospital Ulm (Germany) in accordance with the laws and the permission of the local ethical committees. Post-mortem diagnosis of Alzheimer’s disease was carried out according to the NIA-Reagan Criteria [6, 37]. All procedures were conducted in accordance with the laws and the permission of the local ethical committees. Further detailed methods and information on cases are given in the Supplementary Information.

Aβ aggregation kinetics by Thioflavin T (ThT) and Congo Red (CR) binding assays were performed as described previously [26]. Morphology of the aggregates was characterized by transmission electron and atomic force microscopy. Further details are given in the Supplementary Information.

Cell viability assays were carried out with human neuroblastoma cells (SK-N-SH), embryonic stem cell (ES)-derived neurons and induced pluripotent stem cell (iPSC)-derived neurons. Further details on cultivation and assay procedures are described in the Supplementary Information.

Post-translational modifications could alter the aggregation, degradation and toxicity of Aβ [26‐28, 30, 43, 57]. Synthetic Aβ peptides phosphorylated on either Ser8 or Ser26 showed faster formation of oligomeric assemblies in vitro [26, 41]. To specifically investigate Ser26-phosphorylated Aβ (pSer26Aβ) species in vivo, we generated phosphorylation state-specific antibodies (Fig. 1a). Double-affinity purified antibody SA6192 was highly specific for Aβ phosphorylated at Ser26 (Fig. 1b). It did not detect Ser8 phosphorylated (pSer8Aβ), pyroGlu-modified (pyroAβ3–42), N-terminally truncated (Aβ3–42), or nitrosylated (3NTyr10-Aβ) Aβ variants (Fig. 1c), while the generic 4G8 antibody which recognizes an epitope between amino acids 17 and 24 of the Aβ domain, detected npAβ (Fig. 1b) and all the tested peptide variants similarly (Fig. 1c). SA6192 did not detect full-length APP or its C-terminal fragments in brain extracts of transgenic mice, suggesting selective phosphorylation of Ser26 after the generation of Aβ (Supplementary Fig. 1a, b). Detection of pSer26Aβ by the SA6192 antibody was efficiently blocked with synthetic pSer26Aβ, but not with synthetic npAβ peptide, further demonstrating the specificity of this antibody (Supplementary Fig. 1c, d). We took advantage of the SA6192 antibody to characterize the deposition of pSer26Aβ peptides in transgenic mouse brains. Western immunoblot analysis of brain extracts from APP/PS1KI transgenic mice showed the presence of pSer26Aβ peptides in water-soluble (predominantly containing extracellular soluble Aβ) and in SDS-soluble fractions (predominantly containing intracellular and membrane-associated Aβ) at 6 months of age (Fig. 1d, e). pSer26Aβ reactivity was not detected in non-transgenic mouse brains. The quantification of the SA6192 immunoreactivity of water-soluble APP/PS1KI transgenic mouse brain extracts using synthetic pSer26Aβ as standard revealed that ~10–15 % of extracted monomeric Aβ is in a phosphorylated state (50 μg protein from water-soluble brain extracts contained ~0.17 ± 0.03 ng of pSer26Aβ and ~1.55 ± 0.09 ng of total Aβ).

In 2-month-old transgenic mice, pSer26Aβ was not detectable by Western immunoblotting (Fig. 1d, e). Interestingly, immunohistochemistry revealed abundant deposition of pSer26Aβ intraneuronally in 2-month-old animals when extracellular plaques were hardly detectable (Fig. 1f). The pronounced intraneuronal reactivity was also detected in older mice in different brain regions. Occasionally, extracellular deposits were also positive for pSer26Aβ (Fig. 1f). Double-labelling with pSer26Aβ and generic Aβ antibodies specifically demonstrate preferential intraneuronal accumulation of pSer26Aβ in the presence of pronounced extracellular plaques (Fig. 1g). As compared to staining with generic Aβ antibodies, pSer26Aβ reactivity was restricted to structures in the core of the plaques in aged transgenic mouse brains (10 months) (Supplementary Fig. 2a). Double-staining also revealed the association of reactive astrocytes in the vicinity of neurons with intracellular pSer26Aβ (Supplementary Fig. 2b). Reactivity of the SA6192 antibody was not observed in control mice (Supplementary Fig. 3c). SA6192 immunoreactivity in transgenic mice was efficiently blocked by synthetic pSer26Aβ, further supporting the specific detection of pSer26Aβ deposits (Supplementary Fig. 3b, c). Additional immunohistochemical staining of 6- and 12-month-old 5XFAD mouse brains also demonstrated intraneuronal accumulation of pSer26Aβ aggregates and few extracellular pSer26Aβ positive plaques (Supplementary Fig. 4). These data demonstrate a unique pattern of deposition of pSer26Aβ that differs from that of other Aβ species, including post-translationally modified variants like pSer8Aβ [26, 29] or pyroglutaminated Aβ [43, 57].

Human AD brains also revealed a specific accumulation pattern of pSer26Aβ. pSer26Aβ could be detected in individual cored-neuritic plaques and partially overlapped with the pattern of antibodies raised against the middle region of Aβ (4G8; epitope 17–24) (Fig. 2a, b; Supplementary Fig. 5a, b). pSer26Aβ was also detected in APP-positive dystrophic neurites in these plaques (Fig. 2c, d). Anti-pSer26Aβ also stained additional material indicating the deposition of pSer26Aβ within extracellular plaques (Supplementary Fig. 5a, c). A considerable number of diffuse plaques were also stained with anti-pSer26Aβ (Supplementary Fig. 5d). By analysis of the distinct plaque-types occurring in the medial temporal lobe, pSer26Aβ-positive material was restricted to diffuse, cored and neuritic plaques as well as subpial band-like amyloid (Supplementary Fig. 5d), whereas fleecy amyloid and presubicular lake-like amyloid was not stained in the cases studied here (Supplementary Table 1). Moreover, staining of pSer26Aβ in amyloid plaques was restricted to the symptomatic AD cases observed here. No pSer26Aβ-positive plaques are observed in pathologically diagnosed preclinical AD (p-preAD) cases.

Notably, pSer26Aβ was also detected inside of neurons that showed no or only faint reactivity with antibodies against APP or generic Aβ (Fig. 2a–d, arrowhead). The diffuse neuronal staining was not only detected in AD, but also in pathological pre-AD and even in control cases. The intraneuronal pSer26Aβ showed the typical morphological pattern of granulovacuolar degeneration (GVD) [48]. The morphological distribution of the pSer26Aβ-positive granules predominantly in neurons of the CA1-subiculum regions of the hippocampal formation fitted with that known for GVD (Fig. 2e–g) [48]. Interestingly, most neurons with GVD lesions also contained abnormally phosphorylated τ-protein (Fig. 2h–j), phosphorylated transactive response DNA-binding protein (pTDP43) and casein kinase 1δ/ε (Supplementary Fig. 5e–h). Notably, CK1 indeed could phosphorylate Aβ at Ser26 (Supplementary Fig. 6a–d), suggesting a phosphorylation of Aβ by CK1 in GVD compartments. Interestingly, pSer26Aβ was consistently detected together with pTDP43 in GVD, even in p-preAD and non-AD control cases (Supplementary Table 2).

Phosphorylation of Aβ at Ser26 alters plasticity of a critical turn region and impairs fibrillization [41]. Accordingly, pSer26Aβ showed strongly reduced binding of Congo Red (CR) and Thioflavin T (ThT) as compared to npAβ (Fig. 3a; Supplementary Fig. 7a). In contrast, phosphorylation at Ser8 strongly increased CR and ThT binding. Kinetic analysis revealed a slight but very rapid increase in ThT binding of pSer26Aβ that did not further increase over time (Supplementary Fig. 7a; inset), indicating rapid formation of smaller assemblies without proceeding to fibril formation (Fig. 3b; Supplementary Fig. 7b). In denaturing (Fig. 3c) and non-denaturing PAGE (Fig. 3d), pSer26Aβ was detected as smaller oligomeric assemblies (i.e., dimers and trimers) already at very short incubation periods that persist even after longer incubation time, consistent with a very rapid self-assembly of this Aβ variant. As already indicated by the CR and ThT binding assays, pSer8Aβ reached a higher aggregation state than npAβ represented by the increased reactivity in the upper parts of the gel (Fig. 3b–d; Supplementary Fig. 7b). Even after longer incubation periods, pSer26Aβ only formed intermediate oligomeric forms migrating between 30 and 80 kDa in SDS-containing denaturing gels (Fig. 3c) and between 30 and 400 kDa in native gels (Fig. 3d). npAβ and pSer8Aβ formed higher oligomeric and fibrillar (<1000 kDa) assemblies (Fig. 3c, d). Transmission electron microscopy (TEM) and atomic force microscopy (AFM) revealed only heterogeneous globular species of various sizes of pSer26Aβ without formation of fibrillar structures, as seen with npAβ peptide (Fig. 3e, f; Supplementary Fig. 8).

To assess the toxicity of the pSer26Aβ in a human neuronal model, we used human neuroblastoma cells, human neurons differentiated from embryonic stem cell (ESC) and induced pluripotent stem cell (iPSC)-derived neural stem cells (lt-NES) cells. In a first set of experiments, the different Aβ variants were added without prior aggregation to human neuroblastoma cells (Fig. 4a) and to differentiated neurons (Fig. 4b). As compared to the non-phosphorylated peptide, Aβ pseudophosphorylated at position 26 (AβS26D) induced increased toxicity in neuroblastoma cells (Fig. 4a) and also in hESC-derived neurons (Fig. 4b). Even at concentrations when npAβ showed no overt toxicity, AβS26D impaired neuronal metabolism comparable to the effect of a tenfold higher concentration of npAβ in both neuroblastoma cells (Fig. 4a) and human ESC-derived neurons (Fig. 4b). To specifically assess the toxic properties of different Aβ variants depending on their aggregation state, we next exposed human iPSC-derived neurons to preformed assemblies of npAβ, pSer8Aβ and pSer26Aβ. Dot blot analysis of the different preformed Aβ assemblies showed significant differences in their immunoreactivity against conformation-dependent anti-oligomer antibodies such as A11 (Fig. 4c) and OC (Fig. 4d; Supplementary Fig. 9) [21, 22]. npAβ and pSer8Aβ assemblies were detected by both A11 and OC antibodies after 2–6 h of incubation. However, both antibodies revealed only very little if any reactivity for assemblies formed by pSer26Aβ (Fig. 4c, d). Native-PAGE analysis showed that the samples from different incubation times contained assemblies of different sizes (Fig. 4e). After 2–6 h of aggregation, samples of npAβ and pSer8Aβ contained oligomers of intermediate size (150–480 kDa). After 6–24 h, npAβ and pSer8Aβ also formed high molecular weight assemblies, which were not detected with pSer26Aβ. Instead, oligomers of intermediate size formed by pSer26Aβ were prominently detected at 24 h of incubation. Monomeric and dimeric Aβ were detected at all time points. However, these forms might also result from dissociation of aggregates, even during native-PAGE conditions. Notably, npAβ and pSer8Aβ variants exerted toxicity only at 6 and 2 h of pre-aggregation, respectively, and lost their toxic activity during extended aggregation (Fig. 4e, f). After longer aggregation periods, toxicity of npAβ and pSer8Aβ was decreased. A similar behaviour was previously observed for pyroE3-modified Aβ [39]. Compared to npAβ and pSer8Aβ, pSer26Aβ exerted strongest toxicity (Fig. 4f). Toxicity of pSer26Aβ was observed after 24 h of pre-aggregation. Interestingly, during this time of incubation, pSer26Aβ also formed intermediate-size oligomers of 150–480 kDa, but no larger assemblies or fibrils (Fig. 4e, f).