Site-specific phosphorylation of tau inhibits amyloid-β toxicity in Alzheimer’s mice

See allHide authors and affiliations

Science  18 Nov 2016:
Vol. 354, Issue 6314, pp. 904-908
DOI: 10.1126/science.aah6205

Tau phosphorylation—not all bad

Alzheimer's disease presents with amyloid-β (Aβ) plaques and tau tangles. The prevailing idea in the field is that Aβ induces phosphorylation of tau, which in turn mediates neuronal dysfunction. Working in Alzheimer's disease mouse models, Ittner et al. found evidence for a protective role of tau in early Alzheimer's disease. This protection involves specific tau phosphorylation at threonine 205 at the postsynapse. A protective role of phosphorylated tau in disease challenges the dogma that tau phosphorylation only mediates toxic processes.

Science, this issue p. 904


Amyloid-β (Aβ) toxicity in Alzheimer’s disease (AD) is considered to be mediated by phosphorylated tau protein. In contrast, we found that, at least in early disease, site-specific phosphorylation of tau inhibited Aβ toxicity. This specific tau phosphorylation was mediated by the neuronal p38 mitogen-activated protein kinase p38γ and interfered with postsynaptic excitotoxic signaling complexes engaged by Aβ. Accordingly, depletion of p38γ exacerbated neuronal circuit aberrations, cognitive deficits, and premature lethality in a mouse model of AD, whereas increasing the activity of p38γ abolished these deficits. Furthermore, mimicking site-specific tau phosphorylation alleviated Aβ-induced neuronal death and offered protection from excitotoxicity. Our work provides insights into postsynaptic processes in AD pathogenesis and challenges a purely pathogenic role of tau phosphorylation in neuronal toxicity.

Alzheimer’s disease (AD), the most prevalent form of dementia, is neuropathologically characterized by extracellular amyloid-β (Aβ) plaques and intracellular tau-containing neurofibrillary tangles (1, 2). Growing evidence suggests that Aβ and tau together orchestrate neuronal dysfunction in AD (3), although their molecular connections remain poorly understood. Aberrant tau phosphorylation is the first step in a cascade leading to its deposition and to cognitive dysfunction (4, 5). Aβ is thought to trigger toxic events, including tau phosphorylation (6). Accordingly, the depletion of tau prevents Aβ toxicity in AD models (79). Aβ-induced neuronal network and synaptic dysfunction is associated with aberrant glutamatergic synaptic transmission (10). N-methyl-d-aspartate (NMDA)–type glutamatergic receptors (NRs) drive glutamate-induced neuronal excitotoxicity (11) and mediate Aβ toxicity by downstream responses that promote neuronal dysfunction (12).

Multiple factors, including p38 kinases, contribute to NR-mediated toxicity (12). Although inhibition of p38α and p38β improves Aβ-induced long-term potentiation deficits (13, 14), it increases hyperexcitability in Aβ precursor protein (APP) transgenic mice (15). However, the specificity of p38 inhibitors remains debatable (16, 17). Other p38 kinases, p38γ and p38δ, have not been studied in AD. To understand the roles of p38 kinases in AD, we induced excitotoxic seizures with pentylenetetrazole (PTZ), an approach widely used for studying excitotoxicity in AD mouse models (8, 9). We used mice with individual deletion of p38α, p38β, p38γ, or p38δ (fig. S1). Surprisingly, only p38γ depletion (p38γ−/−), but not systemic p38β, p38δ, or neuronal p38α (p38αΔneu) knockout, changed PTZ-induced seizures (Fig. 1A and fig. S2). Pan-p38 inhibition in wild-type (WT) mice augmented seizures, similar to the effects of p38γ depletion (fig. S3). Consistent with a postsynaptic role, only p38γ localized to dendritic spines and postsynaptic densities (PSDs) of neurons and was enriched in PSD preparations (Fig. 1B and fig. S4). Hence, of all p38 kinases, only p38γ localized to postsynapses and limited excitotoxicity.

Fig. 1 Depletion of postsynaptic p38γ exacerbates excitotoxicity and deficits in APP transgenic mice.

(A) Reduced seizure latency (linear regression slopes) and increased seizure severity in p38γ−/− mice compared with p38γ+/+ animals injected with 30 mg/kg pentylenetetrazole (PTZ) (n = 10 to 12 mice). (B) Colocalization of p38γ and postsynaptic PSD-95 (arrows), but not presynaptic synaptophysin (Syp), in neurons. Scale bars, 1 μm. (C) Mortality in APP23.p38γ+/+ mice was further augmented in APP23.p38γ−/− animals, whereas p38γ+/+ and p38γ−/− mice survived normally (n = 43 to 62). (D to F) Memory deficits in 4-month-old APP23.p38γ+/+ and APP23.p38γ−/− mice (memory deficits were exacerbated in the APP23.p38γ−/− animals). Morris water maze (MWM) test: (D) Representative MWM path traces to a hidden platform. (E) Increased escape latency and (F) reduced time in the target quadrant during probe trials in APP23.p38γ+/+ and, more so, APP23.p38γ−/− mice compared with p38γ+/+ and p38γ−/− mice (n = 8 to 10). (G to H) Increased (G) spike train and (H) spike frequency in APP23.p38γ+/+ and APP23.p38γ−/− mice but not p38γ+/+ and p38γ−/− mice (n = 6 to 8). (I) Representative phase-amplitude comodulograms of interictal hippocampal local field potential recordings showed reduced and virtually lost cross-frequency coupling (CFC) (~8 Hz) in APP23.p38γ+/+ and APP23.p38γ−/− mice, respectively, compared with p38γ+/+ and p38γ−/− mice. (J) Reduced modulation index in APP23.p38γ+/+ mice and, more so, in APP23.p38γ−/− mice compared with p38γ+/+ and p38γ−/− animals (n = 6 to 8). For (A), (C), (E) to (H), and (J): ****P < 0.0001, ***P < 0.001, **P < 0.01, *P < 0.05; ns, not significant. Error bars indicate SEM.

To test whether p38γ−/− augments Aβ-induced deficits, we crossed p38γ−/− with Aβ-forming APP23 mice. APP23 mice present with premature mortality, memory deficits, neuronal circuit aberrations with epileptiform brain activity, and Aβ pathology (9, 15, 18). Compared with APP23.p38γ+/+ mice, APP23.p38γ−/− animals had increased sensitivity to PTZ-induced seizures (fig. S5). Aβ pathology was comparable in the brains of APP23.p38γ−/− and APP23.p38γ+/+ mice (fig. S6), but p38γ deletion aggravated premature mortality and memory deficits of APP23 mice (Fig. 1, C to F, and fig. S7). p38γ−/− mice showed no deficits and had normal motor function (fig. S8). Electroencephalography showed enhanced spontaneous epileptiform activity and interictal hypersynchronous discharges in APP23.p38γ−/− compared with APP23.p38γ+/+ mice (Fig. 1, G and H, and fig. S9). Hippocampal theta and gamma oscillations and cross-frequency coupling (CFC) through theta-phase modulation of gamma power are measures of network activity related to memory, including in humans (1922). These measures are compromised in APP transgenic mice (15). Compromised spectral power and CFC of APP23.p38γ+/+ mice were significantly more affected in APP23.p38γ−/− recordings (Fig. 1, I and J, and fig. S9). Recordings of p38γ−/− and p38γ+/+ mice were indistinguishable. In summary, p38γ depletion exacerbated excitotoxicity, neuronal circuit synchronicity, mortality, and memory deficits in APP23 mice, without changes in Aβ pathology. In addition, p38γ levels were reduced in aged APP23 and APPNL-G-F mice and humans with AD (fig. S10), further suggesting that the loss of p38γ-mediated neuroprotection may contribute to AD pathogenesis.

To determine whether the Aβ toxicity–limiting effects of p38γ were tau-dependent, we crossed APP23.p38γ−/− with tau−/− mice. The exacerbating effects of p38γ loss on reduced survival, memory deficits, and neuronal network dysfunction of APP23 mice were virtually abolished in APP23.p38γ−/−.tau−/− mice (Fig. 2, A to C, and fig. S11). These data also showed that, compared with APP23 mice, APP23.p38γ−/− animals had aggravated memory deficits that persisted with aging. In contrast, increasing tau levels in p38γ−/− mice [brought about by crossing with nonmutant tau-expressing Alz17 mice (23)] significantly enhanced PTZ-induced seizures in Alz17.p38γ−/− mice (Fig. 2, D to F). Conversely, when compared to tau−/−.p38γ+/+ mice, tau−/−.p38γ−/− animals showed similar protection from PTZ-induced seizures (Fig. 2, G to I). Taken together, the effects of p38γ on excitotoxicity and Aβ toxicity were tau-dependent.

Fig. 2 Tau is required for p38γ-mediated inhibition of Aβ and excitotoxicity.

(A) Normal survival of APP23.p38γ−/−.tau−/− compared with APP23.p38γ−/− and APP23.p38γ+/+ mice (n = 42 to 62). (B and C) Normal memory in 12-month-old APP23.p38γ−/−.tau−/− mice compared with APP23.p38γ−/− and APP23.p38γ+/+ mice (n = 10 to 12). MWM test: (B) Escape latency and (C) time in target quadrant during probe trials. (D to F) (D) Further reduction in seizure latencies, shown by linear regression analysis (E). (F) Further enhanced mean seizure severity following 30 mg/kg PTZ in Alz17.p38γ−/− mice versus those already reduced in p38γ−/− compared with p38γ+/+ and Alz17.p38γ+/+ mice (n = 10 to 12). (G to I) (G) Seizure latencies after 30 mg/kg PTZ were reduced, as shown by linear regression analysis (H), and mean seizure severity was increased in tau+/+.p38γ−/− compared with tau+/+.p38γ+/+ mice (I). However, latencies were markedly increased (G) and severities similarly reduced (I) in both tau−/−.p38γ+/+ and tau−/−.p38γ−/− mice (n = 10 to 12). For all panels: ****P < 0.0001, ***P < 0.001, **P < 0.01, *P < 0.05; ns, not significant. Error bars indicate SEM.

We have previously shown that postsynaptic PSD-95/tau/Fyn complexes mediate Aβ-induced excitotoxicity (9). PSD-95/tau/Fyn interaction was enhanced in Alz17.p38γ−/− animals versus Alz17.p38γ+/+ mice (Fig. 3A and fig. S12). Conversely, no PSD-95/tau/Fyn complexes were isolated from tau−/− and tau−/−.p38γ−/− brains (fig. S12). Increasing p38γ levels compromised PSD-95/tau/Fyn interaction in cells, and expression of a constitutively active p38γ variant (p38γCA) completely abolished this interaction (Fig. 3B and fig. S13). Pan-p38 inhibition stopped p38γ/p38γCA-induced disruption of PSD-95/tau/Fyn complexes (Fig. 3B). PSD-95 copurified more tau and Fyn from p38γ−/− versus p38γ+/+ brains, and even more from APP23.p38γ−/− compared with APP23.p38γ+/+ and p38γ−/− brains (Fig. 3C). Conversely, PSD-95/tau/Fyn interaction was reduced in transgenic mice with neuronal expression of p38γCA (Fig. 3D and fig. S14). PTZ transiently increased PSD-95/tau/Fyn complex formation in p38γ+/+ animals; this effect was even more noticeable in p38γ−/− mice (fig. S12). Fyn-mediated NR2B phosphorylation at Tyr1472 (Y1472) facilitates PSD-95/NR2B interaction (24). Consistent with increased PSD-95/tau/Fyn complex formation, NR2B phosphorylation at Y1472 was increased in p38γ−/− brains (fig. S15). Conversely, cellular expression of p38γ and p38γCA—but not p38αCA, p38βCA, or p38δCA—reduced Y1472 phosphorylation of NR2B (fig. S15). Hence, p38γ regulated PSD-95/tau/Fyn complexes, likely at the level of PSD-95/tau interaction (fig. S16).

Fig. 3 Active p38γ dissociates PSD-95/tau/Fyn/NR complexes.

(A) Immunoprecipitation (IP) analysis. More PSD-95/tau/Fyn complexes were immunoprecipitated from the brains of Alz17.p38γ−/− than Alz17.p38γ+/+ animals, despite comparable total protein levels. Detection of glyceraldehyde-3-phosphate dehydrogenase (GAPDH) confirmed equal loading (n = 6). Numbers at right indicate molecular weight. a.i., arbitrary index. (B) p38γ (WT) and p38γCA (CA) failed to disrupt PSD-95/tau/Fyn complexes in the presence of the p38 inhibitor (n = 6). p38 inhibition reduces phosphorylated active p38 levels (p-p38). (C) More tau, Fyn, and NMDA receptor subunits 1 (NR1) and 2B (NR2B) were immunoprecipitated with PSD-95 from p38γ−/− versus p38γ+/+ brains. This result was further enhanced in APP23.p3−/− mice compared with APP23.p38γ+/+ animals, without changes to total protein levels (n = 6 to 8). (D) Virtually no PSD-95/tau/Fyn complexes were immunoprecipitated from the brains of p38γCA-expressing mice (n = 6). For all panels: Representative blots are shown. Quantification of IP bands relative to PSD-95 IP. ***P < 0.001; **P < 0.01; *P < 0.05. Error bars indicate SEM.

Although p38γ hyperphosphorylates tau during long-term in vitro kinase assays (25), the temporal profile of p38γ-induced tau phosphorylation in acute signaling remains unknown. Short-term in vitro kinase reactions using phosphorylation site–specific tau antibodies revealed phosphorylation at Ser199 (S199), Thr205 (T205), S396, and S404 (fig. S17). Mass spectrometric analysis confirmed these and 14 additional, though low-abundant, sites (figs. S17C and S18 and table S4). Coexpression of p38γ or p38γCA and tau in cells revealed tau phosphorylation (p) at T205, less at S199, and hardly any at S396 or S404 (Fig. 4A). Similarly, T205 (and, less so, S199 and S396) were phosphorylated in p38γCA transgenic mice (fig. S19). pT205 increased after PTZ in p38γ+/+ animals but was virtually abolished in p38γ−/− mice, whereas pS199, pS396, and pS404 were induced in both p38γ+/+ and p38γ−/− mice (fig. S19). Similarly, pT205 was markedly reduced in APP23.p38γ−/− animals compared with APP23.p38γ+/+ mice (Fig. 4B). In primary neurons, pT205 (but not p199) was markedly reduced by pan-p38 inhibition (fig. S20). Taken together, these findings indicate that pT205 was a primary p38γ site in tau.

Fig. 4 Site-specific tau phosphorylation disrupts PSD-95/tau/Fyn interaction and inhibits Aβ toxicity.

(A) Coexpression of p38γ (WT) or p38γCA (CA) with tau showed phosphorylation at T205, less at S199, and virtually none at S396 or S404. Detection of GAPDH confirmed equal loading. (B) Compared with APP23.p38γ+/+ animals, APP23.p38γ−/− mice showed a lack of T205 tau phosphorylation (n = 6). Other sites remained phosphorylated. Graph shows quantification of tau phosphorylation. (C) Coexpression of T205E disrupted PSD95/tau/Fyn IP, whereas T205A tau increased it (n = 6). S199 mutations had no effect. Graph shows quantification of tau/Fyn bound to PSD-95. D, Asp. (D) AAV-mediated expression of WT and T205A, but not T205E or GFP, restored susceptibility of tau−/− to 50 mg/kg PTZ-induced seizures, with reduced latency (linear regression) and higher severity (n = 12). (E) Improved memory in APP23 mice upon AAV-mediated p38γCA expression (APP23.AAVp38γCA). MWM test: (Left) Example traces; (middle) escape latencies; (right) time in target quadrant during probe trials (n = 8 to 10). (F) Rescued CFC in APP23.AAVp38γCA compared with APP23.AAVGFP mice (n = 5 to 6). For (B) to (F): ***P < 0.001; **P < 0.01; *P < 0.05; ns, not significant. Error bars indicate SEM.

Next, we showed that a phosphorylation-mimicking Thr205→Glu205 (T205E) tau variant coprecipitated significantly less with PSD-95 as compared with nonmutant and T205A (A, Ala) tau (Fig. 4C and fig. S21). In contrast, phosphorylation-mimicking mutants of all other identified sites had no effect on PSD-95/tau/Fyn interaction (fig. S18). Microscale thermophoresis and glutathione S-transferase–pulldown in vitro and fluorescence-lifetime imaging microscopy (FLIM)–fluorescence resonance energy transfer (FRET) analysis in live cells confirmed the markedly compromised interaction of T205E tau with PSD-95 (fig. S22). The T205E mutation did not hinder tau/Fyn interaction (fig. S21). Phosphorylation of T205 by p38γCA was required for disrupting PSD-95/tau/Fyn complexes (fig. S21). Hence, p38γ regulated PSD-95/tau/Fyn complexes via phosphorylating tau at T205.

The disruption of NR/PSD-95/tau/Fyn complexes prevents excitotoxicity and Aβ toxicity (9). Hence, phosphorylation of tau at T205 should similarly mitigate neurotoxicity. Aβ caused cell death in WT and T205A neurons but significantly less in T205E tau-expressing neurons (fig. S23). Similarly, neurons expressing p38γ and, more so, p38γCA were significantly more resistant to Aβ-induced cell death than controls (fig. S24). PTZ-induced seizures are reduced in tau−/− mice (8, 9). Adeno-associated virus (AAV)–mediated expression of WT and T205A neurons, but not T205E tau or green fluorescent protein (GFP), in the forebrains of tau−/− mice enhanced PTZ-induced seizures (Fig. 4D and fig. S25). In contrast, expression of p38γCA in WT mice using AAV or in Thy1.2-p38γCA transgenic mice decreased PTZ-induced seizures (fig. S25). AAV-mediated p38γCA expression in APP23 mice rescued memory deficits and network aberrations; the same was true for crossing APP23 with Thy1.2-p38γCA mice (Fig. 4, E and F, and figs. S26 and S27). In summary, the levels of active p38γ kinase and tau phosphorylation at T205 determined susceptibility to excitotoxicity and Aβ toxicity.

Here we have shown that T205 phosphorylation of tau is part of an Aβ toxicity–inhibiting response. This is contrary to the current view that tau phosphorylation downstream of Aβ toxicity is a pathological response (3). However, this finding is in line with the idea that tau is involved in normal physiologic NR signaling events in neurons (12). Finally, we found that tau-dependent Aβ toxicity was modulated by site-specific tau phosphorylation, which inhibited postsynaptic PSD-95/tau/Fyn complexes, revealing an Aβ toxicity–limiting role of p38γ in AD that is distinct and opposite to the effects of p38α and p38β (11, 13, 14).

Supplementary Materials

Materials and Methods

Figs. S1 to S27

Tables S1 to S5

References (2653)

References and Notes

Acknowledgments: We thank the staff of the Biological Resources Centre of UNSW for animal care, E. Hinde for help with FLIM-FRET experiments, D. Sullivan and T. Foo for APOE genotyping of human samples, and T. Saito and T. C. Saido for APPNL-G-F mice. The data from mass spectrometry experiments can be found in the supplementary materials. This work was funded by the National Health and Medical Research Council (grants 1081916, 1037746, and 1003083), the Australian Research Council (grants DP130102027 and DE130101591), Alzheimer’s Association (grant NIRG000070035), Alzheimer’s Australia (grants DGP14-39 and DGP14-95), the NIH (grant R28AA012725), and UNSW Australia. A.I. and L.M.I. are inventors on Australian patent application number APO/2016/900764, submitted by UNSW, which covers increasing p38γ activity to prevent neuronal toxicity.
View Abstract


Navigate This Article