A synapsin Ⅰ cleavage fragment contributes to synaptic dysfunction in Alzheimer's disease

Abstract Synaptic dysfunction is a key feature of Alzheimer's disease (AD). However, the molecular mechanisms underlying synaptic dysfunction remain unclear. Here, we show that synapsin Ⅰ, one of the most important synaptic proteins, is fragmented by the cysteine proteinase asparagine endopeptidase (AEP). AEP cleaves synapsin at N82 in the brains of AD patients and generates the C‐terminal synapsin Ⅰ (83–705) fragment. This fragment is abnormally distributed in neurons and induces synaptic dysfunction. Overexpression of AEP in the hippocampus of wild‐type mice results in the production of the synapsin Ⅰ (83–705) fragment and induces synaptic dysfunction and cognitive deficits. Moreover, overexpression of the AEP‐generated synapsin Ⅰ (83–705) fragment in the hippocampus of tau P301S transgenic mice and wild‐type mice promotes synaptic dysfunction and cognitive deficits. These findings suggest a novel mechanism of synaptic dysfunction in AD.


| INTRODUC TI ON
Alzheimer's disease (AD) is the most common age-related neurodegenerative disease. The major clinical manifestation of AD is progressive cognitive decline. Pathologically, AD is characterized by synaptic dysfunction and the accumulation of amyloid plaques and neurofibrillary tangles. Synaptic dysfunction is one of the major contributors to the symptoms of AD (Bastrikova et al., 2008;Davies et al., 1987;Morrison & Baxter, 2012;Selkoe, 2002). Moreover, synaptic dysfunction is a very early feature of AD and occurs years before the onset of cognitive symptoms. Synaptic degeneration is positively related to the severity of dementia (DeKosky & Scheff, 1990;Jacobsen et al., 2006;Reddy et al., 2005;Scheff et al., 2006;Terry et al., 1991). Several synaptic proteins have been reported to be decreased in brain specimens from AD patients (Bereczki et al., 2016(Bereczki et al., , 2018Goetzl et al., 2016;Qin et al., 2004;Reddy et al., 2005).
However, the molecular mechanisms underlying synaptic dysfunction in AD remain elusive.
Synapsin Ⅰ is a neuron-specific phosphoprotein localized to the cytoplasmic surface of synaptic vesicles (SVs). It acts as the key regulator of SV dynamics in presynaptic terminals. Under resting conditions, synapsin Ⅰ clusters SVs in the reserve pool by interacting with phospholipids and the F-actin cytoskeleton. Upon stimulation, synapsin Ⅰ is phosphorylated at Ser9 and dissociates from SVs. After the stimulus, synapsin Ⅰ is dephosphorylated and reclusters SVs in the bouton (Cesca et al., 2010). Thus, the normal function of synapsin Ⅰ is required for the recycling of SVs in the presynaptic terminal. In AD patients, the expression of synapsin Ⅰ is decreased in the brain, which is accompanied by synaptic dysfunction (Qin et al., 2004), indicating that the dysfunction of synapsin Ⅰ may contribute to synaptic dysfunction in AD.
Mammalian asparagine endopeptidase (AEP) is a lysosomal cysteine protease that cleaves protein substrates on the C-terminal side of asparagine residues . AEP is activated by sequential removal of C-and N-terminal propeptides under acidic conditions (Li et al., 2003). Recently, we reported that AEP is activated in the brain in an age-dependent manner and cleaves amyloid precursor protein (APP) and tau, promoting the deposition of amyloidβ (Aβ) and tau during the onset of AD (Zhang et al., 2014(Zhang et al., , 2015. In this report, we show that AEP cleaves synapsin Ⅰ at N82 both in vitro and in vivo, generating the synapsin Ⅰ (83-705) fragment (designated C83 fragment). This fragment induces synaptic dysfunction both in vitro and in vivo. Overexpression of AEP in the hippocampus of wild-type mice results in the production of synapsin Ⅰ C83 and synaptic dysfunction. Furthermore, the expression of the synapsin Ⅰ C83 fragment in tau P301S transgenic mice or wild-type mice promotes synaptic dysfunction and cognitive impairment.
Hence, our results indicate that AEP-mediated fragmentation of synapsin Ⅰ contributes to synaptic dysfunction in AD.

| Synapsin Ⅰ is a substrate of AEP
To investigate the molecular mechanisms of synaptic dysfunction, we analyzed brain tissues from AD patients by mass spectrometry.
Interestingly, a fragment ending at N82 of synapsin Ⅰ was found in the AD brain ( Figure 1a). Since AEP is the only protease that specifically cleaves on the C-terminal side of asparagine residues, we speculate that this fragment is generated by AEP-mediated cleavage. To verify that synapsin Ⅰ is cleaved by AEP, we performed mass spectrometry with brain tissues from wild-type and AEP knockout (AEP -/-) mice.
The same fragment of synapsin Ⅰ was identified in brain tissue lysates from wild-type mice (Figure 1b). Comparative label-free proteomic analysis of wild-type mice and AEP -/mice revealed that the signal of this fragment was much lower in AEP -/mouse brain extracts ( Figure 1c). These results indicate that AEP may be responsible for the fragmentation of synapsin Ⅰ in the brain. To further investigate the presence of the synapsin Ⅰ C83 fragment in the brain, we generated an antibody that specifically recognizes the AEP-generated C83 fragment of synapsin Ⅰ (anti-synapsin Ⅰ C83 antibody) but not full-length synapsin Ⅰ. Using this antibody, we detected the presence of the synapsin Ⅰ C83 fragment in wild-type mice but not in AEP -/mice by Western blotting and immunohistochemistry (Figure 1d-f), supporting the specificity of this antibody for the AEP-generated synapsin Ⅰ fragment.
To confirm the cleavage of synapsin Ⅰ by AEP, we performed an in vitro cleavage assay by incubating GST-synapsin Ⅰ with kidney lysates prepared from wild-type and AEP -/mice at pH 7.4 or 6.0.
AEP was activated at pH 6.0 and generated an N-terminal synapsin Ⅰ fragment that was detected with the anti-GST antibody and a Cterminal fragment that was detected with the anti-synapsin Ⅰ C83 antibody ( Figure 1g). The enzymatic activity of AEP in kidney lysates was confirmed by an enzyme activity assay ( Figure 1j). GST-synapsin Ⅰ can also be cleaved by brain lysate from wild-type mice, but not that from AEP -/mice at pH 6.0 ( Figure 1h). To further explore the cleavage of synapsin Ⅰ by AEP, we incubated the active AEP enzyme with GST-synapsin Ⅰ for 15 min and 30 min, respectively. Western blot analysis showed that synapsin Ⅰ was fragmented in the presence of AEP in a time-dependent manner ( Figure 1i). To exclude the potential effect of other cellular components in HEK293 cell lysates, we incubated purified GST-synapsin Ⅰ with active recombinant AEP.
As expected, purified GST-synapsin Ⅰ was also potently cleaved by AEP ( Figure 1k). Overall, these results indicate that synapsin Ⅰ is a substrate of AEP.

| AEP cleaves synapsin Ⅰ at N82
To confirm that AEP specifically cleaves synapsin Ⅰ, we cotransfected GST-synapsin Ⅰ and myc-AEP or myc-AEP C189S mutant into HEK293 cells. Wild-type AEP strongly induced synapsin Ⅰ fragmentation, while the AEP C189S mutant, whose protease activity is abolished (Li et al., 2003), could not trigger synapsin Ⅰ cleavage ( Figure 2a). Moreover, this cleavage process was suppressed by the AEP inhibitor AENK, whereas the inactive control AEQK had no effect ( Figure 2b). The enzymatic activity of AEP was validated by a fluorescent substrate cleavage assay (Figure 2c,d). These findings indicate that the synapsin Ⅰ C83 fragment is specifically generated F I G U R E 1 AEP cleaves synapsin Ⅰ in vitro. (a) MS/MS spectrum showing the presence of synapsin Ⅰ (54-82) peptide in brain samples from subjects with AD. (b) MS/MS spectrum showing the presence of synapsin Ⅰ (54-82) peptide in brain samples from wild-type mice. (c) Representative extracted ion chromatograms for synapsin Ⅰ (54-82) peptide from wild-type and AEP -/mouse brain samples. (d) Western blot detection of synapsin Ⅰ C83 fragment in the wild-type and AEP -/mouse brain lysates. (e, f) Immunostaining of synapsin Ⅰ C83 fragments in brain slices of wild-type and AEP -/mice (e), and Syn I C83 optical density (f). DG: dentate gyrus. Scale bar, 20 μm. (g) Synapsin Ⅰ cleavage assay. GST-synapsin Ⅰ was incubated with kidney lysates from wild-type and AEP -/mice at pH 7.4 or pH 6.0 at 37°C for 30 min, respectively. Western blot shows that synapsin Ⅰ was cleaved at pH 6.0 by wild-type kidney lysates when AEP was activated (j) (n = 3). (h) Synapsin Ⅰ cleavage assay. GST-synapsin Ⅰ was incubated with brain lysates from wild-type and AEP -/mice at pH 7.4 or pH 6.0 at 37°C for 30 min, respectively. Western blot shows that synapsin Ⅰ was cleaved at pH 6.0 by wild-type brain lysates. (i) Western blot showing the timedependent cleavage of synapsin Ⅰ by purified activated recombinant AEP (rAEP). (k) Western blot showing the cleavage of purified GSTsynapsin Ⅰ by AEP. Syn I, synapsin I. Data are mean ± SEM; *p < 0.005, **p < 0.001, ***p < 0.001, ****p < 0.0001    F I G U R E 3 AEP is activated and cleaves synapsin Ⅰ in the AD brain. (a) Western blot analysis of synapsin Ⅰ and active AEP in 4-, 8-, 10-, and 14-month-old wild-type mice brain, bar graphs are quantitative analysis of Syn I C83 and AEP. (b) AEP activity assay showing the age-dependent activation of AEP in mouse brain. (c) Western blot detection of synapsin Ⅰ C83 fragment in human brain samples from subjects with AD and age-matched controls, bar graphs are quantitative analysis of Syn I C83 and AEP. (d, e) Immunostaining of synapsin Ⅰ C83 fragments in brain slices of AD patients and age-matched control. Scale bar, 20 μm. (f) Immunostaining showing the colocalization of synapsin Ⅰ C83 fragment with AEP in tau P301S mouse brain slices. Scale bar, 20 μm. (g) Immunostaining of synapsin Ⅰ C83 fragments in slices of wild-type and APP/PS1 mice. Scale bar, 20 μm. (h) Western blot detection of synapsin Ⅰ C83 fragment in hippocampus brain lysates of wild-type and APP/PS1 mice. bar graphs are quantitative analysis of Syn I C83. (i) Immunostaining shows that Aβ induces the transport of AEP from soma to neurite in the primary neuron. Scale bar, 20 μm. (j) AEP colocalizes with synapsin Ⅰ after the neurons were challenged with Aβ oligomers. Scale bar, 10 μm. (k, l) Synapsin Ⅰ C83 fragment was detected by Western blot after the neurons were treated with Aβ oligomers for 24 h. Bar graphs are quantification of full-length Synapsin I (l). (m) The mRNA levels of synapsin Ⅰ detected by RT-PCR in the presence or absence of Aβ oligomers. ns, no significance. Data are mean ± SEM; **p < 0.001, ***p < 0.001, ****p < 0.0001 C o n t r o l A β by AEP. The results of mass spectrometric analysis of brain tissues from AD patients and mice suggest that AEP cleaves synapsin Ⅰ at N82. AEP selectively cleaves its substrates on the C-terminal side of asparagine residues. The N-terminus of synapsin Ⅰ contains 5 asparagine residues (N2, N12, N16, N19, and N82). We generated synapsin Ⅰ point mutants with these asparagines substituted with alanines.
Generation of the synapsin Ⅰ fragment was completely abolished in the N82A mutant, but not in the N2A, N12A, N16A, or N19A mutant ( Figure 2e). These results indicate that AEP specifically cleaves synapsin Ⅰ at N82, generating the C83 fragment.

| AEP is upregulated during aging and generates the synapsin Ⅰ C83 fragment in AD brain
Aging is the major risk factor for AD (Guerreiro & Bras, 2015). AEP is activated during aging as the internal milieu of the brain gradually acidifies (Eugenin et al., 2016;Pirchl et al., 2006;Yates et al., 1990). We investigated the appearance of the AEP-generated synapsin Ⅰ C83 fragment in the mouse brain at different ages. The antisynapsin Ⅰ C83 antibody detected the C83 fragment in brain tissue lysates from 4-, 8-, 10-, and 14-month-old mice, and the amount of the C83 fragment increased with age ( Figure 3a). The AEP activity assay showed that AEP was activated in an age-dependent manner in the mouse brain ( Figure 3b). We further verified the presence of the synapsin Ⅰ C83 fragment in the human AD brain using Western blotting and immunohistochemistry. Synapsin Ⅰ C83 was abundant in human AD brains but was present at much lower levels in the brains of age-matched controls (Figure 3c-e). We also detected the presence of the synapsin Ⅰ C83 fragment in the hippocampal CA1 region in the tau P301S transgenic mice, but little was detected in brain sections from age-matched nontransgenic control mice. Furthermore, immunofluorescence staining showed that AEP colocalized with synapsin Ⅰ C83 ( Figure 3f). CCAAT/enhancer-binding protein β (C/ EBPβ) is the transcriptional factor that promotes the expression of AEP in the brain of AD patients and AD mouse models (Wang et al., , 2019. We found that the levels of C/ EBPβ, phosphorylated C/EBPβ, active AEP, and synapsin Ⅰ C83 were much higher in 5-month-old tau P301S mice when compared with those in age-matched wild-type mice ( Figure S1a,b). Western blot and immunohistochemistry confirmed that synapsin Ⅰ C83 fragment was also abundant in brain lysates from APP/PS1 transgenic mice, a well-established animal model of AD (Figure 3g,h). Therefore, AEP is activated during aging and cleaves synapsin Ⅰ in the brain of tau P301S mice and APP/PS1 mice.
Asparagine endopeptidase is a lysosomal protease that leaks into the cytoplasm in neurodegenerative diseases (

| Synapsin Ⅰ C83 fragment induces synaptic dysfunction in vitro
Synapsin Ⅰ regulates neurotransmitter release by interacting with lipids and the actin cytoskeleton to cluster the SVs in the reserve pool (Cesca et al., 2010). A schematic diagram of synapsin Ⅰ and its cleavage by AEP is shown in (Figure 4a). To assess whether AEP-mediated cleavage of synapsin Ⅰ affects the normal function of synapsin Ⅰ in neurons, we infected primary neurons with adeno-associated viruses (AAVs) encoding EGFP, EGFP-synapsin Ⅰ, and EGFP-synapsin Ⅰ C83, respectively. Synapsin Ⅰ was distributed in a punctate pattern, whereas synapsin Ⅰ C83 was more diffusely distributed ( Figure 4b).
Immunofluorescence staining showed that full-length synapsin Ⅰ colocalized with synaptophysin, a membrane protein localized to SVs, whereas synapsin Ⅰ C83 did not ( Figure 4c). These results imply that the C83 fragment does not bind to SVs as does the full-length protein.
Next, we investigated whether the synapsin Ⅰ C83 fragment

First stimulation
Second stimulation ----** C83 was reduced by 45% compared with that in boutons overexpressing full-length synapsin Ⅰ (Figure 4d,e). Furthermore, the release of FM4-64 was also impaired in the presence of synapsin Ⅰ C83 (Figure 4f). In addition, DiI staining showed that the density of dendritic spines in neurons expressing synapsin Ⅰ C83 was decreased compared with that in neurons expressing full-length synapsin Ⅰ (Figure 4g)

| Overexpression of AEP induces synaptic dysfunction and cognitive impairment in vivo
We next assessed the impacts of AEP on synaptic and cognitive function in vivo. AAVs encoding EGFP and EGFP-AEP were injected separately into the hippocampal area of 2-month-old wild-type mice.
Three months later, we assessed the effect of AEP overexpression on spatial learning and memory via the Morris water maze test.
During the training phase, the latency to find the platform was gradually decreased in all mice, indicating a learning effect. However, the learning ability of mice overexpressing AEP was substantially impaired (Figure 5a

| Synapsin Ⅰ C83 fragment induces synaptic dysfunction and cognitive impairment in tau P301S transgenic mice
We further investigated the impact of the AEP-generated synapsin Ⅰ C83 fragment on synaptic dysfunction and cognitive impairment in tau P301S transgenic mice, a mouse model of tauopathy, which is related to AD and related disorders. The mice express human mutant tau P301S and develop widespread neurofibrillary tangle-like inclusions in the brain, accompanied by synaptic dysfunction and behavioral impairment (Yoshiyama et al., 2007).
We found that the 5-month-old tau P301S mice showed cognitive impairment in the Morris water maze test and Y-maze test  (Fedulov et al., 2007;Nicoll, 2017). We observed that LTP was diminished in mice expressing synapsin Ⅰ C83 compared with mice expressing GFP or full-length synapsin Ⅰ (Figure 6g). The average amplitude of fEPSPs was lower in mice expressing synapsin Ⅰ C83 than in those expressing EGFP or full-length synapsin Ⅰ (Figure 6h). The I/O curve showed that the fEPSP response of Syn I C83 mice was weaker than that of the other groups (Figure 6i). These results indicate that overexpression of the synapsin Ⅰ C83 fragment induces synaptic dysfunction and cognitive impairment in tau P301S transgenic mice. Electron microscopy analysis of mouse hippocampal slices showed that the synaptic density was lower in mice expressing synapsin Ⅰ C83 than in mice expressing GFP or full-length synapsin Ⅰ (Figure 6j,k). In addition, Golgi staining showed that the synapsin Ⅰ C83 fragment In summary, the AEP-generated synapsin Ⅰ C83 fragment induces synaptic dysfunction and cognitive impairment in tau P301S mice.

| Synapsin Ⅰ C83 fragment induces synaptic dysfunction and cognitive impairment in wild-type mice
To investigate the direct impact of synapsin Ⅰ C83 fragment on the synaptic dysfunction and cognitive impairment, AAVs encoding EGFP, EGFP-synapsin Ⅰ, and EGFP-synapsin Ⅰ C83 were injected into the bilateral CA1 hippocampal region of 2-month-old wildtype mice. 2.5 months later, the spatial memory was assessed via the Morris water maze test. During the training phase, the latency to find the platform was gradually decreased in wild-type mice expressing EGFP. However, the learning ability of mice expressing synapsin Ⅰ C83 was substantially impaired ( Figure S3a,b). In the probe trial, mice expressing synapsin Ⅰ C83 showed a decreased time in the target quadrant, indicating that their learning and memory abilities were impaired ( Figure S3c). The swim speed was comparable among the mice ( Figure S3d). In the Y-maze test, mice expressing synapsin Ⅰ C83 spent less time in the new arm ( Figure S3e). Furthermore, the LTP was diminished in mice expressing synapsin Ⅰ C83 compared with mice expressing GFP or full-length synapsin Ⅰ ( Figure S3f). The average amplitude of fEPSPs was lower in mice expressing synapsin Ⅰ C83 than in those expressing EGFP or full-length synapsin Ⅰ ( Figure   S3g). Hence, the synapsin Ⅰ C83 fragment induces synaptic dysfunction and cognitive impairment in wild-type mice.

| DISCUSS ION
In the present study, we identified that the synaptic protein synapsin Ⅰ is a substrate of the cysteine protease AEP. AEP is activated and cleaves synapsin Ⅰ in an age-dependent manner in the mouse brain and human AD brain. The AEP-generated synapsin Ⅰ C83 fragment exhibits a reduced ability to localize at SVs, disrupts vesicle recycling, and causes presynaptic defects. Overexpression of AEP or the AEP-generated synapsin Ⅰ C83 fragment in mice induces synaptic dysfunction and cognitive impairment. Therefore, the synapsin Ⅰ C83 fragment mediates synaptic dysfunction and cognitive impairment.
Neurotransmitters are stored in SVs in the axon terminal. In response to an action potential, SVs fuse with the presynaptic membrane, thus releasing neurotransmitters into the synaptic cleft.
Synapsin Ⅰ is an SV-associated protein that acts as a link between extracellular stimuli and intracellular signaling events. Under resting conditions, synapsin Ⅰ clusters SVs in the reserve pool by interacting with phospholipids and F-actin. Upon stimulation, synapsin Ⅰ is phosphorylated and dissociates from the reserve pool, and SVs move close to the readily releasable pool. Deletion of synapsin Ⅰ induces synaptic dysfunction and cognitive impairment in mice (Corradi et al., 2008;Ryan et al., 1996). We found that overexpression of full-length synapsin I in primary neurons increased the density of dendritic spines. This is consistent with the previous reports that synapsin I promotes the outgrowth of neurites and maintains the physiological function of neurons (Cesca et al., 2010;Fornasiero et al., 2010). However, AEP cleaves synapsin Ⅰ and generates the synapsin Ⅰ C83 fragment, which induces spine degeneration and impairs SV recycling and synaptic transmission.
We found that overexpression of full-length synapsin I led to synaptic dysfunction and cognitive impairment in wild-type mice ( Figure S2), indicating that massive overexpression of synapsin I exerts detrimental effects on normal synapses in vivo. Synaptic impairments precede the formation of tangles in tau P301S mice (Yoshiyama et al., 2007). We found that overexpression of fulllength synapsin I did not exacerbate synaptic dysfunction in tau P301S mice, even though the full-length synapsin I could be

| Antibodies and reagents
Antibodies to the following targets were used:

| Human tissue samples
Postmortem brain samples of AD cases and age-matched control cases were from the Emory Alzheimer's Disease Research Center.
The average ages of the AD patients and control subjects were 65.4 and 63.2, respectively. Hippocampal brain samples were used in this study. The stages of the samples were Braak stage 4-6. All the AD cases were confirmed by pathological diagnosis. The staging of AD pathology was determined using the method described previously (Braak et al., 2006). The study was approved by the Biospecimen Committee.

| Synapsin Ⅰ cleavage assay and AEP activity assay
Synapsin Ⅰ cleavage assay and AEP activity assay were conducted as described previously (Zhang et al., 2014).
Samples were ionized on a hybrid LTQ XL Orbitrap mass spectrometer (Thermo) using a 2.0 kV electrospray ionization voltage from a nano-ESI source (Thermo). After collision-induced dissociation (collision energy 35%, activation Q 0.25, activation time 30 ms) for the top 10 precursor ions, data-dependent acquisition of centroid MS spectra at 30, 000 resolution and MS/MS spectra were obtained in the LTQ, and the charge determined by the acquisition software is z≥2. Dynamic exclusion of peaks already sequenced was for 20 s with early expiration for two count events with the signal to noise >2. Automatic gating control was set to 150 ms maximum injection time or 10 6 counts.
To identify AEP cleavage sites on synapsin Ⅰ, the SageN Sorcerer SEQUEST 3.5 algorithm was used to search and match MS/MS spectra to a complete semitryptic human proteome database (NCBI reference sequence revision 50, with 66, 652 entries) plus pseudo-reversed decoys sequences with a 20 p.p.m. mass accuracy threshold (Elias & Gygi, 2007;Xu, Gao, & Ng, 2009). Only b-and y-ions were considered for scoring (Xcorr) and Xcorr along with ΔCn were dynamically increased for groups of the samples organized by a combination of trypticity (fully or partial) and precursor ion charge state to remove false-positive hits along with decoys until achieving a false-discovery rate (FDR) of <5% (<0.25% for proteins identified by more than one peptide). The FDR was estimated by the number of decoy matches (nd) and the total number of assigned matches (nt). FDR =2*nd/nt, assuming mismatches in the original database were the same as in the decoy database. All semitryptic MS/MS spectra for putative AEP-generated synapsin Ⅰ cleavage sites were manually inspected.

| Primary neuronal culture
Primary mouse neurons were cultured as previously described (Zhang et al., 2014). To explore the effect of synapsin Ⅰ C83 fragments on neurons, neurons cultured 7 days in vitro (DIV) were infected with AAVs encoding GFP, GFP-synapsin Ⅰ, GFP-synapsin Ⅰ C83, respectively. Human synapsin Ⅰ promoter was used to drive neuron-specific gene expression. The virus was prepared by BrainVTA (Wuhan) Co., Ltd. 7 days later, neurons were fixed in 4% paraformaldehyde and tested by immunofluorescence.

| Generation of antibody that specifically recognizes synapsin Ⅰ C83 fragment (anti-synapsin Ⅰ C83 antibody)
To generate the anti-synapsin Ⅰ C83 antibody, the peptide AVKQTTAAAC was used to immunize two rabbits. The rabbits were boosted 4 times with the immunizing peptides with 3-week intervals between injections. The titers against the immunizing peptide were determined by ELISA. The maximal dilution giving a positive response with the chromogenic substrate for horseradish peroxidase was 1:512,000. The immunoactivity of the antiserum was further confirmed by Western blotting and immunohistochemistry.

| Morris water maze test
Five-month-old tau P301S mice were trained with extra maze cues as described previously (Zhang et al., 2014). Each subject was tested four times per day for 4 consecutive days, with a 15 min intertrial interval. If the subjects did not find the platform within 60 s, they were manually guided to it, and they had 15 s learning time. The platform was removed on day 5 and the percentage of time spent in the quadrant was measured over 60 s. All trials were analyzed for latency and swim speed using ANY-Maze software (San Diego Instruments).

| Y-maze test
The Y-maze test was carried out as described previously (Ma MX et al., 2007). Each arm was 40 cm long, 12 cm high, 3 cm wide at the bottom, and 10 cm wide at the top. The arms converged in an equilateral triangular central area that was 4 cm at its longest axis.
The three arms were randomly designated.

| FM dye uptake assay
FM dye imaging was performed as described previously (Gaffield & Betz, 2006). Briefly, to load the VS with the FM 4-64 dye, the neurons were incubated in high-K + Tyrode's solution (90 mM

| FM dye release assay
To load the SVs with the FM 4-64 dye, the neurons were first stimulated with high-K + Tyrode's solution (90 mM

| Electron microscopy of synapses
Synaptic density was detected by electron microscopy as described previously (Zhang et al., 2014). After deep anesthesia, mice were perfused transcardially with 2% glutaraldehyde. Hippocampal slices were postfixed in cold 1% OsO 4 for 1 h. Samples were prepared and examined using standard procedures. Ultrathin sections (90 nm).
were stained with uranyl acetate and lead acetate and viewed at 100 kV in a JEOL 200CX electron microscope. Synapses were identified by the presence of SVs and postsynaptic densities.

| Golgi staining
Mice were deeply anesthetized, and the brain was removed from the skull as quickly as possible. Rapid Golgi staining was performed using a kit (FD Neuro Technologies, Inc, PK401). Briefly, brain tissue was immersed in silver impregnation solution for 2 weeks in the dark, cryoprotected at 4 °C for 72 h in the dark, and then cut into 100μm sections. After sectioning and mounting on gelatin-coated slides, sections were developed and clarified. To measure the spine density, all clearly evaluable areas of 50~100 μm of secondary dendrites from each imaged neuron were used.

| Statistical analysis
All data are presented as mean values ± SEM (standard error of the mean). Statistical analysis was performed using either Student's ttest (two-group comparisons) or one-way ANOVA followed by an LSD post hoc test (more than two groups), and p values <0.05 were considered significant.

ACK N OWLED G EM ENTS
This work was supported by grants from the National Natural Science

CO N FLI C T O F I NTE R E S T
The authors declare no competing financial interests.

DATA AVA I L A B I L I T Y S TAT E M E N T
The authors declare that when there is a reasonable request, the authors shall request all the data contained in this study.