Acetylated tau destabilizes the cytoskeleton in the axon initial segment and is mislocalized to the somatodendritic compartment

cDNA encoding 2N4R human tau was cloned into pEGFP-C1 vector (Clontech). In mApple-tagged tau plasmids, EGFP in the pEGFP-C1 vector was replaced with mApple. Tau mutations (K163/174/180/190Q, K274Q, K281Q, K274/281Q, and K274/281R) were generated with the QuickChange mutagenesis kit (Stratagene). The following plasmids were gifts: GFP-tubulin (Dr. Ron Vale, University of California, San Francisco), GFP-end-binding protein (EB) 1 (Dr. Torsten Wittmann, University of California, San Francisco), and GFP-EB3 (Dr. Niels Galjart, Erasmus MC, Rotterdam).

The murine prion promoter (Mo.PrP) expression plasmid (Mo.PrP.Xho) has been previously described [31, 39]. Human tau WT cDNA (2N4R) or cDNA with A820C (K274Q) and A841C (K281Q) mutations were cloned into the Xho1 site of the Mo.PrP.Xho plasmid. The resulting Mo.PrP-tauWT (tauWT) and Mo.PrP-tauK274/281Q (tauKQ) transgenes were microinjected into fertilized mouse oocytes from the FVB/N genetic background and implanted into pseudopregnant female mice. The founder lines with expression of equivalent levels of tauWT and tauKQ, and higher levels of tauKQ (tauKQ^high) in the FVB/N genetic background were then crossed with C57BL/6 mice purchased from Jackson Laboratory. All mice used for experiments were of mixed FVB/N and C57BL/6 genetic background. Tail DNA from offspring was genotyped by using the following primers: 5’ primer GGAGTTCGAAGTGATGGAAG, 3’ primer GGTTTTTGCTGGAATCCTGG. Both male and female mice were used for experiments. Mice were housed in a pathogen-free barrier facility with a 12 h light-dark cycle and ad libitum access to food and water. All animal procedures were carried out under University of California, San Francisco, Institutional Animal Care and Use Committee-approved guidelines.

Superior temporal gyrus of control and AD brains were obtained from the Mount Sinai NIH Brain and Tissue Repository (NBTR), provided by Dr. Vahram Haroutunian (The Mount Sinai School of Medicine, New York). The brain tissues were from early Braak stages 0–2 and late Braak stages 5–6, and were extracted from patients in ages of 70–103 years.

HeLa cells in Dulbecco’s modified Eagle’s medium supplemented with 10 % fetal bovine serum, 100 U/ml penicillin, and 100 μg/ml streptomycin were grown at 37 °C in 5 % CO₂. Primary cultures were established from hippocampi of Sprague-Dawley rat pups (Charles River Laboratories) on postnatal day 0 or 1. Purified cells (50,000 per 300 μl of neurobasal medium supplemented with B27; Life Technologies) were plated on poly-L-lysine-coated, glass-bottom 35-mm dishes (MatTek). After cells had attached, the medium was replaced. At 6 or 7 DIV, the cells were transfected with Lipofectamine 2000 (Life Technologies) and DNA plasmids mixed 2:1 in OPTI-MEM (Life Technologies). After 30 min, the transfection medium (neurobasal medium with kyneurenic acid) was replaced with conditioned neurobasal medium supplemented with B27.

Mice were transcardially perfused with 0.9 % saline, and the brains were fixed in 4 % paraformaldehyde in PBS for 48 h and then incubated in 30 % sucrose in PBS. For antigen retrieval, coronal brain sections 30 μm thick were cut with microtome and incubated in 10 mM citric acid at 90 °C for 20 min. Floating brain sections were permeabilized and blocked with PBS containing 0.3 % Triton X-100 and 10 % normal goat serum (PBST) at room temperature for 1 h. Sections were incubated first with antibodies against AnkG (N106/36, NeuroMab) and βIV-spectrin (gift from Dr. Matthew N. Rasband, Baylor College of Medicine) at 4 °C overnight and then with Alex Fluor 488- and 555-conjugated anti-mouse and anti-rabbit antibodies (Life Technologies) at room temperature for 1 h. Immunostaining with human brains was performed as described previously [30]. Briefly, cases were selected from the Neurodegenerative Disease Brain Bank (NDBB) at the University of California, San Francisco. 8 μm-thick sections were cut from paraffin blocks and mounted on glass slides. Immunoperoxidase staining was performed using an avidin-biotin complex detection system (Vectastain ABC kit; Vector Laboratories). Slides were pretreated for antigen retrieval by immersion in 10 mM citric acid at 121 °C for 5 min. The primary antibody MAb359 [30] was incubated overnight at 4 °C and biotinylated secondary antibody (Vector Laboratories) was incubated at room temperature for 1 h. Slides were incubated in Avidin/Biotinylated enzyme Complex (ABC) at room temperature for 1 h followed by exposure to 3,3’-diaminobenzidine (DAB) substrate (Vector Laboratories). For immunofluorescence, after antigen retrieval and blocking with 0.1 % Sudan Black solution, the primary antibodies MAb359 [30], CP-13 (gift from Dr. Peter Davies, Feinstein Institute for Medical Research), and AnkG (N106/36, NeuroMab) were incubated overnight at 4 °C. The DyLight 488-conjugated anti-mouse secondary antibody was incubated at room temperature for 1 h. AIS Images were acquired with a Nikon Ti-E spinning-disk confocal microscope and a 60X oil objective. Seven serial optical sections (0.5 μm steps) were projected into a single image to visualize the AIS. ImageJ software (NIH) was used to analyze the intensity and length of the AIS.

MT binding assays in HeLa cells were performed as described [40] with modifications. DNA plasmids of mApple-mutant tau and GFP-WT tau were co-transfected into HeLa cells. To assess the intracellular distribution of WT and mutant tau, HeLa cells at 37 °C in 5 % CO₂ were examined with a Nikon Ti-E spinning-disk confocal microscope and a 60X oil objective 24–48 h after transfection. After conversion of fluorescent RGB images into binary images, ImageJ (NIH) was used to subtract a binary image of WT tau from that of mutant tau. Cytoplasmic tau signals after the image subtraction were calculated as the MT-unbound tau index.

Primary rat hippocampal neurons at DIV 6–7 were co-transfected with GFP-EB3 and mApple-tau; the cells were imaged at 37 °C in 5 % CO₂ 24 h after transfection. Live time-lapse imaging was performed every second for 60 sec with a Nikon Ti-E spinning disk confocal microscope and a 60X oil objective. Movement of GFP-EB1 and GFP-EB3 comets were analyzed with the Matlab software package plusTipTracker [41].

Primary rat hippocampal neurons at DIV 6–7 were co-transfected with GFP-tubulin and mApple-tau, and imaged 24 h after transfection. Before imaging, the AIS was immunolabeled with an antibody against the extracellular domain of neurofascin (A12/18, NeuroMab) at 37 °C for 5 min. After brief rinses with Neurobasal A media, Alexa Fluor 647 anti-mouse secondary antibody was applied at 37 °C for 5 min. For FRAP experiments, we used a Nikon Ti wide-field microscope and a 100X oil objective to examine cells at 37 °C in 5 % CO₂. An ROI (~5 μm) for photobleaching was drawn in the center of the AIS, as judged from anti-neurofascin staining. GFP-tubulin was bleached with a 473-nm laser and fluorescence recovery was monitored with a 488-nm laser. Time-lapse images were taken every second for 60 sec. Images in which photobleaching reduced fluorescence intensity by more than 70 % were analyzed. The fluorescence signal was background subtracted and quantified with ImageJ (NIH).

Time-lapse live imaging during photoconversion was performed as described using photoconvertible mEOS2-tau. Primary rat hippocampal neurons at DIV 6–7 were transfected with mEOS2-tau and imaged 48–72 h later with a Nikon Ti-E spinning disk confocal microscope at 37 °C with 5 % CO₂. A ~30-μm portion of an axon segment ~30 μm from the cell body was subjected to photoconversion with a 405-nm laser (two to three 300-ms exposures), and red fluorescence images were acquired every 30 sec for 30 min. Using ImageJ (NIH), changes in red fluorescence intensity was analyzed in both the cell body and a distal axon ~30 μm away from the photoconversion site. Changes in red fluorescence intensity were normalized to the initial red fluorescence intensity from the photoconversion site right after photoconversion.

Human and mouse brain tissues were lysed in RIPA buffer (50 mM Tris, pH. 7.4, 150 mM NaCl, 1 mM EDTA, 0.5 % Nonidet P-40) with histone deacetylase inhibitors (1 μM trichostatin, 5 mM nicotinamide; both from Sigma), 1 mM phenylmethyl sulfonyl fluoride (Sigma), phosphatase inhibitor cocktail (Roche), and protease inhibitor cocktail (Sigma). Lysates were sonicated and centrifuged at 170,000 g at 4 °C for 15 min and at 18,000 g at 4 °C for 10 min. Supernatants were collected, and protein concentrations were measured by bicinchoninic acid assay (Pierce). Proteins were resolved on 4–12 % SDS-PAGE and transferred to nitrocellulose membranes. After blocking with nonfat dry milk, the membranes were probed at 4 °C overnight with primary antibodies: rabbit monoclonal MAb359 for ac-K274 tau [30], rabbit monoclonal MAb63 for ac-K281 tau [31], mouse monoclonal anti-AnkG (N106/20, NeuroMab), rabbit polyclonal anti-βIV-spectrin (gift from Dr. Matthew N. Rasband, Baylor College of Medicine), mouse monoclonal Tau5 (AHB0042, Life Technologies), mouse monoclonal HT7 (MN1000, Thermo Scientific), mouse monoclonal AT8 (MN1020, Thermo Scientific), mouse monoclonal 12E8 (Prothena Biosciences), mouse monoclonal PHF-1 (gift from P. Davies), and mouse monoclonal anti-GAPDH (MAB374, Millipore). The membranes were then incubated with HRP-conjugated secondary antibodies at room temperature for 1 h. Immunoblots were visualized by enhanced chemiluminescence (Thermo Scientific) and quantified by ImageJ software (NIH).

AIS filtering defects have been observed in primary neurons with familial AD mutations [24] and with Aβ exposure [25]. To examine the AIS in AD brain, we performed immunostaining of AnkG in the frontal cortex of AD patients. Interestingly, we observed reduced levels of AnkG in the AIS in AD brain compared to control brain (Fig. 1a). Immunoblot analyses revealed a significant decrease in levels of AIS cytoskeletal proteins, including AnkG and βIV-spectrin, in AD patients at late Braak stages compared to early Braak stages (Fig. 1b, c; Table 1). In AD brains at late Braak stages, acetyl-K274 and -K281 on tau were detected by the acetyl-lysine-specific MAb359 and MAb63 tau antibodies [31], respectively. Levels of acetyl-K274 were significantly higher at late Braak stages than early stages (Fig. 1b, c). Acetyl-K274 tau detected with MAb359 was highly enriched in intraneuronal tau inclusions in human tauopathy brains (Additional file 1: Figure S1A-C) [30]. Acetyl-K274 tau was localized in the cytoplasm of neurons of FTDP-17 patients (Additional file 1: Figure S1A) and also in the corticobasal bodies of patients with A152T mutation (Additional file 1: Figure S1B), which is linked with increased risk of AD and progressive supranuclear palsy (PSP) [42]. In AD patients, acetyl-K274 tau was localized in neurofibrillary tangles (Additional file 1: Figure S1C). Notably, the decrease in AnkG and βIV-spectrin levels in AD brains correlated with increased acetyl-K274 and -K281 on tau (Fig. 1d), raising the possibility that acetylated tau downregulates AIS cytoskeletal proteins in AD.

Next, we examined whether acetylated tau destabilizes the AIS cytoskeleton using transgenic mice with mouse prion promoter-driven expression of wild-type 2N4R human tau (tauWT) or tau with K274 and K281 mutated to glutamines to mimic acetylation (tauKQ) [31]. In the cortex, tauKQ and tauWT mice had comparable human tau levels, whereas tauKQ^high mice had significantly greater levels of tauKQ transgene expression (Fig. 2a, b). To investigate the effect of tauKQ on the AIS cytoskeleton, we analyzed βIV-spectrin and AnkG immunostaining in the somatosensory cortex of 2-month-old tauKQ^high mice. There was a significant decrease in βIV-spectrin and AnkG present at the AIS in tauKQ^high mice compared to non-transgenic controls (Fig. 2c-e), suggesting that acetylation at K274/281 can lead to destabilization of the AIS cytoskeleton in vivo. The length of the AIS cytoskeleton, which can be modulated by external stimuli [43, 44], did not differ in tauKQ^high and non-transgenic control mice (Fig. 2f). To exclude a possibility that AIS destabilization in tauKQ^high mice is driven by overexpression of tau transgene per se and is independent of acetyl-tau mimicking mutations, we further analyzed the AIS proteins in 10–12-month-old transgenic mice expressing equivalent levels of tauWT and tauKQ [31] (Fig. 2e). There was a significant reduction of both βIV-spectrin and AnkG intensity in the cortex of tauKQ mice compared to tauWT mice in 10–12 months (Fig. 2g-i). In the aged tauKQ mice, the AIS length measured by AnkG labeling was shorter than in tauWT mice (Fig. 2j), suggesting that prolonged expression of acetyl-tau mimic also compromises the integrity of the AIS besides reducing the levels of AIS proteins.

Acetylation at K274 and K281 in the MT-binding domain of tau occurs in human AD brains (Fig. 1b, c). Posttranslational modifications of tau in the MT-binding domain can affect its binding affinity for MTs [45‐47]. To examine the effect of acetylation at K274/K281 on tau binding to MTs, we mutated K-to-Q to mimic acetyl-lysine or K-to-R to block acetylation. Substitution of lysine (K) to either glutamine (Q) or arginine (R) has been widely used in a number of studies to determine effects of acetylation of specific lysine residues on protein functions of histones, transcription factors, and enzymes [48‐52]. Importantly, this lysine substitution faithfully represents the functional impact of acetylation/deacetylation on tau [31, 53, 54]. To assess the binding affinity, we analyzed cells expressing mApple-tauWT and GFP-tau mutants in a competitive MT-binding assay [40]. In HeLa cells co-transfected with mApple-tauWT and GFP-tubulin, tauWT co-localized with MTs visualized by GFP-tubulin, indicating attachment of WT tau to MTs (Additional file 1: Figure S2A). If a tau mutant has a lower affinity for MTs, it exhibits a more diffuse distribution than tauWT and an increased unbound tau index which is calculated by subtracting the fluorescence signal of tauWT from the tau mutant. The unbound tau index for tauK274/281Q was significantly higher than tauWT, whereas tauK274/281R showed similar binding affinity as tauWT (Additional file 1: Figure S2B, C), suggesting that acetylation at K274/281 reduces tau binding to MTs. Binding of individual mutant tauK274Q did not change, and that of tauK281Q was modestly reduced. K-to-Q mutations outside the MT-binding domain (K163/174/180/190Q; 4KQ(N)) did not affect the unbound tau index, as expected (Additional file 1: Figure S2C). The difference in the amount of unbound cytoplasmic tau is unlikely due to tau expression levels, as tauK274/281Q and tauWT were expressed at similar levels (Additional file 1: Figure S2D), and the unbound tau index did not correlate with tau levels (p = 0.19, Pearson correlation analysis).

Tau isoforms with different MT-binding affinities differ in their effects on the dynamic behavior of MTs [55, 56]. To determine whether acetylation at K274/281 affects MT dynamics in living cells, we tracked the movement of individual MTs by using GFP-tagged ending-binding (EB) proteins that bind to MT plus-ends [57]. The rates of movement were calculated in an unbiased fashion with plusTipTracker, which faithfully tracks EB comets from time-lapse images of living cells [41].

We co-transfected primary rat hippocampal neurons at 6–7 DIV with GFP-EB3 and mApple-tau constructs. EB comets were imaged 24 to 48 h after transfection, and MT dynamics were analyzed (Fig. 3a, b). The rate of movement was significantly higher in neurons expressing tauKQ than in those expressing tauWT (Fig. 3c). The levels of tauWT and tauKQ did not differ (Fig. 3d), and the rates of movement of EB comets did not correlate with tau levels (p = 0.37, Pearson correlation analysis).

The stability of neuronal MTs varies depending on the cellular compartment. MTs in the AIS are highly stabilized by modifications and bundling and function as a barrier to maintain the polarized distribution of tau in axons [20, 23, 58]. Since the acetyl-tau mimic increased MT dynamics in neurons, we next examined the effect of tau acetylated at K274/281 on the stability of the MTs in the AIS by FRAP. We also suspected that the compromised integrity of βIV-spectrin and AnkG induced by the acetyl-tau mimic (Fig. 2c-J) could contribute to altering the stability of MTs in the AIS because submembranous cytoskeletal proteins are connected to MTs in the AIS [20, 59].

Rat hippocampal neurons at 6–7 DIV were co-transfected with GFP-tubulin and mApple-tau constructs; 24 h later, we labeled the AIS of live neurons with an antibody against the extracellular domain of neurofascin, located in the plasma membrane of the AIS [59]. The neurofascin antibody delineated the AIS during live imaging (Fig. 4a), enabling us to photobleach GFP-tubulin in a segment of the AIS. Monitoring of the GFP-tubulin signal for 1 min after photobleaching showed a significantly faster fluorescence recovery rate in cells expressing tauKQ than in those expressing similar levels of tauWT (Fig. 4b, c), consistent with destabilization of MTs in the AIS by acetylated tau.

Since the AIS is implicated in restricting tau distribution in the axon [23], we suspected that weakening of the AIS barrier by acetylated tau could lead to the mislocalization of axonal tau. Photoconvertible tau constructs were used to monitor the movement of tau in rat hippocampal neurons. At DIV 6–8, the cells were transfected with mEOS2-tau construct that turns red upon UV illumination. Tau was targeted for photoconversion in an AIS segment ~30 μm away from the cell body, and influx of tau into the somatodendritic compartment was monitored (Fig. 5a; Additional file 2: Movie 1, Additional file 3: Movie 2). As judged from the increase in fluorescence intensity, the influx was much greater with tauKQ than tauWT or tauK274/281R (tauKR) (Fig. 5b, c). The higher influx rate suggests that acetylation at K274/281 enabled tau to bypass the sorting barrier in the AIS. In the distal axons, cells expressing tauWT and tauKQ showed similar time-dependent increase in fluorescence intensity (Fig. 5d), suggesting that both can freely redistribute in the absence of the AIS barrier. Since hyperphosphorylation of tau can also contribute to somatodendritic mislocalization of tau [14, 15, 23, 60, 61], we analyzed the phosphorylation status of tauKQ. Acetyl-mimicking mutations at K274/281 did not alter phosphorylation at S262/S356 (12E8) and S202/T205 (AT8) in the cortex of tauKQ-expressing mice. Phosphorylation at S396/S404 (PHF1) was even significantly reduced in tauKQ mice compared to tauWT mice (Additional file 1: Figure S3A, B). These data support that acetylation at K274/281 induces somatodendritic mislocalization of tau by mechanisms that are independent of increased tau phosphorylation.

To test whether stabilization of MTs could restore the AIS barrier function and prevent mislocalization of acetylated tau, we treated the neurons expressing tauKQ with a low dose of epothilone D (EpoD), an MT stabilizer, to reduce MT dynamics [36, 37]. Treatment with EpoD prevented the influx of axonal tauKQ into the somatodendritic compartment (Fig. 5b, c). EpoD also modestly slowed down the movement of tauKQ to the distal portion of axon (Fig. 5d). These findings support the importance of MT stability in controlling axonal sorting of tau.