|Home | About | Journals | Submit | Contact Us | Français|
Tau toxicity has been implicated in the emergence of synaptic dysfunction in Alzheimer’s disease (AD), but the mechanism by which tau alters synapse physiology and leads to cognitive decline is unclear. Here, we report abnormal acetylation of K274 and K281 on tau, identified in AD brains, promotes memory loss and disrupts synaptic plasticity by reducing postsynaptic KIBRA (KIdney/BRAin protein), a memory-associated protein. Transgenic mice expressing human tau with lysine-to-glutamine mutations to mimic K274 and K281 acetylation (tauKQ) exhibit AD-related memory deficits and impaired hippocampal long-term potentiation (LTP). TauKQ reduces synaptic KIBRA levels and disrupts activity-induced postsynaptic actin remodeling and AMPA receptor insertion. The LTP deficit was rescued by promoting actin polymerization or by KIBRA expression. In AD patients with dementia, we found enhanced tau acetylation is linked to loss of KIBRA. These findings suggest a novel mechanism by which pathogenic tau causes synaptic dysfunction and cognitive decline in AD pathogenesis.
Alzheimer’s disease (AD) causes gradual memory loss associated with synapse deterioration. Tauopathies, including AD and frontotemporal lobar degeneration with tau inclusions (FTLD-tau), are characterized by the formation of neurofibrillary tangles (NFTs) in the brain composed of hyperphosphorylated tau, a microtubule-associated protein. Cognitive decline coincides with the deposition of NFTs in AD progression (Giannakopoulos et al., 2003); however, the presence of NFTs is not sufficient to cause memory loss in FTLD-tau transgenic mice (Santacruz et al., 2005; Sydow et al., 2011; Van der Jeugd et al., 2012). Although it is widely accepted that tau plays a significant role in AD pathogenesis, the entity of the toxic species and the mechanism by which tau causes cognitive decline in AD are still unclear.
Fine-tuning of synaptic strength in response to neuronal activity is critical for cognitive processes, such as learning and memory. Synapses are particularly vulnerable to toxicity in AD, and synapse loss is correlated with cognitive decline (DeKosky and Scheff, 1990). In FTLD-tau mouse models, mutant tau alters the efficacy of synaptic transmission and disrupts plasticity linked to behavioral deficits (Hoover et al., 2010; Warmus et al., 2014; Yoshiyama et al., 2007). Under pathological conditions, tau is missorted from axons into dendritic compartments and postsynaptic spines (Zempel and Mandelkow, 2014). In cultured neurons, mislocalization of an FTLD-tau mutant in postsynaptic spines depletes glutamate receptors (Hoover et al., 2010), and amyloid-β (Aβ) oligomers induce missorting of tau into dendrites and spine retraction (Zempel et al., 2010). Moreover, postsynaptic targeting of Fyn kinase by tau exacerbates excitotoxicity in an AD mouse model (Ittner et al., 2010). These findings support a key role for tau-mediated pathogenesis at postsynaptic sites; yet, how tau affects the synaptic mechanisms underlying the encoding of memory is largely unknown.
The activity of p300 acetyltransferase is enhanced in human AD brain (Aubry et al., 2015), and p300 acetylates tau (Min et al., 2010). Human tauopathy brains display a marked pathological enhancement of tau acetylation (Grinberg et al., 2013; Irwin et al., 2013; Irwin et al., 2012; Min et al., 2015; Min et al., 2010). Acetylation of proteins can alter their function, modify their interactions with binding partners, and affect their stability (Spange et al., 2009). Likewise, acetylation affects tau in different ways. It inhibits tau binding to microtubules (Cohen et al., 2011), promotes tau accumulation by preventing its degradation (Min et al., 2015; Min et al., 2010), and enhances tau oligomerization and aggregation (Cohen et al., 2011). These studies support the notion that aberrantly acetylated tau (ac-tau) is pathogenic, but whether ac-tau induces neuronal dysfunction underlying cognitive decline is unknown.
The KIBRA gene was first linked to human memory performance in a genome-wide screen for single nucleotide polymorphisms (SNPs) (Papassotiropoulos et al., 2006). Predominantly expressed in memory-associated regions of the brain, KIBRA is enriched at postsynaptic sites (Schneider et al., 2010; Zhang et al., 2014). It interacts with components of postsynaptic glutamate receptor complexes (Makuch et al., 2011), actin regulatory networks (Duning et al., 2008; Kremerskothen et al., 2003; Kremerskothen et al., 2005), and kinase signaling pathways (Buther et al., 2004; Vogt-Eisele et al., 2014). Mice deficient in KIBRA have impaired hippocampal synaptic plasticity and memory deficits (Makuch et al., 2011). Intriguingly, KIBRA SNPs have been implicated in increased risk for late-onset AD (Burgess et al., 2011; Corneveaux et al., 2010; Rodriguez-Rodriguez et al., 2009).
Here we report that brains from AD patients with dementia exhibit significantly higher levels of tau acetylated on lysine-274 (K274) and lysine-281 (K281). To define the effects of the aberrant acetylation of these lysines on brain function and cognition, we generated transgenic mice using the murine prion promoter to drive expression of either wild-type human tau (tauWT) or human tau with lysine-to-glutamine substitutions to mimic the charge and structure of acetylated K274 and K281 (tauKQ). We investigated the effects of tauKQ on synaptic plasticity and memory encoding and established the dysregulation of postsynaptic mechanisms involving activity-dependent actin polymerization and AMPA receptor (AMPAR) trafficking. Our study also provides a mechanistic link between aberrant tau acetylation, KIBRA deficiency, and dementia in AD.
To investigate the acetylated lysine residues on tau in tauopathy, we immunoprecipitated tau from homogenized human AD brain and rTg4510 mouse brain, a FTLD-tau model that exhibits pathological tau accumulation (Santacruz et al., 2005), and performed analysis by reverse-phase liquid chromatography coupled to tandem mass spectrometry. K274 was acetylated in rTg4510 mouse brain (Figure 1A), in agreement with acetylated K274 (ac-K274) observed in human AD brain (Grinberg et al., 2013). A novel acetylation site on K281 was identified in human AD brain (Figure 1B). K274 and K281 are in the microtubule binding domain of tau. Two monoclonal antibodies, MAb359 and MAb63, were generated using an ac-tau peptide containing amino acids 264–287 as the antigen (Figure 1C). In HEK293 cells, MAb359 and MAb63 showed strong immunoreactivity for tauWT only in the presence of the acetyltransferase p300 (Figure 1D). A single mutation of lysine to arginine for either K274 (K274R) or K281 (K281R) that blocks p300-induced acetylation eliminates the signal from MAb359 and MAb63, respectively (Figure 1D). Blocking acetylation of other lysines did not affect the immunoreactivity of these antibodies for ac-tau (Figure S1A). These results established the specificity of MAb359 for ac-K274 and MAb63 for acetylated K281 (ac-K281).
To assess tau acetylation in an AD-related mouse model, we established a transgenic mouse that expresses tauWT driven by the murine prion promoter, and crossed it with human amyloid precursor protein (hAPP)-J20 mice that have high hAPP/Aβ levels (Mucke et al., 2000). Modest levels of ac-K274 and ac-K281 were detected in hippocampus of 6-month-old tauWT mice, and were significantly elevated in doubly transgenic mice that also express hAPP (Figures 1E–1G). To investigate ac-tau levels in human brains, we first confirmed the antibody specificity in a competition assay, in which MAb359 and MAb63 immunoreactivity was blocked by pre-incubation with recombinant human tau acetylated by p300 (Figure S1B and S1C). MAb359 and MAb63 immunoreactivities were markedly higher in the superior temporal gyrus of AD patients with a Clinical Dementia Rating (CDR) score of 5—the most severe dementia cases—than those that were CDR 0 with no cognitive impairments (Figures 1H–1J). In mildly demented cases (CDR 0.5), the amounts of ac-K274 were similar to non-demented cases, whereas ac-K281 levels were significantly greater (Figures 1H–1J). Thus, aberrant acetylation of K274 and K281 is upregulated by high Aβ/hAPP levels and associated with cognitive impairments in AD.
To explore how ac-K274 and ac-K281 could contribute to cognitive deficits in AD, we generated transgenic mice expressing human tau with lysine to glutamine mutations at K274 and K281 (tauKQ) to mimic acetylation. Since glutamine resembles the chemical structure and neutral charge of acetyl-lysine, this residue substitution is widely used to study the effects of acetylation on specific lysine residues (Hecht et al., 1995; Liu et al., 2012; Scroggins et al., 2007; Wang et al., 2003). As in the tauWT mice, the murine prion promoter was used to drive tauKQ transgene expression. We established two founder lines: tauKQ mice that express HT7-positive human tau at similar levels as tauWT mice, and a higher expression line called tauKQhigh (Figures 2A and 2B). In situ hybridization revealed that the human tau expression patterns in hippocampus, cortex and cerebellum were similar in tauWT, tauKQ, and tauKQhigh lines (Figure S2). Compared with nontransgenic (ntg) controls, similar levels of astrogliosis were induced in the dentate gyrus of all three lines, as detected with GFAP immunoreactivity (Figures 2C and 2D). Phosphorylation of human tau was detected in tauWT mice (Figures S3A–S3D), consistent with another human tau transgenic mouse line (Andorfer et al., 2003). Phospho-serine 202/205 (AT8) levels on tauKQ were reduced whereas phosho-serine 396/404 (PHF-1) and threonine 181 (AT270) levels were similar to tauWT (Figures S3A–S3D). Caspase-cleaved tau was observed in the hippocampus of tauKQhigh mice, but not in tauWT or tauKQ mice (Figure S3E). Tau was also monitored with MC1 staining, which recognizes a pathological confirmation of tau in FTLD-tau mouse models (Decker et al., 2015; Sydow et al., 2011; Van der Jeugd et al., 2012). Consistent with pathogenicity of ac-K274 and ac-K281, tauKQ mice exhibited significantly higher MC1 immunoreactivity than tauWT mice, which was further elevated in tauKQhigh mice (Figures 2E and 2F).
To determine if ac-tau affects synaptic plasticity in the hippocampus associated with memory loss, field recordings were performed in the dentate gyrus molecular layer of acute hippocampal slices. After theta-burst stimulation (TBS) of the perforant pathway to dentate granule cell (DGC) synapses, LTP in tauWT slices was sustained at the same level as ntg slices for the duration of the recordings (Figure 3A). In contrast, potentiation significantly declined in tauKQ slices by 60 min after LTP induction (Figure 3B). Since tauWT and tauKQ mice have similar levels of human tau (Figures 2A and 2B), these results support the pathogenicity of ac-K274 and ac-K281. Moreover, tauKQhigh slices exhibited a pronounced LTP impairment (Figure 3C).
In basal synaptic transmission, the strength of the perforant pathway inputs and the postsynaptic responsiveness of DGCs in tauWT, tauKQ or tauKQhigh slices were similar to ntg controls (Figures 3D–3F). Perforant pathway inputs also showed normal presynaptic release properties in slices from tauKQ-expressing mice (Figure 3G). Whole-cell patch-clamp recordings of DGCs from ntg and tauKQhigh slices revealed comparable AMPAR- and NMDA receptor-mediated synaptic strength (Figures 3H–3J). Thus, the tauKQ-induced LTP deficit is not likely due to insufficient plasticity induction, but rather to the inhibition of long-term synaptic strengthening.
To determine the functional outcome of ac-K274 and ac-K281, we focused our behavioral tests in tauKQhigh mice. Spatial learning and memory was examined in the Morris water maze test (MWM). During hidden platform training, there was no significant difference between 11–13-month-old ntg and tauKQhigh mice in spatial learning or swim speed (Figures 4A and 4B). In 24 h and 72 h probe trials, ntg mice demonstrated robust spatial memory with a strong preference for the target quadrant (Figures 4C and 4D). In contrast, the amount of time that tauKQhigh mice spent in the target quadrant was not significantly different from chance levels in the 72 h probe trial (Figures 4C and 4D), indicating that tauKQhigh mice were impaired in spatial memory.
Besides spatial memory deficits, patients with mild cognitive impairment (MCI) or AD also exhibit deficiencies in distinguishing between similar experiences (Ally et al., 2013; Wesnes et al., 2014; Yassa et al., 2010). To determine if aberrant ac-tau interferes with the encoding of distinct memories, we performed two different behavioral tests to assess pattern separation in tauKQhigh mice. Mice were first tested in an object-context discrimination task (Jain et al., 2012) in which they learn to associate specific objects with two distinct contexts during a sample phase, followed by a test phase to assess their ability to distinguish the incongruent object within the familiar context (Figure 4E). At 6–7 months old, ntg, tauWT and tauKQhigh mice spent an equal proportion of time exploring the objects in the sample phase (Figure 4F). However, unlike ntg and tauWT mice, which spent significantly more time exploring the incongruent object during the test phase, tauKQhigh mice showed no preference, suggesting that tauKQhigh mice were unable to differentiate between previous experiences (Figure 4G).
We also examined pattern separation by testing contextual discrimination learning with fear conditioning. The ability of mice to learn to distinguish between two highly similar contexts in this test requires DGC function (McHugh et al., 2007; Nakashiba et al., 2012). In the fear acquisition phase, 10–12-month-old ntg and tauKQhigh mice exhibited similar levels of freezing when shocked in context A (Figure 4H). In the subsequent generalization phase, ntg and tauKQhigh mice showed generalized fear between context A and a highly similar context B (Figure 4I). For the next 16 days of contextual discrimination training a foot shock was given in A (non-safe), but not in B (safe) (Figure 4J). Ntg mice learned to discriminate the two contexts with increasing freezing in A, but not in B (Figure 4K). TauKQhigh mice exhibited significantly worse discrimination between A and B than ntg mice (Figure 4K). When mice were tested in two very different contexts C (non-safe) and D (safe), tauKQhigh and ntg mice likewise demonstrated the ability to distinguish between C and D (Figure S4). These results suggest that ac-tau interferes with the encoding of similar experiences as distinct memories.
We next dissected the molecular mechanism underlying the ac-tau-induced deficits in LTP, spatial memory and pattern separation. In cultured hippocampal neurons, induction of chemical LTP by a brief exposure to glycine (300 μM) in the absence of magnesium recruits additional GluA1-containing AMPARs to the postsynaptic surface, resulting in the long-term enhancement of synaptic strength (Lu et al., 2001; Oku and Huganir, 2013). To determine if ac-tau affects the activity-induced trafficking of AMPARs, neurons were co-transfected with mApple and GFP-tagged tauWT, tauKQ or tau with lysine-to-arginine mutations to block the acetylation of K274 and K281 (tauKR). Based on the fluorescence intensity of GFP-tagged tau measured in the cell soma, there were equivalent levels of tauWT, tauKQ and tauKR in neurons (Figure S5). The basal level of GluA1 in spines in neurons expressing tauWT, tauKQ, or tauKR was not significantly different from control neurons (Figures 5A and 5B). In neurons with tauWT or tauKR, chemical LTP increased the surface GluA1 immunostaining in spines similarly to control neurons (Figures 5A and 5B). However, tauKQ blocked the recruitment of AMPARs to spines after chemical LTP (Figures 5A and 5B), suggesting that aberrant levels of ac-tau in neurons inhibits the insertion of postsynaptic AMPARs required for LTP. Consistent with a postsynaptic role for ac-tau in pathogenesis, ac-K274 and ac-K281 tau was present in the postsynaptic density in human AD brain (Figure 5C).
The activity-dependent recruitment of synaptic AMPARs and the maintenance of LTP expression require actin polymerization (Fukazawa et al., 2003; Krucker et al., 2000). In the molecular layer of the dentate gyrus, fluorescently conjugated phalloidin labels F-actin in spines of DGCs (Capani et al., 2001; Fukazawa et al., 2003). Human tau co-localized with F-actin in DGC spines of tauKQhigh mice (Figures 5D). Basal F-actin levels in the molecular layer of tauKQhigh slices were not significantly different from ntg slices (Figures 5E–H). TBS of the perforant pathway elevated DGC F-actin levels in the molecular layer of ntg slices and tauWT slices (Figures S6A–S6D), but not in tauKQhigh slices, suggesting that ac-tau disrupts the activity-dependent polymerization of postsynaptic actin (Figures 5E–5H). To investigate how acetylation affects tau binding to polymerized actin, phalloidin-bound F-actin was precipitated from tauWT and tauKQ mouse brain homogenate. TauWT, not tauKQ, coprecipitated with F-actin (Figure S6E), suggesting that acetylation weakens the interaction between tau and the actin cytoskeleton. Strikingly, treatment of slices from 5–6-month-old tauKQhigh mice with jasplakinolide (JPK), a membrane-permeable drug that promotes actin polymerization, restored LTP to ntg levels (Figure 5I). JPK had no observable effect on LTP in tauWT mice (Figure S6F). These results strongly support that the tauKQ-induced synaptic plasticity impairment is mediated by a deficiency in activity-dependent actin polymerization.
Our data so far demonstrate that ac-tau disrupts postsynaptic signaling involving actin regulatory networks that control AMPAR insertion during LTP. KIBRA is enriched in the postsynaptic density (PSD) (Johannsen et al., 2008), and is linked to late-onset sporadic AD (Burgess et al., 2011; Corneveaux et al., 2010; Rodriguez-Rodriguez et al., 2009). Levels of KIBRA protein (~150 kDa) in the brain of CDR 5 cases were significantly lower than CDR 0 cases, consistent with a link between KIBRA and AD (Figures 6A and 6B). The same KIBRA antibody recognized a specific ~150 kDa band in HEK293 cells overexpressing human KIBRA (Figure S7A). KIBRA levels remained significantly reduced in severely demented cases after normalization to the postsynaptic protein PSD-95, supporting KIBRA loss in synapses (Figures 6C and 6D). Notably, the enhancement of ac-K274 or ac-K281, which coincides with cognitive decline, exhibits a strong correlation with reduced KIBRA protein in AD (Figures 6E and 6F).
We found no difference in the total amounts of KIBRA expressed in the hippocampus of tauKQhigh and ntg mice (Figures S7B and S7C). We next examined postsynaptic KIBRA levels in the phalloidin-labeled spines of DGCs (Figure 6G). Minimal KIBRA immunofluorescence was detected in KIBRA knockout mouse brain (Figure S7D). Compared to ntg mice, tauKQhigh mice exhibited weaker KIBRA immunolabeling overlapping with F-actin puncta (Figure 6H and 6I), whereas KIBRA co-localization with F-actin was normal in tauWT mice (Figures S7E and S7F). These data support that high levels of ac-tau trigger a reduction in postsynaptic KIBRA.
Does aberrant tau acetylation impair plasticity by downregulating KIBRA? Similar to our transgenic mice expressing tauKQ (Figure 3), KIBRA-deficient mice have normal basal synaptic transmission but impaired LTP expression (Makuch et al., 2011). We next determined if enhancing KIBRA expression could re-establish plasticity in tauKQ-expressing neurons. Cultured hippocampal neurons were transduced with KIBRA, which was localized in dendrites and within spines (Figure 7A). In unstimulated conditions, quantification of surface GluA1 immunolabeling in spines did not reveal significant differences among control neurons and neurons expressing KIBRA, tauKQ, or KIBRA with tauKQ (Figure 7B and 7C). After chemical LTP, surface GluA1 immunoreactivity was enhanced in spines of control and KIBRA neurons but AMPAR insertion was blocked in tauKQ neurons (Figure 7C). Co-expression of KIBRA with tauKQ in neurons restored the recruitment of AMPARs to synapses (Figure 7C), confirming that increasing KIBRA levels was sufficient to reverse the AMPAR trafficking deficit triggered by tauKQ.
Our data suggest that the tauKQ-induced plasticity impairment is mediated by loss of activity-dependent actin polymerization. To dissect the role of KIBRA in this process, we tested how KIBRA modulates actin polymerization in spines of tauKQ-expressing neurons. In unstimulated neurons, F-actin levels in spines were not significantly changed by KIBRA or tauKQ (Figures 7D and 7E). As expected, spines on control neurons showed an increase in F-actin after chemical LTP, which was likewise observed in KIBRA neurons (Figure 7E). The activity-induced polymerization of actin at synapses was blocked by tauKQ, but it was completely rescued by the co-expression of KIBRA (Figure 7E). These results support a novel mechanism for pathophysiology in AD in which ac-tau reduces KIBRA levels at synapses, leading to impaired actin-based plasticity and loss of synapse potentiation (Figure 8).
Our findings suggest that aberrant ac-K274 and ac-K281 on tau is a contributing factor to memory loss related to tauopathies, including AD. We showed that dementia in AD is associated with elevated levels of ac-K274 and ac-K281, which correlated negatively with those of KIBRA, a synaptic protein critical for AMPAR trafficking. Mimicking acetylation of these lysines in transgenic mice impairs synaptic plasticity and memory retention, likely due to deficient activity-dependent actin polymerization and AMPAR trafficking. Elevating KIBRA in neurons expressing acetyl-mimicking tau was sufficient to re-establish actin polymerization and AMPAR insertion during plasticity.
We identified acetylation of K274 and K281 on tau in the brain of a tauopathy mouse model and human AD brain, respectively, and developed specific antibodies targeting either ac-K274 or ac-K281. Immunoblots confirmed that levels of K274 and K281 acetylation are greater in AD patients with severe dementia than non-demented cases. Acetylation of K281, but not K274, was increased in patients with mild dementia, which supports a potential divergence in the regulation of acetylation on these residues during pathogenesis. The cause of specific tau hyperacetylation in AD is unknown; however, possible mechanisms include p300 dysregulation (Aubry et al., 2015), loss of sirtuin 1 (Julien et al., 2009), or tau autoacetyltransferase activity (Cohen et al., 2013). In an AD mouse model with human tau, levels of ac-K274 and ac-K281 were elevated, illustrating a link between pathogenic hAPP/Aβ and aberrant tau acetylation. In cultured neurons, Aβ oligomers are sufficient to increase acetylated tau levels (Min et al., 2010). Whether APP or its proteolytic fragments promote tau acetylation in vivo remains to be determined. Interestingly, levels of ac-K270 (homologous to K281 on human tau) and other acetylated lysines on murine tau were not affected by hAPP/Aβ (Morris et al., 2015), suggesting that acetylation of human and murine tau could be modulated differently. Notably, ac-K274, while highly enriched in AD and other tauopathy brains (Grinberg et al., 2013), was not detected in ntg or hAPP/Aβ mice lacking pathogenic tau accumulation, supporting the specificity of ac-K274 as a pathogenic signature for tauopathy.
We developed three lines of transgenic mice that express either tauWT or tauKQ and established a causal role for ac-K274 and ac-K281 in synaptic dysfunction and memory deficits. Electrophysiological analyses revealed LTP deficits in mice expressing tauKQ, but not in mice expressing equivalent levels of tauWT. The strength of presynaptic inputs or postsynaptic responsiveness in basal transmission was not altered by tauKQ, supporting that the expression of plasticity is preferentially affected. Indeed, tauKQ mice are largely normal behaviorally, but they exhibited deficits in hippocampal-dependent spatial memory and dentate gyrus-associated pattern separation. Since acetylation neutralizes the positive charges on lysines, acetylation of K274 and K281 in the microtubule binding domain could have a profound effect on the conformation and aggregation propensity of tau. Acetylation of K280 is associated with aggregation of phosphorylated tau (Cohen et al., 2011), whereas acetylated lysines in KXGS motifs block site-specific phosphorylation and tau aggregation (Cook et al., 2014). With the ac-K274 and ac-K281 mimic, phosphorylated residues on tau were either reduced or unchanged, suggesting that hyperphosphorylated tau does not trigger the physiological and behavioral phenotypes in tauKQ-expressing mice. These tauKQ-induced phenotypes were also unlikely due to astrogliosis that occurred in in all three lines expressing either tauWT or tauKQ. The hippocampus of tauKQhigh mice had caspase-3-cleaved tau fragments that are linked to neurotoxicity in AD (Gamblin et al., 2003; Rissman et al., 2004). However, the C3 fragments were undetectable in the lower expresser tauKQ mice with LTP deficits, suggesting that the C3 fragment is unlikely to be the major culprit of tauKQ-induced plasticity impairment. Pathologically, the accumulation of MC1-positive tau in DGC mossy fibers was significantly higher in tauKQ mice than those expressing similar levels of tauWT. The exact toxic tau species responsible for synaptic and behavioral deficits in tauKQ mice remains to be identified.
Mechanistic dissection revealed that pathogenic ac-K274 and ac-K281 obstructs the activity-dependent modifications in postsynaptic strength. Specifically, ac-K274 and ac-K281 on tau blocks activity-induced F-actin assembly, thereby inhibiting AMPAR insertion and synapse potentiation. In tauKQ-expressing mice, promoting actin polymerization with JPK rescued LTP, suggesting that ac-tau destabilizes F-actin accumulation in the postsynaptic compartment. Tau binds to F-actin (Correas et al., 1990; DuBoff et al., 2012), and regulates F-actin bundling (Cunningham et al., 1997; Fulga et al., 2007). The KQ mutation weakened the interaction of tau with F-actin, raising the possibility that acetylation inhibits tau-actin binding and contributes to actin destabilization during activity-dependent polymerization. Aβ treatment or neuronal activity regulates the interaction of tau with F-actin (Frandemiche et al., 2014) and phosphorylated tau causes abnormal F-actin accumulation in neurons (Fulga et al., 2007). However, whether tau regulates actin dynamics at synapses through a direct interaction remains unclear. The finding that ac-tau inhibits synaptic plasticity is very different from the effect of phosphorylated tau, which is linked with the downregulation of glutamate receptors at synapses and loss of spines (Hoover et al., 2010). On the other hand, the interaction between tau and Fyn kinase can induce NMDAR-mediated excitotoxicity (Ittner et al., 2010). Thus, pathogenic forms of tau can trigger postsynaptic dysfunction through different mechanisms. It will be interesting to determine how these synaptic mechanisms collectively contribute to the progression of cognitive decline in AD.
In AD brains, we found higher levels of ac-tau associated with reduced levels of KIBRA, a critical regulator of AMPARs recycling (Makuch et al., 2011). Interestingly, levels of KIBRA mRNA are upregulated in neurons from AD brain (Corneveaux et al., 2010), which may represent a compensatory response for the loss of KIBRA protein that we observe. Synapse degeneration could account for a portion of the KIBRA lost in the brain. Even so, KIBRA is significantly reduced relative to PSD-95, another postsynaptic protein, indicating that, in addition to synapse loss, another mechanism likely contributes to KIBRA deficiency. Enhanced tau acetylation appears to correlate with the reduced levels of KIBRA, which points toward a role for pathogenic tau in the dysregulation of KIBRA in AD.
Acetyl-mimicking tau reduced postsynaptic KIBRA associated with F-actin in DGC spines. The mechanism by which ac-tau disrupts the synaptic localization of KIBRA remains to be established, but it could be an early event in disease pathogenesis. Later in disease progression, the mislocalization of KIBRA could result in its degradation by another mechanism leading to lower overall KIBRA levels in AD brains at CDR 5. Importantly, driving KIBRA expression in neurons with tauKQ restored actin dynamics and AMPAR trafficking at synapses during LTP. This suggests that a sufficient amount of KIBRA at synapses is essential for the regulation of actin and AMPARs during synaptic plasticity. KIBRA interacts with actin cytoskeletal regulators that are involved in synaptic plasticity, including synaptopodin (Duning et al., 2008), dendrin (Kremerskothen et al., 2003) and PICK1 (Makuch et al., 2011). Given that KIBRA modulates the F-actin rich processes of epithelial cells during cell migration (Duning et al., 2008), actin-based plasticity in spines may rely on similar signaling pathways. KIBRA also interacts with the protein kinase Mζ (PKMζ) (Buther et al., 2004), an active atypical PKC isoform that plays a role in LTP maintenance (Serrano et al., 2005) and memory (Drier et al., 2002). PKMζ accumulates in NFTs in human AD brain (Crary et al., 2006), raising the possibility of a mechanistic link between tau, KIBRA and PKMζ. Taken together, our results demonstrate that acetylated tau in AD disrupts one or more KIBRA-mediated signaling pathways that are crucial for the expression of synaptic plasticity.
Our findings suggest that ac-tau blocks activity-dependent AMPAR trafficking by reducing postsynaptic KIBRA and inhibiting actin dynamics, establishing a novel pathway for tau-mediated synaptic dysfunction. We propose that therapeutic strategies designed to reduce tau acetylation could help restore cognitive function in AD.
Immunoprecipitated tau from brain homogenate was separated on a 1D-SDS-PAGE gel. Subsequent excision from the gel and in-gel digestion with trypsin was performed prior to mass spectrometric analysis by reversed-phase nano-HPLC ESI-MS/MS using a QqTOF QSTAR Elite mass spectrometer (SCIEX). For detailed methods, see Supplemental Experimental Procedures.
Human and mouse brain tissues were homogenized and centrifuged, and the supernatant was collected for immunoblot analyses. A Percoll gradient procedure was used to enrich the postsynaptic density fraction from human brain. For detailed methods, see Supplemental Experimental Procedures.
Transgenic mice were generated by injection of the murine prion promoter-tauWT or -tauKQ transgenes into fertilized mouse oocytes, followed by implantation. hAPP-J20 mice (RRID:MGI_3639711) have been described (Mucke et al., 2000). The rTg4510 mouse line (RRID:MGI_4819951) has been described (Santacruz et al., 2005). KIBRA knockout mice (RRID:MGI_5301656) were generated as described (Makuch et al., 2011). For additional methods, see Supplemental Experimental Procedures.
Immunohistochemistry was performed on 30-μm-thick mouse brain sections. For detailed methods on MC1, KIBRA, and HT7 immunolabeling, Alexa Fluor 488 Phalloidin staining, and image quantification, see Supplemental Experimental Procedures
The object-context discrimination test was performed as described (Jain et al., 2012). The protocol used for context discrimination fear conditioning was adapted from Nakashiba et al. (2012). In the Morris water maze, mice were subjected to 4 days of hidden platform training followed by a 24- and 72-hour probe test. See Supplemental Experimental Procedures for detailed methods.
Field recordings were performed in the dentate gyrus molecular layer of acute horizontal brain slices. TBS was applied to the perforant pathway inputs for LTP recordings. See Supplemental Experimental Procedures for detailed methods.
Human tauWT (2N4R) was cloned into the pcDNA3.1 vector (Invitrogen) and the pEGFP-C1 vector (Clontech). Site-directed mutagenesis using PCR was performed on human tauWT cDNA to engineer the KQ or KR mutants. To make the tauKQ mutant, A820C and A841C mutations were introduced on tauWT cDNA. For the tauKR mutant, A821G and A842G mutations were made. For additional information, see Supplemental Experimental Procedures.
Rabbit monoclonal antibodies, MAb359 (RRID:AB_2561278) and MAb63 (RRID:AB_2561279), were generated by Epiotomics (Mountain View, CA). MAb359 has been described (Grinberg et al., 2013). See Supplemental Experimental Procedures for additional information on the antibodies used.
Primary hippocampal neurons from embryonic rats were maintained in Neurobasal/B27 medium and transfected with Lipofectamine 2000. The chemical LTP protocol used was adapted from Lu et al. (2001). For detailed methods, see Supplemental Experimental Procedures.
Differences between multiple means were analyzed by one-way or two-way ANOVA with Bonferroni multiple comparison post-hoc analyses. Learning in the MWM test was analyzed by a longitudinal mixed effects model. For the context discrimination task, a three-way ANOVA with multiple comparisons using Tukey’s range test to control the overall experimental-wise error rate was used to compare the mean percent time freezing in ntg vs. tauKQhigh mice controlling for context, block and day within block. General linear models were used to compare the slopes and intercepts of learning rates in each mouse type. Pearson’s correlation coefficients were used to evaluate the linear relationship between two variables.
We thank Dr. Vahram Haroutunian (Mount Sinai) for human brain samples, Dr. David Borchelt (University of Florida) for the pPrp vector, Dr. Peter Davies (Albert Einstein) for the MC1 antibody, Drs. Charles McCulloch, Isabel Elaine Allen and Maria Glymour for statistical analyses, Dr. T. Michael Gill and Allyson Davis for advice on behavioral studies, Drs. Kurt Thorn and DeLaine Larsen for microscopy support at the UCSF Nikon Imaging Center, Junli Zhang and the Gladstone Institutes Transgenic Gene-Targeting Core, Christopher Ma, Fengrong Yan, Yuan-Hung Lin and Sainan Liu for technical support, Grant Kauwe, Jorge Palop, and Sumihiro Maeda for discussions, Gary Howard for editorial review, and Latrice Gross and Erica Nguyen for administrative assistance. This work was supported by the NIH (1R01AG036884 and R01AG030207 to L.G.; NS40251 and NS062413 to L.M.E.; 5F32AG043301-02 to T.E.T.), the Tau Consortium (to L.G.), and the Stephen D. Bechtel Jr. Foundation. Behavioral data were obtained with the help of the Gladstone Institutes’ Neurobehavioral Core (supported by NIH grant P30NS065780). The Gladstone Institutes received support from a National Center for Research Resources Grant RR18928.
Author ContributionsL.G. and T.E.T. conceived the project. L.G. and T.E.T. designed experiments. T.E.T., P.D.S., S.S.M, S-W.M., Y.L., Y.Z., D.L., I.L., X.C. and B.S. performed experiments. L.M.E., R.L.H., C.W., S-W.M, and R.P. developed experimental protocols, tools or reagents or analyzed data. T.E.T. and L.G. wrote the manuscript.
Publisher's Disclaimer: This is a PDF file of an unedited manuscript that has been accepted for publication. As a service to our customers we are providing this early version of the manuscript. The manuscript will undergo copyediting, typesetting, and review of the resulting proof before it is published in its final citable form. Please note that during the production process errors may be discovered which could affect the content, and all legal disclaimers that apply to the journal pertain.