Tau Acts as a Mediator for Alzheimer's Disease-Related Synaptic Deficits

Abstract

The two histopathological hallmarks of Alzheimer's disease (AD) are amyloid plaques containing multiple forms of Aβ and neurofibrillary tangles containing phosphorylated tau proteins. As mild cognitive impairment frequently occurs long before the clinical diagnosis of Alzheimer's disease, the scientific community has been increasingly interested in the roles of Aβ and tau in earlier cellular changes that lead to functional deficits. Therefore, great progress has recently been made in understanding how Aβ or tau causes synaptic dysfunction. However, the interaction between the Aβ and tau-initiated intracellular cascades that lead to synaptic dysfunction remains elusive. The cornerstone of the two decade-old hypothetical amyloid cascade model is that amyloid pathologies precede tau pathologies. Although the premise of Aβ-tau pathway remains valid, the model keeps evolving as new signaling events are discovered that lead to functional deficits and neurodegeneration. Recent progress has been made in understanding Aβ-PrP^C-Fyn-mediated neurotoxicity and synaptic deficits. Although still elusive, many novel upstream and downstream signaling molecules have been found to modulate tau mislocalization and tau hyperphosphorylation. Here we will discuss the mechanistic interactions between Aβ-PrP^C-mediated neurotoxicity and tau-mediated synaptic deficits in an updated amyloid cascade model with calcium and tau as the central mediators.

Introduction

Alzheimer's disease (AD) was first described by Alois Alzheimer more than a century ago (Alzheimer, 1907), the disease is characterized by progressive memory loss and inability to form new memories. AD is thought to be responsible for 50–60% of dementia cases (Blennow et al., 2006); in fact, there are currently more than 25 million individuals living with AD worldwide (Alzheimer's Association, www.alz.org). Even though cures for AD and related dementias have not yet been identified our understanding of the disease process has expanded greatly since the time of Alois Alzheimer. This review will examine the recent discoveries that reveal how soluble forms of Aβ and tau proteins lead to synaptic deficits.

Shifting Winds: Plaques and tangles give way to soluble Aβ and tau

The brains from AD patients contain amyloid plaques and neurofibrillary tangles, which are considered the two pathological hallmarks of AD; this pathology was initially thought to be responsible for neurotoxicity in AD patients (Alzheimer, 1907; Blennow et al., 2006). Nearly three decades ago, amyloid β-peptide (Aβ) was discovered to be the main component of neuritic plaques (Masters et al., 1985) and the product of proteolytic cleavage of the amyloid-β precursor protein (APP) (Haass et al., 1992). Humans with AD related mutations in APP, presenilin (affecting production of Aβ) or APOE (affecting clearance of Aβ) have been correlated with cognitive deficits (Blennow et al., 2006). Soon after Aβ was found to be the primary component of plaques, the microtubule-associated protein tau in its hyperphosphorylated state was discovered to be the chief component of neurofibrillary tangles (Grundke-Iqbal et al., 1986; Nukina & Ihara, 1986).

Initial research into AD focused on the neuritic plaques and neurofibrillary tangles found at later time points in the disease. Much has been learned about these hallmarks of AD pathology, but the focus of the research community has shifted towards understanding the disease at earlier stages (Selkoe, 2002; Ashe & Zahs, 2010; Crimins et al., 2013). This is due to findings that have indicated that AD has a long preclinical phase preceding cell death and brought the relationship between plaques, tangles, and neurotoxicity into question. In AD the number of neurofibrillary tangles, but not neuritic plaques, has been positively correlated with the severity of dementia (Arriagada et al., 1992). Aβ plaques are found in 40% of elderly adults who have not been clinically diagnosed with dementia (Hulette et al., 1998). Further inquiry has confirmed that many aged persons who do not have dementia and who are aging “normally” show AD pathology (Bennett et al., 2006; Price et al., 2009). Neurofibrillary tangles do not correlate well with neural death as dying neurons often don't contain tangles (Gómez-Isla et al., 1997; Andorfer et al., 2005; Spires et al., 2006). Additionally, the novel mutation APP_E693Δ causes dementia and synaptic deficits without the accumulation of neuritic plaques (Tomiyama et al., 2010). Evidence from AD transgenic mice with increased Aβ levels indicates that cognitive deficits develop before plaques form, if they form at all (Westerman et al., 2002; Lesné et al., 2006). Multiple studies have demonstrated that soluble tau, rather than insoluble tau found in neurofibrillary tangles, is responsible for neurotoxic effects (Santacruz et al., 2005; Oddo et al., 2006). These results downplay the importance of plaques and tangles in AD progression, directing the attention of researchers towards earlier time points in the disease and soluble species of Aβ and tau.

Of Mice and Men: Pushing AD researchers towards the synapse

The creation of transgenic mouse models of AD has advanced the study of the disease, but has lead to new questions as well. These transgenic mice show significant increases in Aβ levels and undergo neuritic plaque formation. However, the degree of widespread neuronal loss is marginal even at ages when behavioral deficits are detected indicating that synaptic dysfunction, not cell death, causes these early behavioral deficits (Games et al., 1995; Hsiao et al., 1996; Bondolfi et al., 2002; Zahs & Ashe, 2010; Rupp et al., 2011). A number of groups researching AD have investigated synaptic changes in the hippocampus, a brain area highly associated with memory. Widespread synapse loss has long been identified in the brain tissue of dementia patients, often via the imaging of the presynaptic terminal marker synaptophysin (Gonatas et al., 1967; Davies et al., 1987; Masliah et al., 1989; Heinonen et al., 1995; Arendt, 2009). In fact, synapse loss has a stronger correlation with cognitive decline than plaque or tangle load in humans with dementia (Davies et al., 1987; Masliah et al., 1989; DeKosky & Scheff, 1990; Terry et al., 1991; Sze et al., 1997; Selkoe, 2002). Loss of synapses has also been identified in multiple AD mouse models as early as 2 months of age (Mucke et al., 2000; Lanz et al., 2003; Hsieh et al., 2006; Perez-Cruz et al., 2011). Baseline synaptic properties (AMPA and NMDA glutamate receptor currents) are altered in AD transgenic mice 3 weeks of age and older (Hsia et al., 1999; Kamenetz et al., 2003; Hsieh et al., 2006; Roberson et al., 2011). At 8 months of age many AD transgenic mice also develop a deficiency in long term potentiation (LTP, a well-developed theory of memory in which synapses are strengthened by a battery of activity) (Chapman et al., 1999; Gureviciene et al., 2004; Malenka & Bear, 2004). The research using AD transgenic mice shows that synapses can be affected at early time points in the disease progression, before profound neuronal loss is observed, suggesting that these changes may be related to the development of dementia.

In AD transgenic mice, synaptic deficits oftentimes do not correlate with Aβ plaque formation, however, deficits often do coincide with elevations in soluble Aβ peptides (McLean et al., 1999; Näslund et al., 2000; Selkoe, 2002; Oddo et al., 2003; Sheng et al., 2012). Furthermore, genetic mutations correlated with AD in humans are associated with an increase in the levels of Aβ or an increase in the proportion of Aβ_1-42 to Aβ_1-40 (two common truncations of the protein found in humans) (Tanzi & Bertram, 2005; Blennow et al., 2006). Recently, research has shifted to understanding the effect of Aβ_1-42 on synapses and the findings have reframed the amyloid hypothesis to focus upon soluble Aβ oligomers (also more descriptively referred to as Aβ-derived diffusible ligands or Aβ micro-aggregates), rather than neuritic plaques (Haass & Selkoe, 2007; Krafft & Klein, 2010; Sheng et al., 2012). One such soluble Aβ oligomer, Aβ*56 has been shown to cause cognitive deficits in mice (Lesné et al. 2006; Cheng et al., 2007). Numerous studies have shown that soluble Aβ oligomers cause the collapse of spines in hippocampal cell- and brain slice-cultures after 5–15 days of treatment (Hsieh et al., 2006; Shrestha et al., 2006; Lacor et al., 2007; Shankar et al., 2007; 2008). Electrophysiological and immunocytochemical studies have shown that treatment with soluble Aβ oligomers leads to deficits in both AMPA and NMDA glutamate receptor signaling and expression at the synapse (Chen et al., 2002; Snyder et al., 2005; Hsieh et al., 2006; Shankar et al., 2007; Zhao et al., 2010). The recapitulation of synaptic changes found in AD transgenic mice by the treatment of cultured neurons with soluble Aβ oligomers validates the hypothesis that soluble Aβ oligomers are responsible for these changes. In our two recent studies, Aβ oligomers induce effects that are similar to those caused by the expression of mutant P301L tau proteins, suggesting that Aβ and tau may share a common signaling pathway (Figure 1).

Figure 1. Pathological tau and Aβ species may share a common downstream pathway that leads to synaptic dysfunction.

A. Representative traces from Hoover et al., 2010. Miniature excitatory postsynaptic currents (mEPSCs) were recorded from 3 week-old cultured rat hippocampal neurons expressing green fluorescent protein (GFP), wild-type tau (WT tau) and mutant P301L tau.

B. Cumulative frequency plots of mEPSC amplitude from the three groups in A.

C. Representative traces from Miller et al., 2014. Un-transfected cultured rat hippocampal neurons were treated with soluble Aβ oligomers for > 3 days and mEPSCs were recorded in both control and treated neurons.

D. Cumulative frequency plots from groups in C. *, p < 0.1; ***, p < 0.001, Kolmogorov-Smirnov test.

Revisiting Amyloid Cascade Hypothesis: Aβ as the trigger and tau as the mediator

Proposed in 1991, the amyloid cascade hypothesis attempted to explain the mechanistic interaction between amyloid and tau pathologies (Hardy & Allsop, 1991). In its original form the amyloid cascade hypothesis laid down the corner stone of the popular model: Aβ → tau → neuronal damage (Figure 2A). In its second version, abnormal increase in intracellular calcium was added as a central piece (Hardy & Higgins, 1992; Figure 2B). Aberrant cleavage of APP or a mutation of APP leads to increased deposition of Aβ. From this point Aβ increases intracellular calcium levels; the hypothesis states that this loss of calcium homeostasis causes two distinct effects: 1) the hyperphosporylation of tau, resulting in neurofibrillary tangles and 2) neurodegeneration (Hardy & Higgins, 1992). A later revision have incorporated synaptic deficits into this cascade (Hardy & Selkoe, 2002), proposing that tau hyperphosphorylaiton is an effect secondary to synaptic deficits and cellular injuries (Figure 2C). In our recent study, Aβ-induced synaptic deficits were blocked by the expression of AP (a phosphorylation blocking mutation) tau (Miller et al., 2014), clearly placing tau phosphorylation before synaptic dysfunction. Therefore, we propose a revision to the amyloid cascade hypothesis in Figure 2D, framing tau hyperphosphorylation as a cellular event between Aβ triggering and final synaptic deficits.

Figure 2. Comparison Of The Amyloid Cascade Hypotheses.

A and B. The initial amyloid hypothesis introduces the central idea that amyloid pathology precedes tau pathology (Hardy and Allsop 1991; Hardy and Higgins 1992).

C. A more recent version has incorporated synaptic deficits into the cascade, putting synaptic deficits before tau hyperphosphorylation.

D. In our model, tau hyperphosphorylation and tau mislocalization precedes synaptic deficits at a very early stage. However, tau mislocalization may play a permissive role at a later stage (see Figure 3 and Ittner et al., 2010). Tauopathy increases in the amount of extrasynaptic NDMA receptors (Hoover et al., 2010), which may lead to neurodegeneration. Although we only see postsynaptic effects in our experimental system, pre-synaptic deficits may also participate in AD pathogenesis (Hermann et al., 2013; Huang et al. 2013; Russell et al., 2012).

Clinicians seeking to develop treatments for AD have focused on Aβ interventions due to years of research indicating that Aβ mediates AD pathogenesis (Hardy & Selkoe, 2002; Tanzi & Bertram, 2005). Unfortunately, the results of drug trials that target Aβ have not yielded substantive results (Gilman et al., 2005; Green et al., 2009). This fact has led researchers to broaden the range of proteins targeted, one such protein is tau (Golde et al., 2010). A growing body of literature now implicates tau as both necessary and sufficient to drive cognitive impairment in AD mouse models and ostensibly AD in humans. When a P301L mutant tau transgene is turned off for a period of time mice recover from the cognitive deficits found while the transgene is active, indicating the necessity of the mutant tau for deficits (Santacruz et al., 2005). The pathology found in transgenic tau mice is exacerbated by increased Aβ levels (Götz et al., 2001; Lewis et al., 2001; Bolmont et al., 2007). In the 3xTg-AD mouse expressing the APP_Swe, tau_P301L, and PS1_M146V transgenes, Aβ pathology is detected before tau pathology suggesting the possibility of a causal relationship between the two protain (Oddo et al., 2003; Santacruz et al., 2005). In this mouse severe cognitive deficit are not observed until tau pathology is detected (Oddo et al., 2004; Billings et al., 2005). Interestingly, cognitive deficits in 3xTg-AD are only mitigated when both Aβ and tau, not Aβ alone, are reduced via immunotherapy (Oddo et al., 2006). Furthermore, reduction of tau in an AD mouse expressing human APP prevented cognitive deficits without affecting elevated Aβ plaque levels (Roberson et al., 2007). Still more evidence of the mediatory role of tau has also been found in humans: the presence of neurofibrillary tangles is a strong indicator of the whether Aβ pathology is correlated to cognitive deficits (Bennett et al., 2004).

Tau protein binds to microtubules and plays a role in cytoskeletal development and regulation in neurons; the protein's phosphorylation state is controlled by kinases and phosphates such as GSK3β and PP1 respectively (Buée et al., 2000; Avila et al., 2004). Pathology outside of neuritic plaques and neurofibrillary tangles, e.g. Lewy bodies and vascular infractions, has been found in many patients with clinical AD diagnoses – it is estimated that between 1/3–1/2 of AD patients have a mixed pathology (Kovacs et al., 2008). Furthermore, patients with mixed pathology have a much higher chance of experiencing dementia than those who exhibit pathology consistent with only one disease (Langa et al., 2004; Fotuhi et al., 2009; Schneider, 2009). Interestingly, alterations in tau are found in a number of other neurodegenerative diseases. Tau inclusions and hyperphosphorylation are also found in frontotemporal dementia, corticobasal degeneration, Pick's disease, and other dementias (Buée et al., 2000; Avila et al., 2004). Tau hyperphosphorylation has also been implicated in Parkinson's disease; many Parkinson's patients experience dementia in addition to motor disturbances (Ishizawa et al., 2003; Aarsland et al., 2008; Reichmann, et al., 2009; Svenningsson et al., 2012). The ubiquitous role of tau in multiple neurodegenerative disorders clearly supports the central role of tau in the cellular cascades that lead to dementia.

The above multiple lines of research all emphasize the role of tau as a downstream mediator of Aβ-induced dementia, supporting the cornerstone of amyloid cascade: the Aβ-tau-neuronal damage pathway (see reviews by Reitz, 2012; Zahs & Ashe, 2013). However, consistent with mainstream thought at that time, the initial form of amyloid cascade argued for the central roles of insoluble deposits of Aβand tau (Hardy & Allsop, 1991). Although it is still controversial which species of Aβ and tau are the main culprits for dementia in AD (see reviews by Karran et al., 2011; Wischik et al., 2013; Zahs & Ashe, 2013), as discussed earlier recent studies favor the roles of soluble species of Aβ and tau over its insoluble deposits in AD progression. Curiously, interstitial tau proteins have also been shown to affect the Aβ-tau cascade although the cellular mechanism remains to be determined (Yamada et al., 2011). Therefore, the first and second forms of amyloid cascade hypothesis may need some modifications even though the central premise that Aβ initiates the cellular events leading to dementia remains valid.

As synaptic deficits attracted greater attention in the field of AD research, a revised version of amyloid cascade was proposed (Hardy & Selkoe, 2002). The updated hypothetical cascade placed progressive synaptic and neuritic injuries before tau pathologies (Figure 2C). However, due to our evolving understanding of the roles that tau plays in AD pathogenesis, the temporal sequence of Aβ initiation, synaptic deficits, tau hyperphosphorylation, tangle formation and neuronal death is still being debated (Morris et al., 2011). Our recent studies indicate that tau mediates early synaptic deficits before overt neurodegeneration and formation of tau tangles (Hoover et al., 2010). Furthermore, Aβ oligomer-induced synaptic deficits are blocked in cultured neurons expressing AP (phosphorylation blocking) tau; this tau construct may act as a dominant negative modulator, blocking the mislocalization of endogenous tau (Miller et al., 2014). In both of our studies, AMPA receptor-mediated excitatory synaptic responses were impaired when the density of dendritic spines had not yet been altered and no significant neuronal loss had yet occurred. Therefore, at least some elements of synaptic deficits such as loss of postsynaptic AMPA receptors are mediated by tau hyperphosphorylation at a very early stage of the disease, putting tau before synaptic deficits in our revised cascade (Figure 2D)

Calcium and tau serve as the central pieces of multiple signaling loops in the entangled Aβ-tau cascade

One of the biggest challenges to the amyloid cascade hypothesis comes from studies using AD mouse models. Many transgenic mice that specifically overproduce Aβ proteins do not demonstrate robust tau pathologies such as tau tangles or severe neuronal loss (Games et al., 1995; Hsiao et al., 1996; Sturchler-Pierrat et al., 1997; Holcomb et al., 1998; Mucke et al., 2000; Jankowsky et al., 2001). Additional tau, presenilin or other human transgenes are needed to produce the full spectrum of the AD hallmarks including amyloid plaques, neurofibrillary tangles, chronic inflammation and neuronal loss in transgenic animals (Oddo et al.,2003; Padmanabhan et al., 2006; Colton et al.,2008; Wilcock et al., 2008; Cohen et al., 2013). Another big challenge for the amyloid cascade hypothesis comes from recent clinical trials (see review by Yoshiyama et al., 2013). Aβ removal, reduction ofβ production and active immunization with Aβhave shown limited efficacy and do not prevent tau pathology or neurodegeneration (Holmes et al., 2008; Green et al., 2009).

To address these challenges, many competing hypotheses have been proposed, including the inflammation hypothesis (see review by McGeer & McGeer 2013), mitochondrial cascade hypothesis (see review by Swerdlow et al., 2010),capillary dysfunction hypothesis (see review by Østergaard et al., 2013) and cholesterol hypothesis (see review by Castello & Soriano 2013). However, the competing hypotheses often propose distinct parallel signaling pathways that directly or indirectly interact with the core Aβ-tau signaling pathway. Therefore, the central premise of classical amyloid cascade remains valid. To the present, there are at least four parallel signaling pathways that affect intracellular calcium and tau, which subsequently lead to neuronal death, loss of dendritic spines and loss of AMPA receptors (Figure 3). If inflammation, mitochondrial dysfunction, vascular impairment or abnormal cholesterol affects any one of these parallel pathways, they will induce or facilitate the pathological progression of AD.

Figure 3. Multiple parallel signaling pathways for Aβ- and tau-mediated neruodegeneration and synaptic deficits.

Red arrows denote the classical Aβ-tau cascade (Pathway 1). GSK3β affects both tau phosphorylation and Aβ production. Blue arrows denote addition roles of presenillin (Pathway 2). Purple arrows denote the recently characterized Aβ-PrP^C-Fyn pathway (Pathway 3). Green and orange arrows denote the dual roles of tau. It may bring Fyn to dendritic spines to allow Aβ-induced neurotoxicity (Pathway 4) or directly impair synaptic function (Pathway 5). Other cellular events such as ER stress, AMPK activation and spastin activation may also be involved. Aβ may also affect neuronal circuits by modulating inhibitory synaptic transmission (Verret et al., 2012).

In the classical amyloid cascade pathway (Figure 3 Pathway 1, red arrows; see reviews by Karran et al., 2011 and Yoshiyama et al., 2013), toxic Aβ oligomers increase the phosphorylation of tau by activating protein kinases GSK3β, CDK5 and ERK; the phosphorylated tau or other toxic tau species may activate multiple cellular events, which eventually lead to synaptic dysfunction and neuronal degeneration (see review by Morris et al., 2011). In addition to increasing the production of Aβ by modulating r-secretase activity, changes in presenilin may also cause lysosome acidification, ER stress and intracellular calcium dysregulation, which subsequently lead to neuronal death, presynaptic dysfunction and deficits in synaptic plasticity (Figure 3 Pathway 2, blue arrows; Ho & Shen, 2011; Shen, 2013; McBrayer & Nixon, 2013; Endres & Reinhardt, 2013; Fonseca et al., 2013). Many recent studies have emphasized the manifold roles of ER stress in multiple neurodegenerative diseases (see review by Endres & Reinhardt, 2013). In a landmark paper, cellular prion protein (PrP^C) has been identified as a receptor for Aβ-oligomers and is required for Aβ-induced inhibiton of long-term potentiation (Laurén et al., 2009). Subsequent studies reveal that mGluR5 bridges the association between PrP^C and Fyn, allowing the Aβ- PrP^C complex to activate Fyn and NMDA receptors, increasing intracellular calcium and inducing synaptic deficits (Figure 3 Pathway 3, purple arrows; Um et al., 2013; Um & Strittmatter, 2013 Larson et al., 2012). Changes in calcium can cause tau hyperphosphorylation by activating calpain and CDK5 (Shukla et al., 2012; Nikkel et al., 2012). By playing a retrograde permissive role, tau hyperphosphorylation in turn allows the Aβ-PrP^C-Fyn complex to impair postsynaptic function by bringing Fyn to dendritic spines during tau mislocalization; the activated Fyn subsequently increases NMDA receptor activity and enhances Aβ-mediated neurotoxicity (Figure 3 Pathway 4, dark green arrows; Ittner et al., 2010). Both Pathways 3 and 4 in Figure 3 lead to overt loss of dendritic spines, suggesting that neurotoxicity is already underway at this point. However, our studies indicate that tau mislocalization may induce calcineurin-dependent loss of AMPA receptors before spine loss at a very early stage of the disease through a different parallel mechanism (Figure 3 Pathway 5, orange arrows).

Tau mislocalization to dendritic spines

Some researchers have proposed that the tau-mediated synaptic deficits are presynaptic as tau is typically localized to the axon. Deficits in axonal transport caused by the treatment of cultured neurons with soluble Aβ oligomers are dependent on the presence of tau (Vossel et al., 2010); however, even a three-fold increase in cellular tau is not sufficient to cause decreases in axonal transport rate implying that tau works in concert with other factor to cause presynaptic deficits (Yuan et al., 2013). A recent line of evidence suggests that tauopathies may originate postsynaptically. Tau is normally found in a gradient along the length of neurons with the highest concentrations of tau found in the axonal compartment of healthy neurons, but mutations in tau or overexpression of human tau cause the gradient to reverse and result in the misclocalization of tau to the somatodendritic compartment (Papasozomenos & Binder, 1987; Götz et al.,1995; Spittaels et al., 1999; Kins et al., 2001; Brandt et al., 2005; Gendron & Petrucelli, 2009). Many models of neurodegeneration demonstrate this pathogenic reversal of the tau protein gradient. A truncation of tau that prevents dendritic tau localization was found to prevent cognitive deficits in an AD transgenic mice without affecting increased Aβ levels (Ittner et al., 2010). Further research has demonstrated that soluble Aβ oligomers cause mislocalization of tau to dendrites and that only neurons demonstrating mislocalization display decreased spine density (Zempel et al., 2010).

As a further step, our group has shown that the P301L model of tauopathy demonstrates tau mislocalization not just to the dendritic compartment, but to dendritic spines as well and that this mislocalization is tau-phosphoylation dependent (Hoover et al., 2010). In our accompanying publication, we extend these findings to a model of AD, showing that Aβ oligomers cause a similar mislocalization of tau to dendritic spines (Figure 4) (Miller et al., 2014). By blockading tau phosphorylation we were able to inhibit the mislocalization of tau to dendritic spines and restored Aβ oligomer induced decreases in AMPA glutamate receptor signaling (Miller et al., 2014). This new study provides new evidence of tau as a mediator between Aβ initiation and final synaptic dysfunction.

Figure 4. Changes in either Aβ or tau can lead to tau mislocalization to dendritic spines.

A. Cultured rat dissociated hippocampal neurons were co-tranfected with both DsRed to label dendritic spines and GFP-tagged tau proteins to detect tau mislocalization. Left and middle: overlay images from neurons expressing DsRed and GFP-tagged wild-type tau with or without exposure to Aβ oligomers for 3 days. Right: an untreated neuron co-expressing DsRed and P301L mutant tau proteins.

B. Either Aβ treatment or P301L tau expression significantly increased the abnormal presence of tau proteins in dendritic spines.

C. Neurons were cultured from transgenic negative mice (TgNg), transgenic mice expressing the APP_Swe mutation or transgenic mice expressing P301L tau. The neurons were co-tranfected with DsRed and GFP-tagged wild-type tau.

D. Only the expression of APP_Swe significantly causes tau mislocalization whereas the expression of P301L tau does not drive wild-type tau proteins to dendritic spines. It suggests that Aβ initiates the cascade whereas P301L mutation makes the mutant tau proteins more vulnerable to mislocalization than wild-type tau proteins. This result argues that Aβ is the initiator of a process mediated by tau. ***, p < 0.01, ANOVA.

Proposed upstream mechanisms of tau mislocalization: GSK3β and other kinases

Dendritic spines are highly dynamic structures that process the majority of excitatory neurotransmission in the brain (Cingolani & Goda, 2008; Patterson & Yasuda, 2011). Various forms of plasticity, including the most studied forms of LTP result from signaling that occurs in spines (Malenka & Bear, 2004; Matsuzaki et al., 2004; Lee et al., 2009). The signaling that occurs in spines is highly compartmentalized due to long, thin necks that connect the bulbous head of the spine to the dendrite and the variety of protein to protein interactions that occur in spines (Byrne et al., 2011). If tau can become mislocalized to the spine it can interact with the myriad of signaling pathways that control functional and structural plasticity. We will first discuss possible upstream mechanisms that could cause tau to enter spines (Figure 5). Phosphorylation has been shown to be necessary for the entry of tau into dendritic spines and for synaptic deficits; conversely, mutation of those same 14 sites to mimic phosphorylation causes mislocalization to spines and is sufficient to cause synaptic deficits (Hoover et al., 2010). Recent findings indicate that those phosphorylation sites are also necessary for soluble Aβ oligomers to induce tau mislocalization to dendritic spines, impairing synaptic function (Miller et al., 2014). Therefore, any cellular events that lead to tau hyperphosphorylation may cause tau mislocalization to dendritic spines, impairing synaptic function.

Figure 5. Hypothetical mechanisms for Aβ initiation of tau-mediated synaptic deficits.

A. possible initiators of dementia: A1. mutation of tau residue 301 from proline to leucine; A2. Swedish double mutation of APP at residues 670, lysine to asparagine, and 671, methionine to leucine; A3. extracellular soluble Aβ oligomers such as dimers, trimers and higher order species such as Aβ*56 (the mechanism of creation and the crystalline structures of these pathological species remains unknown) may initiate the cascade from the axonal or dendritic compartments; and A4. intracellular Aβ.

B. Intracellular Aβ may cause presynaptic dysfunction. However, it is still controversial whether presynaptic release is increased or decreased in AD (Busche et al., 2008; Huang et al., 2013).C. initiators A1–A3 result in tau hyperphosphorylation and mislocalization to dendritic spines (postsynaptic) possibly mediated by GSK3β activity.

D. Once tau is mislocalized to dendritic spines D1. calcineurin activity is increased; D2. calcineurin dephosphorylates GluR1 residue S845; D3. causing AMPA glutamate receptor internalization resulting in a decrease in excitatory glutamatergic neurotransmission.

E. Interstitial soluble proteins of Tau (ISP-tau) might also be involved (Yamada et al., 2011).

Hyperphosphorylation of tau has been found in tissue from humans with AD and other diseases that demonstrate tau pathology (Grundke-Iqbal et al., 1986; Buée et al., 2000; Avila et al., 2004; Wang et al., 2013). The presence of hyperphosphorylated tau in human CSF has been shown to predict future AD in patients with mild cognitive impairment (Buerger et al., 2005; Hampel et al., 2005; Hansson et al., 2006).These phenomena have been recapitulated in mouse models of AD and tauopathy, often at time points prior to the presence of neurofibrillary tangles (Mucke et al., 2000; Santacruz et al., 2005; Boekhoorn et al., 2006; Schindowski et al., 2006; Roberson et al., 2007; De Felice et al., 2008). Recent eveidence has shown that tau hyperphosphorylation is prevalent in the PS1Δ9/APP_Swe mouse model of AD (Cohen et al., 2013). Using phosphatases to decrease tau phosphorylation has been shown to recover its physiological function (Wang et al., 1996). Tau phosphorylation has been correlated with detrimental effects at synapses in both AD and tauopathy transgenic mice (Steinhilb et al., 2007; Bittner et al., 2010; Hoover et al., 2010; Zempel et al., 2010).

Glycogen synthase kinase-3 (GSK3; especially the isoform GSK3β) and cyclin dependent kinase 5 (CDK5) are the two most widely-studied protein kinases that phosphorylate tau (reducing its binding affinity to microtubules) although other kinases can phosphorylate tau including PKA, CaMKII, CK1 and ERK (Lovestone et al., 1994; Wagner et al., 1996; Wang et al., 2013; Yoshiyama et al., 2013). Importantly, activation of GSK3β also increases the production of Aβ peptides (Ly et al., 2013; Phiel et al., 2003). The dual roles of GSK3β make its inhibitors strong drug candidates for AD therapy (Yoshiyama et al., 2013; Figure 3, red arrows). There is compelling evidence that GSK3 is highly involved in the abnormal phosphorylation of tau in diseased states (Bhat et al., 2011). GSK3 activity is stimulated by exposure of cultured cells to Aβ and tau has been shown to be hyperphosphorylated via this pathway (Takashima et al., 1996; Wang et al.,1998; Amadoro et al., 2011). Indeed, pharmacological inhibition of GSK3 rescues cells from death in AD mice, and Aβ induced reelin deregulation causes an increase of GSK3 dependent tau phosphorylation (Cuchillo-Ibáñez et al.,2013 Noh et al., 2013). Inhibition of GSK3 blocks the induction of LTD but not LTP or depotentiation although it remains unclear whether tau phosphorylation is involved (Fitzjohn et al., 2008). Proteins involved in other neurodegenerative diseases, including α-synuclein may play a role in promoting the phosphorylation of tau by GSK3, suggesting that this effect may not be specific to AD (Haggerty et al., 2011; Kawakami et al.,2011; Irwin et al., 2013). The upstream causative factors of GSK3β-mediated tau phosphorylation remain elusive. Receptors for advanced glycation endproducts (RAGEs) have been proposed to activate GSK3β and aggravate Aβ-induced cognitive deficits (Li et al., 2012; Chen et al., 2013; Figure 3 Pathway 1). These findings implicate GSK3 as a likely candidate to carry out the hyperphosphorylation of tau found in AD and other neurodegenerative diseases, supporting our hypothetical model in Figure 5.

Several lines of evidence suggest that either Aβ or tau pathology may result from the misfolding of one or both of the proteins in a prion-like manner (Nussbaum et al., 2013; Zahs & Ashe, 2013). Some researchers have posited that the aggregation of Aβ that is the initial pathogenic step in the amyloid cascade hypothesis is due to prion-like misfolding (Eisele, 2013). Our research does not address whether Aβ aggregates are formed though seeding by other misfolded Aβ peptides, we do have evidence that this effect does not apply to tau phosphorylation induced mislocalization. Unpublished data from our lab indicates that seeding is not involved in tau mislocalization to dendritic spines. We cultured hippocampal neurons from tau_P301L mice and then transfected the neurons with wild-type tau tagged with GFP. Wild-type tau did not mislocalize to dendritic spines, even though the neurons also expressed P301L tau (Figure 4C–D).

Several steps in the pathway between Aβ oligomers and tau-mislocalization are currently unknown. It is well established that there is a mismatch between tau and amyloid beta pathology in patients with AD (Delacourte et al.,2002; van Helmond et al., 2010;). One possible mechanism by which this mismatch may arise is through a signaling pathway whereby Aβ initiates dysfunction through cell surface receptors on the axons of neurons, resulting in a dendritic (postsynaptic) effect by tau (Figure 5A). Aβ probable receptor for Aβ oligomers is PrP^C (Larson & Lesne, 2012; Figure 3 Pathway 3, purple arrows). Furthermore, the precise phosphorylation sites necessary for tau mislocalization and tau-mediated synaptic dysfunction have not been worked out in detail.

Proposed downstream mechanisms: dual roles of tau mislocalization

Our recent study provides a direct link from tau hyperphosphorylation and mislocalization to Aβ induced synaptic dysfunction (Miller et al., 2014). However, the signaling events occurring between tau mislocalization and synaptic dysfunction remain unknown (Figure 5D). One candidate is calcineurin (aka PP2B), a phosphatase that has been implicated in long-term depression (LTD), which is a weakening of synapses. Calcineurin plays an important role in AMPA glutamate receptor dynamics and spine loss via dephosphorylation of AMPA glutamate receptors (Mulkey et al.,1994; Dell'Acqua et al., 2006; Miller et al., 2012). The similarities between LTD and synaptic deficits associated with AD are notable (Hsieh et al., 2006 Sheng et al., 2012). A correlation between changes in calcineurin activity, Aβ levels, and cognitive decline in humans was recently discovered (Abdul et al., 2009). Inhibition of calcineurin has been found to ameliorate cognitive deficits and/or neuropathology in AD transgenic mice (Dineley et al., 2007; Taglialatela et al., 2009). Calcineurin has also been implicated in loss of synaptic glutamate receptors in an AD transgenic mice (Wu et al., 2010) and a number of studies using soluble Aβ oligomers to induce synaptic deficits (Chen et al., 2002; Snyder et al., 2005; Hsieh et al., 2006; Shankar et al., 2007). Our recent study has shown a correlation between the mislocalization of tau to dendritic spines and the necessity of calcineurin for soluble Aβ oligomer-induced decreases in AMPA receptor signaling (Miller et al., 2014). A large gap in our understanding of calcineurin's role in AD is the process of its activation. Calcineurin is activated by small increases in the concentration of Ca²⁺. In the case of LTD, low frequency stimulation is sufficient to cause activation of calcineurin and deficits (Dell'Acqua et al., 2006). The role of calcium in AD has been well studied over the years (Green, 2009). Increases of Ca²⁺ and abnormal Ca²⁺ homeostasis due to a disruption of spine calcium compartmentalization in the diseased state have been reported (Kuchibhotla et al., 2008; Wu et al., 2010; Zempel et al., 2010). The NMDA glutamate receptor, a prominent site of Ca²⁺ entry at the synapse, has been implicated in synaptic deficits caused by soluble Aβ oligomers (Kamenetz et al., 2003; Shankar et al., 2007; Decker et al., 2010; Wu et al., 2010; Li et al., 2011; Kessels et al., 2013). Research indicates that loss of NMDA glutamate receptors can be initiated by increased Aβ levels (Kamenetz et al., 2003; Snyder et al., 2005; Hsieh et al., 2006; Shankar et al., 2007) and tau mislocalization due to tau mutation (Hoover et al., 2010). Although these results implicate NMDA glutamate receptors in AD, a recent article has suggested that their Ca²⁺-related activity is not necessary for synaptic deficits (Kessels et al, 2013). Calcium dysregulation may also lead to ER and mitochondrial stress, which may initiate various cell death pathways (Fonseca et al., 2013). Other studies hold that increases in intracellular calcium may lead to increases in the presence of Aβ species due to interaction with APP cleavage, resulting in positive feedback as Aβ causes more calcium dysregulation (Zeiger et al., 2013).

A variety of other proteins and signaling pathways have been implicated in the synaptic deficits caused by Aβ. Caspase-3 plays an important role in apoptosis pathways and is elevated in humans with AD (Gervais et al., 1999; Stadelmann et al., 1999). Originally researchers thought that its activity was simply related to cell death found in AD, but recent research has associated caspase-3 with synaptic plasticity in AD. Caspace-3 activity is increased in the dendritic spines of Aβ transgenic mice and is associated with causing calcineurin-dependent synaptic deficits independently of cell death (D'Amelio et al., 2011). Genetic knockout or inhibition of caspase-3 rescues cognitive and synaptic deficits in an AD model (D'Amelio et al., 2011) and LTP deficits caused by soluble Aβ oligomers (Jo et al., 2011). This cell death pathway may be mediated by calcium dysregulation due to ER stress resulting in the activation mitochondrial cell death pathways (Endres & Reinhardt, 2013; Fonseca et al., 2013).

Fyn, AMPK and Spastin have been implicated in synaptic deficits found in AD (Ittner et al., 2010; Mairet-Coello et al., 2013; Zempel et al., 2013; Laurén, 2014). In contrast to our studies of tau mutations and Aβ oligomers, the activation of the aforementioned pathways causes significant loss of dendritic spines. Furthermore, in the study by Ittner et al., PSD95-binding NMDA receptors are increased whereas, in our experimental system, synaptic clustering of NMDA receptors is reduced (Hoover et al., 2010). As illustrated in Figure 3, tau may play dual roles in progressive impairment of synaptic function throughout the long process of AD development. Initially, tau acts in a provocative role, at a very early disease stage, impairing synaptic function directly by causing the loss of glutamate receptors (Hoover et al., 2010). Later in the disease progression, tau takes a permissive role: the abnormal trafficking of tau to dendritic spines may bring Fyn to the spines allowing Aβ-PrP^C-mediated neurotoxicity to occur (Ittner et al., 2010).

Conclusion

Our understanding of Alzheimer's disease has advanced greatly over the last two decades as the research focus has been shifted from descriptive pathological investigation to functional and mechanistic studies. Recent studies have emphasized the importance of detecting early functional deficits in AD, the roles of soluble Aβ oligomers and cellular mechanisms underlying synaptic deficits. Through coordinated efforts by scientists in diverse fields, a consensus has begun to emerge in the support of the concept that Aβ initiates the onset of AD while tau mediates the subsequent functional deficits at an early stage of the disease. The phosphorylation-dependent mislocalization of tau to the dendritic compartment and to dendritic spines of neurons may play a dual role. At a very early stage, it may directly suppress postsynaptic transmission by inducing calcineurin-mediated AMPA receptor internalization. At a later stage, the missorted tau may bring Fyn to spines, enhancing NMDA receptor activity and allowing Aβ-PrP^C-mediated neurotoxicity to occur. Interventions focused on abrogating tau mislocalization and hyperphosphorylation in AD may be a more effective strategy for the treatment of AD than present strategies targeting Aβ oligomers as tau-centered treatments may inhibit multiple parallel pathways that lead to neuronal loss and synaptic deficits long after Aβ initiation of the toxic cascade.

Abbreviations

AD

Alzheimer's disease

APP

amyloid precursor protein

Aβ

amyloid beta

GSK3β

glycogen synthase kinase 3 beta

LTD

long-term depression

LTP

long-term potentiation

PS1

presenilin-1

TgNg

transgenic negative