Amyloid-β and Tau in Alzheimer’s Disease: Novel Pathomechanisms and Non-Pharmacological Treatment Strategies

INTRODUCTION

In this article, we are focusing on the last five years of work originating from our laboratory. Our research is based on the assumption, that amyloid-β (Aβ) and tau, the major proteinaceous constituents of the two hallmark lesions of Alzheimer’s disease (AD), the amyloid plaques and the neurofibrillary tangles, not only constitute biomarkers, but in fact initiate and drive the disease process, presenting these molecules as appropriate targets for therapeutic intervention. We discuss our recent mechanistic work that contributes to a better understanding of how Aβ ‘talks’ to tau and how tau causes neurodegeneration. This includes a new mechanism of how neuronal tau accumulates in the somatodendritic domain as the disease unfolds. For much of our work we rely on transgenic mouse models with either Aβ or tau deposition. We also used these mice to develop a novel therapeutic approach termed ‘scanning ultrasound’ (SUS) that builds on ultrasound research going back several decades, and demonstrated that both Aβ and tau aggregates can be effectively cleared with ultrasound, restoring neuronal functions. Regarding tau, we combined SUS with a therapeutic anti-tau antibody fragment to achieve synergistic therapeutic effects. We will put our work into perspective by comparing it to ongoing treatment strategies, in particular vaccinations. Finally, we will discuss how we believe the new pathomechanistic insight will influence therapeutic strategies. We anticipate that combinatorial approaches will lead to better therapeutic outcomes. In this context, ‘combinatorial’ not only refers to the simultaneous targeting of two pathologies (such as Aβ and tau), but also to the combination of a drug with a non-pharmacological procedure.

PATHOMECHANISMS OF TAU

Tau belongs to a family of microtubule-associated proteins that also includes MAP2 and MAP4. These proteins share repeat motifs with which they bind to microtubules. AD brains are histopathologically defined by extracellular amyloid plaques containing the peptide Aβ and intracellular neurofibrillary tangles containing the microtubule-associated protein tau. Aβ is derived by proteolytic cleavage from the larger amyloid-β protein precursor, AβPP. In AD, Aβ is thought to accumulate both because of an increased production and an impaired clearance [1]. The process toward plaque formation involves oligomerization and fibrillization of Aβ. A similar process is known to occur for tau that becomes hyperphosphorylated (p-tau) before forming fibrillar aggregates [2]. Tau pathology in the absence of Aβ deposition is prevalent in several other diseases that are collectively termed tauopathies and includes frontotemporal lobar degeneration (FTLD).

A crucial role for p-tau in the neurotoxicity and degeneration observed in AD and related tauopathies has been demonstrated by us and others, in part by using transgenic mouse models that, in an age-dependent manner, recapitulate major aspects of the human pathology [3]. A role for aging was also demonstrated when pR5 mice expressing P301L mutant human tau found in familial FTLD were back-crossed onto a senescence-accelerated SAMP8 background. We found that this exacerbated the pre-existing pathology that characterizes the tau transgenic mice, presenting this novel strain as a tool to screen for disease-modifying factors [4]. Within the limitation of a mouse model, strains such as Tau58-2/B that express the P301S mutation of tau also recapitulate neurological deficits of distinct tauopathies, such as the behavioral variant of frontotemporal dementia. By assessing Tau58-2/B mice in a comprehensive behavioral test battery, we found that the tauopathy mice showed age-dependent signs of impulsivity and decreased social exploration and executive dysfunction. The deficit in executive function was first limited to decreased spatial working memory, but with aging this was extended to impaired instrumental short-term memory, presenting the mice as a suitable model to test therapeutic interventions for the amelioration of this tauopathy variant [5].

An interesting observation can be made when one addresses the fate of distinct p-tau epitopes in mice. We had previously found in the pR5 mice, that whereas pathological p-tau epitopes such as AT180 (T231) or AT270 (T175/T181) become increasingly phosphorylated in vulnerable areas such as the hippocampus and the amygdala as the disease progresses, the AT8 epitope (S202/T205) goes through a biphasic stage: For example, in the CA1 region, at 3 months, AT8 staining is faint. Once the mice reach 6 months, AT8 staining intensity increases. At 20 months, a remarkable change in the AT8 pattern becomes evident, as staining is now being confined to just a few neurons with rich arborization, and this staining is very intense [6]. Taking this further, in collaborative work, we found that in this second phase, when these neurons undergo intense, fibrillar changes, phosphorylation of additional pathological serine/threonine epitopes such as AT100 (T212/S214) was also massively increased [7]. This was associated with an activation of the tyrosine kinase Pyk2 (also known as Ptk2b) (that has since been identified as an AD risk gene [8]) as well as its putative substrate GSK3β via tyrosine phosphorylation, which may explain the massive pathological phosphorylation of tau in this second stage [7].

Tau transgenic mice were useful in identifying pathomechanisms that affect a range of cellular functions which is not surprising considering that tau is a scaffolding protein interacting with many proteins in an isoform-dependent manner [9]. Multiple aspects of mitochondrial function are impaired by pathological tau [10], as we already showed in 2005 for the oxidative phosphorylation system [11]. More recent collaborative work revealed a major role for tau in impairing mitochondrial fission by preventing the efficient recruitment of Drp1 onto mitochondria, leading to elongated mitochondria [12, 13]. The pR5 mice were also useful for identifying another pathomechanism that involves nuclear depletion and cytoplasmic accumulation of the nuclear factor SFPQ (also known as PSF) [14]. Interestingly, loss of SFPQ function has since been shown to alter the tau isoform ratio and cause an FTLD-like phenotype [15].

Tau in also known to impair synaptic activity as has been shown in numerous studies; however, tau’s impact on neuronal excitability has received remarkably little attention, although it has been reported that the removal of tau reduces network hyperexcitability [16]. In very recent work, we have shown that p-tau induces a more depolarized threshold for action potential initiation and reduces firing in hippocampal CA1 neurons of tau transgenic mice, which was rescued by the suppression of transgenic tau. Furthermore, in primary hippocampal neuronal cultures, we revealed that this reduction in neuronal excitability resulted from the relocation of the axon initial segment (AIS) down the axon in a tau phosphorylation-dependent manner. This suggests that a reduction in hippocampal excitability due to a tau-mediated distal re-localization of the AIS contributes to the hippocampal dysfunction observed in tauopathies [17].

An interesting new thread has been added to the field with the notion that tau pathology propagates extracellularly [18, 19]. This notion has its foundation in the distribution of neurofibrillary tangles that follow a distinct pattern through anatomically connected brain regions and the well documented correlation between the severity of tau pathology and the disease progression implies a ‘prion-like’ seeding and spreading mechanism of p-tau [20]. One mechanism by which p-tau can spread is through being packaged into extracellular vesicles (EVs), membranous vesicles 30–1,000 nm in diameter. We have demonstrated in vitro that p-tau is contained within EVs enriched for exosomes isolated from either wild-type mice or rTg4510 mice with a pronounced tau pathology and have shown that the tau within EVs is able to seed the aggregation of endogenous tau in recipient cells in a threshold-dependent manner [21]. Furthermore, we have shown in vivo that transgenic EVs cause increased tau phosphorylation and soluble oligomer formation in a manner comparable to that of freely available proteins in brain lysates in human tau transgenic ALZ17 mice [22]. Another approach to address tau spreading pursued by us was by assessing the spreading of endogenous phosphorylated tau. To generate endogenous seeds, we injected the protein phosphatase 2A (PP2A) inhibitor okadaic acid (OA) unilaterally into the amygdala of wild-type mice and found that this insult rapidly initiated changes in tau phosphorylation, solubility, and aggregation at anatomically distant sites. More specifically, we detected protein aggregation via thioflavin-S at the injection site and in the cortex of both injected and contralateral hemispheres, which was not induced in tau knock-out mice. Together, this suggested to us that tau phosphorylation can be both a primary response to an insult, and a secondary response communicated to non-exposed brains regions [23]. Taken together this and the work of others demonstrates that extracellular vesicles can transmit tau pathology, indicating a role for extracellular vesicles in the transmission and spreading of tau pathology.

HOW Aβ DRIVES TAU PATHOLOGY: A CENTRAL ROLE FOR THE KINASE FYN

Synaptic degeneration precedes neuronal loss in AD, and not surprisingly, AD has been termed a synaptic failure [24]. Furthermore, Aβ is believed to drive tau pathology which presents the challenging question as to how these molecules actually interact, considering that Aβ is released into the extracellular space, whereas tau that early in development is distributed throughout the neuron, becomes enriched in the axon with neuronal maturation [25]. In addressing this question, we have shown previously that tau is also found in dendrites, albeit at lower levels as in the axon, where it is required to target the kinase Fyn to the spines, mediating Aβ toxicity [26]. More specifically, we found that Fyn phosphorylates the NMDAR subunit NR2b which facilitates the recruitment of PSD95 to form an excitotoxic complex through which Aβ exerts its toxicity. Others have also contributed to this concept [27–29]. Interestingly, distinct forms of Aβ lead to specific phosphorylation events as shown in a collaborative effort for Aβ*56 that activates CaMLIIα which is associated with increased site-specific phosphorylation (S202, S416) and missorting of tau [30]. Localization of tau itself to dendritic spines is phosphorylation-dependent as has been shown by expressing pseudophosphorylated forms of tau [31, 32]. We have also used the genome-editing tool TALEN to generate Tau-mEOS2 knock-in mice which showed that Tau-mEOS2 followed a proximo-distal gradient in axons and a subcellular distribution similar to that of endogenous Tau in neurons obtained from wild-type mice. This was abolished, when either hWT-Tau or hP301L-Tau was overexpressed—a situation resembling that in disease where tau levels are also increased and distort the physiological distribution of tau [33].

It is generally assumed that hyperphosphorylated tau in the axon detaches from the microtubules and passes through the AIS, which serves as a diffusion barrier for physiologically phosphorylated tau, before accumulating in the cell body and dendrites, a process that is partly mediated by Aβ [34, 35]. However, again the question of compartmentalization arises and we asked ourselves whether Aβ may employ a mechanism other than relocalization of tau to account for the massive accumulation of tau in the somatodendritic compartment. Indeed, we identified an additional, and as we believe, more cogent mechanism that involves local Aβ-mediated protein translation of tau in the somatodendritic domain [36]. More specifically, we found that this activation occurred through a signaling cascade that involves Fyn, the serine/threonine-directed kinase ERK as well as the ribosomal protein S6, and the activation of this cascade is associated with an increased phosphorylation of tau at multiple residues. Together, these findings reveal de novo protein synthesis of tau in the somatodendritic compartment, mediated by Aβ, as a novel pathomechanism in AD.

Aβ CLEARANCE AND MEMORY RESTORATION: ESTABLISHING SCANNING ULTRASOUND (SUS) AS A NON-PHARMACOLOGICAL AND NON-INVASIVE THERAPEUTIC STRATEGY

The blood-brain barrier (BBB) is a selective structure that protects the brain parenchyma from circulating factors and restricts access of pathogens and immune cells to the brain [37]. However, this also means that the BBB presents a significant challenge for AD therapeutics, as the vast majority of potentially effective drugs are blocked from accessing the brain and engaging with target molecules in the brain. Repeated high doses of therapeutic molecules are currently required to achieve efficacy following systemic injection in animal studies which poses a significant challenge for translation into humans. This highlights the need for better delivery strategies to reduce both the cost and dose of treatment [38]. One potential way to achieve this goal is the use of ultrasound to transiently open the BBB to access the brain, a method that has been explored by us.

Ultrasound is a type of mechanical energy that is defined as the acoustic wave propagation in a medium at frequencies exceeding the range of human hearing, i.e., above 20 kHz. Different from visual light and other electromagnetic waves such as radio waves, microwaves, or x-rays, acoustic waves can penetrate solids and liquids and bounce back from impediments or when encountering abrupt changes. This explains their suitability for imaging light-impenetrable objects non-destructively. Because of the inherent diffraction limit of the resolution for any kind of wave [39], sound in the normal human hearing range, i.e., with a wavelength above 10 cm, can only resolve large objects. To obtain a higher resolution, a higher acoustic frequency is needed, as is the case for ultrasound that in the medical space is routinely used as an imaging modality for diagnostic applications, primarily in the fields of obstetrics and cardiology, but also for examining the abdomen and musculoskeletal system. In this situation, ultrasound waves are transmitted from the transducer into the patient and then received as echoes by the same transducer, as the wave is partially reflected at tissue interfaces. Important for what we are reviewing here, ultrasound has been explored in recent years as a treatment modality for brain diseases [40].

In order to manipulate the BBB for targeted drug or gene delivery, non-thermal ultrasound can be used to capitalize on the interaction between ultrasound and microscopic bubbles of gas (microbubbles) in tissue or fluids [41]. Microbubbles might pre-exist in tissue, but damaging acoustic pressure is required to generate the necessary cavitating microbubbles [42]. Therefore, preformed, commercially available microbubbles are being used to ensure biological effects even at low acoustic pressures [43]. These microbubbles are routinely used for contrast-enhanced ultrasound imaging. They are biologically inert and have a gas core encapsulated by a thin shell of lipid or polymer. Ultrasound causes them to cavitate, i.e., to expand and contract, resulting in vessel wall displacement [44–46], a process termed by us ‘obicodilation’. Displacement causes a transient opening of endothelial tight junctions because of the disintegration of the associated junction complexes. This transiently facilitates transport across the BBB [47].

Several studies have examined obicodilation as a means to specifically target AD pathology. In one such study, an Aβ-specific antibody was shown to reduce Aβ pathology in TgCRND8 mice, coupled with magnetic resonance imaging (MRI) monitoring using the contrast agent gadobutrol. Application of MRI-guided focused ultrasound to four locations in the right hemisphere reduced Aβ pathology relative to the corresponding areas in the untreated contralateral hemisphere [48]. Neither injection of the antibody nor sonication alone was effective. In a follow-up study, Aβ pathology was reduced even in the absence of a therapeutic antibody [49]. The effect was, however, very modest, suggesting that obicodilation is best used as a delivery tool for peripherally administered anti-Aβ antibodies [49]. Nevertheless, in the absence of an antibody, bilateral sonication of the hippocampus in TgCRND8 mice once per week for 1 month led to a 20% reduction in plaques, restored spatial working memory, and increased hippocampal neurogenesis [50].

An alternative to targeting a small, defined area with ultrasound is to move the ultrasound beam stepwise over the entire skull, thereby opening the BBB throughout the brain, an approach developed by us and termed scanning ultrasound (SUS) [51]. We applied this strategy to two large cohorts of Aβ-depositing and cognitively impaired APP23 mice [26], in the absence of any therapeutic agent. The mice were sonicated in the presence of microbubbles (i.e., obicodilated) once per week for a total of 6–9 weeks. This resulted in a two-fold reduction in plaque burden, and an up to five-fold decrease in monomeric and oligomeric Aβ species. Of note, this reduction is comparable to what is routinely achieved by Aβ-targeted vaccination trials. We also performed an extensive safety study that suggested to us that SUS is a safe method to transiently open the BBB. There were neither ‘dark’ neurons as revealed by Nissl staining, nor edemas or erythrocyte extravasation as shown by hematoxylin and eosin staining. Using the acid fuchsin stain, we found no evidence for ischemic damage. We further investigated the nuclear localization of NFkB, a marker of excessive, chronic inflammation, and the astrocytic marker GFAP and again, found no adverse effect of SUS treatment. In fact, this adds to the extensive safety literature that is already available for many species up to even macaques [40]. Safety has also been demonstrated in a recent clinical trial that used obicodilation to deliver a chemotherapeutic antibody to brain tumors [52]. Importantly, SUS treatment of APP23 mice not only reduced the Aβ pathology significantly, but also restored memory functions to wild-type levels, as shown with three complementary tests. As an underlying clearance mechanism, activation of microglia and uptake of Aβ into their lysosomes was identified, possibly mediated by blood-borne factors that entered the brain through the BBB and stimulated the dormant microglia [51].

In order to obtain additional insight into safety, we performed patch-clamp recordings from hippocampal CA1 pyramidal neurons in wild-type mice 2 and 24 hours after a single SUS treatment, and one week and 3 months after six weekly SUS treatments, including sham treatments as controls. Mice that received multiple SUS/sham treatments were, after aging for one week or 3 months following the final treatment, 6 and 9 months old, respectively, when the electrophysiological recordings and dendritic analysis were performed. In both treatment regimes, no changes in CA1 neuronal excitability were observed in SUS-treated neurons when compared to sham-treated neurons at any time-point. For the multiple treatment groups, we also determined the dendritic morphology and spine densities of the neurons from which we had recorded. The apical trees of sham-treated neurons were reduced at the 3-month time-point when compared to one week; however, surprisingly, no longitudinal change was detected in the apical dendritic trees of SUS-treated neurons. In contrast, the length and complexity of the basal dendritic trees were not affected by SUS treatment at either time-point. The apical dendritic spine densities were reduced, independent of the treatment group, at 3 months compared to one week. Collectively, these data suggest that ultrasound can be employed to prevent an age-associated loss of dendritic structure without impairing neuronal excitability [53]. What has not been determined is how SUS affects dendritic morphology in old mice and whether behavioral read-outs will be affected, and more generally, whether ultrasound could be used as a cognition enhancement tool in healthy people.

TAU CLEARANCE AND BEHAVIORAL IMPROVEMENT: ENHANCING VACCINATION STRATEGIES BY USING ULTRASOUND AS A DELIVERY TOOL

With the exhaustive evidence that now supports a critical role for pathological tau in AD and related tauopathies and considering the recent failure of many anti-Aβ therapeutics in clinical trials, therapies targeting tau have been rapidly increasing [54]. To promote the clearance of p-tau, we and others have generated antibodies specific for phosphorylated tau epitopes shown to be elevated in AD [55–59].

While passive immunization with the majority of these antibodies has demonstrated a reduction in tau pathology, these have been modest and behavioral improvements have only been achieved in some instances [58, 60]. For example, we have previously targeted the PHF1 (S396/S404) p-tau epitope, by both active and passive vaccination [61, 62]. This revealed some efficacy as determined for biochemical and histological read-outs, but no improvement in behavioral readouts. Rather than targeting pathological epitopes, reducing total tau levels therefore may be more beneficial especially as multiple studies have demonstrated that genetic ablation of endogenous tau does not cause behavioral or neuroanatomical abnormalities [26, 63]. This suggested that therapeutics designed to reduce total tau will be well tolerated. Recently, several groups have employed tau-lowering strategies at the mRNA and protein level to reduce total tau levels in the neuron, thereby reducing the progression of the disease. Antisense oligonucleotides (ASOs) were shown to successfully reduce tau expression in the PS19 tauopathy mouse model resulting in a significant decrease and even reversal of p-tau pathology as well as inhibition of hippocampal and neuronal loss, reversed tau seeding and reduced deficit in survival and nesting behavior [64]. Furthermore, this study provided additional evidence in nonhuman primates, that ASOs targeting monkey tau were highly efficacious at reducing endogenous tau mRNA and protein throughout the brain, spinal cord, and cerebrospinal fluid.

Pan-tau antibodies, on the other hand, can effectively reduce tau at the protein level. Intravenous injection with antibody 43D to the amino-terminal domain of tau (tau epitope 6–18) not only reduced tau pathology, but also Aβ pathology in the 3xTg AD mouse model, demonstrating for the first time that a tau therapeutic can also promote the clearance of Aβ. When compared to another anti-tau antibody, 77E9, specific for tau 184–195, 43D proved more effective at reducing tau pathology, rescuing cognitive deficits and ameliorating Aβ pathology [65]. This was also demonstrated with the anti-tau antibodies HJ8.5 (tau 25–30), HJ9.4 (tau 7–13), and HJ9.3 (tau 306–320) which were all demonstrated to block tau seeding in vitro, but only HJ8.5 and HJ9.4 rescued contextual fear deficits in mice [66]. Furthermore, peripheral administration of HJ8.5 to human patients with tauopathies and to human tau transgenic mice increased plasma tau levels in a dose-dependent manner [67]. Reduced tau uptake was also observed in an epitope-dependent manner with anti-tau antibodies Tau13 (N-terminal), 6C5 and HT7 (mid-domain) and Tau46 (C-terminal), whereby the N-terminal and mid-domain antibodies successfully prevented the uptake of tau species whereas the distal C-terminal specific antibody had little effect [68]. It is therefore believed that the site targeted by tau antibodies, rather than affinity, is important for reducing pathological tau in vivo and that targeting the N-terminus is expected to have the greatest effect. Taken together, the results from these recent studies present total tau as an alternative target to pathological tau for the treatment of AD and related tauopathies (see our recent review: [69]).

The mechanism by which antibodies reduce tau levels is still unclear, however, as the vast majority of anti-tau antibodies have not been detected intraneuronally, it has been suggested that they may engage extracellular tau and prevent tau seeding and spreading. To investigate whether antibody-mediated microglial activation and subsequent phagocytosis of the tau-antibody complex is required to reduce tau pathology, studies have investigated anti-tau antibodies with mutant Fc regions or antibodies which completely lack the Fc region all together. In a study which compared a full-effector version of an anti-tau antibody to an effector-less version, generated by mutating the Fc region of the antibody, it was found that the effector function is not required for efficacy in vivo and that, although full-effector function anti-tau promotes microglial uptake of extracellular tau, it also elicits microglial release of pro-inflammatory cytokines which is potentially deleterious to neurons [70]. This suggests that effector-less antibodies may not only be a more effective approach for targeting tau, but also a safer one.

An alternative approach to rendering antibodies effector-less is to remove the Fc region altogether through the generation of either fragment antigen binding (Fab) or single chain fragment variable (scFv) antibodies. This also reduces the size of the antibody, increasing tissue penetration and may therefore facilitate transfer across the BBB and neuronal membranes, allowing intraneuronal targeting of tau. As tau is predominantly localized within neurons this may achieve greater therapeutic outcomes. We recently explored the ability of an anti-tau scFv in reducing pathological tau in the pR5 tau transgenic mouse model [71]. In our choice of antibody specificity, we were guided by earlier work from our team that had indicated that the 2N isoform of tau is particularly linked to disease [9, 72]. We isolated a 2N tau specific full-length antibody with a high affinity and specificity for 2N tau and converted it into an scFv format. We designed a preclinical study with four treatment arms and firstly found, that a tau isoform-specific scFv, RN2N, which binds to amino acids 84–97 of full length tau, was capable of inhibiting p-tau formation at N-terminal epitopes, thereby reducing overall pathological tau levels and improving behavioural outcomes after passive immunization in the pR5 mice. In the study we provided evidence that the antibody fragment or scFv prevents GSK3β-mediated phosphorylation of epitopes in the N-terminal half of tau. In agreement with our study, Ising and colleagues treated P301S transgenic mice with the HJ8.5 scFv and achieved a marked decrease of p-tau accumulation in the hippocampus of the mice by preventing the seeding of extracellular tau [73]. This demonstrates that antibody fragment binding of tau is sufficient to prevent it from undergoing hyperphosphorylation, aggregation and spreading, without the additional requirement for effector function. This is particularly advantageous in terms of safety as effector-less antibodies overcome a potentially dangerous inflammatory response in the brain [74].

In our study, we further demonstrated that four SUS treatments were sufficient to yield a significant reduction in tau pathology, and by combining SUS and the RN2N antibody fragment, we not only achieved increased histological, but also increased behavioral improvements. Importantly, using fluorescently labelled RN2N, we found that SUS not only caused an increased uptake by the brain, but moreover that the antibody fragment was effectively taken up into neurons where the tau damage occurs, and could be visualized not only in the cell body, but also proximal and distal dendrites [71]. This demonstrates that SUS can clear a pathology that, different from the Aβ pathology, is mostly intracellular. Moreover, this study presents SUS as an efficient method to deliver drugs (including different antibody formats) past the BBB into the brain and its cellular constituents.

An alternative approach to achieve therapeutic concentrations of antibodies in the brain is to use viral vectors such as the adeno-associated virus (AAV) vector. Recently, the genes encoding the anti-p-tau monoclonal antibody, PHF1, were delivered directly into brains of P301S mice. In contrast to previous studies using passive immunization with the same antibody, hippocampal antibody levels achieved after AAV delivery were ∼50-fold higher, achieving marked (≥80–90%) reductions in hippocampal tau pathology [75]. AAV-mediated delivery was also demonstrated with a gene encoding an anti-tau scFv in the P301S mice [73]. Although in both studies direct intracerebral injection of the AAV was conducted, a delivery route which is less than ideal for clinical trials, studies aimed at optimizing the vector capsids to efficiently and widely transduce the CNS following intravenous injections have been conducted [76]. Furthermore, ultrasound has been used to successfully enhance the delivery of intravenously delivered AAV across the BBB to achieve widespread gene expression in the brain [77], demonstrating that the technique can additionally be used to facilitate gene therapy approaches for treatment of AD.

CONCLUDING REMARKS: PREDICTIONS

The last years have seen several changes in tau research. Whereas an initial focus has been on serine/threonine-directed phosphorylation (and consequently the kinases and phosphatases that regulate this post-translational modification), tyrosine-directed kinases (such as Fyn and Pyk2) and phosphatases (such as STEP [78]) will be gaining more attention in coming years. Similarly, it can be anticipated that mechanisms in causing tau accumulation that are not driven by phosphorylation and subcellular relocalization will increasingly be explored, and research will be extended to neurodegenerative diseases with protein aggregation other than AD and FTDP-tau. Another shift has occurred with regards to studying the compartment in which tau causes damage. There will be an increasing appreciation that pathological tau impacts all aspects of mitochondrial function. Its role in nuclear functions is also increasingly being explored [79, 80], with more work anticipated to be done in the near future. An integration of how tau impairs the electrophysiological properties of neurons at the synapse and the AIS is still missing, and how this is tied to the release of tau into the extracellular space [81]. Here we do think that more will be learned about the ways in which tau aggregates are being packaged and how the contents is released, is taken up, and interacts with tau in recipient cells. That there is widespread tau seeding activity in AD at early Braak stages has only been shown recently [82].

It may also be that the roundworm C. elegans with its unique capabilities will be more utilized as an animal model complementing rodent work. In C. elegans, protein with tau-like repeat-1 (PTL-1) is the sole homolog of tau. We had found in collaborative work that PTL-1 regulates both neuronal and organismal aging [83, 84]. Furthermore, we found that PTL-1 deficient worms are hypersensitive to oxidative stress and are defective in the nuclear accumulation of the transcription factor SKN-1 in response to stress [85]. Interestingly, in mammals, the SKN-1 homolog Nrf2 has been shown to be involved in the autophagic degradation of p-tau [86]. This highlights the possibility to effectively link rodent and worm studies to gain deeper insight into pathogenic mechanisms. More work will finally go into an understanding how Aβ and tau ‘talk’ to each other. We are still far away from understanding the signaling cascades and integrating the different pathomechanisms that have been claimed to have a role in AD.

As we continue to learn more about the molecular mechanisms underlying the pathogenesis of AD, novel therapeutic targets and strategies for the clearance of specific populations of Aβ and tau will be identified. An example of this is the unpublished work from Karen Duff’s group which employs the use of the neuropeptide PCAP to activate the proteasome specifically in dendrites, thereby reducing dendritic tau levels only [http://www.alzforum.org/news/conference-coverage/new-explanation-dendritic-tau-its-made-there]. Another example is the use of gamma frequency (20 to 40 Hz, i.e., at the other end of the spectrum compared to ultrasound) that has been shown to attenuate amyloid load in mouse models [87]. Despite the success of therapeutics delivered via traditional mechanisms in pre-clinical animal models, we envision that the translation of AD therapeutics into future human clinical trials will employ non-invasive, efficient delivery systems to ensure adequate concentrations of therapeutics are reached in the brains of human patients. This will not only reduce the cost of treatment to an amount which will be sustainable by a country’s health system, but will also reduce the number of treatments for the patient. We also envision the use of combinatorial drugs, that for example target both tau and Aβ. As far as tau is concerned there will likely be a shift from targeting its posttranslational modifications and the enzymes in charge to simply lowering tau levels. Together, we anticipate a better integration of basic and translational research in order to achieve better health outcomes.