Tau: enabler of diverse brain disorders and target of rapidly evolving therapeutic strategies

Abstract

Several lines of evidence implicate the protein tau in the pathogenesis of multiple brain disorders, including Alzheimer’s disease, other neurodegenerative conditions, autism, and epilepsy. Tau is abundant in neurons and interacts with microtubules, but its main functions in the brain remain to be defined. These functions may involve the regulation of signaling pathways relevant to diverse biological processes. Informative disease models have revealed a plethora of abnormal tau species and mechanisms that might contribute to neuronal dysfunction and loss, but the relative importance of their respective contributions is uncertain. This knowledge gap poses major obstacles to the development of truly impactful therapeutic strategies. The current expansion and intensification of efforts to translate mechanistic insights into tau-related therapeutics should address this issue and could deliver better treatments for a host of devastating conditions.

Graphical Abstract

Summary Figure: Potential tau pathomechanisms. Developing effective tau-targeting therapeutics will require a better understanding of how exactly tau contributes to Alzheimer’s disease and other disorders of the central nervous system. Potential mechanisms likely fall into the three broad categories shown. However, the relative pathogenic impact and overall importance of individual mechanisms have yet to be defined in truly disease-relevant contexts and may differ among diseases and even patients. The blue box on the right indicates tau activities that do not directly mediate but indirectly promote or facilitate pathogenic processes.

Summary sentence:

Expanding efforts to develop tau-targeting therapeutics will shed light on this enigmatic protein and could yield better treatments for multiple brain disorders.

REVIEW SUMMARY

Background:

The microtubule-associated protein tau has been implicated in the pathogenesis of Alzheimer’s disease and a range of other neurodegenerative disorders (called “tauopathies”). As the number of people with tauopathies is rising in aging populations across the world, interest in the fundamental biology of this protein and in the development of tau-targeting treatments have been expanding rapidly. Recent insights into the complexity of this intrinsically disordered protein suggest that tau is a worthy but challenging target whose multifaceted nature will likely require a multipronged therapeutic approach. Derived from a single gene by alternative splicing, six major isoforms of tau have been identified in the human brain. In addition, tau is subject to many different posttranslational modifications, further indicating that it may be regulated by multiple processes and may participate in diverse functions.

Advances:

Tau is widely presumed to stabilize microtubules. However, the experimental reduction or ablation of tau in vivo does not alter many neural properties and processes that likely depend on microtubules, including neuronal integrity, axonal transport, synapse formation, and complex brain functions. Although tau reduction seems to have minimal effects on otherwise unmanipulated brains, it can prevent or diminish aberrant cell signaling, neural network dysfunctions (e.g., epileptic activity) and behavioral alterations caused by diverse disease processes, which suggests that tau activities are needed for other pathogenic triggers to cause these derangements. In addition to this “enabling bystander” role, tau’s interactions with a large number of other proteins can cause adverse gains of function, which are associated with—and possibly caused by— the formation of abnormal tau structures and assemblies. Because abnormal forms of tau trigger a plethora of pathomechanisms, targeting individual downstream mechanisms may have limited therapeutic impact, unless the relative pathogenic importance of the specific mechanism has been well established in experimental models that allow for conclusive validation of cause-and-effect relationships. Although much attention has focused on the abnormal aggregation of tau in tauopathies and on the ability of tau “seeds” to spread from neuron to neuron, internalization of propagating tau does not appear to impair neuronal survival or brain functions. Moreover, tau reduction prevents or diminishes neural network dysfunction and behavioral abnormalities also in disease models that do not have abnormal tau inclusions, which suggests that there is more to tau than aggregation and propagation. A promising diversification of tau-targeting therapeutic strategies is beginning to address this complexity. Lowering overall tau levels may have the greatest potential, as this strategy bypasses the unresolved questions of which forms of tau and which downstream mechanisms are most detrimental in any given condition.

Outlook:

Many efforts to develop better treatments for neurodegenerative diseases have failed, in large part because of an inadequate understanding of disease mechanisms and, perhaps, because too many fundamental knowledge gaps, alternative interpretations of data, and methodological complexities did not receive the attention they deserved. This Review highlights important gaps in the understanding of tau and the methodological advances needed to fill them. It also pinpoints obstacles that could complicate the translation of tau-related scientific discoveries into better therapeutics and offers pragmatic strategies to overcome these challenges. Despite the extraordinary progress that has been made to date, the main physiological functions that tau fulfills in the adult and aging brain remain to be defined. Another critical objective is to develop better experimental models and technologies to rigorously compare different tau species and pathomechanisms, particularly their relative impacts on neuronal functions and survival in vivo. For the development of truly informative biomarkers and effective therapeutics, it will be critical to rigorously differentiate between associations and cause-and-effect relationships. Until the main drivers of neuronal dysfunction and demise have been identified for Alzheimer’s disease and other conditions in which tau has a causal or enabling role, it seems prudent to focus on pragmatic strategies, such as overall tau reduction, while also expanding efforts to further validate the importance of more specific targets and approaches. Investigational approaches to lower overall tau levels include tau-targeting antisense oligonucleotides, which have advanced into a clinical trial for early Alzheimer’s disease, and the development of small-molecule drugs that can modulate the production or degradation of tau. The most desirable tau-targeting therapeutics would be efficacious across diverse tauopathies, as well as affordable, easy to access, and well tolerated when administered over long periods of time to fragile groups of people who likely take multiple other medications.

Introduction

The microtubule-associated protein tau (MAPT) is an intrinsically disordered protein that is most abundant in neuronal axons (1). Its interaction with microtubules is intermittent and rather short lived and has been characterized as a rapid “kiss-and-hop” process (2). Neurons can release tau and tau fragments into the extracellular milieu, a process that is enhanced by increased neuronal activity (3). Tau is also present, but less abundant, in neuronal cell bodies, nuclei and synaptic specializations, as well as in glia and other cell types in the brain and peripheral organs (1, 4). In humans, six major tau isoforms are generated in brain cells from a single MAPT gene by alternative splicing of exons 2, 3 and 10 (Fig. 1A). These isoforms differ in their number of N-terminal inserts (zero, one, or two) and microtubule-binding repeat domains [three (3R) or four (4R)] (1, 5), as well as in their subcellular distributions (6, 7) and protein interactomes (8), suggesting involvement in distinct functions. Tau undergoes various posttranslational modifications (Fig. 1B) (1, 9–12), indicating that it may be regulated by multiple pathways and may participate in diverse biological processes. Low body temperature (13) and high dietary salt intake (14) increase tau phosphorylation. The biological purpose of these and other tau modifications is uncertain.

Fig. 1. Tau isoforms and posttranslational modifications.

(A) Six major isoforms of tau are produced in brain from the MAPT gene through alternative splicing. N1 and N2 denote N-terminal inserts; R1 to R4 indicate microtubule-binding repeat domains, whose differential inclusion results in the generation of 3R versus 4R tau. Noncoding exons are shown in white; coding exons not expressed in brains or humans are in light grey; coding exons in human brain are in light blue. aa, amino acids. (B) Physiological posttranslational modifications identified by mass spectrometry of tau isolated from brain tissues of healthy adult WT mice. Note that some amino acid modifications could compete with each other (e.g., ubiquitination and acetylation). K, Lys; R, Arg; S, Ser; T, Thr; and Y, Tyr. [Adapted from figure 1a of (9)]

Although the main functions of tau also remain to be defined, there is no question that it is involved in a variety of common and disabling brain diseases. Intraneuronal accumulations of insoluble, filamentous tau aggregates underlie the formation of neurofibrillary tangles, a pathological hallmark of Alzheimer’s disease (AD) and multiple other neurodegenerative disorders (“tauopathies”) (1, 15). Besides these highly organized tau assemblies, nonfibrillar tau species, including tau oligomers and fragments, may also contribute to neurological disease (1, 16, 17). A causal role of tau in brain diseases is suggested by positive correlations between neurofibrillary tangle burden and brain dysfunctions, autosomal dominant mutations in MAPT that cause neurodegenerative disorders, and neuronal deficits caused or enabled by tau in experimental models (1, 11, 15, 18–22).

Intraneuronal accumulation of abnormal tau species is required for the postmortem diagnosis of AD. Although the heterogenous group of less common non-AD tauopathies by definition also have “tau pathology,” they differ from AD and from each other with respect to many other pathological and clinical features. Notably, a large number of very rare MAPT mutations (mostly within or adjacent to the microtubule-binding repeat domain of tau) can each cause non-AD tauopathies—in particular, frontotemporal lobar degeneration (FTLD)—but they never cause AD (1, 15, 23, 24). By contrast, the rare Ala¹⁵²→Thr (A152T) variant of tau seems to increase the risk for both AD and non-AD tauopathies (25, 26). These associations between specific changes in the amino acid sequence of tau and distinct neurodegenerative disorders merit further exploration, especially because animal and cell culture models expressing human tau with any one of several MAPT mutations that cause FTLD in humans (henceforth referred to as “FTLD-tau”) are widely used to study the pathogenic role of tau in AD and to develop better treatments for this condition. What if the pathogenic roles of tau in AD and FTLD differed fundamentally? Therapeutically targeting FTLD-tau might fail in AD, the most frequent cause of dementia, unless the targeted mechanism is at work in both types of disorders.

A particular challenge in this regard is the plethora of potential pathomechanisms that have been identified for (mostly FTLD-associated) tau. Identifying the mechanism that is most critical in AD has been made difficult by the lack of unquestionably AD-relevant experimental models that allow rigorous comparison of different tau species and pathomechanisms, particularly their effects on neuronal functions and survival. Notably, reducing levels of endogenous wild-type (WT) tau has beneficial effects in experimental models of autism (18), depression (19), epilepsy (21, 22, 27), and stroke (20), even though these models do not develop intraneuronal accumulations of abnormal tau aggregates, indicating that there is much more to tau than aggregation.

Putative physiological functions

Tau’s interaction with microtubules raises at least two possible roles: that tau supports the functions of microtubules and that microtubules support the functions of tau. It is widely presumed that tau’s main function is to stabilize microtubules. However, this dogma has been challenged, if not refuted, by a growing body of evidence. Early experiments carried out in cell-free conditions or using microinjection of tau into non-neuronal cells lacking endogenous tau suggested that tau stabilizes microtubules and regulates their dynamics (28–30). More recent studies refuted the hypothesis that tau stabilizes microtubules in primary rat neurons while yielding additional evidence that tau regulates microtubule dynamics (31). Whether genetic ablation of tau has a long-term impact on the neuronal expression of other microtubule-associated proteins (MAPs) under otherwise normal circumstances is uncertain. Western blot analysis of brain tissue from a small number of mice suggested increased brain levels of MAP1A in Mapt^−/− mice during early postnatal stages compared with WT controls (32). In mice maintained on a high-fat diet throughout life, hippocampal levels of MAP1A, MAP1B and MAP2 were increased in adult Mapt^−/− mice but either unchanged or decreased in aged Mapt^−/− mice (33). In contrast to tau ablation, MAP1B ablation causes defects in the formation of neuronal layers and axon tracts during development and this phenotype is markedly exacerbated by concomitant tau ablation (34), suggesting that tau can partially fulfill or assume developmental MAP1B functions in the absence of the latter molecule. Microtubule-stabilizing drugs reduced tau pathology, behavioral deficits, and neurodegeneration in transgenic mice overexpressing FTLD-tau (1, 35, 36), but whether these agents did so by compensating for a loss of tau function, counteracting an adverse gain of tau function, or another mechanism is uncertain. Moreover, such drugs have failed in clinical trials for AD and other tauopathies (37).

Notably, genetic ablation and acute knockdown of tau did not change axonal transport, which critically depends on microtubules, in cultured primary neurons (38, 39) or in mouse optic nerve axons in vivo (40). These findings seem difficult to reconcile with the notion that tau has a major role in stabilizing microtubules, at least in axons. Furthermore, with the possible exceptions discussed below, genetic ablation of tau during early development is well tolerated in rodents (18, 21, 22, 41–44), and so is acute knockdown of tau in the brains of young adult rodents and non-human primates (27, 45). No compelling evidence could be found to show that tau reduction causes derangements of microtubules or alterations in microtubule dynamics that affect neuronal functions or integrity in young, adult, or aging mammalian brains.

By contrast, mounting evidence suggests that tau helps regulate multiple signaling pathways. Tau promotes insulin-induced tyrosine phosphorylation of the insulin receptor substrate 1 (IRS-1) and inhibits activation of phosphatase and tensin homolog (PTEN) (46). Ablation of these functions in Mapt^−/− mice reduces anorexigenic effects of insulin in the hypothalamus (46), which may contribute to the moderate age-dependent obesity of these mice (42, 46). Conversely, neuronal accumulation of FTLD-tau enhanced insulin responsiveness in transgenic mice (47), a process that might contribute to weight loss observed in people who have or are at risk for AD (48). Yet, tau ablation prevented synaptic and cognitive deficits in a streptozotocin-induced model of type 1 diabetes (49).

Several studies have explored whether tau has roles in synaptic physiology and plasticity. Deficits in long-term potentiation were identified in the hippocampal CA1 region of one tau-deficient (Mapt^−/−) model (50) but not in the dentate gyrus of another (51). Similarly, some groups observed reductions in long-term depression in Mapt^−/− mice (46, 52), whereas others did not (50), possibly because of differences in mouse ages and experimental protocols. Reductions in long-term depression might counteract processes that abnormally increase this activity in AD (53).

Mapt^−/− mice have normal excitatory postsynaptic transmission in the hippocampal CA1 region and dentate gyrus (44, 51). However, they have an increased frequency of inhibitory postsynaptic currents in dentate granule cells (51), which may contribute to their increased resistance to epilepsy of diverse causes (18, 21, 22, 27, 44, 51). The potential roles of tau in the activity of excitatory versus inhibitory neurons and synapses merit further investigation.

Tau also seems to be involved in the generation, maintenance or regulation of brain rhythms. Compared with WT controls, Mapt^−/− mice have alterations in various neural network oscillations (41, 54, 55). Some of these alterations may be related to the abnormal sleep structure that has been observed in these mice, which includes decreased non-rapid eye movement (NREM) sleep time, a higher number of state transitions between NREM sleep and wakefulness, and shortened sleep bouts (55). Sleep deprivation increases tau levels in cerebrospinal fluid (CSF) and brain interstitial fluid (56). Additional studies are needed to determine whether this process reflects a role of tau in sleep regulation and whether it promotes tau pathology in AD and other conditions associated with sleep disturbances (57).

The changes in brain rhythms and the resistance of Mapt^−/− mice to epilepsy may reflect, at least in part, roles of tau in neurogenesis (58) or in neuronal migration and maturation (59). However, although Mapt^−/− mice have slightly smaller brains than WT controls (18), there appear to be no studies demonstrating obvious changes in the overall brain anatomy and cytoarchitecture of these mice. It could be interesting to further explore the effects of tau reduction on neurogenesis and connectivities among brain regions and specific neuronal populations, especially in models of autism in which tau reduction prevented excessive brain growth (megalencephaly) and behavioral abnormalities (18).

Reports of marked behavioral deficits in Mapt^−/− mice suggesting an essential role of tau in locomotor functions and in learning and memory (33, 60) could not be replicated by others (18, 21, 22, 42, 44, 51, 61). In humans, the chromosome 17q21.31 microdeletion syndrome causes intellectual disability, epilepsy, and reduced muscle tone (62). Because MAPT resides at this location, it could be tempting to speculate that tau reduction contributes to the impairments in these patients. However, this deletion includes other genes, and mutations in one of them (KANSL1, encoding KAT8 regulatory NSL complex subunit 1) are sufficient to cause the syndrome (62), making a causal role of tau reduction unlikely. Indeed, tau reduction has beneficial effects in a variety of disease models, as described further below.

Taken together, these findings suggest that tau’s most important physiological functions in the central nervous system (CNS) have yet to be defined. In light of tau’s multiple isoforms and many posttranslational modifications, it is tempting to speculate that these functions may differ among cell types, subcellular compartments, and perhaps even life stages.

Roles in neurological and psychiatric disorders

In principle, tau may be involved in the pathogenesis of CNS diseases in at least three ways. A disease process might (i) block a critical tau function (loss of function), (ii) change tau in ways that turn it into a pathogenic effector or mediator (gain of function), or (iii) require the presence of tau in essentially unaltered form (enablement) (Fig. 2; also see the summary figure).

Fig. 2. Physiological but disease-enabling functions of tau.

(A) Simplified conceptual diagram highlighting how targeting tau could block the development or progression of disease, even when tau is neither a driver nor a direct mediator of the pathogenic cascade. (B) Tau regulates interactions among the tyrosine kinase Fyn, N-methyl-D aspartate receptors (NMDARs), and postsynaptic density protein 95 (PSD-95) (44), as well as between PSD-95 and synaptic Ras GTPase-activating protein 1 (SynGAP1), which negatively regulates synaptic extracellular signal-regulated kinase (ERK) activation (20). Although these tau activities may normally form part of the complex network of regulators that support synaptic and neuronal activities, they may also allow pathological overstimulation of glutamate receptors to result in “excitotoxicity,” a process tau reduction has been shown to counteract (4, 20). P, phosphorylation; MEK, mitogen-activated protein kinase kinase. [Adapted from figure 3 of (4) with permission from Elsevier] (C and D) In mouse models, the presence of endogenous WT tau allows other processes to overactivate the PI3K-Akt-mTOR signaling pathway, which in turn promotes the development of autism-like behaviors, including excessive stereotypic movements, social interaction deficits and cognitive inflexibility (18). (C) Partial or complete genetic reduction of tau prevented excessive stereotypic movements (time spent engaged in such movements in seconds) in Scn1a^RX/+ knockin mice, which carry a sodium channel mutation that causes autism and epilepsy in humans. By contrast, tau reduction did not alter this behavior in mice without this mutation. Numbers of mice per group are indicated in bars. ****P < 0.0001 vs. WT (Scn1a^+/+Mapt^+/+) controls or as indicated by brackets [two-way analysis of variance (ANOVA) with Holm-Sidak test]. ns, not significant. Error bars indicate SEM. [Reprinted from figure 1A of (18) with permission from Elsevier] (D) Because PTEN counteracts PI3K, tau’s ability to inhibit PTEN activation (46) may enable normal activity of the PI3K-Akt-mTOR pathway under physiological conditions. However, when pathological processes overactivate PI3K, as in some autism spectrum disorders, tau’s constraint of PTEN becomes maladaptive and allows for overactivation of the downstream signaling cascade, promoting the development of autism symptoms through diverse anatomical and pathophysiological mechanisms. Tau reduction counteracts this process by releasing the activity of PTEN and possibly also by blocking neural network dysfunctions. PIP2, phosphatidylinositol 4,5-biphosphate; PIP3, phosphatidylinositol 3,4,5-triphosphate; PI3K, phosphatidylinositol 3-kinase; Akt, protein kinase B; mTOR, mammalian target of rapamycin; TSC1/2, tuberous sclerosis 1 and 2; S6, ribosomal protein S6; ASD, autism spectrum disorder. [Reprinted from figure 3 of (18) with permission from Elsevier]

Early studies raised the then-plausible possibility that sequestration of tau into intraneuronal aggregates might prevent tau from stabilizing microtubules, leading to neurodegeneration (63). In light of the more-recent studies discussed above, this and other loss-of-function mechanisms now seem unlikely. By contrast, evidence continues to mount for pathogenic gains of function of tau. Intraneuronal accumulation of aggregated tau correlates well with clinical disease manifestations in AD, FTLD, and other tauopathies (1, 11, 15, 64). In addition to this “guilt by association”, interventions that block neuronal tau accumulation reduce functional deficits and neuronal loss in experimental models (1, 11, 45, 65, 66).

Another possibility is that physiological tau functions enable brain diseases by allowing other factors and processes to alter signaling pathways (Fig. 2). An analogy that might help illustrate this concept in general is the role of the C-C chemokine receptor type 5 (CCR5) in the pathogenesis of AIDS. Because HIV-1 uses CCR5 to enter T lymphocytes, genetic loss of this co-receptor prevents the development of HIV-1 disease (67). Thus, although CCR5 is not an active effector or mediator of HIV-1-dependent immunopathogenesis, it clearly enables this process in critical ways. In a similar vein, physiological functions of tau may enable neural network dysfunction and behavioral abnormalities caused by other pathogenic drivers in models of autism, depression, epilepsy, and stroke (18–22, 27, 44). Consistent with this notion, there is no evidence for abnormal tau aggregation or propagation in these models. Tau reduction also prevents epileptiform activity and some other functional abnormalities in human amyloid precursor protein (hAPP) transgenic mice that show no evidence of abnormal tau accumulation but simulate several other aspects of AD (9, 22, 51). In primary neuronal cultures, tau reduction does not alter axonal transport under baseline conditions but effectively prevents axonal transport deficits caused by oligomeric assemblies of amyloid-β (Aβ) peptides (38, 39), which have also been causally implicated in the pathogenesis of AD (53).

The analogy to CCR5, genetic deficiency of which is well tolerated (67), may also help explain why tau reduction has such limited adverse effects under physiological conditions and such profound protective effects under pathological conditions. Under baseline conditions, redundancies in the distinct biological processes in which CCR5 and tau participate may minimize the consequences of their ablation. By contrast, pathological processes that more specifically depend on the presence of these molecules such as HIV-1 entry into T cells or Aβ-induced deficits in axonal transport, can be blocked by the ablation or reduction of CCR5 or tau, respectively.

The “enablement” and “adverse gain-of-function” mechanisms are not mutually exclusive and, indeed, may form vicious cycles that drive progression of AD and possibly other disorders. For example, tau enables epileptogenesis in models of diverse diseases (18, 20–22, 27, 44), and chronic epileptic activity in humans is associated with the development of tau pathology, which may contribute to cognitive decline in these patients (68). Furthermore, overexpression of tau can cause epileptiform activity in transgenic mice (69–71). Possibly related to this modulation of network hypersynchrony, tau reduction lowers, whereas tau overexpression increases, the power of brain oscillations in the low frequency range (41). Such changes in brain rhythms could alter cognitive functions (72). In addition, increased power of low frequency oscillations predicts progression to AD in people with subjective cognitive impairments (73), whereas increasing the power of higher-frequency oscillations reduces cognitive deficits and counteracts AD pathology in animal models (74, 75), suggesting that changes in brain rhythms could have symptomatic as well as disease-modifying consequences.

Pathogenic mechanisms of abnormal tau

Because tauopathies are defined by the accumulation of abnormal tau aggregates, many studies have focused on tau misfolding and aggregation (1, 15, 76–78). However, the most pathogenic forms of abnormal tau have not yet been conclusively identified and may differ across tauopathies (15, 77, 78) and among individual patients with AD (76). Notably, even when specific tau alterations are tightly associated with disease states, the biological meaning of such alterations is difficult to discern without further experimental analysis. For example, tau that accumulates in brains of people with distinct tauopathies has disease-specific structural features and can be enriched for distinct isoforms of tau (15, 78). However, whether these tau structures and isoform differences causally contribute to the respective tauopathies or result from specific disease processes that occur in one condition but not others remains to be determined. Similar uncertainties about causality apply to many posttranslational modifications of tau. For example, hyperphosphorylation of tau can be pathogenic or protective, depending on the specific sites targeted (79). These complexities are relevant not only to the development of tau-related therapeutics, but also to the interpretation of tau-related biomarkers, an important and rapidly evolving aspect of this field.

Increased binding of radioligands to detect tau aggregates in brains of live people by positron emission tomography (64, 80) and increased levels of phospho-tau species in CSF (81) and blood (82, 83) are becoming increasingly useful in the differential diagnosis of tauopathies. In AD, these biomarkers also correlate with disease progression (64, 81, 83). How well they can predict clinically meaningful benefits in trials of tau-related therapeutic strategies remains to be determined. Notably, preclinical evaluations of therapeutic interventions in FTLD-tau-overexpressing mouse models have revealed thought provoking dissociations between the intraneuronal accumulation of tau aggregates on one hand and neural dysfunctions and neurodegeneration on the other. For example, reducing overall tau levels decreased behavioral deficits despite the persistence of intraneuronal tau aggregates (66), and reducing the RNA binding protein T cell intracellular antigen 1 (TIA1) increased the formation of neurofibrillary tangles while reducing neurodegeneration, behavioral deficits, premature mortality, and tau oligomer formation (84). These results are consistent with other findings, indicating that oligomeric assemblies and fragments of tau are more bioactive than highly organized aggregates such as tau fibrils and that small, nonfibrillar tau species can contribute to functional neural impairments both within neurons and from the extracellular space (12, 16, 17, 69, 85, 86).

Most putative pathomechanisms of tau have been identified in experimental models in which individual isoforms or fragments of tau are overexpressed at rather high levels and studied in isolation from other factors that tau interacts with in the aging human brain (1, 11, 12, 87). Genomic alterations caused by transgene insertions may contribute to tauopathy-like phenotypes (88) but have been excluded in only some of the most widely used animal models in this field (89). Further highlighting that no individual experimental model is perfect, the half-life of tau is much shorter in cultured human and rodent neurons than in human and rodent brains (90), and the relative abundance of tau isoforms differs markedly between immature neurons in vitro and mature neurons in the adult or aging brain (91, 92). Thus, it is critical to scrutinize all available evidence, ideally from human studies and multiple experimental models, before embarking on the costly development of treatments that target individual forms of abnormal tau or specific pathomechanisms. The following discussion focuses on potentially targetable pathomechanisms that are supported by in vivo evidence and whose manipulation in experimental models reduced not only putative disease-relevant morphological or biochemical alterations but also neuronal dysfunctions or loss.

In experimental models, abnormal tau can impair diverse subcellular compartments, including neuronal axons, presynaptic terminals, and postsynaptic specializations (1, 4, 11, 12, 93). It seems unlikely that a single pathomechanism accounts for all of these changes. For example, cleavage of tau at Asp³¹⁴ by caspase-2 causes an enrichment of FTLD-tau in dendritic spines and disrupts synaptic and cognitive functions, presumably by reducing the trafficking or anchoring of AMPA-type glutamate receptors (17), the main transducers of rapid excitatory transmission in the mammalian brain (94). Whether this cleavage is also critical for tau-dependent impairments of other subcellular compartments remains to be determined.

Diverse lines of evidence suggest that abnormal tau can disrupt proteostasis mechanisms (25, 95–97), which are critical for neuronal survival and communication (98). For example, in transgenic mice overexpressing FTLD-tau, the reduction of cognitive deficits by activation of the cyclic adenosine monophosphate-protein kinase A pathway has been related to the prevention of tau-induced impairments of the 26S proteasome (99). Enhancing autophagy by pharmacological or genetic manipulations reduced tau pathology and neuronal loss in zebrafish that express A152T-variant human tau (25).

Another intriguing possibility is that abnormal tau promotes the activation and increased expression of transposable elements (100, 101), a process that may involve the intranuclear depletion of piwi-interacting RNAs (100) and could contribute to neuronal dysfunction and degeneration through a variety of mechanisms, including formation of DNA double-strand breaks and aberrant changes in gene expression (100–103). The suppression of neurodegeneration and locomotor abnormalities in a fruit fly tauopathy model by treatment with a reverse transcriptase inhibitor (100) provides additional support for the potential importance of this mechanism.

Acetylation of tau can promote the formation, accumulation and cell-to-cell propagation of abnormal tau species (104, 105). Acetylated, hyperphosphorylated FTLD-tau may contribute to neural dysfunctions by altering the neuronal axon initial segment, a process that may involve interactions between tau and the microtubule end-binding protein 3, and by changing neuronal excitability and neural network activity (106, 107). Acetylated tau can also disrupt kidney and brain expressed protein (KIBRA)-mediated signaling at synapses (12, 108).

Abnormal tau may also cause neuronal dysfunction and degeneration through aberrant activities of the innate immune system (109). Intraneuronal accumulation of FTLD-tau triggers neuronal phagocytosis by microglia through a process that involves secretion of an opsonin, a molecule that binds to and enhances phagocytosis of a target antigen or cell, and production of nitric oxide (110). Deletion of the complement C3a receptor reduced neuroinflammation, synaptic deficits, and neurodegeneration in transgenic mice overexpressing FTLD-tau (111). In similar models, ablation of inflammasome components reduced tau pathology as well as learning and memory deficits (112), whereas overexpression of the cytokine interleukin 1β increased neuroinflammation and tau pathology (113). Coexpression of apolipoprotein (apo) E4, the most common genetic risk factor that strongly increases AD risk (114), enhanced neurodegeneration in transgenic mice overexpressing FTLD-tau, and pharmacological deletion of microglia reduced both tau pathology and neurodegeneration in this model (115).

Considerable attention has been paid to the transcellular propagation (or “spread”) of tau aggregates from neuron to neuron (15, 65, 77, 109, 116, 117), which is enhanced by neuronal overexcitation (118, 119). However, there does not appear to be any conclusive published demonstration that the propagating tau impairs brain functions or the survival of neurons that take it up from the extracellular space, even in transgenic mice whose neurons overexpress an abundance of abnormal tau and, thus, might be prone to the seed-dependent formation of toxic tau aggregates. Indeed, injection of various tau species promoted tau aggregation and propagation in mouse models overexpressing human WT or FTLD-tau, but no neuronal loss or degeneration was observed even 15 months after injection (120–122). In primary cultures, aggregates of a fluorescently tagged phosphomimetic of tau efficiently propagated from neuron to neuron but did not impair the survival, presynaptic terminals, basic electrophysiological properties, or activity of neurons (123). Whether the axonal transport deficits observed in this study resulted from neuronal overproduction of the engineered tau species or uptake of its propagating aggregates is unclear.

Notwithstanding these caveats, extracellular tau may impair neural functions independently of cell-to-cell propagation. For example, when added to the medium of cultured human neurons, N-terminal fragments of tau released from neurons of AD patients caused neuronal hyperactivity and increased production of Aβ (124). Moreover, even brief exposure of mouse brain slices to extracellular human tau oligomers impaired long-term potentiation (86).

Despite these multiple potential pathomechanisms, it is far from clear how exactly abnormal tau causes or contributes to neuronal dysfunction and degeneration, in large part because available methods and experimental models have not yet been optimized to (i) rigorously and efficiently compare the relative pathogenicity of specific forms of tau in truly disease-relevant contexts and (ii) selectively modulate these tau species in vitro and in vivo.

Interactions with other pathogenic factors

Tau aggregation in transgenic mice overexpressing FTLD-tau is enhanced by coexpression of α-synuclein (125), a protein that accumulates in neurons in Parkinson’s disease, dementia with Lewy bodies, and many cases with AD (126, 127). In FTLD-tau transgenic mice, tau accumulation is also increased by injecting Aβ peptides into the brain (128) and by neuronal overexpression of APP bearing autosomal dominant mutations that cause early onset familial AD and elevate cerebral Aβ levels (129). By contrast, overexpression of tau does not increase Aβ accumulation in such models, as compared to singly transgenic APP mice (129). However, ablation or reduction of endogenous, non-aggregated WT tau prevents or diminishes Aβ toxicity in cultures of rodent primary neurons (130), as well as several (but not all) synaptic, network, and cognitive alterations in different lines of APP transgenic mice (22, 44, 51, 131, 132). These results raise the possibility that even physiological forms of tau can enable or promote disease processes under otherwise pathological circumstances, possibly by allowing other factors such as APP or some of its metabolites to alter signaling pathways that tau normally enhances or constrains (Fig. 2).

Tau also has bidirectional relations with apoE4. Compared with the more common apoE3 isoform, apoE4 increases AD risk, lowers age of disease onset, and promotes hyperphosphorylation of endogenous WT tau in induced pluripotent stem cell (iPSC)-derived human neurons and in transgenic mice (114, 133–135). Conversely, ablation of endogenous tau protects transgenic mice expressing human apoE4 fragments against age-dependent cognitive decline (136), which suggests a mutual interdependence between these pathogenic factors. Neurons of apoE4 carriers produce more apoE fragments than those of apoE3 carriers and these fragments impair mitochondria as well as the cytoskeleton (114, 133, 134). Compared with apoE2 and apoE3, apoE4 also exacerbates tau pathology and neurodegeneration in transgenic mice overexpressing FTLD-tau, whereas ablation of apoE or of microglia reduces these abnormalities (115, 137). Thus, apoE may be required for the development of these tau-dependent abnormalities, and microglia may have a key role in this copathogenic interaction.

Therapeutic approaches

Most tau-related therapeutic strategies that have advanced into clinical trials were designed to target select tau assemblies or posttranslational modifications presumed to be pathogenic, as judged from clinicopathological associations and, in some instances, additional evidence for causal contributions in preclinical models (11, 37, 138). Many of these trials were discontinued because of toxicity or lack of efficacy (37). As with other drug trials that have yielded negative results in AD or other neurodegenerative disorders, it is possible that they were initiated too late in the course of the disease, that target engagement was inadequate in the most relevant sites in the brain, or that the targeted form of tau does not actually contribute much to clinically meaningful functional impairments.

In light of these challenges, it is encouraging that the range and number of tau-targeting approaches have been steadily increasing after many anti-Aβ treatments failed in late-stage clinical trials for AD. More than 20 tau-related therapeutic strategies are undergoing clinical trials, including small-molecule inhibitors of tau acetylation, aggregation, O-GlcNAcylation (GlcNAc, N-acetylglucosamine), or phosphorylation, as well as active and passive immunizations against tau, and microtubule stabilizers (37).

Predicting the outcome of these trials is made difficult by the lack of conclusive answers to the following key questions. Are the most pathogenic forms of tau misfolded monomers, fragments, or multimers, and are they fibrillar or nonfibrillar? Do these pathogenic tau species act intracellularly or extracellularly? What are the roles of tau’s many posttranslational modifications and isoforms in health and disease? Does cell-to-cell spreading of tau impair neuronal survival or brain functions? What are the most impactful downstream mechanisms that medicate tau’s contributions to the development and progression of brain diseases? Could they be blocked without substantial side effects?

The importance of these questions is highlighted by the following hypothetical examples. If tau species that cause the greatest amount of neuronal dysfunction were produced by the affected neurons themselves and acted primarily within their presynaptic terminals or dendritic spines, tau-targeting antibodies that remain in the extracellular space and cannot reach these intracellular compartments would probably not be able to reduce or prevent functional deficits in clinical trials (139). In a similar vein, if the formation and propagation of tau aggregates in human tauopathies were merely coincidental with or a parallel consequence of an unidentified upstream process that is the main driver of neuronal degeneration and brain dysfunction, inhibiting tau aggregation or propagation might yield negative outcomes in clinical trials.

Among therapeutic strategies that are in preclinical development or undergoing clinical evaluation, those that could lower overall tau levels may, thus, have the greatest potential, as this approach bypasses the unresolved questions of which forms of tau and pathomechanisms are most critical in any given condition (Fig. 3). Preclinical proof-of-concept for the benefits of overall tau reduction comes from the suppression of regulatable tau-encoding transgenes, which diminished neural dysfunctions putatively caused by FTLD-tau (66) or A152T-variant tau (41), as well as from the genetic ablation or partial reduction of endogenous mouse tau, which reduced neural network dysfunctions and behavioral abnormalities in mouse models of AD, autism, depression, epilepsy, and stroke (18–22, 27, 44). Notably, even partial (~50%) reduction of tau was effective in several of these studies, and there is no evidence that this level of tau reduction impairs CNS functions in animal models.

Fig. 3. Tau-focused therapeutic strategies.

Approaches that target the upstream components (bold) of the depicted cascade may have the greatest therapeutic potential because the downstream components fan out into a variety of potential targets whose relative pathogenic impacts are uncertain.

Although gene editing approaches are rapidly advancing toward clinical applications, a more readily translatable approach to reducing overall tau levels in humans is the intrathecal injection (via lumbar puncture) of antisense oligonucleotides ASOs that promote the degradation of MAPT mRNAs. This approach, or intracerebroventricular injection of such ASOs, effectively lowers overall tau levels in brains of rodents and nonhuman primates and reduces chemically induced epilepsy in WT mice (27) as well as behavioral abnormalities and neurodegeneration in transgenic mice overexpressing FTLD-tau (45). Intrathecal injections of tau-targeting ASOs were well tolerated in nonhuman primates (45) and are undergoing a phase 1 safety assessment in people with early AD (ClinicalTrials.gov identifier NCT03186989). ASOs can also be used to modulate alternative splicing, a strategy that could help lower the abundance of the most pathogenic isoforms of tau (140), which may differ across tauopathies (1, 15).

An ASO that targets survival of motor neuron2, centromeric (SMN2) precursor mRNA (pre-mRNA) has shown substantial clinical benefits in spinal muscular atrophy (141), a neurodegenerative disease of childhood, and encouraging results seem to be emerging from a trial of an ASO that targets mutant huntingtin for Huntington’s disease (142). However, because ASOs are large, negatively charged molecules that are subject to degradation, bringing ASOs across the blood-brain barrier and maintaining efficacious levels in brain tissues require repeated lumbar punctures, which are suboptimal for the prolonged treatment of chronic brain diseases in older people. Thus, developing effective tau-lowering small-molecule drugs that can be given orally is an important objective. So far, this has been a difficult feat to accomplish, at least in part because decreases in tau levels are a sensitive and early indicator of neurotoxic off-target effects. Consequently, neurotoxic compounds can be false-positive confounders of screens for tau-lowering drugs (143).

In principle, overall tau reducers might act by directly or indirectly reducing the transcription of MAPT or the translation of MAPT mRNA or enhancing the degradation of MAPT (pre-)mRNA or tau protein (Fig. 3). Several types of drugs were selected or developed to reduce specific forms of tau but may also be able to lower overall tau levels. A relevant strategy at preclinical stages of development is the identification of small molecules that specifically target MAPT pre-mRNA (144).

In cultures of primary rodent neurons or iPSC-derived human neurons, tau levels can be reduced by treatment with statins at concentrations that are not toxic in vitro (145, 146). However, it is unknown whether high enough brain concentrations of these drugs could be achieved to adequately reduce tau levels in vivo and whether such concentrations could be tolerated.

Acetylation can compete with ubiquitination and thereby counteract tau degradation (147). Genetic ablation or pharmacological inhibition of the lysine acetyltransferases p300 and CREB-binding protein (CBP) reduces tau acetylation and total tau levels, as well as tau pathology, neurodegeneration and behavioral deficits in transgenic mice (12, 105). Salsalate, a non-acetylated dimer of salicylic acid used to treat arthritis, reduces tau acetylation in the brain (12) and is undergoing a phase 1 trial for mild to moderate AD (NCT03277573).

Enhancing autophagy by pharmacological or genetic manipulations reduced tau pathology and neuronal loss in zebrafish that express A152T-variant human tau (25). A recent study showed that tau degradation is regulated by the farnesylated protein Rhes, a guanosine triphosphate (GTPase) in the Ras family (148). Treatment with lonafarnib, an inhibitor of farnesyltransferase used in cancer treatment, reduced brain atrophy, tau pathology, neuroinflammation, and behavioral deficits in transgenic mice overexpressing FTLD-tau (148).

Another strategy to reduce tau levels involves the generation of proteolysis-targeting chimeras (PROTACs)—molecules linking a target recognition domain with a ubiquitin ligase that helps shuttle the target molecule into protein degradation pathways, particularly the proteasome (149). Such compounds also hold promise for the targeted degradation of tau and are undergoing preclinical evaluation (149, 150).

These strategies highlight an encouraging diversification of the tau-related drug development pipeline, which will hopefully yield better therapeutics for tauopathies and other disorders in which tau has a causal or enabling role. The most important needs (Fig. 4) in this exciting area of translational neurobiology include: (i) affordable anti-tau therapeutics that can be administered safely and easily for long periods of time to fragile populations who are middle-aged or older and likely to take other medications for aging-related comorbidities; (ii) drugs that effectively and safely reduce as many disease-promoting tau species as possible in the most vulnerable cell types and subcellular compartments; and (iii) biomarkers that reliably detect target engagement in the brain of live people and, ideally, correlate well with clinically meaningful functional outcomes.

Fig. 4. Tau-related challenges and needs.

In principle, several of the trailblazing discoveries reviewed here could lead to the development of better treatments for AD and other tauopathies. However, many efforts to develop treatments for brain diseases have failed, in large part because of an inadequate understanding of the pathobiology of these complex disorders and, perhaps, because too many fundamental knowledge gaps, alternative interpretations of data, and methodological complexities did not receive the attention they deserved. Addressing the scientific knowledge gaps and technological challenges highlighted in this Review could accelerate and enable the development of truly impactful tau-related diagnostics and therapeutics. PTM, posttranslational modification.

Conclusion

The multifaceted protein tau raises fascinating scientific questions, and its importance from a public health perspective is increasing as the prevalence of tauopathies continues to surge around the world. Tau may fulfill different functions in different cell types, subcellular locations and circumstances, and may contribute to different diseases through diverse mechanisms. However, its most important functions and pathomechanisms remain to be defined. Adverse gains of function may turn tau into a direct mediator of neurotoxic effects, and even some of its physiological activities may allow disease processes to destabilize neural networks in ways that contribute to behavioral abnormalities and cognitive decline. Developing more effective treatments for tauopathies and other conditions involving tau will likely require a more definitive understanding of its pathobiology. Until this important goal has been reached, lowering overall tau levels would seem to be the most pragmatic and promising therapeutic approach. This strategy has shown beneficial effects in diverse disease models and a safety profile that has allowed related therapeutics to advance into clinical trials.