Beyond microtubule regulation: the multifaceted roles of tau in neuronal function and dysfunction

Abstract

The neuronal microtubule-associated protein tau is involved in several neurodegenerative diseases, collectively referred to as tauopathies. In these diseases, it aggregates and plays a central role in triggering cell death. Tau was originally identified as a factor that co-purifies with microtubules. Since then, it has been shown that, in addition to microtubules, it also interacts with many other molecules, is present in multiple cellular compartments, and influences various signaling pathways. Despite its diverse interactions and its potential function as a neuronal interaction hub, the exact role of tau in a neuron remains unclear even after decades of tau research. A better understanding of the physiological role of tau is therefore a prerequisite for the rational development of therapeutic strategies to manipulate its expression, modification, or interactions. Here, we discuss the current state of knowledge regarding the microtubule-related activities of tau, its propensity for self-assembly and aggregate formation, and its non-microtubule-related interactions. The aim is to develop a framework for understanding how tau acquires toxic properties during disease progression and how this knowledge can be used for the rational development of tau-targeted drugs.

Introduction

An entire group of neurodegenerative diseases is named after the microtubule-associated protein tau – the tauopathies, of which Alzheimer’s disease (AD) is the most prevalent [1, 2]. A common feature of tau in these diseases is the dysregulation of posttranslational modifications and the formation of abnormal tau aggregates, culminating in the formation of neurofibrillary tangles (NFTs). In AD, tau pathology is accompanied by the formation of senile plaques composed of aggregated amyloid-β peptides. However, here too, tau pathology correlates much better with sites of neurodegeneration and cognitive decline than plaque formation [3]. This suggests that the development of tau pathology is the trigger for cell death, and a logical consequence for the development of therapeutics is therefore to reduce tau aggregation or modulate the posttranslational modifications of the tau protein. However, tau is not an easy target molecule, and it has been shown to have many interaction partners and the potential to modulate various cellular processes. A thorough understanding of its diverse functions is therefore key to the development of tau-targeted therapies with minimal side effects, especially since much of the literature on tau continues to incorrectly describe its primary function as a stabilizer of axonal microtubules.

Tau localization

Tau is not a single protein, but rather consists of a collection of proteoforms produced by a single gene located in humans on chromosome 17q21 containing 16 exons [4]. Tau is predominantly expressed in neurons of the central and peripheral nervous system, where its various proteoforms arise through alternative splicing and a variety of posttranslational modifications. The size of tau isoforms ranges from 352 amino acids in the fetal central nervous system to 758 amino acids in most peripheral neurons such as the dorsal root ganglia and certain neuronal populations of the CNS such as cerebellar and retinal ganglion cells [5, 6].

The first clue to the diverse roles of tau came from its subcellular distribution (Fig. 1). Early immunocytochemical studies showed that tau is restricted to axons [7]. It was later found that the monoclonal antibody used in this study (Tau-1) bound to an unphosphorylated tau epitope and that total tau was distributed in multiple cellular compartments, including the somatodendritic compartment of neurons [8]. This already suggested that the localization of tau strongly depends on its posttranslational modification. Indeed, tau has since been found to be localized to the neuronal plasma membrane [9, 10], the nucleus [11, 12], the presynapse [13], and postsynaptic spines [14]. Some of these localizations depend on posttranslational modifications. For example, plasma membrane-associated tau is predominantly dephosphorylated [15], and phosphorylation of sites flanking the microtubule-binding region prevents this interaction [16]. Nuclear localization also depends on the phosphorylation state of tau [12]; moreover, nuclear tau is hyperacetylated at Lys174 [17]. Hyperphosphorylated tau has been observed to accumulate in intact dendritic spines [18]. Presynaptic localization appears to be associated with liquid-liquid phase separation of tau [19], a process in which biomolecules within a cell spontaneously form liquid-like condensates, which is enhanced by increased tau phosphorylation [20].

Fig. 1. Tau localization in and between neurons.

Schematic representation of the presence of tau in various subcellular compartments as well as outside of cells.

Furthermore, tau has been found not only within cells but also in tunneling nanotubes (TNTs), filamentous membranous structures that bridge and connect cells [21]. Tau is also secreted by neurons into the extracellular space [22], likely also under physiological conditions, as secretion is dependent on neuronal activity [23]. On the other hand, tau can also be taken up by neurons, and the low-density lipoprotein receptor 1 (LRP1) has been identified as required for the internalization of soluble tau by endocytosis [24].

The localization of tau at multiple sites inside or outside the neuron makes it a potential versatile biomarker or therapeutic target beyond AD and opens up potential relevance for tau-targeted therapies.

Microtubule binding and self-assembly

As the acronym tau, for “tubulin-associated unit,” suggests, the conventional view is that tau mainly functions as a structural microtubule-associated protein (MAP). Indeed, tau was originally identified as a factor that co-purifies with tubulin through repeated cycles of polymerization and depolymerization [25]. The region of tau responsible for microtubule binding has been identified as an 18-amino acid repeat in the carboxy-terminal half of the protein, with one repeat representing the basic tubulin-interacting unit [26]. Depending on the tau isoform, tau contains three (3 R Tau) or four (4 R Tau) of these repeats in a sequential arrangement. In a near-atomic model of microtubule-tau interaction based on cell-free cryo-electron microscopy experiments, a continuous stretch of a single repeat region spans three tubulin monomers and binds across both intra- and interdimer-tubulin interfaces [27]. Thus, the microtubule-binding region of a 4-repeat isoform forms a continuous stretch spanning four tubulin dimers on a single microtubule protofilament (Fig. 2, left). Live-cell imaging experiments suggested that an additional carboxy-terminal pseudorepeat region strongly enhances microtubule interaction. This region contains an 18-amino acid motif that is highly conserved in the Tau/MAP2/MAP4 family of structural MAPs [28]. Under conditions of high microtubule density, such as those found in the axon, the microtubule interaction of tau thus comprises the classic three or four microtubule-binding repeats plus an additional pseudorepeat region following the repeats. The microtubule interaction of the pseudorepeat region is also supported by nuclear magnetic resonance data of tau obtained in the presence or absence of tubulin [29].

Fig. 2. Microtubule binding and self-assembly of tau.

Schematic representation of the dynamic microtubule interaction of a putative 3D structure of the longest CNS tau isoform (4 R tau) via its microtubule-binding region (blue). Tau binds to a single microtubule protofilament composed of lined-up tubulin dimers (light orange and dark orange). The amino-terminal projection domain (green) protrudes from the microtubule surface upon tau binding. Self-assembly of tau also involves the microtubule-binding region (right). An electron micrograph of a tau filament from an aggregation-prone recombinant tau construct (TauΔK280) is shown below. It shows a single filament structure with a common helical turn period, similar to the cylindrical filaments of paired helical filaments. Structural representations were created using the open-source version of PyMOL (The PyMOL Molecular Graphics System, version 3.1.0, Schrödinger, LLC). The aggregate structure was adapted from the paired helical filament of tau from the primary age-related tauopathy brain model (PDB ID: 7NRQ).

The interaction of tau with axonal microtubules is remarkably dynamic. Tracking single tau molecules in neurons has shown that tau binds to a single microtubule binding site for only milliseconds and binds to 20–30 different binding sites on the same or neighboring microtubules within a second - a behavior we have termed “kiss-and-hop” [30]. This behavior likely explains why tau, despite a comparatively high concentration in the micromolar range, does not affect the rate of axonal transport in vivo [31]. Model calculations assuming a continuous distribution indicate that axonal microtubules are covered with approximately 70 tau molecules per micrometer, corresponding to a mean distance of approximately 15 nm between the centers of each tau molecule [32]. In the case of a less dynamic interaction of tau with microtubules, one would expect at least some steric hindrance to motor protein-dependent transport on microtubules due to the extensive tau coverage.

The short interaction of tau with microtubules is also relevant for another reason. Microtubules exhibit dynamic instability, a process in which self-assembling microtubules stochastically alternate between phases of growth and shortening [33]. In cell-free assays, tau kinetically stabilizes microtubule dynamics by increasing the rescue frequency - the rate at which a depolymerizing microtubule switches back to a growing state [34] - and stabilizes the formation of straight protofilaments [29]. However, it is often claimed in the literature that tau stabilizes microtubules in neurons, although virtually no experimental evidence exists for this. Indeed, the observation that acute or chronic knockdown of tau does not affect microtubule stability in cell culture or in vivo argues against a microtubule-stabilizing function of tau in the axon [35, 36]. It is also known that tau binds preferentially in the distal region of the axon, where dynamic rather than stable microtubules predominate [37, 38]. In fact, it has even been shown in cultured neurons that after tau depletion the stable fraction of microtubules increases [39]. Thus, while tau promotes microtubule polymerization in neurites [30], this does not lead to significant stabilization of axonal microtubules. Therefore, the dynamic interaction of tau with microtubules likely provides a means by which tau can modulate axonal microtubule polymerization without promoting their stability.

Posttranslational modifications play an important role in regulating the tau-microtubule interaction. It has been known from in vitro data for over 40 years that phosphorylation influences tau’s ability to promote microtubule assembly and that dephosphorylated tau is generally more active in this process [40]. Tau possesses over 80 potential phosphorylation sites, and many of these sites that regulate tau’s microtubule interaction are located in regions flanking the microtubule-binding region. These are also the majority of sites where tau exhibits increased phosphorylation in paired helical filaments (PHFs), abnormal filamentous tau aggregates isolated from patients with AD [41, 42]. Phosphorylation can occur through proline-directed protein kinases such as glycogen synthase kinase 3 (GSK3), cyclin-dependent protein kinase 5 (CDK5), p38 mitogen-activated protein kinases (p38 MAPK), and dual-specificity tyrosine phosphorylation-regulated kinase 1A (DYRK1A). Non-proline-directed protein kinases, such as MAP/microtubule affinity-regulating kinase 4 (MARK4), and tyrosine protein kinases, such as members of the SRC family kinase (SFK), are also involved. Given their role in tau phosphorylation, these kinases may also contribute to the development of AD and are therefore considered potential drug targets (for a recent review, see [43]).

Of functional relevance with regard to the interaction of tau with microtubules appears to be tau phosphorylation at the proline-directed Ser/Thr sites targeted by the protein kinases GSK3β and CDK5. In non-neuronal cells, coexpression of GSK3β and tau reduced tau’s binding to microtubules [44] and tau phosphorylation by CDK5 decreased the affinity of tau to microtubules in mouse brain lysates [45], likely due to increased phosphorylation at several sites flanking the microtubule-binding region of tau (particularly phosphorylation at Ser199, Ser202, Ser404, Thr231, Ser235, Ser396, Ser400 and Ser412 [46]). Furthermore, some of the GSK3β-phosphorylation sites are primed by CDK5 phosphorylation, suggesting that the activity of CDK5 would further increase the phosphorylation of tau by facilitating subsequent phosphorylation with GSK3β [47]. However, net phosphorylation is a result of the balance between phosphorylation and dephosphorylation, and protein phosphatase 2A (PP2A) is a major phosphatase of tau and itself interacts with tau and microtubules in a phosphorylation-dependent manner [16, 48], which increases the complexity of the regulation of tau phosphorylation and its influence on the tau-microtubule interaction. While increased phosphorylation of many tau-binding sites by these kinases reduces tau interaction with microtubules in vitro, the functional relevance of altered interaction with microtubules in the axonal compartment is less obvious. Due to the high concentration of tau-binding sites on axonal microtubules, increased phosphorylation appears to have more subtle effects on tau-microtubule interaction. Indeed, a pseudophosphorylated tau construct, in which the complete phosphorylation of a total of 10 disease-associated phosphorylation sites flanking the microtubule-binding region of tau was mimicked by serine/threonine-glutamate mutations (many of them also phosphorylation sites of GSK3β and CDK5), showed only a moderate reduction in the average residence time at a single microtubule-binding site in axon-like processes from 38 ms to 27 ms [28].

This may indicate that tau phosphorylation has a more significant effect on another potential interaction mediated by the microtubule-binding region of tau: tau self-assembly. High-resolution structural data of tau filaments from a patient with AD indicate a filament core composed of protofilaments encompassing the microtubule-binding region and part of the adjacent carboxy-terminal pseudorepeat region, adopting a combined cross-β/β-helix structure [49] (Fig. 2, right). Increased phosphorylation of tau promotes aggregation, at least in vitro [50]. Recently, cell-free experiments have shown that phosphorylation by the proline-directed kinase GSK3β, in particular, promotes the aggregation of full-length tau [51], highlighting the link between regulation of tau-microtubule interaction and aggregation. However, it remains unclear how dynamic microtubule interaction, changes in tau phosphorylation, and self-assembly of tau into dimers, oligomers, and higher-order structures are related in an axonal context, as higher tau aggregates predominantly form in the somatodendritic compartment. Given the presence of extracellular tau and corresponding neuronal uptake mechanisms, incorporated tau from other neurons might also contribute to intracellular tau dynamics. Using live-cell imaging with an aggregation-prone tau mutant reported in cases of tauopathies (TauΔK280), we observed that increased tau aggregation is associated with decreased microtubule interaction (and vice versa when using a tau aggregation inhibitor). Interestingly, self-assembled soluble tau species - and not only larger tau filaments - also showed reduced interaction with microtubules, suggesting a direct link between the extent of tau self-assembly and tau-microtubule interaction [52].

While tau phosphorylation has been the most studied, other posttranslational modifications can also influence tau’s interaction with microtubules and tau aggregation. Acetylation at Lys280 in the microtubule-binding region has been shown to inhibit tau-dependent microtubule assembly and promote tau aggregation [53], acetylation at Lys174 caused tau accumulation and slowed down tau turnover [54], caspase-3-mediated cleavage in the very carboxy-terminal region of tau increased its interaction with microtubules [55], and O-GlcNAcylation appears to differentially modulate microtubule binding and tau aggregation depending on the tau isoform [56], thereby increasing the complexity of the interaction of the different tau proteoforms with microtubules and tau self-assembly.

Tau as a neuronal interaction hub

Tau belongs to a family of closely related structural MAPs that share a conserved carboxy-terminal domain that encompasses the microtubule-binding repeat region [57]. Besides tau, the family includes the neuronal MAP2 and the non-neuronal MAP4. MAP2 and tau are encoded by sister genes derived from a later common ancestor to MAP4. MAP4 emerged from a non-vertebrate ancestor with a similar microtubule-binding region [58]. Together with its sister MAP2, tau evolved by duplication of an ancestral gene from early chordates at the beginning of vertebrate evolution [58] (Fig. 3A). Although tau and MAP2 share an evolutionary highly conserved microtubule-binding region, both proteins exhibit different compartmentalization in neurons: While tau is enriched in the axonal compartment, MAP2 is mainly localized in the somatodendritic area of mature neurons. This suggests a role for the amino-terminal part of both proteins in mediating compartment-specific interactions and functions. The evolutionary origin of tau from a common precursor with MAP2 is reflected in its bipartite structure. While the carboxy-terminal part, containing the microtubule-binding region, is positively charged at physiological pH, the N-terminal projection region is highly acidic (Fig. 3B, top).

Fig. 3. Evolution and bipartite structure of tau.

A Schematic representation of gene duplication at the beginning of vertebrates, which generated the genes for tau and MAP2 from a common precursor. B Bipartite structure of tau in terms of electrostatic potential, predicted disorder values, and functional interactions of the microtubule-binding region (blue) and the projection domain (green). The electrostatic potential surfaces were calculated using the Adaptive Poisson-Boltzmann Solver (APBS) (https://github.com/Electrostatics/apbs-pdb2pqr/releases). The surfaces were visualized in PyMOL as colored electrostatic maps to highlight charged regions. The electrostatic potential is color-coded from red (negative) to blue (positive) at physiological pH. The disorder was predicted using IUPred2A (https://pubmed.ncbi.nlm.nih.gov/15769473/) and indicated by light to dark values, representing low to high disorder prediction, respectively.

Structurally, tau belongs to the class of intrinsically disordered proteins (IDPs) [59]. IDPs contain intrinsically disordered regions that lack clearly defined three-dimensional structures and are characterized by great conformational flexibility, providing a large interaction surface. The bipartite structure of tau is also evident in the degree of intrinsic disorder, which is significantly higher in the N-terminal region of tau compared to the more conserved carboxy-terminal part (Fig. 3B, middle). A look at the changes in disorder during vertebrate evolution is also revealing. While the microtubule-binding region shows no obvious change in disorder, the disorder of the N-terminal region increases sharply from jawless animals to amphibians and mammals [60]. Tau’s sister protein, MAP2, does not show a comparable increase in disorder, suggesting a divergent evolution of these two MAPs.

In addition to microtubules, tau has been shown to interact with many proteins, suggesting that it is involved in various processes (Fig. 3C, bottom). Proven interaction partners are heat shock proteins [61, 62], synaptic vesicle proteins [13, 63], and phospholipid-binding proteins of the annexin family [63, 64]. Systematic information on the tau interactome in human neurons was recently obtained by proximity labeling of induced pluripotent stem cell (iPSC)-derived neurons [65]. More than 240 putative tau-interacting proteins were identified, and almost half of these were detected exclusively in neurons with amino-terminal or carboxy-terminal APEX mapping. This suggests distinct binding partners of the microtubule-binding region of tau and its amino-terminal projection domain. In addition to active zone docking proteins, the N-APEX-tau interactome included proteins involved in postsynaptic organization, suggesting both pre- and postsynaptic functions of tau. This finding is also consistent with data on tau localization, where tau was found in the presynapse, at postsynaptic spines, and also at the plasma membrane (Fig. 1). Nuclear localization was also confirmed by the detection of RNA- and DNA-binding proteins.

Direct information on the specialization of different tau regions has been obtained from the identification of interaction partners, whose binding to specific tau regions has been mapped. The amino-terminal projection region was mostly associated with lipid-binding and synaptic vesicle proteins (an overview of the mapped binding partners can be found in [60]). With regard to tau’s enrichment in the axonal compartment, the binding of tau to annexins may be important. Tau interacts via its extreme N-terminus, encoded by the first exon (E1), with the core domain of AnxA2 in a Ca²⁺-induced open conformation and also interacts with AnxA6 [64], a protein involved in membrane repair of dystrophic neurites [66]. Competition by overexpression of an E1-containing construct reduced tau’s axonal enrichment in primary neurons. The interaction of tau with the peptidyl-prolyl isomerase Pin1 may also be of functional importance under physiological and pathological conditions. Pin1 recognizes phospho-Thr212-Pro213 within the proline-rich region flanking the microtubule-binding region of tau at its amino-terminal side [67]. Dephosphorylation of tau at CDK5-mediated sites is enhanced by Pin1, suggesting an involvement of the Pin1-Tau interaction in the regulation of tau phosphorylation [68]. Interestingly, a tau phosphorylation-dependent interaction with 14-3-3 proteins has also been observed [69]. 14-3-3 proteins belong to a family of phosphoserine- and phosphothreonine-binding proteins and are involved in important cellular processes, such as apoptosis, signal transduction, energy metabolism, and protein transport. This finding indicates that tau modulates crucial cellular signaling pathways in a phosphorylation-dependent manner. Similar to the amino-terminal region of tau, the extensive interactions of 14-3-3 proteins are facilitated by intrinsically disordered protein regions. This suggests a partnership between two intrinsically disordered proteins as pivotal proteins in the modulation of cellular signaling pathways [70]. In contrast, the carboxy-terminal region of tau has been primarily implicated in binding to heat shock proteins, microtubules, and actin filaments, as well as - as previously mentioned - in tau self-assembly.

Thus, the bipartite structure of tau with respect to its charge distribution and the distinct presence of intrinsically disordered regions reflects its bipartite interactome and putative functions. However, it is also important to note that the cellular function of tau remains unclear even after decades of research on tau. So far, only the statement that tau has a relevant function seems certain, because all vertebrates whose genomes have been sequenced so far contain at least one tau gene (zebrafish even have two paralogous tau genes [71]).

It should also be noted that the majority of studies have focused on the low-molecular-weight tau isoforms containing 352 - 441 amino acids, which are expressed in the majority of CNS neurons. In addition, a long tau isoform termed “Big Tau” is expressed in adult PNS neurons and certain regions of the CNS, such as the cerebellum [5]. Big Tau contains an additional large exon (exon 4a), which increases the N-terminal projection region by 251 amino acids in humans. Compared to the N-terminal domain of the low-molecular weight tau isoforms, exon 4 A shows only low sequence homology between species [72]. The enlarged projection region might increase the spacing between microtubules in the axons of PNS neurons, thereby increasing axon caliber and stability. It is currently unclear whether and how the presence of the additional sequence affects the interactome of tau, particularly with regard to its non-microtubule-related functions.

Insights from tau knockdown studies

Analyzing the consequences of tau knockout or knockdown in animal models could help unravel the physiological functions of tau protein. However, studies have yielded inconsistent results, particularly in mouse models with genetic reduction or complete ablation of the MAPT gene (which encodes tau) (Table 1). Based on early cell studies, in which reduced tau expression by antisense oligonucleotides inhibited neurite outgrowth and axon specification in primary neurons, one would have expected that tau knockout mice would be inviable [73, 74]. However, Harada et al. reported a viable tau knockout mouse in the 1990s that exhibited only mild phenotypes such as altered microtubule organization in small-caliber axons [35]. Notably, no significant impairment of axon elongation was observed in the cultured neurons from these mice. Subsequent tau knockout models were also viable and showed only a slight developmental delay in neurite maturation [75].

Table 1.

Overview of the main results of studies with tau knockdown or knockout animals.

Tau Knockout / Knockdown	Cell Culture Type	Key Findings	Reference
tau deficient mice	embryonic hippocampal cultures	Delayed maturation as measured by axonal and neuritic extensions	[75]
tau deficient mice	hippocampal cultures	Axonal extension not affected	[35]
tau deficient mice		Altered microtubule organization in some small-caliber axons	[35]
tau deficient mice (B6.129 × 1-Mapttm1Hnd/J)		Prevention of behavioral symptoms of autism in two mouse models simulating different causes of this disorder	[103]
tau deficient mice (B6.129 × 1-Mapttm1Hnd/J)		Prevention of deficits in spatial learning and memory after repeated mild frontal impact	[130]
tau deficient mice (B6.129 × 1-Mapttm1Hnd/J)		Prevention of deficits in synaptic transmission and plasticity in hAPP-transgenic mice	[77]
tau deficient mice (B6.129 × 1-Mapttm1Hnd/J)		Prevention of stress-driven suppression of neurogenesis and decrease of mTOR)/glycogen synthase kinase-3β (GSK3β)/β-catenin signaling in the adult hippocampus	[110]
tau deficient mice (B6.129 × 1-Mapttm1Hnd/J)		Prevention of modulation of adult hippocampal neurogenesis by stimulation and stress	[107]
tau deficient mice (B6.129 × 1-Mapttm1Hnd/J)		Blocking of amyloid-β-induced cognitive impairments	[76]
tau deficient mice (B6.129 × 1-Mapttm1Hnd/J)		No protection against amyloid-β-induced damage to long-term synaptic plasticity and memory or against amyloid deposition	[78]
tau deficient mice (B6.129 × 1-Mapttm1Hnd/J)	primary cortical neurons	Reduction of α-synuclein phosphorylation by exogenous α-synuclein preformed filaments	[79]
tau deficient mice (B6.129 × 1-Mapttm1Hnd/J)		Reduction in the spread of α-synuclein pathology in vivo	[79]
tau deficient mice (B6.129 × 1-Mapttm1Hnd/J)		No prevention of motor deficits by 6-hydroxydopamine (6-OHDA); no prevention of 6-OHDA-induced loss of dopaminergic nerve endings in the striatum	[80]
tau deficient mice (B6.129 × 1-Mapttm1Hnd/J)	primary cortical neurons	No reduction in the formation of fibril-induced α-synuclein inclusions	[81]
tau deficient mice (B6.129 × 1-Mapttm1Hnd/J)		No reduction in the formation of fibril-induced α-synuclein inclusions	[81]
tau deficient mice (Sv129B/6 tau − /−)		Apparent hyposmic phenotype correlates with α-synuclein accumulation and impaired autophagy in the olfactory bulb of 7-month-old tau knockout mice. α-synuclein accumulation and impaired autophagy also in the substantia nigra of 15-month-old tau knockout mice with motor deficits	[131]
tau deficient mice (Sv129B/6 tau − /−		Increased reflectivity at short wavelengths, determined by retinal hyperspectral imaging, similar to an α-synuclein overaccumulation model	[132]
tau deficient mice (B6.Cg-Tg(MAPT)8cPdav Mapttm1(EGFP)Klt/J)		Severe deficits in LTP, but not LTD; indistinguishable motor skills, exploratory and anxious behavior; however, impaired contextual and cue-based fear conditioning	[82]
tau deficient mice		Selective deficit in LTD	[83]
tau deficient mice (as in Dawson et al., on Bl6/129sv mixed background and Bl6 background)		Reduced tyrosine hydroxylase-positive nigral neurons and impaired motor function	[84]
tau deficient mice (B6.129 × 1-Mapttm1Hnd/J)		Anxiety-related behavior; impaired contextual and cue-based fear memory	[87]
tau deficient mice		Mild motor deficits and reduced tyrosine hydroxylase in the substantia nigra of middle-aged mice. Morris water maze deficits and loss of hippocampal acetylated α-tubulin and excitatory synaptic proteins in aged mice	[85]
bilateral stereotaxic injections in 7-month-old C57Bl6/SJL wild-type mice with AAV-ShRNATau		Impaired motor coordination and spatial memory	[86]
Mapt siRNA injected perineurally at the injury site directly after nerve crush in rats		Suppressed Schwann cell migration after sciatic nerve injury	[88]
tau deficient mice (B6.Cg-Tg(MAPT)8cPdav Mapttm1(EGFP)Klt/J)		Increased proliferation and reduced migration of Schwann cells	[88]
tau deficient mice (B6.129 × 1-Mapttm1Hnd/J)		Reduced response to acute noxious agents; reduced C-fiber density in the sciatic nerve; hypomyelination of myelinated fibers (Aδ fibers)	[89]
tau deficient mice (B6.129 × 1-Mapttm1Hnd/J)		Increased number of degenerating myelinated fibers and reduced conduction properties of the sciatic nerve in aged mice	[133]
tau deficient mice (B6.129 × 1-Mapttm1Hnd/J)		Upregulation of BAF-57, a protein involved in neuron-specific gene repression	[91]
tau deficient mice (B6.Cg-Tg(MAPT)8cPdav Mapttm1(EGFP)Klt/J)		Reduction of anxiety-like behavior and reduced fear expression by aversive conditioning; complex altered brain transcriptome and phosphoproteome with only minor behavioral changes	[92]
tau deficient mice (B6.129 × 1-Mapttm1Hnd/J)		Hyperglycemia and glucose intolerance in early life	[134]
tau deficient mice		Exclusion of murine tau leads to accelerated tau aggregation in old age	[135]

While some studies even suggested positive effects of tau ablation in models of neurological and neuropsychiatric disorders, as well as an alleviation of spatial learning deficits after mild traumatic brain injury, others failed to find a protective effect. For example, some models of amyloid-β pathology found protection against synaptic dysfunction and cognitive impairment [76, 77], while others found no protective effect against amyloid-β deposition or deficits in synaptic plasticity [78]. Similar inconsistencies have been observed in models of Parkinson’s disease. While tau ablation reduced α-synuclein pathology in vitro and in vivo in some studies [79], other experiments reported no benefit in alleviating the loss of dopaminergic neurons or the formation of α-synuclein inclusions [80, 81].

The assessment of neuronal function using electrophysiological recordings has also been controversial. One group observed significant deficits in long-term potentiation (LTP), but not in long-term depression (LTD) [82]. Another study, however, reported opposite electrophysiological findings and showed selective impairment in LTD [83]. Conflicting results have also been published regarding behavioral studies. One study reported intact motor skills, exploratory behavior, and anxiety-like phenotypes [82]. In contrast, several studies showed impaired motor function in tau knockout mice [84–86], while others reported anxiety-related behavior and deficits in contextual and cued memory for fear [87]. A more consistent phenotype across multiple tau knockout models was deficits in peripheral nerve myelination [88–90], which may indicate a robust role for the big tau isoform in the peripheral nervous system.

These seemingly contradictory results are likely due to several factors, including differences in the genetic background of the mouse strains used (e.g., original background vs. C57BL/6 J backcrosses), the age of the mice at the time of study, and the specific method of generating the tau knockout mice. Despite accounting for these variables, inconsistencies remain. The divergent results underscore the complexity of evaluating the role of tau and the need for cautious interpretation of results from tau knockout models. Some discrepancies could also be due to compensatory mechanisms of gene expression in a knockout scenario. Previous research has already suggested a role for tau in regulating gene expression [91] (which would also be consistent with its nuclear localization), and recent transcriptomic and phosphoproteomic analyses of tau knockout mice reveal complex alterations in gene expression and signaling pathways in the brain [92]. It should also be considered that many of these effects may also be species-dependent, and modulation of tau expression in the human brain can have different consequences.

Tau in disease: acquisition of toxic properties

It is somewhat surprising that the knockdown or loss of a promiscuous binder such as tau protein, which interacts with microtubules and components of various signaling pathways, appears to have only subtle effects in a systemic context. This makes it unlikely that tau’s central contribution to the development of neurodegenerative diseases is due to a loss-of-function effect. The literature still frequently claims the presence of a sequence of events whereby tau exhibits increased phosphorylation during disease, leading to a loss of microtubule interaction and, consequently, to microtubule destabilization and disassembly, which then leads to the death of affected neurons. However, evidence for such a sequence of events is lacking for several reasons: first, tau does not appear to stabilize microtubules in a neuronal context, and second, the loss of tau does not significantly impair neuronal survival.

An alternative hypothesis is much better supported experimentally: disease-modified tau develops toxic properties that lead to neuronal degeneration (Fig. 4). Initial evidence for this “gain of toxic properties” hypothesis comes from experiments in which disease-like modified tau was expressed in neuronally differentiated PC12 cells and terminally differentiated human CNS model neurons [93]. In this study, a pseudohyperphosphorylated human tau construct was used that mimics disease-like permanent, highly stoichiometric tau phosphorylation. The tau construct exerted a cytotoxic effect with the induction of apoptotic cell death and caspase-3 activation. Remarkably, toxicity occurred without detectable higher protein aggregates, suggesting that potentially misfolded soluble tau species and not higher filamentous tau aggregates represent the toxic species.

Fig. 4. Gain of toxic properties of tau during neurodegenerative diseases.

Schematic representation of the increase in toxic properties of tau as a consequence of pathological conditions. Tau dysfunction and the formation of soluble tau oligomers are a primary event in tauopathies, followed by the formation of higher tau aggregates, neuronal degeneration, and microtubule disassembly as a result of impaired communication and signaling. The gain of toxic tau properties is initiated by various pathological conditions, including amyloid-beta oligomers or inflammatory processes [136].

Indeed, several studies suggest that dysfunction of soluble tau or soluble tau oligomers represents the most toxic species in tauopathies, far more toxic than larger, fibrillar aggregates. In one study, small toxic oligomers, generated by sonication, rather than stable fibrillar structures, were found to be highly cytotoxic [94]. Furthermore, cognitive, synaptic, and mitochondrial abnormalities were observed when tau oligomers were injected into the brain of wild-type mice [95, 96] and brain-derived tau oligomers from various tauopathies have been shown to influence neuronal function, gene regulation, and disease progression [97]. The toxicity of soluble tau species, which precedes the formation of larger aggregates, is also supported by the finding that neuronal loss in AD patients exceeds the distribution of neurofibrillary tangles [98] and that increased formation of tau oligomers occurs long before the formation of higher tau aggregates [99]. On the other hand, at least in animal models, NFT-bearing neurons have been shown to be functionally intact in vivo [100], suggesting that the simplistic correlation between the development of higher tau aggregates and region-specific neuronal loss is incorrect. Rather, the data suggest that the formation of NFTs could be a rescue mechanism for affected neurons to reduce the amount of toxic soluble tau species by forming biologically inert aggregates.

The hypothesis of a “gain of toxic properties” regarding tau is also supported by several cell and animal experiments in which a reduction in tau levels provides protective effects (see also Table 1). These include very diverse diseases such as amyloid-beta-induced tau pathology [101, 102], behavioral symptoms of autism [103], developmental epileptic disorder [104], stress-related brain pathology [90], and chronic pain [105], suggesting that tau is actively involved in the development and persistence of various diseases and is involved in many cellular processes. For example, increased stress hormone levels have been shown to promote tau secretion and transneuronal tau spread [106], while exposure to chronic pain appears to trigger tau misfolding and aggregation [105].

Tau knockout studies have indicated that tau may also be involved in adult hippocampal neurogenesis (AHN) (Table 1; [107]). While the occurrence and relevance of AHN in humans are still controversial (see, for example, the review by [108]), there is evidence that impairments in AHN may be a feature of early AD [109]. Chronic stress reduced hippocampal neurogenesis in a mouse model in a tau-dependent manner [110], and accumulation of phosphorylated tau in GABAergic interneurons impaired AHN [111], confirming a contribution of tau to hippocampal neurogenesis under physiological and pathological conditions.

Given the diverse roles of tau in neuronal function, it is likely that the acquisition of toxic properties during disease progression is also multifaceted. The most obvious candidate for impaired function is microtubule regulation, as microtubules are the primary interaction partner of tau. However, given the central role of microtubules in various cellular processes, it is difficult to distinguish the primary effects of impaired microtubule interaction of tau from secondary effects mediated by other factors and altered signaling mechanisms. The interaction of tau with microtubules is remarkably dynamic, as demonstrated by the kiss-and-hop interaction with axonal microtubules [30]. Since this dynamic interaction is necessary for tau to not impede microtubule-dependent transport, the acquisition of toxic properties may result in impaired dynamics of microtubule interaction and thus impaired axonal transport. Indeed, cleavage of tau at the extreme carboxy terminus by caspase-3 generates a truncated tau version (TauC3) that exhibits a longer residence time on microtubules, leading to reduced transport efficiency and the development of dendritic atrophy [55]. Interestingly, the proportion of tau cleaved at the caspase-3 site doubles in the hippocampus of senescent mice, and TauC3 is also increased in AD patients. This suggests a toxic gain-of-function mechanism in which posttranslational modification of tau alters the dynamics of tau-microtubule interactions, thus leading to axonal transport defects and neuronal degeneration.

Alterations in the balance of different tau isoforms may also contribute to microtubule-dependent toxicity. The adult human brain contains equal amounts of tau isoforms with four (4 R Tau) or three (3 R Tau) repeats of microtubule-binding domains, which differ by the presence or absence of exon 10 (E10). An imbalance in the 3 R/4 R ratio on either side can cause tauopathies, as known from disease-associated mutations affecting E10 splicing [112, 113]. Indeed, an imbalance of tau isoforms impaired axonal transport in human neurons [114], and modulation of tau isoform imbalance reduced tau pathology and cognitive decline in mice [115].

Several other potential mechanisms of tau toxicity have been identified that are likely related to microtubule-independent tau interactions. These include alterations in tau-dependent postsynaptic targeting of the Src kinase Fyn, which attenuate amyloid-β toxicity [14], alterations in synaptic vesicle association, which cause tau-induced synaptic toxicity [13], and tau oligomer-induced synaptic and mitochondrial dysfunction [116]. Taken together, these data suggest that the increase in tau’s toxic properties is associated with a potential disruption of the diverse tau interactions. These include, on the one hand, interactions of the microtubule-binding carboxy-terminal half (tau-microtubule interaction and tau-tau association) and, on the other hand, interactions of the amino-terminal projection domain (membrane components, synaptic proteins and members of signaling cascades). On the other hand, it is worth mentioning that hyperphosphorylated tau can also mediate antiapoptotic effects by stabilizing β-catenin, a dual-function protein involved in the regulation and coordination of cell-cell contacts and gene transcription [117]. This suggests complex consequences of altered tau interactions that should be considered in drug development.

Finally, it should also be considered that the toxic properties of pathologically altered tau are not limited to individual neurons but can also spread between neurons via multiple, non-exclusive mechanisms (for a review, see [118]). These mechanisms can include unconventional secretion directly across the plasma membrane, which can be facilitated by the interaction of tau with plasma membrane components, as well as conventional secretion mediated by ectosomes or exosomes. Cell-to-cell transfer via tunelling nanotubes has also been described (see Fig. 1). Tau oligomers can then reach the cytosol of the recipient neuron, allowing template-directed seeding into misfolded pathological tau conformations and the induction of cellular toxicity. Such prion-like spread of pathology through protein misfolding might play a role in the patterned disease progression in tauopathies, as originally described by the Braaks [119], and might contribute to the pathological heterogeneity of tauopathies due to distinct structural conformations, known as tau strains [120].

Toward tau-targeted drugs

Tau’s diverse interactions make it difficult to target those interactions that may be most critical for the development of neurodegeneration. Therefore, it may be more promising to reduce the alterations in tau’s posttranslational modifications that occur in pathological conditions. Increased phosphorylation at selected sites (hyperphosphorylation) or other posttranslational modifications that can influence tau’s conformation and its various interactions could play a crucial role. However, general inhibition of important tau kinases such as GSK3, a ubiquitously expressed serine/threonine kinase with a variety of substrates, carries the risk of severe side effects in chronic diseases, as tauopathies would require long-term treatment. For example, the competitive GSK3 inhibitor tideglusib reached phase II clinical trials, but showed no clinical benefit for AD patients under acceptable safety conditions [121]. CDK5 inhibitors are also being investigated, and some preclinical studies have shown promising results [122]. However, these inhibitors are also likely to have side effects, as CDK5 is involved in many neuronal processes.

Another posttranslational modification of tau that may play an important role in disease pathogenesis is tau acetylation. Tau acetylation at Lys174 is elevated early in AD brains, and an acetylation-mimicking mutant slowed tau turnover and induced cognitive deficits in vivo [54]. Inhibition of acetyltransferase p300-induced tau acetylation with the prescription drug salsalate/salicylate increased tau turnover and reversed tau-induced memory deficits. This suggests that targeted inhibition of tau acetylation could be a therapeutic strategy. However, the histone acetyltransferase p300 regulates the transcription of many genes, and general inhibition of p300 is likely to have side effects. Recently, a more specific approach was pursued to target acetylated Lys174 on tau using specific antibodies. Indeed, the treatment reduced tau pathology in a tauopathy mouse model, supporting a crucial role of Lys174 acetylation in disease pathogenesis [123].

Another, more specific, tau-targeting approach could be to prevent the formation of soluble tau oligomers. In this regard, it might be advisable to target the formation of soluble tau aggregates without affecting highly aggregated filaments, which can serve as a biologically inert waste bin for the cell to remove the more toxic soluble tau conformers. Since the same tau region is involved in microtubule interaction and self-association of tau, a drug that inhibits tau-tau binding should, if possible, not impair tau microtubule interaction to avoid undesirable side effects. Using a quantitative live-cell imaging assay, we recently developed a corresponding cellular screening system and were able to identify candidate factors that prevent tau aggregation and restore physiological tau-microtubule interaction [52, 124].

One approach that could also target extracellular tau, which may be involved in tau spread, could be an immunization approach with antibodies targeting various tau epitopes. Several phase II clinical trials have produced mixed results, and antibodies such as semorinemab or bepranemab did not slow disease progression. However, other antibodies are still being investigated [125]. Approaches aimed at reducing total tau expression may also be promising, as the total amount of tau in the temporal neocortex appears to be pathologically elevated in patients with AD [55]. Therefore, normalizing tau levels in patients with antisense drugs could be useful, as this would also have the potential to influence tau within cells. A Phase 1b study demonstrated that treatment of patients with mild AD with the antisense drug BIIB080 was safe and without serious side effects, and reported a reduction in total tau concentration in the CSF of more than 50% [126]. However, it remains unclear how such treatment affects intracellular tau levels and in which brain regions a reduction occurs.

Given the multifaceted role of tau in neuronal function and dysfunction, approaches aimed at long-term downregulation of tau in cells should be conducted with caution, as vital intracellular signaling mechanisms may be affected, leading to adverse effects that extend beyond microtubule-related functions. A second lesson could be to utilize a polypharmacological approach to simultaneously modulate multiple targets such as microtubule dynamics, kinase and phosphatase regulation, and synaptic activity in affected neurons. Last but not least, it should be considered that although tau has been extensively studied in the nervous system, it has also been found in non-neuronal tissues such as the heart, skeletal muscle, lung, kidney, and testis [127]. In mammalian renal epithelial cells, it may influence cytoskeletal organization and renal metabolism [128] and it may contribute to spermatogenesis in the testis [129]. The physiological relevance of tau in these peripheral tissues is largely unknown; however, it cannot be ruled out that a general influence of drugs on tau could also lead to adverse effects in non-neuronal tissues.

Author contributions

LB: Investigation, Writing - Original Draft and Revision; KT: Investigation, Writing - Original Draft; NIT: Investigation, Writing - Original Draft; RB: Conceptualization, Writing - Original Draft and Revision, Supervision. The authors declare that no generative AI or AI-assisted technologies were used in the preparation of this manuscript.

Competing interest

The authors declare no competing interests.