Tau oligomerization on microtubules in health and disease

When it comes to the microtubule-associated protein (MAP), tau, an internet search or a typical review article reveals a central dogma about its role in human disease. The dogma generally states that under physiological conditions, tau binds along the microtubule lattice, providing stability to the polymerized microtubule structure¹. However, in disease, hyperphosphorylation (~3–4 fold increase in phosphorylated residues^2,3) of tau results in its dissociation from microtubules, causing destabilization and subsequent loss of microtubules within diseased neurons. Subsequently, the dissociated hyperphosphorylated tau self-associates into filamentous aggregates that are the hallmarks of advanced tauopathies and neurodegeneration¹. Early work showed that tau alters the parameters of microtubule dynamic instability, reducing catastrophes and slowing depolymerization rates, implying tau acts to stabilize microtubules^4,5. Additionally, microtubule density within axons of tau knockout mice appeared lower. However, the magnitude of this effect was variable between mouse lines (~12–40%), and limited to the proximal regions of only small caliber axons⁶. Microtubule density in dendrites was not affected by loss of tau. Nonetheless, the prevailing notion of tau as a microtubule stabilizer in neurons has been widely accepted in the literature. However, it is important to stress that tau binding does not abolish dynamic instability. In addition, subsequent studies have challenged this dogma, revealing that tau is not the potent stabilizer it was once thought to be^7,8. We note here that the definition of “stabilize” may not be uniform in the literature, but the term is typically taken as counteracting the loss of total microtubule mass in neurons. In this respect, the definition of tau as a ‘stabilizer’ of microtubules would mean that the failure of tau activity would result in the overall loss of microtubule polymer within cells. It is worth noting that neurons contain many other MAPs that interact with, and stabilize or alter, microtubule dynamics, raising further questions about why the specific loss of tau would lead to overall microtubule destabilization in cells⁹. While this topic has been comprehensively reviewed elsewhere¹, it raises a crucial question: if the primary function of tau is not to stabilize microtubules, then what is it? From a microtubule biologist’s perspective, in order to understand the conversion of tau to its pathological state, one must first understand the normal physiological behaviors of the molecule on microtubules. Here, we review our recent work on the behaviors of tau on microtubules and the potential roles it may play in neurons. We also speculate as to how these processes may go awry when tau forms pathological oligomers.

Cooperative assembly of tau on microtubules

The interaction of tau with microtubules has been intensively studied since its discovery in the 1970s¹⁰. Early studies using classical biochemical methods revealed non-cooperative binding to microtubules ^11,12. However, evidence of a more complicated interaction has also been reported¹³. In contrast to bulk biochemical experiments that report averaged molecular behavior, we and others have used single molecule fluorescence to study the dynamic behavior of tau molecules associated with microtubules^14–16. Findings from these studies proposed that tau binds cooperatively to the microtubule lattice via two modes: one driven by self-association between tau molecules and one dictated by through-lattice allostery.

Tau exists in two kinetically distinct states on the microtubule lattice. Individual molecules of tau electrostatically interact with the negatively charged C-terminal tails (CTTs) of tubulin, allowing for rapid diffusion of tau along the lattice¹⁴. Tau can also self-associate and form more stable clusters of molecules that envelop the microtubule surface both in vitro and in vivo^14–17. These clusters have been termed ‘condensates’, ‘islands’ and ‘envelopes’ and we refer to them as envelopes herein. Tau envelope formation requires the pseudo-repeat domain located C-terminal to the classically defined microtubule-binding repeats (MTBRs). There are six isoforms of tau, and although these studies were primarily performed with the longest isoform, 2N4R, alternative splicing of tau does not appear to drastically alter envelope formation¹⁴. A tau construct that lacks the MTBRs but contains the pseudorepeat region does not bind robustly to microtubules alone, but co-enriches with wild-type 2N4R tau envelopes on the microtubule¹⁴. Therefore, tau self-association is not only a pathological phenomenon, but also occurs under physiological conditions along microtubules^17,18. Intriguingly, self-association is not the only driver of tau oligomerization on the microtubule. We found that the binding of tau to the microtubule also alters the underlying tubulin lattice structure, making the lattice more hospitable to the binding of adjacent tau molecules¹⁷. The microtubule lattice can exist in an expanded or compacted form, which has classically been defined by the nucleotide bound to the exchangeable site in β-tubulin, with GTP lattices being expanded and GDP lattices being compacted^19,20. It is now clear that motors and MAPs recognize, and can even induce, either a compacted or expanded lattice state. For example, kinesin-1 and CAMSAP binding drives expansion of a compacted GDP lattice^21–23, while tau and MAP2 association induce and maintain lattice compaction¹⁷. This dual mode of cooperativity therefore allows tau to stake out territory on the microtubule and protect the underlying compacted lattice state. Importantly, heterogeneity of tubulin nucleotide state, defects in the lattice, and microtubule lattice spacing has been reported in vivo, posing the hypothesis that MAP binding may be one way to control the compaction state, and therefore downstream effector interactions with the lattice^24–28.

The Role of Tau Envelopes: Competition with MAPs, Motors, and Severing Enzymes

Tau is a member of a MAP family that also includes MAP2 and MAP4. We have reported that, unlike the primarily neuronal MAPs, tau and MAP2, the ubiquitously expressed MAP4 does not form envelopes along an expanded microtubule lattice¹⁷. Tau and MAP2 are thought to have arisen from a duplication of MAP4 during early vertebrate evolution²⁹, raising an important question: why did tau and MAP2 evolve the ability to form envelopes that can induce a compacted lattice state? One reason may be to delineate access to specific subsets of microtubules by orthogonal MAPs, and therefore regulate processes such as microtubule motor transport and microtubule severing (Figure 1A–C). Both in vivo and in vitro work has demonstrated that tau competes with MAPs such as MAP7, MAP6, and MAP4^7,17,30,31 (Figure 1C). These results suggest that MAP competition may play a role in controlling the distribution of MAPs within cells. Cooperative assembly of MAPs into envelopes provides an elegant molecular mechanism to spatially bias where specific MAPs perform their functions within a dense cytoplasmic environment.

Figure 1: Roles of tau envelopes on microtubules.

(A) Model of multiple 2N4R tau molecules binding to microtubules and inducing compaction of the underlying lattice structure. Purple tubulin heterodimers indicate a compacted lattice, while grey/black tubulin heterodimers indicate an expanded lattice. Tau envelopes inhibit kinesin-1 motility on microtubules, while the dynein-dynactin complex is more effective at navigating tau envelopes. Models based on alphafold predictions for 2N4R tau (aa 1–441) and KIF5B (aa 1–880), and PDB structure 7Z8F for dynein-dynactin-BicDR⁴⁷. (B) Model of a 2N4R tau envelope inhibiting the microtubule severing activity of spastin. Spastin is excluded from the tau envelope and its severing activity is spatially restricted to areas of the microtubule lacking tau. Model based on alphafold prediction for 2N4R tau (aa 1–441) and PDB structure 6PEN for spastin⁴⁸. (C) Model depicting a MAP envelope composed of 2N4R tau molecules and MAP2c, which also induces a compacted lattice, but excluding MAP4 and MAP7^17,30. Model based on alphafold predictions for 2N4R tau (aa 1–441), 5R-MAP4 (aa 1–979), 3R-MAP2 (aa 1–467), and MAP7 (aa 1–749).

Several studies have shown that tau is a general inhibitor of motor transport, reducing the landing rate for all transport motors: kinesin-1, kinesin-2, kinesin-3, and the processive dynein-dynactin complex^{14,15,30–34} (Figure 1A). The effects on kinesin motors are more pronounced than on the dynein motor, due in part to the overlapping binding sites of tau and kinesin motor domains on the microtubule lattice¹⁴. In contrast, the much smaller dynein microtubule binding domain does not overlap with that of tau, enabling the dynein-dynactin complex to navigate tau envelopes more effectively¹⁴. In addition, kinesin-1 prefers an expanded microtubule lattice, opposite to that induced by tau envelope formation^17,21. Therefore, microtubules that are enveloped by tau are likely designated for purposes other than anterograde transport.

Tau envelopes also effectively inhibit microtubule severing by spastin and katanin^14,16,35, indicating an additional physiological function of cooperative envelope formation in spatially regulating microtubule turnover (Figure 1B). Indeed, missorting of tau from axons leads to spastin-induced microtubule disruption, a potential contribution to tau pathology^35,36. Therefore, we propose that cooperative tau envelope formation provides a molecular mechanism for spatially segregating the activities of a diverse range of orthogonal MAPs in neurons. It is logical to conclude that disruption of normal tau envelope formation in cells could have pathological implications by interfering with the precise localization and therefore function of other microtubule effectors.

Role of Phosphorylation in Tau Function and Pathology

Phosphorylation of tau is clearly correlated with tau aggregation in tauopathies^37,38, but the effects of site-specific phosphorylation on the physiological functions of tau remain unclear. Further studies are necessary to determine if hyperphosphorylated tau completely dissociates from microtubules as stated in the dogma of the field, or if hyperphosphorylation tunes the cooperative binding behavior of tau. The effects of phosphorylation on tau envelope formation are currently unclear, but we note that substantially more baculovirus-expressed tau, which is known to be phosphorylated, is necessary to form envelopes in vitro as compared to nonphosphorylated tau expressed in bacteria^14,16. This, combined with prior work, indicates that phosphorylation may affect the affinity of tau for microtubules, but considering the high concentration of tau that has been reported in the axon (~ 3–6 μM)³⁹, it is less clear that phosphorylation would completely abolish tau binding to microtubules in neurons. It could be that phosphorylation tunes microtubule affinity or possibly the cooperativity of tau envelope formation to alter the dynamics of tau envelopes within cells. Sequestration of tau into aggregates or filaments may be another pathway for tau dissociation from microtubules under pathological conditions.

Understanding Tau Function in Pathology

As stated above, the general dogma has been that tau dissociation from microtubules leads to a decrease in microtubule mass within neurons, leading to general disruptions of microtubule-based processes and eventual cell death. If tau does indeed entirely dissociate from microtubules either due to hyperphosphorylation, sequestration into aggregates or filaments, or due to the competition observed between MAPs, it is likely that other abundant MAPs within the neuron would bind and stabilize the microtubule, and designate those tracks for other purposes^7,30. For example, MAP7 may bind in tau’s absence and designate that microtubule for anterograde cargo transport by kinesin-1³⁰, thereby interfering with the balance of axonal transport. It is therefore difficult to imagine the loss of tau from a microtubule would lead to complete depolymerization of that microtubule when numerous other stabilizing MAPs with high microtubule-binding affinities are present in axons (MAP7, MAP1A, MAP1B, MAP4)⁹. Even if tau is not a potent microtubule stabilizer, it is possible that loss of tau envelopes along the microtubule could lead to increased severing by katanin or spastin. It has yet to be tested how the other abundant MAPs affect severing by these enzymes, but it is possible that tau offers a protection that other MAPs do not. However, prior work has revealed that spastin and katanin severing actually increases microtubule mass by producing seeds from which new microtubules grow^40–42. Therefore, increased microtubule severing may paradoxically increase microtubule mass in the absence of tau. This line of logic suggests that a potential role of tau envelopes is to in fact limit microtubule mass in neurons (by limiting severing), in direct opposition to the current dogma!

We must also consider how pathogenic tau filaments might alter microtubule-based processes. First, tau filamentation may act as a ‘sink’ to deplete tau monomers from the cytosol. Since envelope formation is concentration dependent¹⁴, this may negatively impact tau envelope formation on microtubules. Secondly, it is still unknown if tau filaments bind microtubules. Cryo-EM structures of tau filaments purified from brain tissue from patients with different tauopathies reveals exposed microtubule-binding repeats, along with the N- and C-terminal flanking regions that, due to the disordered and flexible nature of these regions, may still be capable of binding to microtubules, albeit with lower affinity than non-aggregated tau^43–46. Since tau filaments are made from thousands of copies of tau, tau filaments might be expected to be highly multi-valent microtubule binding structures. If tau filaments do indeed bind to microtubules, there are a plethora of new research avenues to explore. It seems logical to hypothesize that tau filament association with microtubules could be a pathological event that disrupts microtubule homeostasis leading to cytotoxicity. Additionally, how might filament binding differ from cooperative tau envelope formation and would filaments affect the compaction state of the lattice? How might tau filaments differentially affect motors, severing enzymes and other MAPs, including non-aggregated tau? We suggest that it should be determined if tau filaments bind microtubules, then a comprehensive analysis must be performed. Despite decades of research, tau continues to stymie efforts to fully elucidate its physiology and pathology. The discovery of cooperative tau envelopes provides an exciting new framework in which to re-examine tau biology as it relates to the microtubule cytoskeleton, which we believe may be the key to understanding both tau physiology and pathology.