Tau - It’s not what you think

Tau

Tau (see Glossary) was discovered in the 1970s as a protein that co-purifies with tubulin from brain preparations and promotes microtubule assembly in the test tube [1]. Since then, tau has become one of the most studied proteins in the biological sciences because of its relevance to the normal functioning of neurons (and a few other cell types) and because of its relevance to several neurodegenerative diseases [2, 3]. Tau consists of multiple splice variants, over 80 potential phosphorylation sites, and domains that have been documented to interact with various proteins and structures [3, 4]. Tau is considered to be enriched in axons, where it binds microtubules, but tau is also present at lower levels in the neuron’s soma and dendrites, where it is generally not microtubule-bound [5–7]. There is broad consensus that tau is multi-functional, contributing to microtubule-related features of nerve cells such as microtubule-actin interactions, motor-driven transport of organelles along microtubules, and flexural rigidity of microtubules, and also contributing to non-microtubule features of these cells, such as properties of dendritic spines [8–15]. Tau’s many phosphorylation sites are regulated by an array of kinases and phosphatases, which in turn are regulated by signaling cascades relevant to various cellular functions [16–18].

Although it has many functions, tau is best known as a protein that confers stability to the microtubules of the axon. Decades of research have led to the claim that the primary function of tau is to stabilize microtubules in the axon. On the basis of this broadly accepted claim, some disease researchers are even seeking to use microtubule-stabilizing drugs to treat diseases in which tau dissociates from microtubules [19–21]. The idea that tau is a stabilizer of microtubules has its basis in test tube studies and overexpression studies conducted in non-neuronal cells, both of which demonstrate that at sufficient levels and under certain conditions, tau can indeed stabilize microtubules [22–24]. But is this actually the case in the axon?

Despite the prevailing dogma on tau, there have been reasons to be skeptical. Microtubule researchers have long observed that proteins with microtubule-binding sites that may not physiologically stabilize microtubules can stabilize them if overexpressed in cells or if exposed in excess to microtubules in vitro. In addition, it has been known for over two decades that tau is most enriched on the especially labile microtubules that occupy the distal region of the axon and the growth cone [6, 25]. Finally, the normal on-off rate of tau’s association with the microtubules in the axon is 40 milliseconds [26], which seems rather fast for a protein that supposedly stabilizes the microtubule against disassembly. For these and other reasons, a fresh look at tau’s role in microtubule stability in the axon is long overdue.

Microtubule stability in the axon of cultured neurons

In the early 1990s, it was reported that individual microtubules in the axon of cultured rat sympathetic neurons consist of a stable domain toward the minus end of the microtubule and a labile domain toward the plus end of the microtubule [27, 28]. The labile domain was shown to depolymerize rapidly in the presence of nocodazole, a drug that binds free tubulin subunits, while the stable domain was shown to depolymerize much more slowly. These studies took advantage of the fact that post-translational detyrosination (and also acetylation) of alpha tubulin usually accumulates in the more stable microtubule fraction of cells, leaving the more labile fraction richer in tyrosinated (and unacetylated) tubulin (see Box 1). Using immuno-electron microscopy, it was shown that the labile domain labels heavily for tyrosinated tubulin, while the stable domain shows no labeling at all. This domain structure was later also demonstrated with immunofluorescence microscopy, using a strong extraction protocol that enables individual domains to be visualized as they splay away from the axonal microtubule bundle [28]. In neurons treated with nocodazole for just 15 minutes, almost all of the labeled microtubule profiles were no longer present, demonstrating a tight correlation between staining for tyrosinated tubulin and the identity of the microtubule profile as a labile domain. In cultured rat sympathetic neurons, roughly half the microtubule mass of the axon was determined to be stable and the other half labile. The half-lives for depolymerization in the presence of nocodazole were a few minutes for the labile domains and over 100 minutes for the stable domains [27, 29]. Similar findings were later obtained in studies on cultured rat hippocampal and cortical neurons, but with a greater fraction of the microtubule mass being labile [30]. The domain structure of axonal microtubules is important throughout the life of the neuron because each domain, by virtue of its differences in composition, interacts differently with molecular motors, microtubule-severing proteins, and proteins called +TIPs that interact preferentially with the plus end of rapidly elongating microtubules. Individual microtubules in adult neurons also consist of stable and labile domains, as was shown several years ago in studies on intact nerves from animals [31].

The current tau dogma would predict greater levels of tau associated with the stable domain of the microtubule compared to the labile domain, with the stable domain becoming less stable when tau is depleted. However, when a tau antibody was microinjected into cultured neurons to sequester tau in the cell body prior to axon outgrowth, the microtubules in the tau-free axons that rapidly grew from these neurons were no more less stable than the microtubules in control axons [32]. Despite this work calling into question tau dogma, only recently have there been follow-up studies, with the results suggesting that axonal microtubules become more stable in the absence of tau.

Tau and microtubule stability

A new study [33] set forth to ascertain the effects of depleting tau on axonal microtubule stability in cultured rat cortical or sympathetic neurons using more quantitative and multi-faceted approaches than those used in the earlier work [32], with special attention to the effects on stable and labile microtubule domains. After four days of tau depletion by pan-tau-isoform RNA interference, the axons were not markedly different in their length or morphology from control counterparts, but contained less microtubule mass, per unit length of axon. Tau dogma would predict that this microtubule loss was due to destabilization of the microtubules, meaning that the stable domains of the microtubules become less stable. However, quantitative immunofluorescence revealed that the tau-depleted axons contained higher levels of detyrosinated tubulin, per unit length of axon, than controls axons, despite having less microtubule mass. In addition, far fewer labile (i.e., tyrosinated-tubulin-rich) domains of microtubules can be visualized in tau-depleted axons relative to control axons. These observations suggest that the microtubule diminution after tau depletion is not from stable domains becoming less stable but rather is due to the partial loss of the labile domains. In fact, the remaining portions of the labile domains become more stable as tau is lost.

The kinetics of microtubule loss in the presence of nocodazole were consistent with the results on microtubule detyrosination. More microtubule mass was stable in tau-depleted axons than controls with the remaining portion of the labile fraction not as labile (i.e., more stable) than the labile fraction of controls. Live-cell imaging of rates of microtubule assembly also revealed that that labile domains had become less dynamic (i.e., less labile) than control counterparts.

Live-cell imaging also revealed dramatically fewer microtubule assembly excursions in tau-depleted axons compared to control axons. Each stable domain should theoretically be able to assemble a new labile domain from its plus end, and there are no fewer stable domains after tau depletion than before. There would only be fewer assembly excursions if tau has an additional role in not only protecting the labile domain from being stabilized but also in promoting its assembly. In fact, the ability of tau to promote microtubule assembly was its original hallmark [34], even before the notion of it stabilizing microtubules became so prevalent.

When studying microtubule dynamics in non-neuronal cells or in vitro with limited accessory proteins present, the stability of the microtubule is often defined as how long it assembles before it undergoes a bout of disassembly. Thus, one might wonder, in considering these results, if tau provides some level of stability to the labile domain in the sense that tau enables the labile domain to grow longer than it otherwise would in the absence of tau. However, in the absence of tau, the labile domain is shorter, more detyrosinated, less dynamic and less sensitive to nocodazole. Thus, by these criteria, the labile domain is more labile in the presence of tau than in the absence of tau, which disqualifies tau’s influence on the labile domain as a stabilizing effect. The reason for this, as explained later, is that in the absence of tau, other proteins can influence the labile domains of the microtubules in ways that these proteins do not when tau is present (Figure 1A).

Tau’s distribution on axonal microtubules

If it is keeping the labile domain labile, tau should theoretically be concentrated on the labile domain. It was known for decades that tau is enriched on the labile domains of microtubules that extend into the distal region of the axon, but whether the same is true of microtubules along the axon had yet to be explored. Visualizing this directly is problematic because stable and labile domains are staggered along the length of the axon, with individual microtubules taking on a range of lengths, and because the microtubules are spaced so closely together that individual microtubule domains cannot be resolved by conventional epi or confocal fluorescence techniques. As an indirect approach, it was reasoned that if there is more tau on labile domains than stable domains, then the ratio of microtubule-associated tau to microtubule mass in the axon should be lower after depolymerization of the labile domains with nocodazole. Indeed, such studies revealed that there is at least twice the density of tau on the labile domains compared to the stable domains. Moreover, newly assembling labile domains arising after rinsing out the nocodazole were especially rich in tau. Collectively, these findings are consistent with tau rapidly binding and concentrating on the labile domain of the microtubule as it assembles, with far less tau on the stable domain (Figure 1B).

How does tau prevent labile domains from being stabilized?

Might tau be enriched on labile domains to prevent genuine microtubule-stabilizing proteins from binding to the labile domain? If this is true, tau is not a stabilizer of the microtubule but a preventer of microtubule stabilization by other such proteins. Several years ago, it was reported that certain microtubules in lamellar regions of the axon are rich in doublecortin but deficient in tau, suggesting that these two microtubule-associated proteins may somehow exclude one another from the microtubule lattice [35]. However, total doublecortin levels in the axon are probably insufficient to account for a major shift in microtubule stability when tau is depleted, and indeed, investigation into this matter showed that doublecortin is not responsible for the greater stability of microtubules in the absence of tau [33]. Moving on from doublecortin, MAP6 was considered a better candidate for conferring stability on tau-depleted axonal microtubules, because, unlike tau, MAP6 is richer on stable domains of axonal microtubules than labile domains [25] and because a number of studies had already documented that MAP6 has the appropriate properties to stabilize microtubules in cells [36]. In addition, MAP6, originally known as STOP (“stable tubule only protein”), was shown many years ago to slide along the lattice of the microtubule [37]. Most of the MAP6 in cells is not microtubule-associated but becomes microtubule-associated when cells are exposed to cold, in order to make the microtubule stable to cold-induced depolymerization [38]. Might the axon respond to tau depletion similarly to how it responds to cold by inducing greater levels of MAP6 to bind to the labile microtubule domains, perhaps in part by sliding out from the stable domain to bind to the labile domain? In support of this idea, not only was there more MAP6 bound to microtubules because of tau depletion, there were also higher levels of total MAP6 in the neuron [33].

Like tau depletion, depletion of MAP6 by siRNA resulted in loss of axonal microtubule mass. However, virtually everything else that was assayed in these experiments gave the inverse of the results of tau depletion. There was a diminution in the stable microtubule fraction, with the labile microtubule fraction becoming more labile (i.e., less stable) than control counterparts [33]. However, MAP6 cannot be the only microtubule stabilizer in the axon because the remaining portion of the stable microtubule fraction was no less stable after MAP6 depletion. A long list of other proteins has been implicated with microtubule stability in the axon [39], but whether any of the others share MAP6’s inverse relationship with tau remains an open question. One provocative aspect of these results is that loss-of-tau phenotypes in neurons might be just as attributable to more MAP6 binding to the microtubules as less tau.

Relevant to the possibility that MAP6 is not alone in its inverse relationship with tau is recent work on a microtubule-associated protein called MAP7 (also called ensconsin), which is known to interact with kinesin-1, drawing it to the microtubule to enhance motile events along the microtubule [40, 41]. These studies demonstrate competition between tau and MAP7 for the microtubule lattice wherein tau binds abundantly first and then is gradually displaced by MAP7 [42]. MAP7 might be a genuine microtubule-stabilizing protein [40, 41], but even if it is not, these findings suggest a pattern (which includes doublecortin as well) by which tau impedes the binding of various different microtubule-associated proteins (with different microtubule-binding domains) to the microtubule lattice.

A new model

Understanding tau’s role in regulating the stability properties of axonal microtubules depends on the inverse role that tau has with genuine stabilizing proteins, such as MAP6. This kind of relationship would not likely be recapitulated in studies involving overexpression of tau in non-neuronal cells or in test tube studies – unless of course the experiments were designed specifically to test for such a relationship. In the context of the axon, it seems reasonable to posit a new model in which tau enables microtubules to assemble long labile domains while preventing them from being stabilized by proteins. How might this work? The competition between tau and genuine stabilizers probably cannot be explained by different affinities for the same binding sites on the microtubule, because tau and these other proteins do not share the same kinds of microtubule-binding sites [43, 44]. It may be that tau’s expression levels in neurons and tau’s strong affinity for microtubules in the axon combine to overpower the capacity of genuine stabilizers to accumulate in sufficient numbers to stabilize the labile domain. When the labile domain assembles, tau immediately interacts with the lattice, associating and dissociating at a rapid rate, and this keeps down the number of molecules of MAP6 and other potential stabilizers that can bind to the labile domain. When a signaling event causes phosphorylation of tau’s microtubule-binding sites, less tau is available to associate with the microtubule and hence genuine stabilizers such as MAP6 can then associate with the labile domain in higher numbers, rendering it more stable. This could be a partial stabilization, or it could lead to the labile domain becoming stabilized to the point of becoming part of the stable domain. Once the genuine stabilizers get a foothold, then tau becomes inefficient at outcompeting them, such that the stable domain does not revert to being labile. Interestingly, this model for tau and MAP6 (shown in Figure 2) is similar to a scenario described many years ago for MAP1b in which a phospho-variant of MAP1b binds to microtubules in the distal region of the axon to keep the non-phosphorylated version of MAP1b from stabilizing the microtubules [45]. Sequestration of MAP1b in the cell body of the neuron by the identical antibody injection approach as described earlier for tau demonstrated destabilization of microtubules in the distal region of the axon [46], consistent with a role for MAP1b as a genuine microtubule stabilizer.

Taken together, these results are relevant to emerging ideas about how various proteins such as traditional microtubule-associated proteins and molecular motor proteins might compete for association with the microtubule. These ideas, based mainly on studies conducted in vitro, suggest that the interaction of a particular protein with the microtubule might alter its lattice in such a way as to make it more or less amenable to the binding of other proteins [42]. This process, called “lattice gating,” can also make the lattice more amenable to additional binding of the same protein, leading to high densities of that protein on discrete regions of the microtubule. This has been shown for tau, with vivid demonstration of tau “condensations” on individual microtubules [47].

Relevant also to these ideas are ground-breaking studies recently published at the near-atomic level, demonstrating via cryo-electron microscopy that tau binds along microtubule protofilaments, tethering together tubulin dimers [48]. The new model discussed here suggests that, at least in the context of microtubules in the axon, these interactions probably promote microtubule assembly without stabilizing the microtubule, perhaps because of the rapid on-off rate of tau. It is not difficult to fathom, however, that such tethers could stabilize the microtubule under less physiological conditions, thus leading to misleading results in overexpression studies.

Adult neurons, disease and therapy

The most urgent next step is testing whether tau’s role in adult neurons is the same as indicated by the work on developing neurons in culture. There are differences that may be relevant, such as developing neurons express almost exclusively the 3R isoform of tau, which has three repeats of the microtubule-binding domain, whereas adult neurons also express the 4R isoform, which has four repeats of the microtubule-binding domain [49, 50]. There is also greater complexity to the phosphorylation status of tau in adult neurons, as well as the levels and complexity of other microtubule-related proteins with which tau may have inverse or redundant effects. In addition, features of the microtubules themselves might cause them to interact differently with tau and these other proteins; for example, polyamination of tubulin (which itself confers stability to microtubules) (See Box 1) is sparse in developing neurons but abundant in adult neurons [51]. Even so, the abundance of microtubule stabilizers in adult axons puts an even higher onus on tau to maintain a robust labile microtubule fraction. Some support for this view comes from work on post-mortem brain tissue from humans with tauopathies, in which the microtubule mass is diminished relative to normal brain samples, but more highly detyrosinated and/or acetylated [52–55]. This is consistent with the expectations from new findings discussed here, and the opposite of the expectations of tau dogma.

In tauopathies, the detachment of tau from the microtubule is generally considered to cause the loss of microtubule mass, as opposed to the toxic properties of the abnormally phosphorylated tau. Interestingly, when tau expression is experimentally reduced in adult mouse brain, the mouse displays memory loss and synaptic dysfunction, suggesting a major role for tau in maintenance of these brain functions [56]. This may suggest that the reduction and partial stabilization of labile domains of axonal microtubules is causative of these symptoms in patients with tauopathy. A popular idea in recent years has been that the microtubule loss in Alzheimer’s disease and other tauopathies can be corrected by treatment with low levels of microtubule-stabilizing drugs [21]. While preclinical studies using such drugs on mouse disease models have produced some encouraging results, the studies discussed here on tau suggest that the mechanistic basis for such an approach needs re-thinking. Figure 3 shows one possible cascade (but not the only one) for how this might occur, starting with beta amyloid. Restoring lost labile microtubule mass may prove to be a more appropriate strategy than stabilizing microtubules. Whether low levels of microtubule-stabilizing drugs can substitute for tau in this regard remains to be seen. Germ line tau knockout mice are interesting to think about, in this context, because they appear to be relatively asymptomatic until old age [57], which suggests that compensatory mechanisms have prevented dramatic abnormalities in microtubule stability or tyrosination status from taking place in these animals, at least until old age. Perhaps understanding how other proteins compensate for the absence of tau in the younger tau knockout mice might prove useful toward the development of therapies for tauopathies. Along these lines, proteins that normally pare back labile domains of microtubules in the axon (Box 2) may be good targets for future drug discovery [30, 33, 58–63]. Another possibility is to pursue therapies that promote the binding of tau to microtubules, such that even under compromised tau conditions, available tau can be recruited to restore the labile microtubule domains [64, 65].

Concluding Remarks

It is remarkable that after hundreds of research papers from investigators around the world, there is still much to learn about tau, with dogma waiting to be challenged and surprises yet to be revealed. Questions remain (see Outstanding Questions Box) such as whether similar results to those of the cell culture studies discussed here will be found in adult neurons in vivo. Tau’s distribution on the stable and labile domains of the microtubules needs to be visualized directly, without the use of drugs, presumably by immuno-electron microscopy and/or expansion microscopy. Lattice gating as a mechanism to account for the competition between tau and genuine stabilizers makes sense but will require testing with quantitative techniques of greater resolution. Perhaps most pressing is what all of this means for treatment of tauopathies such as Alzheimer’s disease.