van Eersel J, Ke YD, Liu X, Delerue F, Kril JJ, Götz J, Ittner LM.
Sodium selenate mitigates tau pathology, neurodegeneration, and functional deficits in Alzheimer's disease models.
Proc Natl Acad Sci U S A. 2010 Aug 3;107(31):13888-93.
Please login to recommend the paper.
To make a comment you must login or register.
I am very enthusiastic about the paper by Ittner et al. for several reasons. First, it confirms the highly protective effects of tau reduction we observed in hAPP-J20 mice (Roberson et al., 2007 and Palop et al., 2007) in another APP transgenic line with a solid AD-like phenotype and on an independent tau knockout strain. As in our lines, tau reduction rescued memory and longevity in APP23 mice without changing Aβ levels or plaque loads. This kind of reproducibility underlines the robustness of the tau reduction effects and is reassuring to me, especially in light of a recent report suggesting that tau ablation changes Aβ levels and plaque loads in opposite directions and has adverse effects in the Tg2576 model (Dawson et al., 2010).
Second, while the biological functions of tau have so far been explored primarily in axons, Ittner et al. discovered an interesting new mechanism by which tau may modulate synaptic function and neuronal excitability in dendrites. Third, this mechanism involves the tyrosine kinase Fyn, which we showed sensitizes APP transgenic mice to Aβ-induced neuronal, synaptic and cognitive deficits (Chin et al., 2004 and Chin et al., 2005). Fourth, consistent with our hypothesis that tau reduction protects against Aβ by preventing neuronal overexcitation (Roberson et al., 2007), targeted perturbation of the NR/PSD-95 interaction, which prevents excitotoxicity, also prevented premature mortality and memory deficits in APP23 mice.
Taken together, these findings strongly suggest that modulating tau, its interaction with Fyn, or key proteins involved in or affected by this interaction may be of therapeutic benefit in AD. It remains possible, though, that other mechanisms also contribute to the excito-protective effects of tau reduction, including changes in presynaptic terminals or in the axonal transport of cargoes supporting synaptic functions.
In this manuscript, Ittner and colleagues showed that tau has a role in Aβ toxicity, which may be different from the role of tau on microtubules. Interaction of tau and Fyn is required for stabilizing the NR2/PSD95 complex. Reduction of tau, or interfering with the interaction of tau and Fyn, rescued the premature death and memory deficit in the APP Tg mouse. The results are very interesting, and suggest tau as an attractive drug target for AD therapy.
The physiological role of tau has been thought of as microtubule stabilization. However, the tau gene-deficient mouse did not show much evidence of brain dysfunction. Recently, the results of crossbreeding tau-deficient mice with the GSK3β overexpression or the APP overexpression mouse were reported. Reduction of tau level rescued both the impairment of LTP caused by GSK3β overexpression, and the memory deficits caused by APP overexpression (Gomez de Barreda et al., 2010; Roberson et al., 2007). These reports and the paper by Ittner et al. suggest that tau may have some roles in the synapse in addition to stabilizing microtubules. The extent of memory impairment in APP Tg mice depends on hippocampal LTP level. Tau is involved in synaptic plasticity, and deficiency of tau rescues LTP in the APP Tg mouse. I am not sure whether the destabilization of NR2/PSD95 in the tau knockout mouse can explain the attenuation of LTP level in the APP Tg mouse.
There are two different tau-deficient mouse lines. One is a simple tau gene knockout, and the other involves the replacement of exon 1 of the tau gene by EGFP cDNA. The former two reports used the tau gene knockout mouse, and Ittner used the EGFP mouse as a tau-deficient mouse. We may need to give careful consideration to this difference.
The important result by Ittner et al. that post-synaptic targeting of the Src kinase Fyn depends on tau should also be relevant to p75-mediated Aβ toxicity. The observed prevention of Aβ toxicity in APP23 mice with absent or truncated tau could, in part, be due to diminished p75 activity since Src kinases are required for p75 activation by Aβ aggregates (Egert et al., 2007).
I agree with Lennart Mucke, Akihiko Takashima, and Michel Goedert that this is a major opus by Ittner and Goetz and coworkers, and will become seminal in the long-standing question of how amyloid and tau are related to each other in the pathogenic processes in AD. The amyloid-tau relation is central by definition, as well as pathologically diagnostic for AD. Moreover, I approach the age where the matter becomes personally more and more important to be solved sooner rather than later. The issues at hand have separated "baptists" and "tauists" for too long, and for no apparent reason. I, at least, have adhered to both convictions over the last 20 years without too much negative consequences. I therefore welcome the Ittner study also in this respect.
Whether Fyn is "the" missing link in AD needs, and deserves, careful consideration, but this study will undoubtedly impact the field for some time to come. The data presented were dug out of an impressive number of cellular and mouse models by a wide range of technologies. Typical for the better studies is that they stir up more questions and have many implications that I have yet to come to full terms with, given the mass of finer details uncovered by these findings. Here, I planned to restrict myself to aspects and questions closest to our own scientific interests, i.e., the synaptic effects of tau and its phosphorylation in vivo (by GSK3 mainly), and the relation to amyloid in various mouse models, as we recently reviewed (Jaworski et al., 2010).
First, the evidence for NR2b phosphorylation by Fyn is convincing, but I wondered about the distinction between "electric" signals at the synapse (graphically not too well depicted in Fig. 7), namely, the classic ESPCs that are normal, as opposed to the "excitotoxic" signals that are strongly affected. What are these latter signals, and what is their function in wild-type mice under normal physiological conditions? These are described for wild-type mice in Figure 7, panel A—but perhaps the caption for that panel should be "APP23 mice." The data imply at least two pools of NR2bs in the same synapse, containing sub-subtypes, truncated, or otherwise modified or adapted subunits, extra-synaptic or tethered.
Teleologically, the most active synapses that produce the most Aβ peptides (for the sake of simplicity, I used the term "Aβ" throughout to indicate all molecular forms and complexity of all the amyloid peptides) must then also be the most vulnerable to this novel mechanism. How do active synapses counteract the inherent Aβ-mediated excitotoxicity? Is this the price to pay for an "LTP-ed" synapse?
A most mind-troubling issue, even an enigma, for me, is why the protein tau remains labeled by some of the most ardent tauists as "an axonal protein," c.q. specific axonal marker? Because we are interested in the role of tau in those brain regions that are struck in AD, we deal with glutamatergic synapses that are located on dendritic spines. I permanently force my Ph.D. students and coworkers to search the literature for evidence that tau is present not only pre-synaptically in axons, but also post-synaptically in dendritic spines—so far unequivocal proof is lacking. We do see, without many problems, mouse protein tau in dendritic spines in primary neuron cultures, and definitely in transfected neurons expressing human tau. Apparently, extending this to mouse brain in vivo, i.e., demonstrating that tau in spines in wild-type mouse brain sections, is technically (too) demanding, but must be done sooner than later….
In another vein, and raising further questions, is the notion that microtubules are not permanently but dynamically based in dendritic spines (Jaworski et al., 2009). This leads us to another dissociation: If protein tau is present in dendritic spines, which we never doubted, it cannot be bound to microtubules. That leaves it free to become bound to the PSD, proposed and implicated here by Ittner et al., or become sequestered to the actin network—that other important post-synaptic scaffold in spines.
A most intriguing question, raised but discussed somewhat "en passant" by Ittner et al., is, How does protein tau get into the dendrites and post-synaptic compartments, i.e., spines, in the first place? Besides the "piggy-back transport of Fyn," the findings highlight a major “normal” physiological function of protein tau in dendrites, which could potentially become more important for neuroscience than for the AD field (this thought illustrates nicely the narrow boundary between physiology and pathology, a concept instilled in my brain in my early Ph.D. student days!). There is, nevertheless, an important pathological flipside even to this dendritic tau: the dramatic neuropil tauopathy in AD, which according to experts is quantitatively an order of magnitude more important than neurofibrillary tangles in the soma (Mitchell et al., 2000). Is that exclusively axonal and pre-synaptic—or what is the contribution of dendritic post-synaptic compartments? In this respect, the statement by Ittner et al that "...levels of tau in the dendritic compartment are much lower than in axons,..." needs to be taken cautiously and verified by proper quantitative methods—if at all possible.
We used an AAV-tau.255 vector to express the same truncated version of tau, which, in contrast to full-length tau.4R and tau.P301L, was not neurotoxic. The lack of toxicity correlated with the fact that tau.255 was largely retained to the cell soma, as opposed to full-length tau that located to dendrites (see, e.g., Fig. S5D in Jaworski et al., 2009). We proposed that the missing microtubule binding domain (MTBD) and the inherent lack of transport over microtubules (MTs) prevented tau.255 from reaching the post-synaptic compartment in the dendritic spines. Interestingly, we observed increased phosphorylation of tau residues that constitute epitope AT180, i.e., S231/S235, which align closely with the 7th PXXP motif (P233-K-S-P235) that binds Fyn (Lee et al., 1998; confirmed here by Ittner et al.). Hardly accidental is then the fact that the AT180 epitope is typically generated by GSK3, which is the subject of our current efforts to understand the contribution of these kinases to the physiology and pathology of APP and protein tau (Terwel et al., 2008).
Importantly, in our AAV-based model, even wild-type tau was neurotoxic, causing rapid and extensive degeneration of CA1/2 pyramidal neurons in wild-type mice, i.e., in absence of human amyloid peptides, and without formation of large tau aggregates in soma or neuropil. Unfortunately, Ittner et al. did not analyze or discuss the eventual neurodegeneration they might have observed in their combined mouse models. Combining the novel data with previously published data leads me to conclude that Aβ causes excitotoxicity and provokes seizures, eventually causing premature death, but does not cause neuronal cell death—an observation made in many single transgenic APP mouse models.
The novel, documented NR2b-Fyn-mediated mechanism depends on transport of Fyn by endogenous murine protein tau into the post-synaptic compartment of dendritic spines. As stated above, the transport of Fyn appears to be "piggy-backed" on protein tau into dendritic processes, which inevitably could or even must result in phosphorylation of tau at Y18 (Lee et al., 2004). Does this phosphorylation come into play in the mechanism? Moreover, Fyn is transported by full-length tau over the dendritic MT system involving the tau-MTBD domain, which implicates this is not a novel function for tau, but "a novel route." This raises again other questions as the MT system in dendrites is not polarized like the axonal MT system. How tau manages to overcome the dendritic "traffic jam" is open for experimentation, which I am sure some of us are doing already.
I agree with Akihiko Takashima that GSK3 must be accounted for in the amyloid-tau equation. I beg to disagree with him regarding the statement that "Reduction of tau rescued…the impairment of LTP caused by GSK3β overexpression…." as LTP was not measured, but the spatial cognition task in the water maze was (Gomez de Barreda et al., 2010). Moreover, the data show that tau-/- did not rescue the GSK3-imposed phenotype, as the Tet/GSK3β mice did not differ significantly from Tet/GSK3β+tau-/- mice (Fig. 2 in Gomez de Barreda et al., 2010).
With regard to GSK3, I commented on this forum on a closely related study on amyloid-induced axonal transport problems involving NMDAR implicating GSK3 in the mechanism (Decker et al., 2010). Obviously, amyloid, tau, and GSK3 play on the same team more often than not, both at the physiological or pathological side of the brain.
Technically, I am not convinced that an issue remains in terms of tau-/- mice. I agree fully with Lennart Mucke that Ittner et al. reassures us that APP-induced deficits are mitigated in a tau-/- background, even in different parental strains (Roberson et al., 2007; Ittner et al., 2010). Science is more often than not reproducible and therefore pleasing and rewarding!
Allow me to conclude by stating that all commentators agree with Ittner and colleagues that protein tau is an attractive study target—and a promising drug target for primary and secondary tauopathies.