The building blocks of Alzheimer's

Deposits of amyloid-β peptide (Aβ) and tau are the pathological hallmarks of Alzheimer's disease (AD). Treatments aimed at Aβ production, clearance, or aggregation are all in clinical trials. However, interest in tau as a target has been muted, partly because tau pathology seems to occur downstream of Aβ (1–4), making it uncertain whether tau-directed therapeutics would prevent Aβ-induced impairments. Also, tau is posttranslationally modified in AD (5–8), and debate continues about which modifications should be targeted. Reducing overall tau levels might be an alternative approach (9). As tau haplotypes driving slightly higher tau expression increase AD risk (10), reducing tau levels might be protective. Therefore, we determined the effect of reducing endogenous tau expression on cognitive deficits in transgenic mice expressing human amyloid precursor protein (hAPP) with familial AD mutations that increase Aβ production.   

We crossed hAPP mice (11) with Tau^–/– mice (12) and examined hAPP mice with two (hAPP/Tau^+/+), one (hAPP/Tau^+/–), or no (hAPP/Tau^–/–) endogenous tau alleles, compared with Tau^+/+, Tau^+/–, and Tau^–/– mice without hAPP (13). Tau reduction did not affect hippocampal hAPP expression, and conversely, hAPP did not affect hippocampal tau levels (fig. S1). The six genotypes showed no differences in weight, general health, basic reflexes, sensory responses, or gross motor function.

To test learning and memory, we used the Morris water maze. In the cued version, mice learn to find the target platform using a conspicuous marker placed directly above it. At 4 to 7 months of age, Tau^+/+, Tau^+/–, and Tau^–/– mice learned quickly, but as expected (14, 15), hAPP/Tau^+/+ mice took longer to master this task (Fig. 1A; P < 0.001). In contrast, hAPP/Tau^+/– and hAPP/Tau^–/– mice performed at control levels.

Panel A

This figure shows how long it took mice from all six possible genotypes to escape from the water maze, i.e., to find the platform and crawl up on to it.

“Cued platform” means that there was a flag or visible marker on top of the platform so that mice could easily see it and swim to it, once they got the idea that a platform existed and that it was possible to find it and use it.

Panel B

In this case the figure shows how long it took the same mice to escape from the water maze when a hidden platform was used. The mice had to remember where the platform was previously and use their spatial memory to navigate the pool and find the platform.

Note that all six genotypes of mice learned more slowly on the hidden platform than on the visible platform (Panel A). Why do you think that is? Which group looks to be the worst? What do the statistics tell us?

Panel C

After the water maze training, the researchers take the platform out of the pool altogether; this is called a probe trial.

Twenty-four hours after 3 days of training with a hidden platform, the mice are placed in the pool with no platform and their swim path is recorded, as shown here in a bird’s eye view. The quadrant of the pool that used to contain the platform is shown in gray, and the exact location where the platform used to be is shown in blue.

Mice with normal amounts of tau and no hAPP gene spent most of their time swimming in the quadrant where the platform used to be. Mice with normal amounts of tau and the hAPP gene swam all over the pool, showing little preference for the “correct” quadrant. Mice with no tau and the hAPP gene swam almost entirely in the “correct” quadrant, much like healthy mice, showing they remembered where the platform was.

Panel D

This is another way of looking at the data from the probe trial in Panel C. The graph shows how many times mice crossed the exact location where the platform used to be and compares this across genotypes.

All mice without the hAPP gene cross the platform location (the Target) significantly more times than they cross a similar area in any other quadrant. Mice with no tau and the hAPP gene do this, too, but not quite as well as the mice without hAPP. (The differences in crossed areas—shown as bar height—are not as pronounced.)

Why is it useful to look at the data from the same trial in two different ways like this (Panel C versus Panel D)?

Panel E

This figure shows data from another probe trial, this time given 72 hours after 5 days of training on the hidden platform.

What differences do you see between this trial and the probe trial given 24 hours after 3 days of training (Panel D)? Why do you think there are these differences?

Answers

Panel B: Mice learn the visible task more quickly because it is easier. There is no memory involved, just swimming toward a visible object. hAPP/Tau^+/+ mice look to be the worst at this task, and hAPP/Tau^+/- perform the next worst. Statistics tell us that hAPP/Tau^+/+ mice are significantly worse at this task than mice without the hAPP gene when looking at all four days of training.

Panel D: Looking at the same data in different ways might help us see details or aspects that could be missed by just one method of analysis. If the difference between the groups is real, then the effect should be seen any way you look at it (i.e., how many seconds the mouse spends at the platform location versus how much distance it travels at the platform location).

Panel E: Here, all genotypes cross the target a bit more often than they did 24 hours after 3 days of training (Panel D). hAPP/Tau^-/- cross the target more frequently 72 hours after 5 days of training than 24 hours after 3 days of training (Panel E versus D), and hAPP/Tau^+/- also cross the target significantly more after more days of training (Panel E versus D). hAPP/Tau^+/+ mice may cross the target a little more than after 5 days of training, but this difference is not significant, showing that they still have not learned the task and do not remember the location of the platform.

These differences are due to the increase in training so that, in the end, even slightly learning-impaired mice (hAPP/Tau^+/-) were able to remember and learn the task.

The more difficult hidden-platform version of the water maze demands spatial learning. Mice without hAPP learned this task over 3 days of training regardless of tau genotype, whereas hAPP/Tau^+/+ mice showed no evidence of learning until days 4 and 5 (P < 0.001; Fig. 1B). Notably, hAPP/Tau^+/– mice were less impaired than hAPP/Tau^+/+ mice (P < 0.02), and hAPP/Tau^–/– mice did not differ from controls without hAPP (Fig. 1B). Probe trials, in which the platform was removed and mice were given 1 min to explore the pool, confirmed the beneficial effect of tau reduction (Fig. 1, C to E). In an initial probe trial 24 hours after 3 days of training, hAPP/Tau^+/+ mice had no apparent spatial memory of the platform location, crossing the target platform location no more than they crossed equivalent areas in nontarget quadrants (Fig. 1D). However, hAPP/Tau^–/– mice, similar to mice without hAPP, did cross the target platform location more often (P < 0.01; Fig. 1D). After two additional days of training, hAPP/Tau^+/– mice also had more target than nontarget crossings (P < 0.01), whereas hAPP/Tau^+/+ mice still showed no spatial learning and memory (Fig. 1E). Thus, the tau reduction gene dose-dependently ameliorates Aβ-dependent water maze learning and memory deficits.

Increased exploratory locomotor activity is seen after entorhinal cortex lesions and may reflect deficits in spatial information processing (16); hAPP mice show similar hyperactivity (15). hAPP/Tau^+/+ mice were hyperactive in the Y maze (P < 0.001; Fig. 2A), a new cage (P < 0.05; Fig. 2B), and the elevated plus maze (P < 0.001; Fig. 2C). In contrast, these abnormalities were not seen in hAPP/Tau^+/– and hAPP/Tau^–/– mice (Fig. 2, A to C). To determine whether the benefits afforded by tau reduction were sustained, we examined older mice. Hyperactivity persisted in hAPP/Tau^+/+ mice and remained absent in hAPP/Tau^–/– mice at 12 to 16 months (P < 0.05; Fig. 2D).

Panel A

The Y Maze is a simple Y-shaped maze with three connected chambers for the mice to explore.

Hyperactive mice will explore the maze more and enter chambers at a higher rate during the 6-minute trial than nonhyperactive mice.

hAPP/Tau^+/+ mice entered new chambers significantly more than any of the other genotypes of mice, showing that lower levels of tau seems to prevent this abnormal behavior in hAPP mice.

Panel B

All six genotypes of mice were allowed to explore an empty cage for 5 minutes. hAPP/Tau^+/+ mice explored and spend significantly more time being active than the other genotypes of mice. This indicates that lower levels of tau seem to prevent this abnormal hyperactive behavior in hAPP mice.

Panel C

The elevated plus maze is a cross-shaped maze that is suspended a few feet off the ground and has two arms that have high walls and two arms that are open, without walls.

Hyperactive mice will explore the maze more and cover more distance during the 10-minute trial than nonhyperactive mice. hAPP/Tau^+/+ mice explored the maze and traveled significantly more distance than the other genotypes of mice.

This indicates that lower levels of tau seem to prevent this abnormal hyperactive behavior in hAPP mice.

Panel D

This panel shows almost the same data as in Panel C, except these mice are middle-aged instead of young.

As in Panel C, the hAPP/Tau^+/+ mice seem to be hyperactive and cover significantly more distance in the maze during the trial than the other groups.

This shows that lowering the amount of tau in hAPP mice prevents hyperactive symptoms in both young mice and in middle-aged mice.

Panel E

These survival curves show how many mice of each of the six genotypes have died at any given age.

By 6 months of age, only 85% of hAPP/Tau^+/+ mice remain alive, but this premature death is not seen in either the hAPP/Tau^+/- or hAPP/Tau^-/- mice, showing that reducing tau in hAPP mice increases their life span.

Premature death of unclear etiology was also observed in hAPP mice (P < 0.005; Fig. 2E) (17, 18). Again, both hAPP/Tau^–/– and hAPP/Tau^+/– mice were protected from this early mortality. Thus, tau reduction prevented major Aβ-dependent adverse effects in hAPP mice. We examined several plausible mechanisms by which tau reduction might exert protective effects and we eventually discovered an unexpected role for tau.

We first ruled out the possibility that tau reduction altered Aβ levels or aggregation. Tau reduction did not alter hAPP expression (fig. S1), soluble Aβ_1-x or Aβ_1-42 levels, or the Aβ_1-42/Aβ_1-x ratio (fig. S2). In addition, hAPP/Tau^+/+, hAPP/Tau^+/–, and hAPP/Tau^–/– mice had similar plaque load at 4 to 7 months (fig. S3) and 14 to 18 months (Fig. 3, A and B). We also found no effect of tau reduction on levels of Aβ*56, a specific Aβ assembly linked to memory deficits (19) (fig. S4). Thus, the beneficial effects of reducing tau were observed without detectable changes in Aβ burden, suggesting that tau reduction uncouples Aβ from downstream pathogenic mechanisms.

Panel A

The brains of hAPP mice with all three levels of tau were sliced and stained with a chemical that binds to amyloid plaques and fluoresces green when hit with a certain wavelength of light under the microscope.

By both looking at the images and analyzing the total area of the brain that contains plaques (graph on the right), we can see that there are no significant differences in the amount of plaques in the three genotypes seen here. (Mice without the hAPP gene have no plaques and are not shown there).

Panel B

Here, the brains of hAPP mice with all three levels of tau were stained with an amyloid plaque–binding chemical that turns dark brown.

Using the same analysis methods as in Panel A, we can see that there are no significant differences in the amount of plaques in the three genotypes.

Panel C

The brains of hAPP mice with all three levels of tau were sliced and stained with two chemicals (i.e., “double-labeled”). One chemical glows red and stains neurons with abnormal or misshapen dendrites or axons, whereas the other chemical stains amyloid plaques green.

By looking at the images and by analyzing the amount of abnormal or “dystrophic” neurons (graph on the right), we can see that there are no significant differences in the amount of abnormal neurons in the three genotypes.

Panel D

The brains of hAPP mice with all three levels of tau, as well as mice without hAPP, were sliced and stained with a chemical that binds to newly sprouted axons, showing where they are.

All mice with the hAPP gene, regardless of their tau levels, have abnormal sprouting axons in a specific layer of the hippocampus where they are not supposed to be (labeled “oml” in the images). Moreover, reducing levels of tau doesn’t appear to have any effect on this.

Quantifying the density of axonal sprouting shows the same thing (graph on the right).

Next, we looked for abnormal forms of tau that might act as downstream effectors of Aβ in hAPP/Tau^+/+ mice. Major AD-related phosphorylation sites in human tau are conserved in murine tau, including those phosphorylated by proline-directed kinases, such as glycon synthase kinase (GSK)–3β and cdk5, or by microtubule affinity–regulating kinase (MARK). Changes in murine tau phosphorylation at these sites are easily detected, for example after brief hypothermia (20) (fig. S4). However, in hippocampal homogenates of 4- to 7-month-old hAPP/Tau^+/+ mice, we did not find changes in tau phosphorylation at proline-directed kinase sites, including Thr¹⁸¹, Ser²⁰², Thr²³¹, and Ser^396/404, or at the primary site for MARK, Ser²⁶² (fig. S5). Generation of neurotoxic tau fragments has also been implicated as a mechanism of Aβ toxicity (21). Tau-deficient primary neurons are resistant to Aβ-induced degeneration (3,22), apparently because Aβ toxicity in vitro involves production of a 17-kD tau fragment (21). We confirmed the presence of a 17-kD tau fragment in lysates of Aβ-treated primary neurons, but found no abnormal tau proteolysis in hippocampal homogenates from hAPP mice (fig. S6), suggesting that the neuroprotective effects of tau reduction in the two systems are mechanistically different. The relative lack of modified tau also distinguishes our model from transgenic lines overexpressing tau with mutations that cause frontotemporal dementia, but not AD, in humans (2, 4, 23). In our study, reduction of endogenous, wild-type tau protected hAPP mice against Aβ-dependent cognitive impairments, and this did not involve the elimination of a large pool of tau with typical AD-associated modifications. Our experiments do not rule out the possibility that another type of tau modification, or a small pool of modified tau in a restricted subcellular compartment or cellular population, could play a role downstream of Aβ.

To begin addressing this possibility, we stained brain sections from Tau^+/+ and hAPP/Tau^+/+ mice with phospho-tau antibodies. We saw little difference overall between Tau^+/+ and hAPP/Tau^+/+ mice in phospho-tau immunoreactivity, but we did observe scattered phospho-tau–positive punctae in dystrophic neurites surrounding amyloid plaques (fig. S7). We thus wondered whether the benefits of tau reduction in hAPP mice could relate to prevention of neuritic dystrophy, which may contribute to AD-related cognitive decline (24). Despite the differences in their behavior, hAPP/Tau^+/+, hAPP/Tau^+/–, and hAPP/Tau^–/– mice had similar amounts of neuritic dystrophy (Fig. 3C). Thus, tau is not required for the formation of plaque-associated dystrophic neurites. Given that tau reduction prevented behavioral deficits but not neuritic dystrophy, these may represent parallel, rather than causally linked, disease manifestations, or tau reduction may act downstream of neuritic dystrophy.

Tau has a well-characterized role in axonal outgrowth (12), so we tested whether tau reduction prevented the aberrant sprouting of hippocampal axons observed in AD (25) and hAPP mice (18). Similar degrees of sprouting were observed, regardless of tau genotype (Fig. 3D). Thus, although tau reduction affected important outcome measures related to Aβ-induced neuronal dysfunction, not all effects of Aβ were blocked.

Excitotoxicity is implicated in the pathogenesis of AD (26, 27). Consistent with the increased incidence of seizures in AD patients (28), TgCRND8 hAPP mice are more susceptible to the γ-aminobutyric acid type A (GABA_A) receptor antagonist pentylenetetrazole (PTZ) (29). Using a similar paradigm, we found that hAPP/Tau^+/+ mice were also abnormally sensitive to PTZ, with 20% suffering fatal status epilepticus at a dose that was not lethal to mice without hAPP (P < 0.05). Tau reduction prevented this effect, as no hAPP/Tau^+/– or hAPP/Tau^–/– mice died. Seizures in hAPP/Tau^+/– and hAPP/Tau^–/– mice were less severe and occurred at longer latencies than in hAPP/Tau^+/+ mice (P < 0.01; Fig. 4, A and B).

Panel A

Mice of all six genotypes were injected with a chemical that induces seizures.

Mice with lower amounts of tau had less severe seizures than mice with normal amounts of tau. This was true for mice with and without the hAPP gene.

In Tau^-/- mice, those without the hAPP gene had significantly less severe seizures than those with hAPP, showing that hAPP does to some extent affect how susceptible mice are to seizures.

Panel B

This panel looks just at hAPP mice that were injected with the seizure-inducing drug.

Mice with lower amounts of tau had a delayed seizure onset compared to mice with normal amounts of tau.

Panel C

This panel looks just at non-hAPP mice that were injected with the seizure-inducing drug.

Mice with lower amounts of tau had a delayed seizure onset compared to mice with normal amounts of tau.

Panel D

Here, mice with the three different levels of tau (all without the hAPP gene) were injected with a different seizure-inducing drug.

Lower levels of tau made mice less susceptible to seizures, requiring a higher dose to elicit the type of seizures seen in Tau^+/+ mice. The lower the amount of tau, the more resistant to seizures the mice were.

Tau reduction also increased resistance to PTZ in hAPP-nontransgenic mice, lowering seizure severity and delaying seizure onset (P < 0.01; Fig. 4, A and C). To confirm that tau reduction could reduce aberrant neuronal overexcitation, we challenged mice with excitotoxic doses of the glutamate receptor agonist kainate. As expected, intraperitoneal injection of kainate dose-dependently induced seizures in Tau^+/+ mice (Fig. 4D). In contrast, Tau^+/– and Tau^–/– mice were resistant to kainate across a range of doses (P < 0.05; Fig. 4D). Thus, tau modulates sensitivity to excitotoxins and may be involved in regulating neuronal activity. The excitoprotective effect of tau reduction in mice without hAPP is more likely related to a physiological function of tau than to the removal of a pathological form of the protein. Sensitization of neurons to Aβ by physiological forms of tau could explain why tau reduction is effective in hAPP/Tau^+/+ mice despite their lack of obvious tau modifications.

Our findings raise the possibility that tau reduction could protect against AD and other neurological conditions associated with excitotoxicity. Of course, the therapeutic implications of our findings must be interpreted with caution. First, there are differences between the mouse model and AD, including the absence of substantial neuronal loss or neurofibrillary pathology in hAPP mice. The contribution of these abnormalities to AD-related cognitive impairment, relative to the role of reversible Aβ-induced neuronal dysfunction that is modeled in hAPP mice, remains to be determined (30). Second, microdeletions of chromosome 17q21 encompassing the tau gene are associated with learning disabilities in humans (31), although abnormalities in these individuals may relate to insufficiency of other genes in the region, such as the corticotropin-releasing hormone receptor gene, which is implicated in neuropsychiatric disease (32). We found no adverse effects of tau reduction on health or cognition in mice, and the evidence that even partial tau reduction robustly protected mice from Aβ and excitotoxic agents highlights its potential benefits.

Supporting Online Material

www.sciencemag.org/cgi/content/full/316/5825/750/DC1

Materials and Methods

Figs. S1 to S7

References

References and Notes

1. R. Tanzi, L. Bertram, Cell 120, 545 (2005).

2. J. Lewis et al., Science 293, 1487 (2001).

3. M. Rapoport, H. N. Dawson, L. I. Binder, M. P. Vitek, A. Ferreira, Proc. Natl. Acad. Sci. U.S.A. 99, 6364 (2002).

4. S. Oddo, L. Billings, J. P. Kesslak, D. H. Cribbs, F. M. LaFerla, Neuron 43, 321 (2004).

5. C. X. Gong, F. Liu, I. Grundke-Iqbal, K. Iqbal, J. Neural Transm. 112, 813 (2005).

6.  K. Iqbal et al., Biochim. Biophys. Acta 1739, 198 (2005).

7. W. H. Stoothoff, G. V. Johnson, Biochim. Biophys. Acta 1739, 280 (2005).

8. I. Khlistunova et al., J. Biol. Chem. 281, 1205 (2006).

9. C. A. Dickey et al., FASEB J. 20, 753 (2006).

10. A. J. Myers et al., Hum. Mol. Genet. 14, 2399 (2005).

11. L. Mucke et al., J. Neurosci. 20, 4050 (2000).

12. H. N. Dawson et al., J. Cell Sci. 114, 1179 (2001).

13. Materials and methods are available as supporting material on Science Online.

14. J. J. Palop et al., Proc. Natl. Acad. Sci. U.S.A. 100, 9572 (2003).

15. D. T. Kobayashi, K. S. Chen, Genes Brain Behav. 4, 173 (2005).

16. O. Steward, J. Loesche, W. C. Horton, Brain Res. Bull. 2, 41 (1977).

17. K. K. Hsiao et al., Neuron 15, 1203 (1995).

18. J. Chin et al., J. Neurosci. 24, 4692 (2004).

19. S. Lesné et al., Nature 440, 352 (2006).

20. E. Planel et al., J. Neurosci. 24, 2401 (2004).

21. S. Y. Park, A. Ferreira, J. Neurosci. 25, 5365 (2005).

22. T. Liu et al., J. Neurochem. 88, 554 (2004).

23. K. SantaCruz et al., Science 309, 476 (2005).

24. R. B. Knowles et al., Proc. Natl. Acad. Sci. U.S.A. 96, 5274 (1999).

25. J. W. Geddes et al., Science 230, 1179 (1985).

26. M. P. Mattson, Nature 430, 631 (2004).

27. J. W. Olney, D. F. Wozniak, N. B. Farber, Arch. Neurol. 54, 1234 (1997).

28. J. C. Amatniek et al., Epilepsia 47, 867 (2006).

29. R. A. Del Vecchio, L. H. Gold, S. J. Novick, G. Wong, L. A. Hyde, Neurosci. Lett. 367, 164 (2004).

30. J. J. Palop, J. Chin, L. Mucke, Nature 443, 768 (2006).

31. J. R. Lupski, Nat. Genet. 38, 974 (2006).

32. F. Holsboer, Curr. Opin. Investig. Drugs 4, 46 (2003).

33. We thank M. Vitek and H. Dawson for tau knockout mice; P. Seubert and P. Davies for antibodies; H. Solanoy, X. Wang, and Y. Zhou for technical assistance;C. McCullough for advice on statistics; D. McPherson and L. Manuntag for administrative support; and G. Howard and S. Ordway for editorial review. L.M. received consulting fees from Merck and honoraria from Amgen, Elan, and Pfizer. E.D.R. received consulting fees fromRinat Neuroscience. Supported by NIH grants AG011385 and AG022074 (L.M.), MH070588 (K.S.L), andNS054811 (E.D.R); the Giannini Foundation (E.D.R.); the S.D. Bechtel Jr. Young Investigator Award (E.D.R.); and the NIH National Center for Research Resources grant RR18928-01.