Formation of Neurofibrillary Tangles in P301L Tau Transgenic Mice Induced by Aβ42 Fibrils

See allHide authors and affiliations

Science  24 Aug 2001:
Vol. 293, Issue 5534, pp. 1491-1495
DOI: 10.1126/science.1062097


β-Amyloid plaques and neurofibrillary tangles (NFTs) are the defining neuropathological hallmarks of Alzheimer's disease, but their pathophysiological relation is unclear. Injection of β-amyloid Aβ42 fibrils into the brains of P301L mutant tau transgenic mice caused fivefold increases in the numbers of NFTs in cell bodies within the amygdala from where neurons project to the injection sites. Gallyas silver impregnation identified NFTs that contained tau phosphorylated at serine 212/threonine 214 and serine 422. NFTs were composed of twisted filaments and occurred in 6-month-old mice as early as 18 days after Aβ42injections. Our data support the hypothesis that Aβ42 fibrils can accelerate NFT formation in vivo.

Transgenic mice that express P301L mutant human tau form abnormal tau-containing filaments in brains (1, 2). These filaments have striking similarities with the NFTs of several human neurodegenerative diseases, including Alzheimer's disease (AD) and frontotemporal dementia with parkinsonism linked to chromosome 17 (FTDP-17), but their numbers are considerably lower than these commonly found in human disease (3). To determine whether β-amyloid can accelerate NFT formation, we injected synthetic Aβ42 fibrils into the somatosensory cortex and the hippocampus of 5- to 6-month-old P301L tau transgenic mice (4) and nontransgenic littermates (5–7). For the control peptide, we used the reversed sequence, Aβ42-1, derived from the identical source (6). Aβ42 fibrils were generated by incubation at 37°C with shaking and were confirmed by electron microscopy (Fig. 1, A and B) (5, 6). Aβ42 fibrils were stable in vivo in both P301L transgenic and wild-type control mice and were readily detectable at least until 45 days after the injections (Fig. 1C). As expected, brain amyloid deposits were accompanied by reactive astrogliosis at both the injection sites (Fig. 1D) and the amygdala (Fig. 1E) (8); these were seen in both Aβ42- and in control-injected transgenic mice and persisted for at least 45 days after injection. This reaction may be related to the fact that neurons in the amygdala project to the injection sites, as shown by retrograde transport of Texas red–conjugated dextran from the injection site in the somatosensory cortex to cell bodies in the amygdala (Fig. 1F) (8).

Figure 1

(left). Stereotaxic injections of Aβ42 fibrils into mouse brains. Electron microscopy confirmed that aggregated synthetic Aβ42formed fibrils (A), whereas a reverse control peptide, Aβ42-1, did not (B). Suspensions of Aβ42 fibrils or Aβ42-1 were injected stereotaxically into the somatosensory cortex of one hemisphere, as well as the CA1 region of the hippocampus of the contralateral hemisphere. Antibody 6E10 revealed that Aβ42 fibrils were present at the injection sites for at least 45 days after the injection (C). As staining controls, amyloid plaque-containing APPsw transgenic (36) and human AD brain sections were included (not shown). Aβ42 and Aβ42-1 increased the occurrence of reactive astrocytes both around the injection site (D) and in the amygdala (E). Reactive astrocytosis persisted for at least 45 days. Stereotaxic application of Texas red–coupled dextran into identical coordinates stained a subset of neurons in the amygdala, confirming that these projected to the injection sites (F). [(C) to (F)] Seven-month-old P301L female analyzed 45 days after injection. Bars: 0.4 μm [(A) and (B)]; 50 μm [(C) to (F)].

Eighteen days after the injections of Aβ42, Gallyas silver impregnation (9) revealed numerous NFTs (Fig. 2, A to E), along with neuropil threads and degenerating neurites (Fig. 2C) in the amygdala of P301L, but not wild-type, mice. Occasional NFTs were also present in the parietal cortex (Fig. 2D). The NFTs in mice (Fig. 2E) were very similar to those in AD brains stained in parallel by the same protocol (Fig. 2F). Moreover, the neuropil threads were similar to those known in AD (10–12). A subset of Gallyas-positive NFTs in the mice was also stained with thioflavin-S, consistent with the histopathology of AD (Fig. 2G). Immunoelectron microscopy identified many AT8-positive tau filaments in somatodendritic localizations of neurons within the basolateral amygdala of Aβ42-injected P301L mice (13) (Fig. 2, H to J). The filaments had a width of 20 to 25 nm and a periodicity of 90 nm and are best described as twisted ribbons. In human carriers, the P301L mutation causes predominant expression of four repeat (4R) isoforms, with a small amount of wild-type 3R isoforms, resulting in 15-nm-wide twisted filaments with a periodicity of greater than 130 nm (14). Because mice endogenously express only 4R tau isoforms, and the transgene was designed to express 4R human P301L tau, the filaments observed here contained no 3R tau. Importantly, the human intronic FTDP-17 mutations that reduce the formation of 3R tau also cause twisted ribbons composed mainly of 4R tau. It is therefore possible that the relative amounts of 3R and 4R isoforms contribute to the ultrastructural morphology of the filaments.

Figure 2

(right). NFTs in P301L tau transgenic mice induced by Aβ42 fibrils. Gallyas silver impregnations of NFTs in the amygdala of Aβ42- (A) and Aβ42-1-injected (B) P301L tau transgenic mice. Aβ42 fibrils induced the Gallyas-positive formation of numerous NFTs and neuropil threads in the amygdala and, occasionally, the cortex as early as 18 days after the injection (C to E). NFTs in mice were very similar to those found in brains obtained from AD patients; these slides were stained in parallel by the same protocol (F). A subset of NFTs was also stained by thioflavin-S (G). Immunoelectron microscopy revealed the presence of many twisted AT8-positive tau filaments in the basolateral area of the amygdala of Aβ42-injected P301L mice (H to J). (A) Eight-month-old P301L male analyzed 40 days after injection; (B) 6.5-month-old P301L female analyzed 40 days after injection; [(C) and (D)] Six-month-old P301L female analyzed 18 days after injection; (E) 5.25-month-old P301L male analyzed 21 days after injection; (F) Human 86-year-old female AD patient; (G) Seven-month-old P301L female analyzed 45 days after injection. Bars: 25 μm [(A) to (C)], 12.5 μm [(D) to (G)], 800 μm (H), 100 μm (I), 50 μm (J).

Quantitative analyses revealed five times more Gallyas-positive NFTs in the Aβ42-injected P301L mice than in Aβ42-1- or uninjected P301L mice (Fig. 2, A and B, andFig. 3). Cross-sectional time-course analyses of NFT formation showed initial NFTs 18 days after Aβ42injection, with further increases in numbers (n = 58) at least until 60 days after the injection. NFT formation in both hemispheres in the Aβ42-injected P301L mice did not vary with gender (females: 23 ± 21; males: 23 ± 4;n = 7, P = 0.86, Mann-WhitneyU test). In contrast, Aβ42-1-injected P301L males developed few NFTs and P301L females, no NFTs, at 6 to 8.5 months of age. This difference was statistically significant (females: 0; males: 5.8 ± 1.9; n = 7, P < 0.01, Mann-Whitney U test). Importantly, the presence of the tau mutation was necessary for NFT formation because homozygous transgenic mice expressing human wild-type tau at tau levels similar to or exceeding those of P301L mice (2) failed to develop NFTs in response to Aβ42 either at 6 or 12 months of age.

Figure 3

(left).Gallyas-positive NFTs in the amygdala. NFTs were counted on day 22 after injection in Aβ42-, Aβ42-1-, and uninjected P301L transgenic mice, nontransgenic littermate controls, and transgenic mice expressing wild-type human tau (37). The mean age (months ± SD) at the time of analyses is indicated. Gallyas-positive NFTs were counted according to (8) and represent the sum in 20 standardized frontal sections comprising both the ipsilateral and the contralateral amygdala. Mann-WhitneyU test: P = 0.007 (two-tailed exact significance) comparing Aβ42- with Aβ42-1-injected P301L mice.

An unexpected finding was the spatial separation of the site of Aβ42 injection and remote NFT formation in the amygdala, with no significant differences between the ipsilateral and the contralateral amygdala (11.4 ± 10.13 and 9.4 ± 8.0;n = 7, P = 0.058, Wilcoxon Signed Ranks Test). This finding suggests the possibility that damage to presynaptic terminals or axons of neurons that project to the injection site caused NFT formation in the respective cell bodies. The anatomical separation of amyloid deposition and NFT formation is therefore consistent with Aβ42-induced axonal damage and, possibly, impaired axonal transport of tau (15). We confirmed that the affected neuronal population in the amygdala projected to the cortical injection sites by showing retrograde transport of Texas red–conjugated dextran from the injection sites to the amygdala (Fig. 1F). Other mechanisms of somatodendritic accumulation of tau are less likely: First, we excluded a direct exposure to Aβ42 fibrils of the cell bodies in the amygdala by immunohistochemistry. Second, increases in synthesis of tau protein are unlikely, as indicated by the absence of axonal dilatations or spheroids in amygdala neurons. Third, a diffusible toxic factor would hardly explain the failure of neurons adjacent to the injection sites to develop NFTs. Moreover, selective vulnerability of the amygdala for NFT formation is suggested by doubly transgenic mice expressing both mutant APP and P301L tau (16). In human patients with AD, an anatomical separation of amyloid plaques and NFTs is frequently found, with amyloid deposits around synapses and NFTs in the respective cell bodies of projection neurons (17). In addition, the amygdala is among the most vulnerable areas affected early by NFT formation in human patients (18). High vulnerability of the amygdala in our P301L mice is supported by the fact that neurons in the amygdala expressed similar levels of the transgene as compared with cortical or hippocampal pyramidal neurons, yet these developed hardly any NFTs (2).

Although our experiments did not formally address the involvement of astrocytes and microglia in NFT formation, activation of these cells alone was not sufficient for NFT formation because Aβ42and Aβ42-1 similarly activated astrocytes and microglial cells, both around the injection sites and in the amygdala (Fig. 1, D and E) (19).

The formation of NFTs in AD is associated with hyperphosphorylation and conformational changes of tau (20–22). To determine whether the Aβ42-induced NFT formation in P301L mice was associated with altered phosphorylation and conformation of tau, we used antibodies directed against abnormal phospho-epitopes (R145d, pS422, AT100, TG3) (23–25), hyperphosphorylated epitopes (AT8, S199P, AT180, 12E8, AD2, PHF1) (23, 26–31), as well as conformation-dependent antibodies (TG3, MC1) (22, 24), using standard procedures (8, 32).

Whereas several antibodies, including AT8, detected phosphorylated tau throughout the brains of P301L mice independently of the injections, R145d/pS422 and AT100 directed against phospho-epitopes S422 and S212/T214, respectively, specifically detected NFTs and neurons only in response to Aβ42 (Fig. 4, A to F). The spatial distribution pattern of these abnormally phosphorylated forms of tau was identical to that observed by Gallyas stainings and occurred, again, predominantly in the amygdala (Fig. 4, A to C). Costaining revealed that neurons stained by R145d/pS422 were also stained by AT100 (Fig. 4, D to F). Neither R145d nor AT100 immunostained any cells in nontransgenic mice. The specificity of R145d, pS422, and AT100 immunoreactivity for Aβ42-associated abnormal phosphorylation was exceptional because these antisera revealed few signals in uninjected or Aβ42-1-injected P301L tau transgenic mice, and none in transgenic mice expressing wild-type human tau (19). Moreover, all Gallyas-positive NFTs were also stained by R145d, as indicated by sequential immunofluorescence and Gallyas silver impregnation protocols, strongly suggesting that the NFTs in P301L mice contained S422-phosphorylated tau. Semiquantitative analyses revealed that about one-half of the R145d-positive neurons (Fig. 4, G and I) were Gallyas-positive (Fig. 4, H and J), and R145d stained these neurons generally more intensely than cells without NFTs.

Figure 4

(right). Abnormal phospho-epitopes of tau induced by Aβ42 fibrils. The R145d epitope S422 was not phosphorylated in the hippocampus and cortex (A), but was specifically induced by Aβ42fibrils in the amygdala [(B); higher magnification: (C)]. Double immunofluorescence staining with R145d (tau phospho-epitope S422) and AT100 (phospho-epitope S212/T214) revealed that R145d-positive neurons in the amygdala were AT100-positive [(D) and (E); merge: (F)]. About half of the R145d-positive neurons (G and I) bore Gallyas-positive NFTs (Hand J), and generally these included the neurons that were most intensely stained by R145d. In contrast to R145d and AT100, the AT8 epitope was phosphorylated in many neurons of the hippocampus, cortex (K), and amygdala [(L); higher magnification: (M)]. [(A) to (F), (L) and (M)] Six-month-old P301L male analyzed 18 days after injection; [(G) to (J)] Seven-month-old P301L female analyzed 45 days after injection. Bars: 100 μm [(A), (B), (K), (L)]; 25 μm [(C) and (M)]; 40 μm [(F) and (J)].

Together, the result obtained with immunostaining is consistent with the possibility that phosphorylation of epitopes Ser-212/Thr-214 and Ser-422 is tightly associated with NFT formation. Our data extend previous findings that Aβ42 induced tau phosphorylation in vitro and in vivo at the AT8 and 12E8/Ab31 epitopes (33, 34): In our P301L mice, tau was phosphorylated at these epitopes, even in the absence of injected Aβ42. Therefore, these epitopes may be necessary but were not sufficient for NFT formation in P301L mice. By using R145d/pS422 and AT100, we found that Aβ42 injections were followed by phosphorylation of tau at S212/T214 and S422, suggesting a role of these epitopes in NFT formation.

In summary, our data establish that Aβ42 fibrils can significantly accelerate NFT formation in P301L mice and provide further support for the hypothesis that β-amyloid can be a causative pathogenic factor. Our data do not exclude the possibility that other factors can also induce NFT formation in brain, in view of the many tauopathies associated with NFT formation in the absence of β-amyloid plaques (3, 35). Our data show that, in transgenic mice, the interaction of β-amyloid with the P301L mutation was required for NFT formation—neither β-amyloid nor the mutation alone was sufficient to generate high numbers of NFTs. Moreover, the mice generated here provide an in vivo assay to determine whether amyloid-lowering therapies such as Aβ vaccination are effective in preventing NFT formation in vivo.

  • * These authors contributed equally to this work.

  • To whom correspondence should be addressed at the Division of Psychiatry Research, University of Zürich, August Forel Strasse 1, 8008 Zürich, Switzerland. E-mail: goetz{at}

  • To whom correspondence should be addressed at the Division of Psychiatry Research, University of Zürich, August Forel Strasse 1, 8008 Zürich, Switzerland. E-mail: nitsch{at}


Stay Connected to Science

Navigate This Article