Truncations of tau protein at aspartic acid421 (D421) and glutamic acid391 (E391) residues are associated with neurofibrillary tangles (NFTs) in the brains of Alzheimer disease (AD) patients. Using immunohistochemistry with antibodies to D421- and E391-truncated tau (Tau-C3 and MN423, respectively), we correlated the presence of NFTs composed of these truncated tau proteins with clinical and neuropathologic parameters in 17 AD and 23 non-AD control brains. The densities of NFTs composed of D421- or E391-truncated tau correlated with clinical dementia index and Braak staging in AD. Glutamic acid391 tau truncation was prominent in the entorhinal cortex, whereas D421 truncation was prominent in the subiculum, suggesting that NFTs composed of either D421- or E391-truncated tau may be formed mutually exclusively in these areas. Both truncations were associated with the prevalence of the apolipoprotein E ɛ4 allele. By double labeling, intact tau in NFTs was commonly associated with D421-cleaved tau but not with E391-truncated tau; D421-cleaved tau was never associated with E391-truncated tau. These results indicate that tau is not randomly proteolyzed at different domains, and that proteolysis occurs sequentially from the C-terminus to inner regions of tau in AD progression. Identification of NFTs composed of tau at different stages of truncation may facilitate assessment of neurofibrillary pathology in AD.
The progressive accumulation of pathologic neurofilamentous inclusions in vulnerable neurons is associated with the formation of neurofibrillary tangles (NFTs), neuritic plaques, and neuropil threads in Alzheimer disease (AD) (1, 2). The major constituent of NFTs is the microtubule-associated tau protein, which abnormally assembles into paired helical filaments (PHFs) (3, 4). Accumulation of NFTs along the entorhinal cortex (ERC)-hippocampal formation-isocortical pathway reflects a progressive pattern of neurofibrillary pathology in the brains of AD patients (1, 5); most clinicopathologic studies have found that NFT accumulation in AD patients correlates with clinical parameters of dementia (6-10).
The abnormal processing of tau (3, 4, 11-14) in AD results in the formation of fibrillar and insoluble aggregates within affected neurons. One pathologic tau modification considered to play a role in PHF formation includes the abnormal phosphorylation of specific sites (15-19). Immunohistochemical studies using several antibodies such as AT8 (which recognizes tau phosphorylation at S202/T205) and AD2 (which recognizes tau phosphorylation at S396/404) have demonstrated that phosphorylation of certain tau epitopes occurs early in the formation of NFTs and neuropil threads (20-22). On the other hand, biochemical analyses of native PHFs purified from the brain of AD cases indicate that tau cleaved at glutamic acid391 (E391), which is specifically recognized by monoclonal antibody MN423 (23, 24), is an important component of the pronase-resistant PHF core (24, 25). Indeed, truncation at E391 is associated with most of the neuropathologic hallmarks in AD brain tissue (26-29).
Several studies have documented another truncation of tau protein at the position aspartic acid421 (D421) in the carboxyl terminus; this truncation is generated by the activity of caspases (30-34). Cells transfected with tau truncated at D421 showed alterations in cytoskeletal elements and apoptosis (32, 34). Moreover, when cultured neurons were exposed to the β-amyloid peptide, a D421-truncated tau product was generated, probably as a result of activation of the apoptotic cascade (30). Using the monoclonal antibody Tau-C3, D421 truncation of tau protein has also been found to be associated with the neurofibrillary pathology of AD (30). This truncation is found not only in AD brains because recent studies have demonstrated that D421-truncated tau is also present in the brains of cases of Pick disease (35-37). These truncated forms of tau display increased rates and extents of polymerization in vitro compared with wild-type full-length tau, suggesting a role for tau truncation in NFT formation (30, 38, 39).
We have previously proposed a model of tau processing during NFT formation and maturation in which the tau molecule undergoes structurally dynamic conformational changes, including misfoldings and truncations (40, 41). In this scheme, sequential D421 and E391 truncations are associated with conformational changes of the tau molecule. Early in this sequence of events, it seems that the intact tau protein adopts aberrant conformations that may expose specific sites that are then susceptible to enzymatic cleavage, resulting in the D421 truncation. In later stages, new conformations may promote further proteolytic activity and the eventual emergence of the E391 truncation. The clinicopathologic significance of the E391 truncation of tau was analyzed in a population-based sample of AD and control nondemented individuals from Cambridge, United Kingdom (42), and the density of lesions composed of E391-truncated tau was found to correlate significantly with dementia severity (7).
To investigate the significance of D421-truncated tau in AD, we correlated clinical and neuropathologic parameters with the presence and distribution of NFTs containing D421-cleaved tau and compared them with those containing E391-truncated tau. Additionally, using double labeling immunofluorescence, we investigated whether D421 and E391 truncations were spatially associated in NFT formation and maturation. We found that the numbers of lesions containing either D421- or E391-truncated tau correlated with the clinical severity of dementia and neuropathologic Braak stage. Based on the present results, we propose a model of NFT formation and maturation in which tau is not randomly proteolyzed at different domains, but that proteolysis occurs sequentially from the extreme C-terminus to inner regions and not in the opposite direction.
Materials and Methods
Samples of brain tissue from AD and age-matched control cases were obtained from the Cambridge Brain Bank Laboratory, Cambridge, United Kingdom. Demographic information on these cases has been reported previously (42-45). In brief, 40 elderly cases (17 AD and 23 nondemented cases) from the Cambridge Project for Long Life (CPLL) collection were studied (Table 1). The cases were classified as AD and nondemented controls based on premorbid clinical assessments and detailed postmortem neuropathologic examination (42-45). The severity of dementia of the AD subjects was determined using the Cambridge Examination for Mental Disorder of the Elderly (CAMDEX) index (46), and cases were classified according to the Braak and Braak neuropathologic staging (BST) (1) (Table 1).
Postmortem brains were sectioned sagittally into 2 hemispheres, and dissected samples were either frozen at −70°C or fixed in buffered 10% formalin for at least 3 weeks. Paraffin-embedded samples of the hippocampus and ERC were sectioned in approximately 10-μm-thick sections and processed for immunohistochemistry and morphometry. Additional tissue samples from AD subjects were also provided by the Northwestern University Alzheimer Disease Brain Bank.
Horseradish Peroxidase Immunohistochemistry With Tau-C3 and MN423 Antibodies
Using the immunoperoxidase technique, CPLL cases were processed for MN423 (23) and Tau-C3 (30) immunolabeling (7). In brief, endogenous peroxidase was inactivated from the samples by incubation in 0.5% hydrogen peroxide (Sigma Chemical Co., St. Louis, MO) in PBS, pH 7.4, for 10 minutes. After 2 hours of incubation with Tau-C3 or MN423 in 0.01% PBS-Triton (Research Organics, Cleveland, OH), a secondary horseradish peroxidase-conjugated anti-mouse secondary antibody (Zymed Laboratories, Inc., San Francisco, CA), diluted in PBS-triton, was added and incubated for 1 hour. Thereafter, a solution of 0.01% hydrogen peroxide combined with 0.06% 3,3′ diaminobenzidine (Sigma Chemical Co.) in PBS (pH 7.4) was used to develop the horseradish peroxidase enzymatic reaction. The reaction was allowed to proceed until faint dark brown coloration was observed in the sections; it was then stopped by washing in PBS. Some paired stained brain samples were counterstained with Cresyl Violet (Cole-Parmer Instrument Company, Vernon Hills, IL). The sections were mounted in DPX (BDH Laboratory Supplies, Poole, UK).
Immunolabeled brain sections were viewed and analyzed under bright-field microscopy using a Reichert 1762Y light microscope (Reichert, Inc., Depew, NY) and a Nikon Eclipse 80i light microscope (Nikon, Inc., Melville, NY). Observations were conducted directly through the 10× eyepiece of the microscope, and different objective lenses were selected (20×/25×/40×) depending on the purpose. For counting, low-magnification lenses were used, but when critical observations to identify NFTs were necessary, a 100× objective lens (numerical aperture, 1.4) was used. Counts of NFTs were performed directly from the microscope field and expressed as density of structures per square millimeter, as described earlier (47). In each case, the density of NFTs along the hippocampal formation and ERC was obtained by evaluating 3 adjacent fields selected randomly in each subarea. Neurofibrillary tangle densities were registered from the ERC layer II (ERCII), CA1, and subiculum to include the major areas comprising the perforant pathway (48, 49). Additionally, bright-field microscopy images were registered with a JVC Tk-C1380 camera (Victor Company of Japan, Ltd., Yokohama, Japan).
For statistical analyses, parametric Pearson r coefficients were used to correlate neuropathologic and clinical parameters. For pairwise combinations, group differences were calculated using parametric Student t-test. Statistical analysis was conducted with Graph Pad Prism statistics software, version 3.0 (GraphPad Software, Inc., San Diego, CA) and SPSS 10.0 for Windows (SPSS Inc., Chicago, IL).
Immunofluorescence With Tau-C3 and MN423 Antibodies
Some brain sections provided by the Northwestern University Alzheimer's Disease Brain Bank and the Cambridge Brain Bank were processed for double immunolabeling with either Tau-C3 or MN423 antibodies in combination with a panel of antibodies that map different residues of the tau molecule, that is, Biotinylated-Tau-46.1, (50) anti-pSer422, and anti-pSer396 (Biosource, Camarillo, CA). The procedure for immunofluorescence was as previously described (40), with some modifications. For 40-μm-thick sections, incubation with the primary antibodies was conducted overnight at 4°C, whereas 10-μm-thick sections were incubated only for 2 hours at room temperature. The corresponding fluorochrome-tagged secondary antibodies were incubated at room temperature for 2 and 1 hour, respectively. For sections counterstained with 0.001% thiazin red (Sigma Chemical Co.), incubations were conducted for 10 minutes and rinsed with distilled water. As previously reported, thiazin red staining facilitates the identification of β-pleated sheet conformation structures such as NFTs and neuropil threads in AD brains (26).
To analyze possible colocalization between Tau-C3 or MN423 and additional antibody epitopes, pairwise combinations were designed to avoid cross-recognition by the corresponding secondary antibodies. When 2 primary monoclonal antibodies were assessed, we combined immunoglobulin G(γ) and immunoglobulin M(μ), which were then visualized with the corresponding appropriate fluorescein isothiocyanate or tetramethylrhodamine isothiocyanate-anti-(γ) and anti-(μ) chain-specific secondary antibodies (Jackson Immuno-Research Laboratories, Inc. West Grove, PA). When the combination of primary antibodies involved 2 immunoglobulins from the same subclass (immunoglobulin G[γ]), 1 of these antibodies was coupled to biotin and visualized in a second-step incubation with Alexa-488/594-streptavidin complex (Molecular Probes, Eugene, OR). In other combinations, we used primary antibodies from different hosts (e.g. rabbit and mouse) to avoid cross-reactions.
Fluorescent samples were analyzed in a Zeiss-510 lasers canning confocal microscope (Carl Zeiss, Inc., Jena, Germany). Optical conditions of the confocal system image acquisition were as previously described (30, 40, 41). In brief, images were scanned from the top of the sample to 15-μm depths through the z axis. Horizontal z sections were collected and analyzed individually for colocalization patterns and projected as superimposed stacks of 2-dimensional images. Images were generated and later analyzed using the Zeiss LSM Image Browser Version 2.30.011 (Carl Zeiss). Additional images were also generated using a TCP-SP2 confocal laser scanning microscope (Leica, Heidelberg, Germany).
Neurofibrillary Pathology Recognized by Tau-C3 and MN423 in AD Brains
As previously reported (41), neurofibrillary pathology, primarily consisting of NFTs (Fig. 1B, C) and neuropil threads (Fig. 1A; arrows), was identified by Tau-C3 immunolabeling in the ERC and hippocampus of AD cases. Most NFTs were intracellular and were composed of compact fibers that were densely stained by Tau-C3 (Fig. 1B, C). For MN423 labeling, 2 populations of NFTs displaying the typical profile of either intracellular flame-shaped NFTs with compact fibers or extracellular NFTs with loose fibers in the neuropil were observed (Fig. 1D and E, respectively). In Figure 1C and F, lower-magnification images show large numbers of NFTs stained with the Tau-C3 and MN423 antibodies, respectively. In most of the cases, Tau-C3 recognized clustered or isolated neuropil threads, whereas MN423 labeling was much less evident in these structures (not shown). The fibrillar assembly of NFTs was corroborated by colocalization of immunofluorescent labeling with the antibody of interest and thiazin red. A large number of thiazin red-positive NFTs were decorated by either Tau-C3 or MN423 antibodies (not shown).
Comparative Density of Tau-C3 and MN423-Immunoreactive NFTs in the Brain of AD Cases and Controls
The density of NFTs was determined in 3 major areas of the hippocampal formation: ERCII, CA1, and subiculum. As shown in Figure 2A, within the CA1 area and the subiculum, there were significantly higher numbers of Tau-C3-immunoreactive NFTs in AD compared with nondemented cases (p = 0.0017 and p = 0.0001, respectively). There was, however, no significant difference between these 2 groups in the ERCII (p = 0.6051). In contrast, in the same cases, higher densities of MN423-positive NFTs were found in AD cases than in controls in all areas analyzed (Fig. 2B; p = 0.0013 for ERCII, p = 0.0002 for CA1, and p = 0.0061 for subiculum).
To determine whether the distributions of Tau-C3- and MN423-labeled NFTs were related, all cases were grouped and analyzed together. As shown in Figure 2C, an inverse correlation of NFTs was found between the density of Tau-C3- and MN423-recognized structures. Neurofibrillary tangles immunoreactive to MN423 predominated in ERCII (significantly higher than those for Tau-C3; p = 0.0108), which is known to be the earliest affected area in AD (1, 5, 51, 52). Conversely, NFTs recognized by Tau-C3 were predominant in areas in the hippocampal formation, such as the subiculum (significantly higher than those for MN423; p = 0.0006), which are affected later in AD. Findings in the CA1 area were intermediate, and the densities of NFTs immunoreactive to Tau-C3 and MN423 overlapped (p = 0.5415; Fig. 2C).
Correlation Between the Density of Tau-C3- and MN423-Decorated NFTs and Neuropathologic Braak Staging
To determine whether the accumulation of NFTs containing truncated proteins (D421 and E391) was related to the progression of the neurofibrillary pathology, as assessed by the BST (1), we performed a correlation analysis between densities of NFTs immunoreactive to Tau-C3 or MN423 and the BST.
As shown in Figure 3A, the accumulation of the NFTs recognized by both Tau-C3 and MN423 antibodies significantly correlated with the BST (r = 0.6122, p < 0.001; and r= 0.634, p < 0.0001, respectively). Moreover, there was a significant correlation of the progressive accumulation of NFTs recognized by Tau-C3 and those recognized by MN423 (Fig. 3B; r = 0.549, p = 0.0002).
In an analysis by area, the density of NFTs from all cases was averaged and correlated with the BST. In all areas analyzed, there was a significant correlation between the density of MN423-labeled NFTs and the BST (Table 2). When the correlation analysis was carried out for NFTs labeled with Tau-C3, however, significant values were obtained in the CA1 area and the subiculum but not in the ERCII (Table 2).
Correlation Between the Density of Tau-C3- and MN423-Recognized NFTs and the Clinical Severity of Dementia (CAMDEX)
To determine whether the accumulation of the NFTs containing truncated proteins (D421 and E391) was related to the clinical severity index (determined by the CAMDEX instrument) in AD, we performed a correlation analysis between the densities of NFTs immunoreactive to Tau-C3 or MN423 and the CAMDEX score. As shown in Figure 3C, the progressive accumulation of NFTs recognized by both Tau-C3 and MN423 antibodies correlated significantly with the clinical index of severity (r = 0.5051, p = 0.0009 and r = 0.5149, p = 0.0007, respectively).
To perform a specific analysis by area, the density of NFTs within each area was correlated with respect to CAMDEX score, and in all areas analyzed, there was a significant correlation between the density of MN423-labeled NFTs and the clinical index of severity (Table 3). When the correlation analysis was conducted on the density of NFTs decorated by Tau-C3, however, significant values were obtained in the CA1 area and the subiculum but not in the ERCII (Table 3).
D421- and E391-Truncated Tau Are Linked to the Prevalence of Apolipoprotein E ɛ4 Allele in AD Cases
Apolipoprotein E (ApoE) genotyping had previously been performed in the CPLL cases (Table 1). We attempted to determine whether the prevalence of the ApoE ɛ4 allele in AD cases is associated with the development of NFTs composed of D421- or E391-truncated tau. The CPLL cases were grouped according to their ApoE genotype, and the densities of Tau-C3- or MN423-recognized NFTs were compared between AD and normal control cases. As shown in Figure 4, the prevalence of the ApoE ɛ4 allele in AD cases was associated with increased numbers of Tau-C3- and MN423-immunoreactive NFTs compared with control normal subjects. Thus, when the ApoE ɛ3/4 genotype was present, the density of these lesions was significantly higher in AD cases compared with controls (p = 0.028 for Tau-C3 and p = 0.029 for MN423; Fig. 4A, B). This difference was not observed, however, when the groups carrying the ApoE ɛ3/3 genotype were compared. Only 1 AD case and 1 control subject were homozygous for the ApoE ɛ4 allele, but the densities of both Tau-C3- and MN423-labeled NFTs in the AD case were considerably higher than in the control case (Fig. 4A, B). The differences are attributed to the ApoEɛ status and not to the fact that we are comparing AD versus normal cases because for the ApoE ɛ3/3 group, which was analyzed in a similar way, there was no difference between AD and controls.
Transient Proteolysis of the C-Terminus of Tau Protein During NFT Maturation
To characterize the order of proteolytic events occurring at the C-terminus of tau protein during the process of NFT formation and maturation, we performed double label confocal microscopy with antibodies directed against cleaved tau (Tau-C3 or MN423) in combination with antibodies to additional phosphoepitopes located at the C-terminus of the molecule in the AD samples. In early-stage NFTs, colabeling of D421-truncated tau and tau with an intact C-terminus (recognized by Tau-C3 and biotinylated-Tau-46.1 antibodies, respectively) was commonly found (Fig. 5A-F). Additionally, D421-truncated tau also coexisted with phosphorylation at S396 (not shown) and S422 (Fig. 5G-I). Whereas phosphorylation at S422 precludes cleavage at D421, it is possible that labeling with the Tau-C3 and the anti-pSer396 might reflect a dual state within a tau monomer.
Because it is impossible for the same tau molecule to carry an intact C-terminus and be truncated at D421, the structures colabeled for D421-truncated and intact tau were considered to be chimeric NFTs in which the 2 distinct molecular species coexist within the same structure (Fig. 5A-I). On the other hand, interestingly, MN423 labeling was never associated with NFTs displaying Tau-46.1 (Fig. 6D-I). The only coexistence found for E391-truncated tau was in association with NFTs carrying tau molecules no longer than the pSer396 residue (Fig. 6A-C). These structures were interpreted as late-stage chimeric NFTs (Fig. 6A-C). In other combinations, we never found colocalization of biotinylated-MN423 with either Tau-13 (N-terminus marker) or Tau-C3 (not shown).
Pathologies Visualized by Tau-C3 and MN423 Antibodies
Posttranslational modifications, for example, phosphorylation, nitration, glycation, truncation, and conformational changes, have been proposed to contribute to the pathologic assembly of tau into PHFs. All of these modifications have been identified by immunohistochemistry within PHFs and NFTs (7, 16, 20, 21, 40, 41, 53). We previously analyzed the antigenicity of NFTs with Tau-C3 and MN423 antibodies, which independently recognize truncations of tau protein at D421 and E391 residues, respectively (7, 29, 40, 41). In the present study, we corroborated the observation that both antibodies independently recognize NFTs within the ERC and the hippocampal formation of AD patient brains. Most of the NFTs recognized by Tau-C3 correspond to the intracellular type that displays firmly packed fibers, sometimes with a well-preserved nucleus. MN423 also immunostained intracellular NFTs, but extracellular lesions with less compacted fibers predominated. This is consistent with the concept that extensive proteolysis of NFTs in the extracellular space favors the occurrence of E391 truncation (7, 29). Although intracellular NFTs are recognized by many antibodies that detect different modifications of tau, including phosphorylation, extracellular NFTs lack most of these epitopes (18, 25, 54-56). Polymeric tau that is extensively truncated at the C-terminus remains as a major component of the PHF core.
Distribution and Comparative Number of NFTs Decorated by Tau-C3 and MN423 in Allocortical Areas of AD and Age-Matched Individuals
As expected from previous studies, the density of NFTs containing E391 truncated tau was significantly greater in AD cases than in controls in all areas analyzed (Fig. 2B). Previous studies reported that lesions containing D421-truncated tau are apparent in different subregions of the temporal lobe of AD (41) and non-AD cases (ranging from BST I to IV) (41, 57), but their numbers have not previously been compared with those in normal age-matched controls. Here, we found that the density of NFTs composed of D421-truncated tau is also higher in AD than in age-matched controls in CA1 and the subiculum but not ERCII (Fig. 2A).
Previous studies of AD have described extensive neuronal loss in ERCII that correlate with the appearance of extracellular NFTs (51, 56, 58), which are composed primarily of E391-truncated tau. Because the Tau-C3 antibody detected intracellular NFTs only, it is not surprising that the density of NFTs stained with that antibody is lower in the ERCII. In control cases, however, this epitope had not appeared, resulting in similar density counts in ERCII in AD and controls (Fig. 2A).
To clarify whether the distribution of NFTs composed of either D421- or E391-truncated tau is mutually exclusive in different areas of the hippocampal formation, NFT counts were analyzed across all cases. As shown in Figure 2C, the ERCII predominantly has NFTs composed of E391-truncated tau, whereas in the subiculum, D421-truncated tau-positive NFTs were predominant. This topographic selectivity seems to be due to the maturation stage of NFTs, which is most advanced in the areas that are initially affected by AD progression (1, 5, 51). In contrast, the subiculum is affected later in disease progression; therefore, pretangle or early-stage NFTs predominate in it (1, 59, 60). The hippocampus is mostly affected at intermediate stages of AD pathology progression, and the prevalence of NFTs in different maturation stages is common in this area (1, 59, 60). In this regard, a recent study analyzing NFT composition in brains of limbic NFT dementia indicates that maturation of these structures is reflected as a shift in their composition from 4-to3-repeat tau isoforms. Mixed populations of NFTs composed by either 3- or 4-repeat tau isoforms were predominant in the hippocampus in that study (61). Our present results also demonstrate variable stages of maturation of NFTs in the hippocampus. As show in Figure 2C, the density of NFTs composed of either D421- or E391-truncated tau overlaps, which indicates that distinct populations of NFTs, composed of tau at different state of proteolysis, coexist in the same area.
Our results also confirm the inverse topographic distribution of NFTs composed of D421- or E391-truncated tau. They imply that proteolytic cleavage of the C-terminus of tau is unidirectional and transiently generated along the regional progression of the neurofibrillary pathology. This progression may be favored by regional vulnerability, that is, ERCII is more susceptible to cell death and early onset of fibrillar pathology (7). This vulnerability is not well understood but has been associated with abnormalities such as reduced mitochondrial metabolism (62), low glutamate receptor levels (63), and dendritic reorganization (64), which might predispose and precede the fibrillar aggregation of tau into PHFs in this area. Furthermore, toxicity of overexpressed recombinant D421-truncated tau has been demonstrated in cultured cells (34), and its presence in NFTs early in this area may therefore contribute to neuron degeneration. The D421-truncated tau-containing NFTs may be quickly transformed and converted into structures that are composed of more advanced E391-truncated tau, which are more numerous in the ERCII (Fig. 2C). In areas affected later in AD, such as the CA1 sector of the hippocampus, cell loss is less pronounced, and a mixed population of NFTs (composed of D421 or E391) may overlap (Fig. 2C). Selective increments of protective factors such as ceruloplasmin (65) and clusterin (66) have been described in this area, and these might confer more resistance to cellular degeneration; thus, NFTs in this area may exhibit more variable stages of maturation. Proteolytic transformation of NFTs in the subiculum is not advanced, so the more predominant truncation of tau protein occurs at the D421 residue.
Correlations With Neuropathologic and Clinical Parameters
The progression of the neurofibrillary pathology in AD has been staged on the basis of NFTs and neuritic plaque spread along the ERC, hippocampal formation, and isocortical areas (1). This hierarchic progression (BST) has been validated by some authors and shown to correlate with clinical manifestations of dementia (45, 67). This is, however, controversial because discrepancies have been observed when BST criteria have been applied to the oldest-old individuals (68-71). In those cases, BST Stages I and II, which are clinically silent in younger adults, are not clinically silent (70); similar results were observed in the Nun study (71). Previously, in the cases used in the present study, however, the numbers and distribution of NFTs visualized by silver staining followed the progression of the fibrillar pathology as assessed by the BST (1, 45).
The present study validates the pathologic significance of the progression of NFTs composed of truncated tau protein in AD and demonstrates that the 2 types of NFTs, composed of either D421- or E391-truncated tau, correlate with the neuropathologic stage of AD, as assessed by the BST (Fig. 3A). We previously reported the same significance for extracellular NFTs immunoreactive to MN423 and, here, now extend our analysis to the early truncation recognized by the Tau-C3 antibody. The results indicate that NFTs carrying these truncations may be reliable pathologic markers of the progression of the neurofibrillary pathology in AD. Indeed, other studies have reported positive correlations of counts of NFTs that are recognized by antibodies to different modifications in tau, including phosphorylation and truncation (7, 16, 29, 72, 73). Because NFT counts are increased during the progression of AD, positive correlations might simply be explained by this increment. In previous studies, however, single NFTs were staged at a particular level of tau processing (40, 41, 54, 74, 75), and immunologic probes such as the antibodies used in the present study may be better predictors of the combined progression of different NFTs during the evolution of AD.
When NFT counts were analyzed with respect to the CAMDEX, we found the same correlations despite the low density of NFTs in the ERCII. The relationship between NFTs decorated by either Tau-C3 or MN423 and cognitive function was significant when a correlation analysis of total counts by case was performed (Fig. 3C). When averaged by case, NFTs decorated by either of the antibodies paralleled the severity of dementia. These results are in agreement with our previous reports on these cases in which tau protein was analyzed biochemically and was found to correlate with the same neuropathologic and clinical parameters (10). Our present results indicate that NFTs carrying both early (D421) and advanced (E391) truncated tau reflect a hierarchic pattern of neurofibrillary pathology progression staged by the BST and imply that the total accumulation of these structures may be an accurate predictor of clinical dementia.
D421-Truncated Tau Is Associated With the Expression of the ApoE ɛ4 Allele in AD
Apolipoprotein E ɛ4 is expressed in both neurons and glia in the human brain, and the role of this protein has been attributed mainly to lipid transport (76). It was proposed several years ago that the prevalence of the ApoE ɛ4 allele in heterozygous or homozygous conditions was a risk factor for the development of late-onset dementia (77-80). One proposed biochemical explanation for this association was based on the ability of the ɛ-4 variant to direct and bind tau to its cytoskeletal compartment (79, 81). Additionally, the ApoE ɛ4 variant displays a low capacity to bind MAP2, another microtubule-associated protein (82). Whether this binding capacity is further reduced due to interaction with other candidate proteins or by additional, pathologic tau modifications such as phosphorylation is not yet understood.
To address this issue, we analyzed the association between the prevalence of the ApoE ɛ4 allele with respect to the neurofibrillary pathology displaying either D421- or E391-truncated tau. Interestingly, as shown in Figures 4A and B, the expression of the ApoE ɛ4 was significantly associated with more NFTs composed of D421- and E391-truncated tau in AD subjects but not in controls. These results suggest that the ɛ4 allele of ApoE may be linked to early and advanced processing of tau in which activation of proteolytic pathways generates a truncated molecule possibly via the abnormal processing of lipids or an abnormal association between these proteins. Analyses of the physical association between ApoE ɛ4 and early-truncated tau may clarify these issues.
Our results are also consistent with previous studies demonstrating that the ApoE ɛ4, but not the ApoE ɛ3-variant, inhibits neurite outgrowth in cultured neurons (83). Likewise, the ApoE ɛ3 variant has been shown to protect against apoptosis in endothelial cells, contrasting with the low capacity evidenced by ApoE ɛ4 (84). Together, these studies suggest different roles for the ɛ4 and ɛ3 variants of ApoE.
Maturation of NFTs in AD Depends on Sequentially Ordered Proteolysis of the C-Terminus of Tau
We previously found that during the progression of AD, NFTs seem to follow a linear evolution involving conformational changes and truncation of tau (40, 41). We proposed that full-length tau is prone to develop a conformational change that facilitates the appearance of early truncation of tau at D421. This truncation seems to promote structural rearrangements in the molecule, and tau may attain a new conformation associated with the loss of the N- and C-terminus. This conformational change would then be transformed by further truncations, resulting in cleavage at E391. This truncation emerges as the limitation of proteolysis of polymeric tau protein within the PHF; the protease-resistant remainder of the filament, called the PHF core, displays E391 truncation (13, 24, 25).
In the present study, we further characterized the progression of the proteolytic events occurring at the C-terminus of tau. As shown in Figure 5, close association was found between D421-truncated and C-terminus intact tau in early NFTs. This may indicate that monomeric D421-truncated tau and full-length tau polymerize together. Alternatively, although assembled of full-length tau, the filament may remain susceptible to the action of proteolytic enzymes such as caspases (85, 86). Whether caspases are capable of acting in polymeric substrates is uncertain at present; we have only found that active caspase 3 coexists with NFTs composed of D421-truncated tau in some neuronal cells (unpublished findings). We found Tau-C3 to colocalize with tau reactive to antibodies directed against both pSer396 (not shown) and pSer422 on NFTs (Fig. 5G-I). Because it is impossible for the same tau molecule to have both an intact extreme C-terminus and a truncation at D421, these have been referred to as chimeric NFTs, that is, NFTs that contain tau molecules in various stages of proteolytic processing. This concept implies that 2 independent molecules at different stages of abnormal processing may coexist in the same structure. Prior to this study, few reports demonstrated the chimeric characteristics of NFTs in AD (57). Therefore, it must be possible for the cleaved and full-length tau monomers to polymerize together or for enzymes to cleave the tau when it is already in a filamentous form. In agreement with our previous report, E391 truncation of tau appears later in the evolution of NFTs after the tau molecule has undergone conformational changes involving self-folding and rearrangement of the N-terminus (40, 41). Interestingly, in the present study, we also demonstrated that truncations do not occur randomly along the C-terminus because we were unable to observe the coexistence of C-terminus intact full-length tau and E391-truncated tau in the same NFT (Fig 6D-I). Conversely, E391-truncated tau failed to show associations with portions of the tau molecule C-terminal to the pSer396 residue. This subtype of chimeric NFTs commonly displayed characteristics of late-stage NFTs and was frequently found at advanced stages in the AD evolution.
The present findings contrast with previous in vitro models in which an early interaction between E391-truncated and full-length tau protein was proposed (87). In vitro polymerization paradigms have also demonstrated that E391-truncated tau polymerization rates were faster than when the C-terminus of the tau molecule was intact (38). We wish to emphasize that we observed no evidence whatsoever of the coexistence of D421 and E391 truncations in the same NFT, that is, these truncations seem to be physically and mutually exclusive. Our model (Fig. 7) proposes that the C-terminus of tau is sequentially and gradually truncated during NFT formation and maturation. The occurrence of early truncation at D421 may be a necessary step that promotes the exposure of different sites that allow subsequent proteolysis. Although no candidate enzyme(s) have been found that result in the E391 truncation, it is possible that truncation at E391 is the result of gradual proteolysis by non-site-specific enzymes, for example, carboxypeptidases (88). It has been recently proposed that tau is a substrate of aminopeptidases, and additional proteolysis of the C-terminus beyond the D421 cleavage site can conceivably be conducted in a similar manner (89, 90). Many studies have suggested that truncations play a role in driving the assembly of tau into PHFs and NFTs (30, 38, 91), possibly by facilitating conformational rearrangements of the molecule and strong interactions that may stabilize the emerging polymeric structure. Our data suggest that, although truncation, especially at E391, may contribute to filament stabilization or increases in assembly rates, it is unlikely to initiate polymerization. The proteolytic processing of the tau protein and the identification of chimeric NFTs may render a more reliable parameter to classify the evolution of the neurofibrillary pathology in AD.
In summary, this study demonstrates that the presence of D421 and E391 truncations correlate significantly with the BST and clinical severity of dementia. Moreover, they are topographically separated throughout the different areas affected in AD brains and not associated with each other during NFT formation and maturation.
The authors thank Drs. Virginia Lee, Peter Davies, and Michel Novak for the use of the Tau-46.1, Alz-50, and MN423 antibodies, respectively; to Dr. R.W. Berry for valuable criticisms; and the Cambridge University Brain Bank, United Kingdom; and the Northwestern University Alzheimer Disease Brain Bank for the access to the tissue used in this study.