Pilot tests of tau ELISAs to analyze the brains of AD patients and non-AD controls.
We developed several tau ELISAs by combining several antibodies against epitopes raging from the N-terminal to C-terminal regions of tau (Fig. 1A). All these ELISAs showed good dose-dependent standard curves using recombinant human tau-441 protein (Supplementary Fig. 1). We then tested samples pooled from the RIPA-insoluble fraction (i.e., GuHCl fraction) of the frontal cortex of 8 individuals with AD or without AD (control) by using these ELISAs. The results are summarized in Fig. 1B. When antibodies against the N-terminal to middle regions of tau were used in ELISA, the control sample showed relatively high levels of tau. In particular, including an antibody against the N-terminal region (i.e., Tau13) showed relatively high tau levels (25–400 ng/mg). In the ELISAs using Tau13 antibody, a mild difference between the control and AD samples was observed; AD sample showed approximately 1.2- to 5-fold higher levels of tau than those in the control sample. On the other hand, when antibodies against the middle to C-terminal regions of tau were used, the control sample showed relatively low levels of tau (below 100 ng/mg). More importantly, when antibodies against the later-middle (i.e., 218–225 a.a. of Tau-441: epitope of Tau5 antibody) to C-terminal regions of tau were combined, substantial differences between the control and AD samples were generally observed; AD sample showed more than 10- to 100-fold higher levels of tau than the control sample. These results suggest that tau ELISAs have different reactivities depending on the epitopes targeted by tau antibodies, which might affect the evaluation of tau pathology in brains of AD patients.
Detailed comparison of tau ELISAs to distinguish the brains of AD patients and non-AD controls.
To confirm our pilot results, we next analyzed individual RIPA-insoluble fraction of frontal cortex of AD and non-AD control brains (n = 60, Supplementary Table. 1) by using representative tau ELISAs that used Tau13 (epitope: N-terminal 20–35 a.a. of Tau-441) or OST (epitope: MTBR 323–363 a.a. of Tau-441) as the capture antibody. The results are shown in Fig. 2A. When Tau13 was used as a capture antibody, the brains of control cases showed relatively higher levels of tau (100–600 ng/mg, also shown in Supplementary Table 2), and the brains of AD patients showed 1.5- to 4-fold increases in tau levels, compared to those in control brains; moreover, this difference was not always significant. When OST was ued as a capture antibody, the brains of control cases showed relatively lower levels of tau, especially when antibodies against the middle to C-terminal regions of tau were used as detection antibodies (less than 100 ng/mg, also shown in Supplementary Table 2). Moreover, consistent with our initial results, ELISAs using antibodies against the later-middle to C-terminal regions showed robust differences between control and AD; the brains of AD patients had over 500-fold higher levels of tau accumulation than the brains of controls (also shown in Supplementary Table 2). To confirm that ELISAs using antibodies against the later-middle to C-terminal regions of tau would be better to distinguish AD patients and non-AD controls, we performed ROC curve analyses. Indeed, when OST antibody was used as a capture antibody and antibodies against the middle to C-terminal regions of tau were used as detection antibodies, the AUC was more than 0.96 (Fig. 2B). In particular, OST-77G7 ELISA showed the best result (AUC = 0.97). On the other hand, when Tau 13 was used as a capture antibody, the AUC was relatively lower, ranging from 0.65 to 0.79, and Tau13-Tau5 ELISA yielded the lowest AUC value (AUC = 0.65). The significant difference in AUC between each ELISA is described in detail in Supplementary Table 3. These results confirm that tau ELISAs using antibodies against the later-middle to C-terminal regions of tau can better distinguish AD patients and non-AD controls.
We also analyzed the RIPA-soluble fraction to determine how far this easily-extractable fraction can distinguish AD and non-AD controls by these ELISAs. We tested OST-77G7 and Tau13-Tau5 ELISAs, which showed the best and worst AUC, respectively, for the RIPA-insoluble fraction. When tested by Tau13-Tau5 ELISA, the control cases and AD patients showed almost similar values without a significant difference (p = 0.1283, Supplementary Fig. 2A). On the other hand, when tested by OST-77G7 ELISA, AD patients showed a significant approximately 1.5-fold increase (median value) in tau levels (p = 0.0053, Supplementary Fig. 2A). Notably, in OST-77G7 ELISA, a strong correlation was observed (r = 0.85) between RIPA-soluble tau and RIPA-insoluble tau, while the RIPA-soluble tau levels in some AD patients overlapped with those of controls (Supplementary Fig. 2B). Indeed, in the ROC curve analysis to distinguish AD patients and non-AD controls, the AUC was 0.72 (Supplementary Fig. 2C), which was weaker than that of the GuHCl fraction. These results indicate that the RIPA-insoluble (i.e., GuHCl) fraction is more suitable than the RIPA-soluble fraction for distinguishing AD patients and non-AD controls by this tau ELISA.
To assess whether other popular extraction methods, such as using Sarkosyl instead of RIPA, give a similar result, we also tested the Sarkosyl-insoluble fraction from a small number of subjects. In OST-77G7 ELISA, AD samples showed significantly higher Sarkosyl-insoluble tau levels than non-AD control samples (p = 0.0003), while in Tau13-Tau5 ELISA, there was no significant difference between these samples (p = 0.1046) (Supplementary Fig. 3A). Notably, tau levels in the Sarkosyl-insoluble fraction measured by OST-77G7 ELISA correlated well with tau levels in the RIPA-insoluble fraction measured by the same OST-77G7 ELISA (Supplementary Fig. 3C); in addition, these results showed good AUC values (AUC = 0.97) for distinguishing AD patients and non-AD controls, similar to those of the RIPA-insoluble fraction (Supplementary Fig. 3D). These results indicate that our findings can be applied to other popular extraction methods, including a method using Sarkosyl.
Correlation of tau ELISA results with NFT neuropathological stage and other AD-related neurodegenerative markers.
To determine whether these tau ELISAs indeed reflect tau accumulation in the brain, we analyzed the correlation between Braak NFT stage and tau levels in the RIPA-insoluble fraction determined by each ELISA. The results are summarized in Table 1. When Tau13 was used as the capture antibody, we observed mild-to-moderate correlations between tau levels by ELISA and Braak NFT stage (r = 0.35–0.67). In particular, tau levels measured by Tau13-Tau5 ELISA showed the lowest correlation (r = 0.35, p = 0.056, Fig. 3A). On the other hand, with ELISAs using OST as a capture antibody, we obtained better correlations with Braak NFT stage, especially when detection antibodies against the later-middle to C-terminal regions of tau were used (r > 0.80), including OST-77G7 ELISA (r = 0.81, p < 0.001, Fig. 3D). These findings indicate that tau ELISAs using antibodies against the later-middle region to C-terminal regions of tau can better reflect the pathological accumulation of tau in the brain.
Table 1
Correlation between tau levels determined by each ELISAs and Aβ, synaptic/ neuronal markers, and other neurodegenerative markers.
| AUC | Braak NFT | Aβ1–40 | Aβ1–42 | GFAP | CD11b | ApoE GuHCl |
Tau ELISA | | r | p-value | r | p-value | r | p-value | r | p-value | r | p-value | r | p-value |
Tau13-HT7 | 0.79 | 0.55 | < 0.001 | 0.39 | 0.017 | 0.23 | 0.753 | 0.16 | 1.00 | 0.12 | 1.00 | 0.34 | 0.058 |
Tau13-Tau5 | 0.65 | 0.35 | 0.056 | 0.21 | 0.942 | 0.26 | 0.438 | 0.09 | 1.00 | 0.01 | 1.00 | 0.21 | 0.941 |
Tau13-77G7 | 0.76 | 0.53 | < 0.001 | 0.42 | 0.008 | 0.46 | 0.004 | 0.19 | 1.00 | 0.10 | 1.00 | 0.37 | 0.029 |
Tau13-Tau46 | 0.76 | 0.50 | < 0.001 | 0.43 | 0.005 | 0.52 | < 0.001 | 0.04 | 1.00 | -0.01 | 1.00 | 0.37 | 0.031 |
OST-Tau13 | 0.84 | 0.67 | < 0.001 | 0.50 | < 0.001 | 0.41 | 0.014 | 0.21 | 0.924 | 0.26 | 0.408 | 0.43 | 0.005 |
OST-HT7 | 0.9 | 0.73 | < 0.001 | 0.67 | < 0.001 | 0.70 | < 0.001 | 0.23 | 0.699 | 0.09 | 1.00 | 0.53 | < 0.001 |
OST-Tau5 | 0.96 | 0.82 | < 0.001 | 0.63 | < 0.001 | 0.50 | < 0.001 | 0.38 | 0.023 | 0.30 | 0.181 | 0.53 | < 0.001 |
OST-77G7 | 0.97 | 0.81 | < 0.001 | 0.65 | < 0.001 | 0.49 | < 0.001 | 0.45 | 0.003 | 0.42 | 0.008 | 0.60 | < 0.001 |
OST-Tau46 | 0.96 | 0.83 | < 0.001 | 0.64 | < 0.001 | 0.64 | < 0.001 | 0.26 | 0.379 | 0.20 | 1.00 | 0.55 | < 0.001 |
We also analyzed the correlation with the levels of Aβ, inflammatory cell markers, GFAP and CD11b, and apoE protein. These results are also summarized in Table 1, and representative results of the correlations of Tau13-Tau5 ELISA, and OST-77G7 ELISA are also shown as graphs in Fig. 3B, C, E, and F. In brief, while both Aβ40 and Aβ42 in the GuHCl fraction were increased in the brains of AD patients (Supplementary Table 2), Aβ40 in particular tended to have a better correlation with tau levels measured by ELISAs that combined antibodies against the middle region to C-terminal regions of tau (r > 0.60), likely because Aβ40 increases during AD progression while Aβ42 reaches a plateau at an early stage (39). Regarding inflammatory markers, we confirmed that the levels of inflammatory cell markers tended to be increased in AD patients compared to controls (GFAP levels were significant, but CD11b levels only exhibited a trend, Supplementary Table 2). The results of the ELISAs using Tau13 as a capture antibody were generally not well correlated with the levels of glial markers (GFAP: r = 0.04–0.19; CD11b: r = -0.01–0.12). On the other hand, the results of ELISAs using OST as a capture antibody tended to be better correlated with the levels of these markers (GFAP: r = 0.21–0.45; CD11b: r = 0.09–0.42). In particular, their correlations with tau levels measured by OST-77G7 ELISA were significant (GFAP: r = 0.45, p = 0.003; CD11b: r = 0.42, p = 0.008). It is known that apoE accumulates on NFTs, especially extracellular NFTs (41, 42), in addition to amyloid plaques. Since accumulated apoE could be evaluated in the GuHCl fraction (38), we analyzed the correlation with apoE levels in the GuHCl fraction, and we observed that the apoE levels were better correlated with tau levels measured by ELISAs using OST as a capture antibody (r = 0.43–0.60), rather than those using Tau13 (r = 0.21–0.37). These results indicate that Tau ELISAs with antibodies against the later-middle to C-terminal regions of tau (especially OST-77G7 ELISA) can better reflect AD-associated pathological changes, including Aβ, inflammatory cells, and apoE accumulation in addition to tau accumulation.
Distinct reactivity of tau antibodies by western blotting analysis
To address the reason why ELISA results are different depending on the epitopes targeted by tau antibodies, we performed western blotting analysis. In the GuHCl fraction, tau was generally detected as the monomer form (50–64 kDa) and aggregated form (> 64 kDa) by western blotting (Fig. 4A). When an antibody against the N-terminal region of tau was used (i.e., Tau13), there was a trend that the monomer form of tau was more clearly visible than its aggregated form. On the other hand, when antibodies against the middle to C-terminal regions of tau were used, the aggregated form was clearly observed in AD patients, especially with antibodies against MTBR to the C-terminal region of tau (i.e., 77G7 and Tau46); these results were confirmed by densitometric analysis (Fig. 4B). These difference in the reactivity of tau depending on antibody recognition of epitopes from the N-terminal to C-terminal regions of tau might explain the observed difference in the ELISA results.
Tau accumulation across brain regions during disease development
To address whether these tau ELISAs can evaluate the pattern of tau spread during AD development, we analyzed multiple brain regions of 18 individuals with different stages of AD (demographic information of each case is shown in Supplementary Table 4). We used OST-77G7 ELISA as one of the best tau ELISAs. Results are shown in each individual case (Fig. 5). Indeed, AD patients showed increased insoluble tau levels in several brain regions, including the limbic areas and neocortical areas, but less remarkable levels in several subcortical areas, including the striatum and thalamus, and cerebellum. On the other hand, individuals with early AD pathology (PSC or NFTC), showed increased insoluble tau levels only in the entorhinal cortex, but not apparent in other brain regions. Interestingly, some of these individuals showed somewhat increased insoluble tau levels in the amygdala and temporal cortex, although the extent of tau levels was much fewer compared to that in the entorhinal cortex. The control groups (Braak NFT stage I) generally showed low insoluble tau levels. Notably, one individual showed apparent insoluble tau levels in the entorhinal cortex (Cont#5). When Tau13-Tau5 ELISA was used, there were no such trends recapitulating the progression of Braak NFT stage (Supplementary Fig. 4). These results indicate that OST-77G7 ELISA can address the pattern of tau spread pattern across brain regions during AD development.