With recent advances in assay development, it is now possible to measure plasma levels of P-tau217 in clinical research, drug trials, and perhaps also in clinical practice in the future. To best interpret the P-tau217 results it is important to understand what plasma P-tau217 levels represent. Although proposed frameworks have largely assumed that soluble P-tau best mirrors tau deposition11, we and others have previously noted that CSF P-tau may be increased also as a function of Aβ deposition7,8 —i.e., provide an indicator of Aβ-related tau pathophysiology that anticipates the development of measurable tau tangle pathology. We therefore hypothesized that plasma P-tau217 levels would be related both to Aβ and tau aggregation. Supporting our hypothesis, we found that both Aβ plaques and tau tangles were independently associated with higher plasma P-tau217 levels, both when measured post-mortem by neuropathological quantification (providing the most sensitive indicator of tau tangle deposition) or in vivo by PET imaging (which may not detect tau tangles until there is more substantial and spatially extensive tangle deposition). Aβ plaques and tau tangles also interacted, so that that highest plasma P-tau217 levels were seen in individuals who had high levels of both amyloid plaques and tau tangles. The independent effects of Aβ load and tau tangle load on plasma P-tau217 were significant, with considerable increases in model R2’s when combining Aβ and tau predictors compared to when using just one of them alone. Even though the results were mainly similar when using either postmortem neuropathology examination or in vivo PET imaging to detect tau tangles, we found that plasma P-tau217 was more strongly associated with tau tangles than Aβ plaques when using neuropathology examination, which was not the case when using PET (compare Fig. 1A with Fig. 4A). This is probably explained by the fact that tau PET imaging does not reliably detect lower amounts of tau aggregates in the brain, as shown in a recent end-of-life study by Fleischer et al evaluating the diagnostic performance of 18F-flortaucipir9. However, even when using neuropathology examination, Aβ pathology had an independent effect on plasma P-tau217 levels, and in cases with more limited tau pathology (mainly restricted to the medial temporal lobe) plasma P-tau217 correlated with Aβ pathology but not tau tangles, although we acknowledge that the methods used for tau quantification in the brain may not be sensitive to the earliest most subtle tau aggregation (Fig. 3). These results are congruent with recent studies showing that both blood and CSF P-tau217 is associated with Aβ PET also in tau PET negative cases7,12. Together, these results imply that plasma P-tau217 is not associated with tau tangle pathology independent from Aβ pathology, including primary age-related tauopathy (PART, see Fig. 3B)1. However, plasma P-tau217 is associated with Aβ pathology, even during the early stages of the disease (Fig. 3A), and it is further strongly associated with more widespread tau tangles in cases with Aβ pathology (Fig. 1C and 4C). More research is needed to understand the molecular and cellular mechanisms behind how early Aβ pathology leads to changes in extracellular P-tau217 levels. Overall, the link between Aβ pathology and increased plasma P-tau217 is in line with recent results from cell models13 showing that phosphorylation of tau is increased in the presence of Aβ.
Plasma P-tau217 also provided statistical mediation (but causal mediation was not tested) for the effects of Aβ load on tau load in both the neuropathology dataset, and for the BioFINDER-2 PET-imaging dataset. Intriguingly, the mediation effect of P-tau217 was especially strong for neuropathological quantification of tangle burden in the neocortex (excluding tangle count densities in entorhinal cortex and hippocampus, Fig. 2C), where plasma P-tau217 explained up to 77% of the effect of Aβ plaques on tangle burden and the direct effect of Aβ plaques on tangles became non-significant when correcting for P-tau217. The in vivo measures with Aβ and global tau PET showed similar results, with considerable mediation for plasma P-tau217 on global tau PET (66%), but less mediation (29%) when using tau PET quantified in the entorhinal cortex or hippocampus. This shows that plasma P-tau217 may be especially important for the spread of tau outside of the medial temporal lobe. Isolated medial temporal lobe tangle pathology (which may even appear in the absence of Aβ pathology, including “PART”14) is not strongly related to increased phosphorylation and/or secretion of tau. When grouping the subjects into three hypothetical “disease stages” (Fig. 6), we found no correlations between plasma P-tau217 and either Aβ PET or tau PET in those that were negative for both Aβ and tau PET (we consider these individuals to not have AD), significant correlations to Aβ PET (but not tau PET) in those who were within the positive range for Aβ PET only (we consider this an early stage of AD), and significant correlations to both Aβ PET and tau PET (but strongest with tau PET) in those who were positive for both Aβ and tau PET (we consider this a later stage of AD). Taken together, these findings are congruent with the hypothesis that Aβ pathology leads to an increased release of soluble P-tau, which in turn is associated with a spread of tau tangle pathology beyond the medial temporal lobe. However, for these tests we note different sensitivities of the biomarkers to detect underlying pathologies could impact their correlations. Figure 8 shows a hypothetical model which integrates the findings from this study and previous studies on biomarkers and development of AD. Additional valuable information could come from Aβ overproducing mouse models that do not develop tau tangles, to characterize the extent to which Aβ could promote tau secretion even in the absence of subsequent tau tangle deposition7,15.
Over the last years, there has been an increased interest in anti-tau treatments for AD16. Our results suggest that therapies targeting P-tau217 may break the link between buildup of aggregated Aβ and tau, and thereby reduce atrophy and cognitive decline in AD (which are both strongly associated with tau pathology17,18). At least one tau-directed treatment (using an anti-tau antibody) has been reported to reduce CSF levels of P-tau217 (conference report19). One intriguing possibility from these findings is that hyperphosphorylated soluble tau may be involved in the pathogenesis of AD, and it may be therefore be a suitable treatment target. Further drug development may incorporate plasma P-tau217 data in early stages to select drug candidates most likely to have beneficial effects on the biochemistry of tau in the brain in AD.
A limitation of this study is the relatively small dataset in the neuropathological cohort. We used the large BioFINDER-2 cohort to validate the neuropathology findings, but we acknowledge that tau PET has limited sensitivity for early stage tau pathology (with isolated tangle pathology in entorhinal cortex and hippocampus)9. We therefore refrained from doing some of the subgroup analyses in BioFINDER-2 (focusing on MTL tau in individuals with negative global tau).
In summary, plasma P-tau217 is independently associated with both amyloid plaques and tau tangles, including associations with amyloid pathology even in cases with restricted tau tangle pathology, and strong associations with tau tangles in amyloid positive cases.