Optimal blood tau species for the detection of Alzheimer’s disease neuropathology: an immunoprecipitation mass spectrometry and autopsy study

Alzheimer’s disease (AD) neuropathological changes include the intracellular accumulation of hyperphosphorylated tau (p-tau) as neurofibrillary tangles (NFTs) and neuropil threads as well as the aggregation of extracellular amyloid beta (Aβ) plaques [17]. Accurate and early detection of AD neuropathological changes during life is critical for timely therapeutic intervention [11]. Analysis of cerebrospinal fluid (CSF) and positron emission tomography (PET) represent gold standard methods for the in vivo detection of AD but are expensive and invasive. Plasma-based assays can now detect various phosphorylation tau sites in blood [9], providing an accessible and cost-effective alternative for disease detection with similar prognostic and diagnostic accuracies [15, 22, 28].

Evaluation of plasma p-tau biomarkers against post-mortem AD pathology is necessary for validating their use for clinical purposes. There are over 80 tau phosphorylation sites that can be abnormally phosphorylated during AD progression [27]. Among these, plasma p-tau181, p-tau217, and p-tau231 have been the most studied due to the availability of specific immunoassays for their quantification. Higher levels of plasma p-tau181 have been associated with increased odds for having autopsy-confirmed AD [20], and to discriminate between AD and non-AD pathology up to 8 years prior to death [13]. Plasma p-tau217 can distinguish between individuals with autopsy-confirmed AD and cognitively unimpaired individuals with better accuracy than alternative plasma and MRI-based measures of AD pathology [21], and has been shown to reflect both amyloid and tau pathologies [14]. Plasma p-tau231 is the earliest blood tau biomarker to increase in relation to AD neuropathological changes [1] and is selectively associated with amyloid plaques [2]. Only two plasma-to-autopsy studies conducting head-to-head comparisons of p-tau epitopes for the detection of AD neuropathology have been published [24, 25], but they examine two to three p-tau variants using immunoassays, which require separate sample preparation and analysis for each targeted epitope. Mass spectrometry (MS) allows for increased specificity for detection of low abundant proteins and can quantify numerous p-tau epitopes simultaneously in a single acquisition [3, 18]. This enables the characterization of the specific sites abnormally phosphorylated at each stage of the disease with a fair head-to-head comparison, providing the information for an effective utilization of blood tau biomarkers in both clinical practice and therapeutic trials.

Here, we used an IP-MS method to examine the association between six p-tau (p-tau181, p-tau199, p-tau202, p-tau205, p-tau217, p-tau231) and two non-phosphorylated (tau195–205, tau212–221) tau species in plasma ante-mortem samples and post-mortem AD neuropathological examination. We compared the accuracy of each tau species for discriminating brain donors with and without autopsy-confirmed AD, including among those with and without dementia during life. Each tau form was examined against Braak staging for NFTs, CERAD neuritic amyloid plaque score, and semi-quantitative ratings of p-tau severity across seven different cortical and subcortical brain regions.

A multitude of studies have demonstrated the association of plasma p-tau with amyloid and tau PET imaging, CSF biomarkers, and cognition, but relatively few have described the relation to autopsy findings. Such studies report results from a single phosphorylation site, and rarely compare multiple p-taus. Here, we determine for the first time, the levels of six phosphorylated and two non-phosphorylated tau species simultaneously quantified by MS in ante-mortem plasma of brain donors. We found that the concentrations of all p-tau and non-phosphorylated tau peptides were increased in neuropathologically confirmed AD, but p-tau217, p-tau205, and p-tau231 were the species with larger dynamic ranges. In particular, p-tau217 was the most accurate biomarker discriminating brain donors by AD and cognitive status and showed the highest associations with amyloid and tau neuropathological ratings. However, evaluation of the levels of the tau peptides with CERAD and Braak classifications indicated that different phosphorylated tau species increase at different stages of the disease. Taken together, these results contribute to the existing literature by not only demonstrating the capability of plasma p-tau to detect AD pathology but also revealing distinctions among p-tau species. We postulate that these differences will be valuable for selecting the most suitable biomarker in different scenarios, i.e., diagnosis, prognosis, or treatment monitoring.

By directly comparing the levels of the tau biomarkers in plasma, we observed that p-tau217, followed by p-tau205 and p-tau231, exhibited the highest fold-changes in AD cases compared with non-AD cases, greater than the other p-tau (181, 199, 202) and non-phosphorylated tau (195–209, 212–221). Those peptides with the largest dynamic ranges also had the greatest capacity to discriminate between neuropathologically confirmed AD and non-AD participants. Plasma p-tau217 was the biomarker with superior performance, followed by p-tau231 and p-tau205, which had similar accuracies. Normalization of p-tau217 and p-tau205 concentrations with their respective non-phosphorylated peptides (p-tau217/212–221 and p-tau205/195–209) rendered a similar accuracy, in concordance [18] and contrast [3] to previous work. When we analyzed the performance of the different tau species to detect AD when brain donors were stratified by dementia status at the time of the blood draw, all biomarkers showed a superior prediction in demented participants (CDR ≥ 1). This is probably due to higher tau levels in blood with more advanced disease. However, the accuracy in those with normal cognition or mild cognitive impairment (CDR < 1) remained high for most tau peptides. The same pattern as neuropathological examination was observed: p-tau217 was the biomarker with highest performance, followed by p-tau205 and p-tau231. These results are in line with previous studies showing that among the currently available plasma biomarkers, p-tau217 concentrations reflect underlying AD pathology with the greatest fidelity as determined by neuropathology [14, 24], PET imaging [21], and CSF biomarkers [12]. Longitudinally, p-tau217 has been shown as the only available blood biomarker with marked amyloid-dependent changes and with increases associated with clinical deterioration and brain atrophy [1]. These studies encompass comparisons with other tau phopshorylations, such as p-tau181 or p-tau231, but not with p-tau205 due to the previous lack of an available assay.

Through the utilization of our MS method, we recently described that plasma p-tau231, p-tau217, and p-tau205 exhibited stronger correlations with PET signals compared to other tau species, but their variations were associated differently with amyloid and tau PET [18]. P-tau231 was influenced more by amyloid, p-tau217 by both amyloid and tau, and p-tau205 primarily by tau. Here, we investigated associations between the plasma tau species and neuropathological scores of amyloid (CERAD) and tau (Braak) accumulation in the brain. We also observed that p-tau217, p-tau231, and p-tau205 displayed higher associations with the neuropathological staging of amyloid and tau, as well as with p-tau severity ratings at autopsy. Interestingly, plasma p-tau231 exhibited the most significant fold-changes with mild β-amyloid plaque density and reached a plateau between moderate and severe scores. Plasma p-tau217 showed the highest raise with moderate amyloid plaque density and at Braak III–IV, and continually increased thereafter. Meanwhile, plasma p-tau205 levels changed the most with severe amyloid plaque score and in Braak V–VI. The remaining peptides displayed their most pronounced changes—although moderate—in late disease stages (CERAD 3 and Braak V–VI). This observation suggests a widespread increase in phosphorylation and overall tau levels as the disease advances significantly.

These findings emphasize that not all tau markers are equal or indicative of the same brain changes. Among them, p-tau217 emerges as the most promising biomarker, exhibiting higher dynamic ranges and superior accuracy. However, other markers can provide valuable insights into different underlying processes. In the successful TRIALBLAZER-ALZ 2 clinical trial, participants were recruited based on their tau PET burden [23], showing the best results in low-medium tau PET population. The plasma p-tau profile characterization of each patient with a method like the one presented herein could assist in defining the target participants for a specific AD treatment.

There are limitations to the current findings. First, we did not explore trends in plasma tau species levels longitudinally. While plasma tau peptides accurately detect AD at autopsy, the clinical significance of a unit increase in raw values remains unclear. In addition, the findings are limited to participants from a single clinical cohort, which introduces the potential for selection bias. The present sample is from a National Institute on Aging-funded ADRC and is most representative of individuals who present to a clinic with concerns regarding their cognitive functioning. This population allows for development and validation of biomarkers, but inferences regarding risk and screening for AD in the general population cannot be made. Furthermore, the sample exhibited demographic homogeneity, with a majority identifying as white. This homogeneity could limit the broader applicability of the results to more diverse populations. We did not include the examination of other potential clinical, genetic, demographic, and social determinants of health variables and how they relate to the studied plasma biomarkers. Prospective population-based studies are warranted to address these knowledge gaps and identify cutoff values that optimize sensitivity and specificity for the detection of AD in a wider range of individuals. The AUC values reported in the present work appear lower than studies analyzing blood-based biomarkers against imaging and fluid measures. In most investigations, PET and CSF markers demonstrate an accuracy of 90–95% when compared to post-mortem. Hence, in the context of in vivo research where p-tau is shown to have an AUC > 0.95, this evaluation is conducted against a “gold standard” that is not precise. Thus, the likely explanation for the AUC values being less than 0.90 in this study is rooted in the evaluation against actual neuropathology. In addition, we are comparing blood biomarkers to end-stage disease—which is an unlikely scenario (e.g., it will be a preclinic or MCI stage). Future studies should address if the overlap cases between AD and non-AD groups would benefit from a two-step diagnostic workflow, as recently proposed [6]. In that case, blood tau biomarkers could serve as a first screening tool for AD pathology (step 1), followed by confirmatory testing with CSF Aβ42/Aβ40 or PET imaging (step 2) only in patients with intermediate risk at the first step. This would reduce the need for confirmatory testing while accurately classifying patients, offering a viable option to centers that do not have access to specialized testing. Lastly, we recognize that while MS techniques provide the potential for multiplexing various phosphorylations, offering a distinct platform for AD staging, they demand more extensive efforts for scalability and implementation in clinical routine than other platforms.

In summary, our findings support plasma p-tau217 as the most promising p-tau species for detecting AD brain pathology. Plasma p-tau231 and p-tau205 may additionally function as markers for different stages of the disease.