Site-Specific Phospho-Tau Aggregation-Based Biomarker Discovery for AD Diagnosis and Differentiation

Differential Detection of Tau Aggregates in AD brains but not in the Controls.

We first investigated the aggregated tau species patterns in postmortem AD brains and age-matching normal controls using two independent anti-tau antibodies: Tau-5 is a mouse monoclonal antibody, generated using amino acid 210–241 epitope of bovine tau as immunogen; A1103 is a rabbit polyclonal antibody that recognizes carboxy-terminal human tau sequence (amino acids 283–441 of 2N4R-tau sequence). With both antibody probes, AD brain tissues demonstrated clear “smear”-type of HMW, tau antibody-detectable species characteristic of typical protein aggregates (Figs. 1A & 1C, highlighted with red rectangles of AD samples), whereas there were very little or no such tau aggregates in the control samples (Figs. 1A & 1C, highlighted with red rectangles of CTL samples). Monomeric tau species as well as low-molecular-weight tau antibody-detectable species (likely partially degraded tau fragments) were observed from both AD and control brain tissues. Separately, we compared AD brain samples with those of rare tauopathies PSP, CBD, and PiD. Interestingly, rare tauopathy brain homogenates showed minimal aggregated tau species, in stark contrast with AD brain samples (Figs. 1B & 1D, highlighted with red rectangles). Control experiments demonstrated primary antibody-specific recognition of protein bands (Fig. S1). These initial differential detections led us to a systematic screening of postmortem AD, MCI, rare tauopathies, and normal control brain samples with a comprehensive panel of site-specific phospho-tau antibodies.

Identification of Novel Site-Specific Phospho-Tau Sites as Potential Biomarkers.

Selection of tau phosphorylation sites were primarily based on the high frequencies of these tau phosphorylation sites in AD patients from recent high-resolution quantitative proteomics map of PTMs on tau isoforms extracted from postmortem AD brain tissues,¹⁵ as well as available site-specific quality tau antibodies. Only significant AD patient frequency p-tau sites were chosen (patient frequency > 10%; Fig. 2). More than twenty such p-tau sites were identified and over a dozen p-tau sites were selected (Fig. 2, highlighted in red) to screen the postmortem brain homogenates for HMW, site-specific p-tau antibody-detectable aggregated tau species with western blotting analysis. These sites include p-tau T181 and p-tau T217 sites that were recently being developed as AD biomarkers for diagnosis and/or tracking disease progression, therefore these sites served as validation controls (for an excellent review on phospho-tau AD biomarkers, see reference Leuzy et al., 2022).^13,21–24

Similar to western blotting analyses of postmortem AD and control subject brain tissues with anti-tau antibodies, our search for candidate p-tau biomarkers was mainly based on the presence/absence of the HMW p-tau aggregate bands comparing AD brains with normal controls. As expected, p-tau T181 (Fig. 3A) and p-tau T217 (Fig. 3E) blots showed the presence of much stronger intensities of “smear-like” HMW protein aggregates from AD brains detected by respective site-specific p-tau antibodies (highlighted by the red rectangles). Separate anti-GAPDH (or anti-actin) western analysis showed an approximate comparable amount of loading controls. In addition to p-tau181 and p-tau217 controls, we identified p-tau198, p-tau202, p-tau231, and p-tau404 sites (Figs. 3B, 3C, 3F, & 3H) that differentiated AD brains from control tissues. However, p-tau205 and p-tau262 sites were not able to discriminate AD from normal controls (Figs. 3D & 3G). We further compared AD brain samples with those from rare tauopathies. All these sites, except S262, had minimal or no HMW p-tau aggregates in rare tauopathies detected by each site-specific p-tau antibody, in clear contrast to those HMW aggregates in the AD brains (Fig. 3; highlighted in the rectangles in each of the right panels). To our knowledge, p-tau198, p-tau202, and p-tau404 have not been systematically evaluated as potential biomarkers to differentiate AD from normal controls and to differentiate AD from rare tauopathies. P-tau231 has been reported as a biomarker for AD diagnosis.^14,21,25

To further rationalize our selection, we performed semi-quantitation of the HMW phospho-tau aggregates western blotting bands (“smears”) by densitometry function in the LICOR Odyssey western blot detection system. Western blotting quantitation showed p-tau198 and p-tau404 as two most promising new markers with highest intensity ratio for AD samples over the controls (Fig. 4A), or AD samples over rare tauopathies subjects (Fig. 4B). p-tau217 and p-tau231 showed similarly high contrasts as those for p-tau198 and p-tau404, however, one of the most studied AD biomarker p-tau181 somehow showed less contrasts.

P-tau S198 is a promising biomarker for AD diagnosis with comparable differentiation power of the existing p-tau biomarkers.

P-tau198 was identified from our antibody screening as a site with one of the biggest contrasts of phosphorylated tau aggregates between AD and control brains, and AD versus rare tauopathies (Fig. 3B & Fig. 4). We therefore selected this site as a proof-of-principle example to demonstrate the feasibility of our approach for new biomarker discovery. We quantified p-tau198 levels in the brain homogenates of temporal cortex in the discovery cohort using ELISA assays. Demographic characteristics and neuropathological diagnosis of this cohort of AD, rare tauopathies, and non-AD patients are described in Tables S1 and S2. Brain p-tau198 levels significantly increased in AD when compared with controls (2.303 ± 0.107 vs. 0.971 ± 0.172 relative units with a 2.37-fold increase and a p value < 0.0001; Fig. 5A). Diagnostic performance to identify AD cases from normal controls yields an AUC of 0.96 (CI_95% = 0.90–1.00) (Fig. 4B). P-tau198 not only readily identified AD cases from normal controls, but also was capable of discriminating AD from PSP and CBD with differentiating power of AUC of 0.98 (CI_95% = 0.94–1.00) for AD vs. PSP and AUC of 0.99 (CI_95% = 0.96–1.00) for AD vs. CBD (Figs. 5A & 5B). AD, PSP, and CBD all belong to 4R tauopathies.

We also compared p-tau198 with p-tau181 and p-tau217 side-by-side, two known biomarkers in current clinical use. P-tau198 showed comparable differentiating power with those of p-tau217 but slight better diagnostic performance than those of p-tau181, such as the AUC measures in AD vs. controls and AD vs. PSP discrimination tests (Figs. 5C–5F). P-tau217 showed differentiating power to identify AD from controls with AUC=0.98 and CI_95% = 0.93–1.00 (Fig. 5F). Average p-tau217 concentrations were 0.81 ± 0.04 relative units in the AD brains and 0.33 ± 0.05 relative units in the control brains, a 2.43-fold increase with a p value < 0.0001 (Fig. 5E). P-tau181 tests showed differentiating power to identify AD from controls with AUC=0.89 and CI_95% = 0.76–1.00 (Fig. 5D). Average p-tau181 concentrations were 2.58 ± 0.17 relative units in the AD brains and 1.39 ± 0.08 relative units in the control brains, a 1.86-fold increase with a p value < 0.0001 (Fig. 5C). Due to low case number of PiD brains, we did not evaluate the diagnostic performance of various p-tau markers related to PiD cohorts.

Phospho-tau S198 is a promising biomarker for MCI diagnosis.

To test the potential of p-tau198 as a biomarker for more challenging MCI early AD diagnosis, we quantified p-tau198 levels in the brain homogenates of temporal cortex in the discovery cohorts of MCI and cognitively normal subjects using ELISA assays.

Brain p-tau198 levels significantly increased in MCI brains when compared with cognitively normal controls (1.18 ± 0.15 vs. 0.75 ± 0.11 relative units with a 1.57-fold increase and a p value < 0.05; Fig. 6A). P-tau198 was capable of discriminating MCI cases from cognitively normal controls with an AUC value of 0.75 and CI_95% = 0.58–0.92 (Fig. 6D). In comparison to two existing site-specific p-tau biomarkers T181 and T217, p-tau217 has somewhat better diagnostic performance with an AUC value of 0.85 and CI_95% = 0.71–0.99 (Figs. 6C & 6F). P-tau181 levels in MCI brains were not significantly elevated in comparison with those of cognitively normal brains (p=0.07; Fig. 6B) and diagnostic ROC curve showed lower discrimination capabilities (AUC=0.68; Fig. 6E). No significant difference was found between APOE4 carriers vs. non-carriers in relation to p-tau198 ELISA levels (Fig. S2A). Correlation analyses showed that there are no significant correlation between PMI vs. p-tau198 levels or age vs. p-tau levels measured by ELISA assays (Figs. S2B & S2C).

In summary, we demonstrated an experimental approach based on misfolded tau aggregation detection for new biomarker discovery. Using site-specific phospho-tau antibodies, such detection is therefore residue specific. We were able to identify multiple potential p-tau biomarkers from an antibody screening effort. The identified p-tau198 was able to not only differentiate AD from non-AD controls and AD from 4R-dominant PSP and CBD subjects with differentiating power comparable to two existing p-tau181 and p-tau217 biomarkers, but more importantly, we showed that p-tau198 is a promising biomarker capable of discriminating MCI from cognitively normal subjects as p-tau217. To our knowledge, there is no well-established biomarkers to diagnose MCI currently in clinical practice.

Postmortem Brain Tissue Collection.

Postmortem brain tissues from non-AD controls, cognitively normal subjects, mild cognitive impairment (or mild impairment) subjects, AD dementia patients, and rare tauopathy patients were collected from the Bryan Brain Bank and Biorepository of the Duke University Medical Center Alzheimer’s Disease Research Center (Duke/UNC ADRC). Demographics, NeuroStatus, Braak staging of AD-tau and other neuropathology diagnosis are described in detail in Tables S1 and S2. All participants were enrolled in the autopsy and brain donation program of the Joseph and Kathleen Pryce Bryan ADRC as previously described.⁴¹ All subjects gave informed consent. All protocols and the use of postmortem brain tissues for research were approved by the Duke University Institutional Review Board. Neuropathological evaluation was performed following published guidelines.^42–45 Clinical diagnosis of dementia, MCI, and cognitively normal was based on clinical dementia rating (CDR). Consensus meeting were held annually to review and update the clinical diagnosis based on contemporaneous NIA-AA criteria.⁴¹

Brain Tissue Handling and Analysis.

Brain tissues were homogenized at 10% (w/v) in lysis buffer (10 mM Tris pH 7.4, 150 mM NaCl, 0.5% Nonidet P-40, 0.5% deoxycholate, 5 mM EDTA) on ice with Fisherbrand 150 hand-held homogenizer. The brain homogenates were further subjected to centrifugation at 1,000 × g for 3 min at 4 °C and the supernatants were collected and stored at −80 °C for western blotting and ELISA analyses (see below).

Western Blotting Analysis.

Brain homogenates (10% w/v) were resuspended in 1x Laemmli sample buffer (no reducing agent, no heating), loaded onto 4–12% Bis-Tris NuPAGE (Invitrogen, Waltham, MA) or home-made12% Tris-Glycine gel. Proteins were transferred to Immobilon FL membrane (Millipore, Burlington, MA) for 30 min at 100 volts. The membranes were then blocked with LI-COR blocking buffer, and probed with the appropriate primary antibodies. Tau-5 (cat# AHB0042, 1:1000 dilution) and p-Tau-Thr181 mAb AT270 (cat# MN1050, 1:1000) were purchased from Invitrogen (Wal-tham, MA). P-tau198 rabbit mAb (cat# ab79540, 1:1000 dilution) was purchased from abcam (Waltham, MA). Tau Rabbit pAb A1103 (cat# AP1103, 1:1000 dilution), p-tau202(cat# AP0765, 1:1000 dilution), p-tau205 rabbit pAb, (cat# AP0168, dilution 1:1000), p-tau217 rabbit pAb (cat# AP1233, 1:1000 dilution), p-tau231 Rabbit mAb (cat# AP0053, 1:1000 dilution), p-tau262 Rabbit pAb (cat# AP0397, 1:1000 dilution), and p-tau404 Rab-bit pAb (cat# AP0170, 1:1000 dilution) were obtained from ABclonal (Woburn, MA). Fluorescent labeled secondary antibodies, IRDye 680LT donkey anti-mouse IgG (cat# 926–68022, LI-COR, Inc. Lincoln, NE) and IRDye 800CW Donkey anti-rabbit IgG (cat# 926–32213 LI-COR, Inc. Lincoln, NE) were diluted to 1:50,000. The immunoreactivity was visualized by Odyssey western blot detection (LICOR, Inc. Lincoln, NE). GAPDH or β-actin was used as loading control. Control western blotting analyses were performed the same as regular analyses except that no primary antibodies were applied to confirm antibody-specific detection.

Semi-quantification of western blotting intensity/area of “smear” results was performed as the following. The immunoreactivity was visualized first and then HMW phospho-tau aggregate band quantification was analyzed by densitometry using Odyssey western blot detection software (LICOR, Inc., Lincoln, NE). During the quantification process, equal rectangle area was set for each sample for normalization. All the background-subtracted density values were normalized with loading control (GAPDH or β-actin). Western blotting exposure time was optimized to avoid signal oversaturation.

ELISA Measurements.

Protein concentrations of the brain samples were measured by BCA assays (Pierce BCA protein assay kit, Thermo Fisher Scientific). 25 μg of total protein per well was used for ELISA assays according to manufacturer’s instruction. ELISA kit for p-tau181(cat# KHO0631) was purchased from Invitrogen (Waltham, MA) and p-tau217 ELISA kit (cat# 59672) was procured from Cell Signaling (Danvers, MA). ELISA kit for p-tau198 was homebrew using Tau (Total) human ELISA kit (Invitrogen, cat# KHB0041) plate coated with total tau capture antibody and p-tau198 rabbit mAb (abcam; cat# ab79540) as detection antibody. The absorbance was measured at OD450 by SPECTRAmax plus plate reader (Molecular Devices, San Jose, CA).

Statistical Analysis.

All data are presented as the mean ± S.E.M and the differences were analyzed with unpaired Student’s t test as implemented within GraphPad Prism software (version 6.0). p values < 0.05 were considered significant. Relative quantification of p-tau198, p-tau181, and p-tau217 in AD patients and control subjects (AD vs. CTL) or AD and rare tauopathy subjects were compared using the statistical software GraphPad Prism 9. The specificity and sensitivity of tau with the specific phosphorylated sites in AD and CTL subjects, or AD and tauopathy subjects were determined based on the area under the curve (AUC) of receiver operating characteristics (ROC) analysis. The 95% confidence interval (CI_95%) of AUC was calculated with Wilson/Brown method.

Correlation Analysis.

Genotypic APOE4 carriers or non-carriers in relation to p-tau198 ELISA levels were analyzed by t-test in GraphPad Prism 9.3.1. The nonparametric Spearman’s correlation coefficients between postmortem interval (PMI) versus p-tau198 levels, or age versus p-Tau198 levels were calculated using GraphPad Prism 9.3.1. Pairs of datasets were input into GraphPad data table. Confidence level was set at 95%.