Plasma p-tau212 antemortem diagnostic performance and prediction of autopsy verification of Alzheimer’s disease neuropathology

Recent advances in blood biomarkers for Alzheimer’s disease (AD), particularly plasma phosphorylated tau (p-tau), enable diagnosis and prognosis beyond what cerebrospinal fluid (CSF) and neuroimaging alone can offer. While plasma Aβ biomarkers show good diagnostic accuracy, their modest dynamic range limits robustness. In contrast, plasma p-tau species (p-tau181, p-tau217, p-tau231) strongly associate with brain Aβ load, tau tangle burden, disease severity across the AD continuum, and predict longitudinal pathology and cognitive change. Plasma p-tau assays also identify abnormal Aβ-PET, predict autopsy-confirmed AD, and are being used in therapeutic trials as proxies for brain amyloid and tau pathology. Head-to-head studies consistently show p-tau217 outperforming p-tau181 and p-tau231. However, some p-tau217 assays bind forms also phosphorylated at threonine-212. The specific contribution of p-tau212 has been unclear. This study developed p-tau212-specific antibodies and a blood-based assay, then evaluated whether plasma p-tau212 is a sensitive and specific biomarker for AD neuropathology and clinical diagnosis, including autopsy-verified cohorts and real-world memory clinic populations, and compared its performance with p-tau217 and other p-tau isoforms.

The paper situates plasma p-tau biomarkers within a growing literature demonstrating strong associations with Aβ and tau pathology, high diagnostic and prognostic accuracy, and utility in clinical trial screening and monitoring. Prior work consistently reports higher performance of p-tau217 compared to p-tau181 and p-tau231, including in cognitively unimpaired individuals across the AD continuum. Plasma p-tau species predict abnormal Aβ-PET, autopsy-confirmed AD, and cognitive decline. Anti-amyloid trials show biomarker dynamics in plasma p-tau relative to amyloid reduction. Despite this, specificity around phosphorylation sites recognized by assays (e.g., cross-reactivity between p-tau217 and neighboring sites like T212) prompted investigation of p-tau212 as an independent biomarker. The study builds on validated immunoassays and mass spectrometry methods for tau PTMs and prior autopsy correlations.

Antibody development and specificity: Two sheep monoclonal antibodies were generated, one specific for p-tau212 and one for p-tau217. Specificity was assessed via direct ELISAs against synthetic peptides and recombinant tau that were non-phosphorylated or phosphorylated at defined sites. The p-tau212 mAb bound only peptides phosphorylated at threonine-212 (and doubly at T212/T217), with no cross-reactivity to T217-only, T181, S214, or T231 epitopes. The p-tau217 mAb bound only T217-phosphorylated peptides with reduced affinity when S214 was also phosphorylated; no binding to T212-only, T181, S214, or T231. Additional mass spectrometry specificity experiments confirmed selective binding to T212-containing peptides. Neuropathology immunostaining: Immunohistochemistry and immunofluorescence were performed on hippocampal, entorhinal, and temporal cortex sections from autopsy-confirmed AD cases (Braak VI). p-tau212 staining was assessed for neurofibrillary tangles (NFTs), neuropil threads, and dystrophic neurites. Co-localization with p-tau217 and AT8 (pS202/pT205) was quantified via confocal microscopy. Assay development and analytical validation: A Simoa HD-X immunoassay for p-tau212 was developed using the p-tau212 antibody as capture and an N-terminal tau antibody (Tau12/Tau2) for detection. Calibration used recombinant full-length tau. Analytical validation followed clinical chemistry consortium recommendations: dilution linearity in plasma and CSF; between-run CVs ~5.3–12.9%; duplicate precision CVs <20% for most samples; spike-recovery 79.6–94.0% (plasma) and 85.3–103.1% (CSF); LLOQ 0.17 pg/mL. Correlation with plasma p-tau217 measured by mass spectrometry was strong (Spearman r = 0.867, p < 0.0001). Cohorts and participants: Total n = 388 across five cohorts. BLSA-Neuropathology plasma cohort (n = 47): autopsy-verified AD (n = 20), ASYMAD (n = 15), non-AD controls (n = 12). UCSD-Neuropathology CSF cohort (n = 67): high ADNC, other pathologies, low ADNC, mixed high ADNC plus other pathologies. Gothenburg plasma discovery cohort (n = 30): CSF-defined AD (n = 16) and biomarker-negative controls (n = 14). Polish cohort (n = 95): paired CSF and plasma classified by Erlangen Score (ES); ES 0–2 controls, ES ≥3 AD. Slovenian memory clinic cohort (n = 149): SCD/SCI, non-AD MCI, AD MCI, AD dementia with CSF-defined Aβ positivity. Biomarkers and measurements: Plasma and CSF p-tau212 measured on Simoa HD-X at University of Gothenburg; plasma p-tau217, p-tau231, p-tau181 measured using validated in-house Simoa assays. Plasma p-tau217 also quantified by immunoprecipitation–mass spectrometry (Orbitrap PRM). Measurements performed blinded to clinical data. Outcomes and pathology staging: Pathology assessed using CERAD neuritic plaque scores, Braak NFT staging, Thal Aβ phases, and ADNC criteria. Clinical diagnoses used DSM-V and MCI criteria; Aβ positivity defined by CSF Aβ42/40 ratio (site-specific cutoffs). Statistical analysis: Data summarized as median (IQR). Normality via Shapiro–Wilk; non-parametric tests for group comparisons (Mann–Whitney; Dwass–Steel–Critchlow–Fligner). ROC AUC with DeLong tests for biomarker comparisons. Spearman correlations for associations with Aβ/tau biomarkers and Braak stage. Post-mortem cohorts analyzed with estimated marginal means adjusting for sex, age, and collection-to-death interval. Multiple comparisons adjusted; significance p < 0.05.

- Antibody specificity: p-tau212 and p-tau217 sheep mAbs were highly specific to their target phosphorylation sites with no cross-reactivity to other sites (T181, S214, T231), confirmed by ELISA and MS. - Brain staining: p-tau212 robustly stained NFTs, neuropil threads, and dystrophic neurites in entorhinal cortex and hippocampus from AD brains. Immunofluorescence showed high co-localization of p-tau212 with p-tau217 and with AT8 (pS202/pT205). Quantitatively, 100% of NFTs and 96% of neuronal threads were stained by both p-tau212 and p-tau217. - Assay performance: The p-tau212 Simoa assay demonstrated good linearity, precision (between-run CV ~5–13%; duplicate CV <20% in 27/29), recovery (plasma 79.6–94.0%; CSF 85.3–103.1%), and LLOQ 0.17 pg/mL. Plasma p-tau212 correlated strongly with plasma p-tau217 by MS (Spearman r = 0.867; p < 0.0001). - Autopsy cohorts (BLSA, UCSD): In BLSA plasma, p-tau212 was significantly higher in autopsy-verified AD vs ASYMAD and controls (p < 0.0001) and increased with CERAD plaque frequency and Braak stage. Estimated fold changes included higher levels in frequent vs sparse plaques and in Braak V–VI vs lower stages (e.g., 3.98× for Braak III–VI vs I–II; p = 0.0077; 2.08× for V–VI vs III–IV; p = 0.0025). In UCSD CSF, p-tau212 increased markedly with higher CERAD plaque scores (approximately 6.1× in moderate, p < 0.003; 12.2× in severe, p < 0.0001) and higher Braak stages. - Discovery/clinic cohorts: • Gothenburg plasma: p-tau212 was 3.7-fold higher in biomarker-positive AD vs biomarker-negative controls (p < 0.0001). Diagnostic accuracy AUC 91.5% (95% CI 79.1–100%), exceeding p-tau181 and p-tau231 (both 86.6%). • Polish paired plasma/CSF: plasma p-tau212 4.06-fold higher and CSF p-tau212 ~10-fold higher in AD vs controls (both p < 0.0001). AUCs: plasma 85.6% (95% CI 75.8–92.6%), CSF 91.5% (95% CI 83.4–100%); difference not significant (p = 0.233). • Slovenian memory clinic: Both plasma p-tau212 and p-tau217 increased from Aβ− SCD/non-AD to Aβ+ AD dementia. AUCs for Aβ− SCD vs Aβ+ AD dementia were 92.5% (95% CI 87.0–97.9%) for p-tau212 vs 95.6% (95% CI 91.7–99.4%) for p-tau217 (DeLong p = 0.232). For Aβ− non-AD MCI vs Aβ+ AD dementia, AUC 89.2% (95% CI 85.8–92.6%) vs 87.8% (95% CI 78.2–97.2%) (DeLong p = 0.819). PPV for Aβ positivity vs SCD: 94.7% (p-tau212) and 98.1% (p-tau217); NPV: 72.4% and 74.2%, respectively. Dichotomized thresholds (3.2 pg/mL for p-tau212; 4.4 pg/mL for p-tau217) showed 83.5% concordance in Aβ status classification. - Associations with pathology and biomarkers: In the Slovenian cohort, plasma p-tau212 correlated with CSF Aβ42/40 (ρ = −0.48, p < 0.0001), CSF p-tau181 (ρ = 0.51, p < 0.001), and CSF total tau (ρ = 0.55, p < 0.0001), comparable to p-tau217. Correlation with Braak stage (UCSD) was ρ = 0.67 for p-tau212 vs 0.59 for p-tau217 (p < 0.0001). Across cohorts, p-tau212 often showed larger fold changes and similar or better diagnostic accuracy than p-tau181 and p-tau231, and performance comparable to p-tau217. - Overall, plasma p-tau212 tracked AD neuropathology at autopsy and clinical stages, with diagnostic performance similar to p-tau217 and superior to p-tau181/p-tau231 in several analyses.

The study demonstrates that phosphorylation at tau threonine-212 is a robust marker of AD neuropathology detectable in blood. Highly specific p-tau212 antibodies stained classical tau pathology in AD brain tissue and co-localized with p-tau217 and AT8, indicating that T212 phosphorylation occurs within neurofibrillary tangles and neuronal threads. The newly developed ultra-sensitive Simoa plasma p-tau212 assay exhibited strong analytical validation and correlated well with mass spectrometry-based p-tau217 measurements. In both autopsy-verified and memory clinic cohorts, plasma p-tau212 distinguished AD from controls or other pathologies, tracked amyloid plaque burden (CERAD), and increased with Braak NFT stage. Its diagnostic accuracy and effect sizes were similar to p-tau217 and generally greater than p-tau181 and p-tau231, aligning with the literature that proximal mid-domain p-tau epitopes (e.g., T217, T212) better reflect AD pathology. Concordant performance across plasma and CSF supports p-tau212 as a peripheral proxy of central AD pathophysiology. In a real-world memory clinic, plasma p-tau212 effectively stratified patients along the AD continuum and identified Aβ positivity with high PPV, indicating potential for first-line screening, triage, and monitoring in clinical care and therapeutic trials. Altogether, the findings address the central question by showing that p-tau212 alone, independent of T217, captures AD-relevant phosphorylation with strong diagnostic and staging utility, reinforcing its relevance for clinical implementation alongside or as an alternative to p-tau217.

This work introduces and validates highly specific antibodies and an ultra-sensitive plasma assay for p-tau212, establishing p-tau212 as a peripherally accessible biomarker that mirrors AD neuropathology. Plasma p-tau212 displays diagnostic accuracy and fold changes comparable to p-tau217 and superior to p-tau181 and p-tau231 in multiple cohorts, correlates with amyloid plaque burden and Braak staging, and co-localizes with established tau pathology markers in brain tissue. These data support the use of plasma p-tau212 for AD diagnosis, staging, and potential trial screening and monitoring. Future directions include validating performance in cognitively unimpaired pre-symptomatic populations and refining low-threshold cutoffs for early detection, expanding evaluation across diverse cohorts and comorbid pathologies, and longitudinal studies to assess disease progression, conversion risk, and response to disease-modifying therapies. Additional work in Down syndrome-related AD (AD-DS) and broader real-world implementation studies is warranted.

The study notes the need for validation of p-tau212 in cognitively unimpaired cohorts with low-threshold positivity to establish pre-symptomatic utility. While multi-cohort and autopsy-verified data strengthen findings, some subgroup sample sizes (e.g., specific autopsy strata) were modest. Further external replication, longitudinal validation, and assessment across diverse populations and comorbid conditions are needed for generalizability and clinical implementation.