Plasma extracellular vesicle tau and TDP-43 as diagnostic biomarkers in FTD and ALS

Frontotemporal dementia (FTD) includes bvFTD, svPPA and nfvPPA, and together with PSP, CBD and ALS forms a disease continuum with overlapping genetics and pathology. Approximately half of bvFTD and most ALS/FTD-ALS cases show TDP-43 proteinopathy, while PSP/CBD and about 40% of bvFTD show FTLD-tau pathology. Tau exists as six isoforms and aggregates can be 3R- or 4R-predominant, depending on disease. There are no disease-modifying therapies for FTD/ALS, partly due to lack of biomarkers that define underlying molecular pathology ante mortem (except in genetic cases). Accurate, early, pathology-specific biomarkers could improve diagnosis and stratify patients for trials. The study aims to test whether plasma EV-based measures of TDP-43 and 3R/4R tau ratio can identify FTLD-TDP versus FTLD-tau pathology across FTD spectrum disorders and ALS, and correlate with disease severity.

Previous studies suggested plasma GFAP/NfL ratios may distinguish FTLD-tau from FTLD-TDP, and investigated TDP-43 (total, phosphorylated, aggregated) in blood/CSF with conflicting results. CSF p-tau181/tau and cryptic exon-derived peptides have been proposed as TDP-43 markers. CSF tau isoforms have been explored to differentiate 3R vs 4R tauopathies, but blood detection of full-length tau is hampered by fragmentation and low concentrations. The authors previously developed a CSF assay (immunoprecipitation–mass spectrometry) to measure tau microtubule-binding region, but lumbar puncture is invasive. EVs can carry pathological tau and TDP-43 and may reflect disease-associated cytoplasmic mislocalization of TDP-43; thus, plasma EVs are a promising minimally invasive source.

Design and cohorts: Two cohorts were analyzed. DESCRIBE cohort (Germany): subcohort 1 (pilot) measured EV 3R/4R tau ratios in HC (n=15), AD (n=23), svPPA (n=17), bvFTD (n=42), PSP (n=44). Subcohort 2 (validation) measured EV 3R/4R tau ratios and EV TDP-43 in HC (n=56), ALS (n=165), bvFTD (n=179), PSP (n=163) and included genetically and/or neuropathologically confirmed cases (37 genetic, 31 autopsy; total 63 unique confirmed). Sant Pau cohort (Spain) served as an independent validation: HC (n=50), ALS (n=65), ALS-FTD (n=58), bvFTD (n=50), PSP (n=41) with 34 genetic cases. Inclusion criteria and clinical assessments followed established diagnostic criteria, with extensive neuropsychological and disease severity scales. Genetic testing included C9orf72 expansions and sequencing panels (details in Methods). Autopsy diagnoses were made using standardized immunohistochemistry and neuropathologic criteria. EV preparation: From 500 µl EDTA plasma (and CSF for subsets), medium-sized EVs (mEVs) were obtained from 10,000g pellets; small EVs (sEVs) were isolated from supernatant by size-exclusion chromatography (fractions 7–10 pooled). EVs were lysed in CHAPS buffers. L1CAM immunocapture was used to enrich putative neuron-derived EVs to examine the distribution of tau and TDP-43 cargo. Assays: Full-length tau in EVs was confirmed by mass spectrometry, enabling distinction of 3R and 4R isoforms. Two sandwich immunoassays were developed/optimized for EV 3R and 4R tau using isoform-specific capture antibodies with HT7 detection. TDP-43 in plasma, sEV and mEV fractions was quantified using Simoa (Quanterix). Plasma NfL was measured by Simoa. Western blotting validated presence of TDP-43 and tau isoforms. Nanoparticle tracking analysis characterized EVs. Assay performance (LLOQ, precision, linearity, matrix effects) was evaluated (Supplementary Table 1). Statistics: Nonparametric comparisons used Kruskal–Wallis with Dunn’s correction. Associations were tested with two-sided Spearman correlations; monotonic regression splines illustrated relationships. ROC analyses (MedCalc) provided AUCs with 95% CIs; sensitivity/specificity were computed. Gaussian mixture modeling defined data-driven cut-offs for sEV 3R/4R tau ratios and sEV TDP-43 (intersections of fitted normal distributions). Potential confounders (age, sex, disease duration) showed no impact on EV biomarkers.

• EV cargo and origin: Full-length tau was detectable in CSF and plasma EVs, enabling quantification of 3R and 4R tau. The majority of plasma EV tau and TDP-43 resided in L1CAM-positive EVs.
• DESCRIBE subcohort 1 (pilot, sEV 3R/4R tau ratio): HC ≈1.16 (IQR 0.99–1.28), AD ≈0.91 (0.57–1.25), svPPA ≈1.00 (0.98–1.11), bvFTD high ≈2.59 (2.02–3.87), PSP low ≈0.18 (0.13–0.29); P<0.001 across key comparisons. ROC AUCs for distinguishing PSP and bvFTD from others were high (e.g., PSP vs HC AUC 0.96; PSP vs AD 0.99; bvFTD vs HC 0.93). In bvFTD, sEV 3R/4R tau ratio positively correlated with NfL (r=0.68, P<0.0001); in PSP it correlated inversely with NfL (r≈−0.48, P=0.001). Ratios associated with clinical/cognitive severity (bvFTD: higher ratio worse; PSP: lower ratio worse).
• DESCRIBE subcohort 2 (sEV 3R/4R tau ratio): PSP had the lowest ratios (median 0.45, IQR 0.34–0.60), significantly lower than HC (0.99), ALS (0.95), bvFTD (1.10–2.28; elevated subset). ROC AUCs for PSP vs HC/ALS/bvFTD were 0.96–0.98. Approximately 50% of bvFTD showed elevated ratios; median sEV bvFTD reported as 2.28 (IQR 1.13–2.4) in elevated subgroup analyses. Ratios were not associated with age, sex or disease duration.
• Pathology-confirmed cases (subcohort 2): sEV 3R/4R tau ratio ≈1 in TDP-43 pathology and non-TDP-43/non-tau groups (similar to HC). PSP/GGT-type tauopathies had low ratios (median 0.42, IQR 0.35–0.60). MAPT mutation carriers had markedly high ratios (median 3.96, IQR 3.81–4.12). Thus, EV tau ratio separates FTLD-tau (MAPT high, PSP/GGT low) from FTLD-TDP and others.
• EV TDP-43 (subcohort 2): Plasma sEV TDP-43 was highest in ALS (median 45.45 pg/ml, IQR 28.88–83.21), elevated in a subset of bvFTD (median 31.25 pg/ml, IQR 14.45–41.09), and low in HC (9.47 pg/ml, IQR 7.63–13.33) and PSP (9.09 pg/ml, IQR 7.73–13.27). AUCs: ALS vs HC 0.99 (0.97–1.00), ALS vs PSP 0.99 (0.98–1.00), ALS vs bvFTD 0.91 (0.88–0.94). EV TDP-43 outperformed plasma NfL for several comparisons. EV TDP-43 correlated with NfL (ALS r=0.67; bvFTD r=0.42; P<0.0001) and with worse clinical performance (ALS: MMSE↓, ALS-FRS↓; bvFTD: MMSE/MOCA↓, FAQ/CDR/NPI-Q/CBI-M worse). Plasma non-EV TDP-43 did not differ between diagnostic groups.
• Pathology-confirmed EV TDP-43: TDP-43 pathology cases had high EV TDP-43 (median 63.95 pg/ml, IQR 42.89–86.63), higher than HC and tau groups (PSP/GGT-type median 2.85; MAPT median 2.86; P<0.00001). Non-TDP-43/non-tau genetic cases (SOD1, FUS, CHCHD10) showed low EV TDP-43 similar to HC. VCP and TBK1 mutation carriers showed unexpectedly low EV TDP-43.
• Combined biomarker separation: In bvFTD, EV TDP-43 inversely correlated with EV 3R/4R tau (r≈−0.50 to −0.71 across cohorts). Plotting EV TDP-43 vs EV tau ratio separated bvFTD into two subgroups: putative FTLD-TDP (high TDP-43, low/normal tau ratio) overlapping with ALS; and putative FTLD-tau (high tau ratio, low TDP-43). PSP and HC formed distinct clusters.
• Cut-offs (Gaussian mixture modeling): DESCRIBE sEV 3R/4R tau: 0.77 and 1.27–1.28; sEV TDP-43: 13.87 pg/ml and 56.18 pg/ml. Sant Pau: tau 0.78 and 1.28; TDP-43 17.85 and 57.34 pg/ml. Nearly interchangeable across cohorts.
• Diagnostic performance using cut-offs: • PSP (low sEV 3R/4R tau <0.77/0.78): DESCRIBE sensitivity 93.25% (88.25–96.58), specificity 95.25% (92.68–97.12); pathology-confirmed PSP/GGT-type: sensitivity 100% (39.76–100), specificity 100% (93.94–100). Sant Pau: sensitivity 100% (91.40–100), specificity 94–94.94%. • MAPT (high sEV 3R/4R tau >1.27/1.28): All MAPT mutation carriers identified with 100% sensitivity/specificity in DESCRIBE; ~38–58% of sporadic bvFTD exceeded upper cut-off (putative tau pathology) depending on cohort. • TDP-43 pathology (sEV TDP-43 >13.87/17.85 pg/ml): DESCRIBE genetic/autopsy-proven TDP-43 cases detected with sensitivity 88.00% (76.13–95.67) and specificity 100% (75.29–100). Sant Pau genetic TDP-43 cases sensitivity 88.89% (70.84–97.65), specificity 85.71% (42.13–99.64). ALS and ALS-FTD identified with high sensitivity/specificity vs HC/PSP (e.g., Sant Pau ALS sensitivity 86.15%, specificity 100%; ALS-FTD sensitivity 96.55%, specificity 100%).

Plasma EV 3R/4R tau ratios and EV TDP-43 levels provide minimally invasive, pathology-specific biomarkers for FTD spectrum disorders and ALS. High EV TDP-43 is characteristic of ALS and bvFTD with TDP-43 pathology, while EV tau ratios identify FTLD-tau: MAPT mutation carriers show high ratios and PSP/GGT-type tauopathies show low ratios. Together, these markers separate bvFTD into two molecularly distinct subgroups (putative FTLD-TDP vs FTLD-tau) and discriminate PSP and ALS from controls and other neurodegenerative groups with high diagnostic accuracy. Both biomarkers correlate with neurodegeneration (plasma NfL) and with clinical, neurological and cognitive measures of disease severity (for example, ALS-FRS, MMSE, MOCA, CDR-SB), supporting their relevance to disease burden and potential utility in monitoring progression and therapeutic target engagement. The low EV TDP-43 in VCP and TBK1 mutation carriers highlights possible biology affecting EV release/sorting. The predominance of EV tau and TDP-43 in L1CAM-positive EVs suggests a CNS origin, though neuron-specificity of L1CAM is debated. Overall, the combined EV TDP-43 and 3R/4R tau ratio approach addresses an unmet need for molecularly informative blood biomarkers in the ALS/FTD/PSP spectrum.

This study demonstrates that plasma EV biomarkers—TDP-43 concentration and the 3R/4R tau isoform ratio—can noninvasively identify underlying molecular pathology across ALS and FTD spectrum disorders. They discriminate ALS and PSP from controls and other diseases with AUCs typically >0.9, and stratify bvFTD into FTLD-TDP versus FTLD-tau subgroups. Biomarker levels correlate with neurodegeneration and clinical severity, indicating potential for disease monitoring and use as surrogate endpoints in trials. Cut-offs derived by mixture modeling were robust across independent cohorts. Future work should expand pathology-confirmed FTLD-tau samples, include preclinical and longitudinal assessments, and test utility across diverse populations and therapeutic interventions. Integration with established AD biomarkers may also help identify AD cases with TDP-43 co-pathology.

• Limited numbers of pathology-confirmed FTLD-tau cases; need more autopsy/genetic confirmations, especially in non-PSP tauopathies.
• Cross-sectional design for most analyses; longitudinal validation for progression tracking is required.
• Group size imbalances (e.g., DESCRIBE subcohort 2) and potential recruitment bias toward less advanced cases.
• EV CNS origin is inferred but not definitively proven; L1CAM EV neuron specificity and association remain debated.
• Floor effects for EV TDP-43 in HC/PSP could occur with small plasma volumes (mitigable by larger starting volumes).
• Certain mutations (VCP, TBK1) showed low EV TDP-43 despite linkage to TDP-43 pathology, possibly reflecting EV biogenesis/release differences.
• Assay transferability and pre-analytical standardization across centers require further harmonization; sEVs performed slightly better than mEVs, but workflows differ.