N-terminal and mid-region tau fragments as fluid biomarkers in neurological diseases

Abstract Brain-derived tau secreted into CSF and blood consists of different N-terminal and mid-domain fragments, which may have a differential temporal course and thus, biomarker potential across the Alzheimer’s disease continuum or in other neurological diseases. While current clinically validated total tau assays target mid-domain epitopes, comparison of these assays with new biomarkers targeting N-terminal epitopes using the same analytical platform may be important to increase the understanding of tau pathophysiology. We developed three total tau immunoassays targeting specific N-terminal (NTA and NTB total tau) or mid-region (MR total tau) epitopes, using single molecule array technology. After analytical validation, the diagnostic performance of these biomarkers was evaluated in CSF and compared with the Innotest total tau (and as proof of concept, with N-p-tau181 and N-p-tau217) in three clinical cohorts (n = 342 total). The cohorts included participants across the Alzheimer’s disease continuum (n = 276), other dementias (n = 22), Creutzfeldt–Jakob disease (n = 24), acute neurological disorders (n = 18) and progressive supranuclear palsy (n = 22). Furthermore, we evaluated all three new total tau biomarkers in plasma (n = 44) and replicated promising findings with NTA total tau in another clinical cohort (n = 50). In CSF, all total tau biomarkers were increased in Alzheimer’s disease compared with controls (P < 0.0001) and correlated with each other (rs = 0.53−0.95). NTA and NTB total tau, but not other total tau assays, distinguished amyloid-positive and amyloid-negative mild cognitive impairment with high accuracies (AUCs 84% and 82%, P < 0.001) matching N-p-tau217 (AUC 83%; DeLong test P = 0.93 and 0.88). All total tau assays were excellent in differentiating Alzheimer’s disease from other dementias (P < 0.001, AUCs 89–100%). In Creutzfeldt–Jakob disease and acute neurological disorders, N-terminal total tau biomarkers had significantly higher fold changes versus controls in CSF (45–133-fold increase) than Innotest or MR total tau (11–42-fold increase, P < 0.0001 for all). In progressive supranuclear palsy, CSF concentrations of all total tau biomarkers were similar to those in controls. Plasma NTA total tau concentrations were increased in Alzheimer’s disease compared with controls in two independent cohorts (P = 0.0056 and 0.0033), while Quanterix total tau performed poorly (P = 0.55 and 0.44). Taken together, N-terminal-directed CSF total tau biomarkers increase ahead of standard total tau alternatives in the Alzheimer’s disease continuum, increase to higher degrees in Creutzfeldt–Jakob disease and acute neurological diseases and show better potential than Quanterix total tau as Alzheimer’s disease blood biomarkers. For progressive supranuclear palsy, other tau biomarkers should continue to be investigated.


T-tau immunoassay development and validation for CSF
For NTA t-tau and MR t-tau, mouse monoclonal antibody targeting aa 159-163 (HT7, #MN1000, Thermo Scientific) and for NTB t-tau mouse monoclonal antibody targeting aa 6-18 (1-100, #816601, BioLegend) were conjugated to magnetic homebrew carboxylated beads (#103207, Quanterix). All detector antibodies were biotinylated in-house with EZ-Link™ NHS-PEG4-Biotin (Thermo Scientific, USA). NTA and MR ttau were measured using a 2-step protocol, where the assay beads, detector antibody and the analyte are first co-incubated, followed by washing of the beads, and incubation with the enzyme. NTB t-tau was measured using a 3-step protocol, where the beads were initially incubated with the analyte, then with the detector, and finally with the enzyme. After another wash, the substrate resorufin β−D-galactopyranoside (RGP; #103159, Quanterix) was added to the beads to gain the measured fluorescent signal. Recombinant non-phosphorylated full-length Tau-441 (#T08-54N, SignalChem) was used as a calibrator in all the inhouse t-tau assays. Three-fold diluted calibrator series in assay diluent (Tau 2.0 diluent, #101556, Quanterix) ranging from 500 to 0.7 pg/ml (MR and NTB t-tau), 56 to 0.1 pg/ml (CSF NTA t-tau) or 18.5 to 0.03 pg/ml (plasma NTA t-tau) were included in each sample plate.
Limits of detection (LOD) and lower limits of quantification (LLOQ) were defined by running 16 blank samples in duplicates and adding three or ten standard deviations to the measured mean value, respectively.
Upper limits of quantification (ULOQ) were set as the concentration of the highest calibration accounting for the used dilution factor. For dilution linearity, two human CSF samples (pooled Alzheimer's disease and pooled controls) were measured with 2-fold dilution series (1:2, 1:4, 1:8) as duplicates. For spike recovery, two CSF samples were diluted 1:2 and spiked 10 pg/ml and 40 pg/ml (NTA t-tau) or 50 pg/ml and 150 pg/ml (NTB and MR t-tau) of exogenous non-P Tau441 (the assay calibrator). Spiked samples were then measured on a same plate with non-spiked CSF and spiked assay diluent. Spike recovery was calculated as: % Recovery = Concentration of spiked sample / (Concentration of non-spiked sample + Concentration of spiked buffer). Precision and accuracy were tested by measuring five replicates of two human CSF samples as duplicates in three consecutive days and calculating the intra-assay precision (variation within run, CV r (%)) and inter-assay precision (variation between runs, CV rw (%)).
With dilution, measured concentrations became lower in all in-house assays, implying that the sample matrix has some interference with the obtained signal. However, when comparing samples with the same dilution, this should not be problematic. Due to extremely high t-tau concentrations, some of the Creutzfeldt-Jakob disease and acute neurological disorders samples were further diluted 1:10, and these measurements were above the ULOQ (2000 pg/ml) and all extrapolated from the standard curve. This should be kept in mind and the concentrations gained from these analyses are likely underestimations and should be interpreted with caution.

T-tau measurements for CSF
All pilot CSF cohort samples measured above the quantification limits of each t-tau assay. From the clinical cohorts, 26/292 (8.9%) CSF samples were below the LOD of NTA t-tau (six controls, nine Aβsubjects with mild cognitive impairment (MCI), one Aβ+ MCI, three non-Alzheimer's disease dementia, four progressive supranuclear palsy, and three Alzheimer's disease), and 5/292 (1.7%) of NTB t-tau (two controls, one Aβ-MCI, one Aβ+ MCI, and one non-Alzheimer's disease dementia). One CSF sample measured with NTB (Aβ-MCI subject) and one sample measured with MR t-tau (Aβ-MCI subject) was excluded due to technical error (no reading). In addition, one CSF sample was finished before measurement with NTB (Aβ-MCI).
All t-tau levels were extremely high in Creutzfeldt-Jakob disease and acute neurological disorders, and most of the samples gave readings above the defined upper limit of quantification (ULOQ) even with the additional 1:10 dilution. Thus, majority of the concentrations were extrapolated from the calibration curve (>2000g/ml) and results should be interpreted with caution. Data presents as median (interquartile range). Significant differences in pairwise comparisons to unspecified dementia (#) and AD (*) are presented. Hepatic encephalopathy and limbic encephalitis were not included in the analysis (n = 1 for both).