4R-specific brain MTBR-tau increases in primary tauopathies

First, frozen brain tissues from 59 individuals with autopsy-confirmed AD, FTLD-tau or FTLD-TDP and three normal control individuals without clinical diagnoses of neurodegenerative diseases were processed for biochemical extraction of insoluble tau and underwent MS analyses (Table 1). The superior frontal gyrus (SFG) at the level of the frontal eye fields and the insular cortex were chosen as regions of interests because of the higher severity of tau aggregation seen in these regions across the clinical spectrum of primary tauopathies ranging from behavioral variant frontotemporal dementia (bvFTD) to Richardson’s syndrome21. MTBR-tau275 (275–280) and MTBR-tau282 (282–290) located within the R2 repeat domain and specific to 4R isoforms were normalized with total tau (t-tau) measured by mid-domain tau peptide 181–190 to account for individual variabilities in t-tau (Fig. 1a). Brain MTBR-tau275/t-tau increased approximately fivefold in CBD (1.078 ± 0.346) and threefold in FTLD-MAPT (0.762 ± 0.333) compared to normal control (0.235 ± 0.053, P < 0.0001, P < 0.0001, respectively) and non-tauopathy FTLD-TDP (0.213 ± 0.066, P < 0.0001, P < 0.01, respectively; Fig. 1b). Interestingly, MTBR-tau275/t-tau increased in CBD and FTLD-MAPT compared to other tauopathies, such as 3R tauopathy, PiD (0.226 ± 0.110, P < 0.0001, P < 0.01, respectively), 4R tauopathies, AGD (0.200) and PSP (0.489 ± 0.229, P < 0.0001, P < 0.05, respectively), and 3R/4R mixed tauopathy, AD (0.319 ± 0.047, P < 0.0001, P < 0.01, respectively). MTBR-tau275/t-tau moderately increased in PSP (approximately twofold) compared to FTLD-TDP (P < 0.01). Brain MTBR-tau282/t-tau had a similar increase profile to MTBR-tau275/t-tau (Fig. 1c) but moderately (2.6-fold) increased in AD (0.915 ± 0.180) compared to FTLD-TDP (0.348 ± 0.133, P < 0.05), which was not observed in MTBR-tau275/t-tau. These results suggest that 4R-specific MTBR-tau species are enriched in the insoluble fraction of SFG/insular cortex brain tissue in a subset of 4R tauopathies, such as CBD and FTLD-MAPT, and moderately increased in a different PSP (4R tauopathy) and AD (3R/4R mixed tauopathy).

Table 1 Demographics and brain MTBR-tau measures of participants in the primary tauopathy cohortFig. 1: 4R-specific insoluble brain MTBR-tau is enriched in CBD, FTLD-MAPT, AD and PSP.

a, Schematic of the quantified peptides of t-tau181–190 and 4R isoform-specific MTBR-tau in the R2 region (gray bars, MTBR-tau275 and MTBR-tau282). The relative abundance of each MTBR-tau was normalized to the t-tau peptide. b,c, MTBR-tau275/t-tau (b) and MTBR-tau282/t-tau (c) were measured in the tauopathy patient’s insoluble brain fractions from the SFG (circle, n = 54) and insula (triangle, n = 8). Both MTBR-tau species were most enriched in CBD (n = 12) and FTLD-MAPT (n = 8). PSP (n = 16) and AD (n = 7) had moderate enrichment. AGD (n = 1), PiD (n = 3) and FTLD-TDP (n = 12) did not change in MTBR-tau275 or MTBR-tau282 compared to normal control (n = 3). The red (n = 9) and blue (n = 1) filled circles indicate AD and PSP copathology, respectively. *P < 0.05, **P < 0.01, ****P < 0.0001. The box plots show the minimum, 25th percentile, median, 75th percentile and maximum. Differences in biomarker values were assessed with a one-way ANOVA. A two-sided P < 0.05 was considered statistically significant and corrected for multiple comparisons using a Benjamini–Hochberg FDR set at 5%.

4R-specific CSF MTBR-tau decreases in primary tauopathies

Next, CSF from 29 normal controls, 5 FTLD-MAPT and 78 autopsy-confirmed cases of AD, primary tauopathies and FTLD-TDP were analyzed for MTBR-tau275 and MTBR-tau282 (Table 2). CSF MTBR-tau275 and MTBR-tau282 concentrations did not separate different tauopathies (Supplementary Fig. 1a,b) due to individual variability in t-tau concentrations (Supplementary Fig. 1c). Therefore, CSF MTBR-tau275 and MTBR-tau282 from truncated tau were normalized by t-tau measured by mid-domain tau 212–221 (Fig. 2a), like the normalization methods previously reported in truncated tau and Aβ isoform measurements20,22. CSF MTBR-tau275/t-tau decreased in CBD (0.00525 ± 0.00117), AD (0.00472 ± 0.00085) and FTLD-MAPT (0.00491 ± 0.00207), compared to normal control (0.00657 ± 0.00078, P < 0.001, P < 0.0001, P < 0.01, respectively) and non-tauopathy control, FTLD-TDP (0.00611 ± 0.00115, P < 0.05, P < 0.01, P < 0.05, respectively; Figure 2b). This decrease was particularly substantial in FTLD-MAPT P301L, which has more typical FTLD pathology than in R406W, which has many features of AD. CSF MTBR-tau275 also decreased in CBD, AD and FTLD-MAPT compared to other 4R tauopathies, AGD (0.00759 ± 0.00013) and PSP (0.00669 ± 0.00091, P < 0.001, P < 0.0001, P < 0.01, respectively), and the 3R tauopathy, PiD (0.00676 ± 0.00138, P < 0.05, P < 0.01, P < 0.05, respectively). CSF MTBR-tau282/t-tau had similar decrease profiles to CSF MTBR-tau275/t-tau (Fig. 2c). Interestingly, CSF MTBR-tau275/t-tau did not change in PSP compared to control or FTLD-TDP even though these ratios increased moderately in the brain.

Table 2 Demographics and CSF MTBR-tau measures of participants in the pathologically confirmed primary tauopathy cohortFig. 2: 4R-specific CSF MTBR-tau decreases in CBD, FTLD-MAPT and AD.

a, Schematic of the quantified peptides of t-tau 212–221, truncation and 4R isoform-specific MTBR-tau in the R2 region (gray bars, MTBR-tau275 and MTBR-tau282). The relative abundance of each MTBR-tau was normalized to the t-tau peptide. b,c, CSF MTBR-tau275/t-tau (b) and MTBR-tau282/t-tau (c) significantly decreased in CBD (n = 18), AD (n = 10) and FTLD-MAPT (n = 5) compared to normal control (n = 29), FTLD-TDP (n = 21) and other FTLD-tau. FTLD-MAPT P301L (red, n = 2), R406W (blue, n = 2) and S305I (green, n = 1) decreased in MTBR-tau/t-tau measurements in this order. *P < 0.05, **P < 0.01, ***P < 0.001, ****P < 0.0001. The box plots show the minimum, 25th percentile, median, 75th percentile and maximum. Differences in biomarker values were assessed with a one-way ANOVA. A two-sided P < 0.05 was considered statistically significant and corrected for multiple comparisons using a Benjamini–Hochberg FDR set at 5%.

To assess if the soluble CSF MTBR-tau/t-tau measures reflected brain tau pathology measured by the paired insoluble brain MTBR-tau/t-tau measures, MTBR-tau/t-tau from antemortem CSF and brain from the same individuals were analyzed for correlation (n = 54; Fig. 3a,b). MTBR-tau275/t-tau and MTBR-tau282/t-tau from the CSF and brain correlated moderately (r = −0.27, P = 0.049 and r = −0.45, P = 0.0006, respectively) across all disease groups (n = 54) and strongly (r = −0.61, P = 0.0004 and r = 0.75, P < 0.0001; Fig. 3c,d) in 4R tauopathies (PSP, CBD and AGD, n = 29). This suggests that 4R-specific MTBR-tau species have inverse correlation in the CSF and brain in 4R tauopathies. In CBD, MTBR-tau275/t-tau and MTBR-tau282/t-tau from the CSF and brain correlated moderately but no statistical significance was obtained (n = 12, r = −0.25, P = 0.43 and r = −0.31, P = 0.33, respectively; Figure 3e,f). One CBD participant who had no cognitive impairment (Clinical Dementia Rating plus National Alzheimer’s Coordinating Center FTLD sum of boxes23 (CDR plus NACC FTLD-SB) = 0) had fewer changes in MTBR-tau275/t-tau and MTBR-tau282/t-tau in both brain and CSF (brain MTBR-tau275/t-tau = 0.321, brain MTBR-tau282/t-tau = 0.499, CSF MTBR-tau275/t-tau = 0.0071, CSF MTBR-tau282/t-tau = 0.0134). This may suggest that the changes of these biomarkers depend on the severity of the disease.

Fig. 3: CSF soluble MTBR-tau correlates with brain insoluble MTBR-tau aggregates.

a,b, MTBR-tau275/t-tau (a) and MTBR-tau282/t-tau (b) from paired CSF and brain inversely correlated in tauopathies, FTLD-TDP and control (n = 54, r = −0.27, P = 0.049, −0.45, P = 0.0006, respectively). c,d, MTBR-tau275/t-tau (c) and MTBR-tau282/t-tau (d) from paired CSF and brain had higher correlations (r = −0.61, P = 0.0004 and r = −0.75, P < 0.0001, respectively) in 4R tauopathies (CBD, PSP and AGD, n = 29). e,f, MTBR-tau275/t-tau (e) and MTBR-tau282/t-tau (f) from paired CSF and brain correlated in CBD (n = 12, r = −0.25, P = 0.43 and r = −0.31, P = 0.33, respectively). The gray shadow represents the 95% confidence intervals for the linear regression.

To assess if CSF MTBR-tau275/t-tau and MTBR-tau282/t-tau decrease with disease stage, the correlations between these biomarkers and duration (the interval between age at onset and CSF collection) were investigated (Table 2 and Supplementary Fig. 2). Average duration across diseases was 5 ± 4 years (n = 81) and 4 ± 1 years for CBD-only (n = 18). In CBD, there were negative correlations between duration and CSF MTBR-tau275/t-tau and MTBR-tau282/t-tau (r = −0.37 and −0.39, respectively) although statistical significance was not observed (P = 0.13 and 0.11, respectively). This result suggests that participants with longer duration who are at later pathological stages have larger degrees of decrease in the CSF MTBR-tau275/t-tau and MTBR-tau282/t-tau biomarkers.

CSF MTBR-tau is reproducible in repeated lumbar punctures

To evaluate reproducibility and stability of the CSF MTBR-tau measurements within the same individual, we examined CSF MTBR-tau275/t-tau in an independent cohort of 25 participants who underwent repeated lumbar punctures (LPs) (3–5 times) within approximately 4 months as a part of an ongoing study examining protein turnover kinetics24 (Extended Data Table 1 and Extended Data Fig. 1). These participants include individuals clinically diagnosed with PSP-Richardson’s syndrome (PSP-RS, n = 7) or corticobasal syndrome (CBS, a clinical syndrome associated with heterogenous neuropathological substrates including AD, CBD, PSP and FTLD-TDP; n = 9), with two participants having autopsy-confirmed CBD. They also included seven MAPT mutation (P301L, R406W and IVS10+16) carriers who were either symptomatic or asymptomatic and two noncarrier family members who are normal controls. The mean coefficient of variation for CSF MTBR-tau275/t-tau in repeated LPs was 12 ± 7%, establishing the high reproducibility and stability of CSF MTBR-tau measures within 4 months.

Consistent with the FTLD-MAPT cases analyzed with the pathologically confirmed cohort, CSF MTBR-tau275/t-tau decreased in FTLD-MAPT mutation carriers in the repeated LP cohort. Interestingly, CSF MTBR-tau275/t-tau decreased in two symptomatic FTLD-MAPT P301L mutation carriers (participant numbers 02 and 03, 0.00381 ± 0.00021) and a symptomatic FTLD-MAPT R406W mutation carrier (number 05, 0.00508) compared to normal control (numbers 01 and 04, 0.00666 ± 0.00027). However, CSF MTBR-tau275/t-tau did not change in FTLD-MAPT R406W mutation carriers who were asymptomatic at LPs (numbers 06 and 07). The FTLD-MAPT variant IVS10+16 promotes the splicing of tau exon 10, resulting in greater production of 4R over 3R isoforms. Indeed, symptomatic FTLD-MAPT IVS10+16 mutation carriers (numbers 08 and 09, 0.00921 ± 0.00053) had 1.38-fold higher CSF MTBR-tau275/t-tau compared to normal controls, indicating that an increase in 4R isoforms is reflected in the CSF.

CSF MTBR-tau275/t-tau decreased in the two participants who were clinically diagnosed as PSP-RS but later were autopsy-confirmed with CBD (numbers 10 (0.00396) and 11 (0.00535)), which is consistent with the pathologically confirmed CSF cohort results. However, CSF MTBR-tau275/t-tau did not change in participants clinically diagnosed with PSP-RS (0.00779 ± 0.00052) or CBS (0.00748 ± 0.00187), who are not yet autopsy-confirmed during the repeated measures studies. Average duration across diseases was 5 ± 3 years (n = 21) and 4 ± 2 for CBS only (n = 9), which were similar to the pathologically confirmed cohort.

CSF MTBR-tau in clinically diagnosed primary tauopathies

To estimate the CSF MTBR-tau biomarker performance in clinically diagnosed primary tauopathies, we measured CSF MTBR-tau275/t-tau in an additional independent cohort of 238 primary tauopathies with single LP (Extended Data Fig. 2). This cohort was previously analyzed for CSF t-tau and phosphorylated tau25 and includes clinically diagnosed cases of AD, sporadic bvFTD, bvFTD secondary to FTLD-MAPT, PSP-RS, CBS and the CBS-PSP continuum26. Individuals with the CBS-PSP continuum are defined as patients who initially presented with CBS but subsequently developed clinical features of PSP-RS as the disease progressed. CSF MTBR-tau275/t-tau decreased in the CBS-PSP continuum and FTLD-MAPT compared to cognitively normal controls (P < 0.05). However, CSF MTBR-tau275/t-tau did not statistically change in either AD or clinically diagnosed CBS compared to control or other tauopathies.

Diagnostic accuracies of CSF MTBR-tau in primary tauopathies

Finally, we examined the diagnostic accuracies of CSF MTBR-tau275/t-tau and MTBR-tau282/t-tau in a pathologically confirmed primary tauopathy cohort. First, CSF t-tau (mid-domain peptide 212–221) and phosphorylated tau (pT217/T217) were examined in primary tauopathies for comparison (Supplementary Figs. 1c and 3a). CSF t-tau increased in AD compared to normal control and PSP (P < 0.05) and can differentiate AD from FTLD-tau (PSP, CBD, AGD, PiD, FTLD-MAPT) with an AUC of 0.794 (Supplementary Fig. 3b). However, CSF t-tau does not distinguish among FTLD-tau. CSF pT217/T217 increased in AD compared to normal control, FTLD-TDP and FTLD-tau (P < 0.0001) and can differentiate AD from FTLD-tau with an AUC of 0.987 (Supplementary Fig. 3c). AD copathology in other neurodegenerative diseases (for example, FTLD-TDP, CBD, PSP) also increased CSF pT217/T217. These results suggest that we can use CSF pT217/T217 to accurately identify individuals with AD pathology, regardless of copathology.

Effect of amyloid on CSF MTBR-tau275/t-tau and MTBR-tau282/t-tau in primary tauopathies were further assessed using AD Thal phase. CSF pT217/T217 strongly correlated with AD Thal phase (r = 0.52, P < 0.0001; Supplementary Fig. 4a). However, CSF MTBR-tau275/t-tau and MTBR-tau282/t-tau did not correlate with AD Thal phase in the whole cohort (r = −0.22, P = 0.06 and r = −0.24, P = 0.04, respectively; Supplementary Fig. 4b,c) or in CBD (r = −0.14, P = 0.60 and r = −0.07, P = 0.78, respectively; Supplementary Fig. 4d,e). These results suggest that CSF MTBR-tau275/t-tau and MTBR-tau282/t-tau decreased in CBD independently from AD copathology.

The diagnostic accuracies of CSF MTBR-tau275/t-tau and MTBR-tau282/t-tau were examined to determine if we can distinguish CBD from control, FTLD-TDP and FTLD-tau as a group and individual tauopathy (Table 3 and Supplementary Fig. 5). CSF MTBR-tau275/t-tau and CSF MTBR-tau282/t-tau can distinguish CBD from normal control, other FTLD-tau (PSP, PiD and AGD), PiD and PSP with AUCs of 0.800–0.889. CBD can be distinguished from FTLD-TDP with AUCs of 0.701–0.770. When AD copathology cases were excluded, CSF MTBR-tau275/t-tau and CSF MTBR-tau282/t-tau can distinguish CBD from PSP with AUCs of 0.859 and 0.886, respectively (Table 3 and Supplementary Fig. 5k,l).

Table 3 Diagnostic accuracies of 4R-specific CSF MTBR-tau to distinguish CBD from FTLD-tau and control

Lastly, we retrospectively assessed CSF MTBR-tau275/t-tau and MTBR-tau282/t-tau by final clinical syndromes in a neuropathologically confirmed cohort to determine if these biomarkers can facilitate antemortem diagnosis of primary tauopathies. The numbers of individuals who had CSF MTBR-tau275/t-tau and MTBR-tau282/t-tau lower than cutoff (0.00563 and 0.01220, respectively; defined in Table 3 to differentiate CBD and PSP) within each clinical syndrome and neuropathological diagnosis are summarized (Extended Data Tables 2 and 3 and Supplementary Fig. 6). CBD (n = 18) had clinical syndromes of either CBS (n = 9), bvFTD (n = 4), nonfluent variant PPA (nfvPPA, n = 4) or PSP-RS (n = 1). Without biomarkers, the diagnostic accuracy of CBD within CBS was 9 out of 15 (60%). Among 15 individuals with CBS, 7 and 10 had lower than cutoff values of CSF MTBR-tau275/t-tau and MTBR-tau282/t-tau and 6 and 7 were CBD, respectively. With these biomarkers, the diagnostic accuracies of CBD within CBS were 6 out of 7 (86%) and 7 out of 10 (70%) for CSF MTBR-tau275/t-tau and MTBR-tau282/t-tau, respectively. Across different clinical syndromes of CBD, 12 out of 18 (67%) and 15 out of 18 (83%) had CSF MTBR-tau275/t-tau and MTBR-tau282/t-tau values lower than cutoff, suggesting that these biomarkers identify CBD regardless of clinical syndromes with up to 83% accuracy. Other applications of these biomarkers include distinguishing CBD from PSP among PSP-RS since 0 out of 11 (0%) and 2 out of 11 (18%) PSP had lower than cutoff for CSF MTBR-tau275/t-tau and MTBR-tau282/t-tau, respectively.