An overview of the tau protein and a summary of the analytical features investigated is shown in Fig. 2. Cohort 1 was subjected to a thorough investigation involving both soluble (TBS) and insoluble (SI) fractions as well as four different antibodies, while cohort 2 was used to verify previous results and the two fractions were therefore investigated with only the HT7 antibody, chosen as representative of the main findings obtained with all antibodies. Results comparison of Cohorts 1 and 2 are presented at the end of this section.

Disease-dependent tau isoforms profiles

Frontal grey matter brain samples were analysed using seven different fraction-antibody combinations (Fig. 2c). In general, the isoform measurements, both in TBS and SI fractions, indicate a higher abundance of 0N and 1N species compared with 2N. When examining the 3R and 4R isoforms, the relative abundance of these species showed similar presence of 3R and 4R in control and AD cases, prevalence of 3R in PiD and 4R in CBD and PSP, especially in the SI fraction. As tau protein was digested with trypsin, the link between the N and R isoforms was lost, but within each group they could be assessed and their relative amounts are presented in Fig. 3 and Suppl. Table 4 for all disease and fraction-antibody combinations and differences between the fractions are shown in Suppl. Figure 4.

Fig. 3

Relative portion of the 0N/1N/2N and 3R/4R isoforms. Each panel represents a fraction-antibody combination as indicated. The individual bars are colour coded for the respective isoforms and grouped according to disease. The black horizontal lines indicate the mean percentage for each isoform. See also Suppl. Table 4 for respective mean values

Controls

N isoforms In TBS, 0 N isoform was slightly more abundant, especially when immunoprecipitated with 77G7, closely followed by 1 N. The 2 N only isoform represented a small percentage of the total tau amount (0 N/1 N/2 N: Tau12 = 52/45/3%, HT7 = 53/45/2%, 77G7 = 57/40/3%, TauAB = 51/47/2%). The same was observed in the SI fraction, with a moderate increase in the levels of 0 N in relation to 1 N and 2 N (0 N/1 N/2 N: HT7 = 57/41/2%, 77G7 = 61/36/3%, TauAB = 53/45/2%).

R isoforms 3R isoform was slightly more abundant than 4R in both TBS and SI. In TBS, IP with Tau12, HT7 and TauAB in approximately a 6/4 ratio in 3R/4R isoforms (3R/4R: Tau12 = 64/36%, HT7 = 59/41%, TauAB: = 62/38%), in contrast with IP with 77G7, which resulted in a nearly 1/1 ratio (3R/4R: 77G7 = 52/48%). In SI, the isoform ratio was virtually the same as in TBS fraction, with a slightly higher abundance of 3R compared with 4R isoform, with 77G7 being closest to a 3R/4R ratio of 1/1 (3R/4R: HT7 = 62/38%, 77G7 = 55/45%, TauAB: = 60/40%).

Alzheimer’s disease

N isoforms In AD TBS fraction, regardless of the antibody used, the abundance of 0N isoform was higher than 1N. This was especially noticeable when using 77G7 antibody for the IP, similar to what was observed in controls. With values of approximately 5%, 2N was the least abundant isoform in TBS (0N/1N/2N: Tau12 = 50/46/4%, HT7 = 51/46/3%, 77G7 = 59/37/4%, TauAB = 53/45/2%). Interestingly, there was a clear, significant increase of 0N vs. 1N and 2N with all antibodies in the SI fraction when compared with TBS (0 N/1 N/2 N: HT7 = 69/29/2%, 77G7 = 73/25/2%, TauAB = 69/29/2%), see Suppl. Figure 5.

R isoforms In TBS 3R was moderately more abundant than 4R, with mean values of around 60%, except when the IP was performed with 77G7, where a 1/1 3R/4R ratio was observed (3R/4R: Tau12 = 65/35%, HT7 = 59/41%, 77G7 = 50/50%, TauAB: = 63/37%). SI HT7 and TauAB showed similar results for 3R/4R isoforms as for their TBS counterparts, which was not the case for 77G7, which resulted in a clear, significant increase in 3R isoform compared with 4R (3R/4R: HT7 = 63/37%, 77G7 = 69/31%, TauAB: = 56/44%), see Suppl. Figure 5.

Progressive supranuclear palsy

N isoforms In TBS, IP with Tau12 and TauAB resulted in slightly higher 1N than 0N. IP with HT7 yielded similar amounts of 0N and 1N, whereas IP 77G7 resulted in more 0N than 1N. Regardless of the antibody used for IP, 2N was the least abundant isoform. (0N/1N/2N: Tau12 = 46/51/3%, HT7 = 48/49/3%, 77G7 = 53/43/4%, TauAB = 44/54/2%). In the SI fraction, IP with all three antibodies resulted in significantly higher levels of 0N, see Suppl. Figure 5. 2N was always the least abundant isoform regardless of the antibody used (0N/1N/2N: HT7 = 54/44/2%, 77G7 = 57/40/3%, TauAB = 55/43/2%).

R isoforms In the TBS fraction, IP with either Tau12 or TauAB resulted in higher levels of 3R, whereas IP with HT7 and 77G7 rendered equal amounts of 3R and 4R isoforms (3R/4R: Tau12 = 64/36%, HT7 = 52/48%, 77G7 = 50/50%, TauAB = 60/40%). In the SI fraction, HT7 pulled slightly higher amounts of 4R than 3R isoform, whereas IP with 77G7 and, in particularly, TauAB resulted in significantly more 4R tau (3R/4R: HT7 = 47/53%, 77G7 = 43/57%, TauAB = 36/64%), see Suppl. Figure 5.

Corticobasal neurodegeneration

N isoforms In the CBD TBS fraction, there were slightly higher levels of 0N isoform, especially when immunoprecipitated with 77G7, compared with 1N and 2N (0N/1N/2N: Tau12 = 52/45/3%, HT7 = 53/45/2%, 77G7 = 60/36/4%, TauAB = 52/47/1%). In the SI fraction the abundance of the 0N isoform was significantly higher (0N/1N/2N: HT7 = 61/37/2%, 77G7 = 67/30/3%, TauAB = 67/31/2%), see Suppl. Figure 5.

R isoforms When it comes to the R isoforms, IP of the TBS fraction with all antibodies except Tau12 resulted in lower levels of 3R compared with 4R, TauAB being the antibody that yielded the largest difference. (3R/4R: Tau12 = 57/43%, HT7 = 40/60%, 77G7 = 43/57%, TauAB = 31/69%). In SI, the 4R prevalence in CBD becomes very apparent with all three antibodies with a significant decrease of 3R (3R/4R: HT7 = 17/83%, 77G7 = 13/87%, TauAB = 8/92%), see Suppl. Figure 5.

Pick’s disease

N isoforms Regarding N isoforms, 0N was generally the most abundant in TBS, especially when immunoprecipitated with 77G7, and except for TauAB, followed by 1N and 2N (0N/1N/2N: Tau12 = 51/46/3%, HT7 = 55/43/2%, 77G7 = 65/31/4%, TauAB = 44/53/3%). In the SI fraction, the predominance of the 0N isoform became more evident compared with 1N and 2N (0N/1N/2N: HT7 = 62/36/2%, 77G7 = 68/29/3%, TauAB = 65/33/2%); however, due to the individiual heterogeneity no changes were significant, see Suppl. Figure 5.

R isoforms For the R isoforms, the TBS fraction of PiD displayed higher levels of 3R compared with 4R, except when immunoprecipitated with 77G7 (3R/4R: Tau12 = 64/36%, HT7 = 58/42%, 77G7 = 46/54%, TauAB = 68/32%). The 3R predominant nature of PiD was significantly more distinguished in the SI fraction, where 3R was remarkably more abundant than 4R (3R/4R: HT7 = 84/16%, 77G7 = 81/19%, TauAB = 88/12%), see Suppl. Figure 5.

Disease dependent non-phosphorylated tau profiles in the human brain

Average profiles are displayed in Fig. 4 showing one disease group per panel (an alternative view with one fraction-antibody combination per panel is shown in Suppl. Figure 6, and scatterplots for each quantified peptide are shown in Suppl. Figure 7). These profiles differed depending on the antibody used for IP and the overall effect is easiest to observe in Suppl. Figure 6. To avoid artificial dents in the curves, the signals of the isoform-specific peptides were summed so that the data point 0N + 1N + 2N is the sum of peptides 45–68 0N, 68–97 1N, and 68–87 2N, and 3R + 4R is the sum of 275–286 3R and 299–317 4R peptides. With the exception of AD, tau profiles in TBS showed a clear dependence on the antibody epitope and the peptide abundance generally decreased with distance to the epitope. For AD, the tau profiles for all four antibodies showed an increased presence of MTBR peptides. Of note, the 25–44 peptide was lower than its two adjacent peptides in most profiles; the reason for this is not clear but it may be caused by partial oxidation of methionine, which accounts for a substantial portion of both the brain-derived and the [U-15N]-protein standard peptide; however, the 6–23 peptide also contains a methionine that also is partially oxidised, making the statement less convincing. This effect has been observed previously [60, 61]. In the SI fraction, tau profiles in tauopathies resulted in high levels of MTBR peptides regardless of the antibody epitope, being especially increased in AD, followed by CBD and PiD, and finally by PSP, which displayed the most level profiles. In control cases, tau profiles in the SI fraction looked comparatively flat (i.e., similar abundance of different tau peptides across the whole tau sequence), except for HT7 and 77G7 which exhibited slightly higher abundance of mid-region and MTBR peptides, respectively.

Fig. 4

Abundance of monitored tryptic peptides along the tau protein chain. Each panel represents a patient’s group profile as indicated. The profiles are mean values for each fraction-antibody combination. Peptides representing different isoforms (0N/1N/2N and 3R/4R) are summed in the main graphs to avoid dents and their mean shares are shown in the respective side panels. Antibody epitope locations are indicated with grey arrows

Controls

TBS-soluble fraction

After IP with the N-terminal Tau12 antibody, the tau profile between 6 and 23 and 3R + 4R was relatively, flat after which it dropped for peptides 354–369 and 407–438, which had similar levels. IP with HT7 (targeting mid-domain tau) and 77G7 (MTBR) resulted in higher abundance of mid-region and MTBR peptides respectively. On the other hand, while the tau peptide profile between 3R + 4R and 407–438 with HT7 was similar to that of Tau12 (i.e., a steep decrease between 3R + 4R and 354–369 where it levelled out until 407–438), this was not the case for 77G7, which decreased more gradually towards the C-terminus. Finally, IP with TauAB yielded the only (almost) flat profile, mostly unchanged throughout the whole tau sequence. Altogether, the analysed TBS tau pools in control cases seemed to be comprised of a mix of full-length tau and fragments containing N-terminal, mid-region and MTBR regions.

Sarkosyl-insoluble fraction

In control SI fractions, tau protein was less abundant than in TBS. Additionally, tau profiles in SI were flatter than their TBS counterparts. IP with HT7, a very subtle increased abundance in mid-region and MTBR peptides was noticeable, especially for the 212–221 peptide. Similarly, with 77G7, most abundant peptides were circumscribed to the mid-region and the MTBR sequence, but this time with the highest abundance shifted toward the MTBR region, where 3R + 4R was slightly higher that the surrounding peptides. Finally, IP with TauAB resulted in a virtually flat tau profile. In summary, the analysed control SI fractions were predominantly comprised of full-length tau and some mid-region and MTBR fragments.

Alzheimer’s disease

TBS-soluble fraction

In the AD TBS fractions, all tau profiles showed elevated abundance of MTBR peptides, regardless of which antibody was used for the IP. When immunoprecipitated with Tau12, the tau profile was characterized by a lower amount of the 212–221 peptide. This was followed by a peak at 243–254 and 3R + 4R peptides, from which the abundance gradually dropped towards the C-terminus, with 407–438 being the least abundant of all. Tau profiles for HT7 and 77G7 were relatively similar. Both displayed a progressive increase toward the MTBR region. While HT7 showed the highest abundance at 243–254 and 3R + 4R peptides, the maximum abundance peak for 77G7 was at 3R + 4R and 354–369. With both antibodies, the C-terminal 407–438 peptide was low; with HT7 its levels were similar to those of Tau12, while for 77G7, 407–438 was somewhat more abundant. Lastly, the tau profile with TauAB was also prominent in MTBR peptides, and very low in N-terminal, mid-region, and C-terminal species. Notable was the pronounced drop for 212–221. Overall, the analysed AD TBS fractions seemed to be comprised of full-length tau and substantial amounts of MTBR tau fragments.

Sarkosyl-insoluble fraction

In the AD SI fraction, IP with all three antibodies resulted in similar tau profiles, which again were characterized by the remarkably prominent abundance of peptides belonging to the MTBR, which were much more dramatically increased than in their TBS counterparts. All three profiles displayed substantially lower levels of N-terminal, mid-region and C-terminal peptides when compared with those belonging to the MTBR. In the mid-region, lower abundance of the 212–221 peptide was seen with HT7 and TauAB, while for 77G7 the more N-terminal peptides showed even lower abundance than the 212–221 peptide. Regardless of epitope, all anti-tau antibody profiles displayed extremely high levels of MTBR peptides, where 3R + 4R and 354–369 were more abundant than 243–254 (for TauAB all MTBR species had similar abundance). In all three profiles, the C-terminal peptide 407–438 was the least abundant tau species, even when immunoprecipitated with TauAB. Thus, AD SI profile was clearly dominated by large amounts of MTBR fragments, also including, to a lesser degree, N-terminal fragments and full-length tau. Overall tau levels in SI fraction from AD were much higher than any other disease group investigated; the abundance of MTBR peptides in AD was 12–72 times higher than in controls and for non-isoform specific MTBR peptides 7–18 times more than in CBD (Suppl. Figure 7 h-k).

In the AD group (both in TBS and SI fractions), 212–221 was typically lower than the adjacent peptides (Fig. 4 and Suppl. Figure 6). It is likely that the observed reduction is due to phosphorylation (see Sect. Phosphorylations below). As aforementioned, the most prominent finding involved the MTBR peptides located between amino acids 243 and 369. Although the effect was visible in the TBS fraction, it was striking in the SI fraction where these peptides were increased more than 10 times even when using antibodies with epitopes outside the MTBR. This effect has also been observed previously [61] and is likely caused by aggregates comprised of several tau peptides belonging exclusively to the MTBR and attached to longer peptides that can be captured by the antibody. Noteworthy is that CBD also exhibited this behaviour, albeit to a lesser degree than for AD (see below).

Progressive supranuclear palsy

TBS-soluble fraction

Among the tauopathies examined here, PSP is the one with a tau peptide brain profile resembling that in the control group the most (Fig. 4 and Suppl. Figure 6). In TBS, tau profiles with Tau12 and HT7 were very similar. Both were rather flat in the N-terminal and mid-region. This was followed by a very subtle increase for MTBR peptides 243–254 and 3R + 4R when immunoprecipitated with Tau12, whereas with HT7 mid-region and MTBR peptides showed similar abundance. With both antibodies, abundances dropped steeply at peptide 354–369, and continued with a slight decrease towards the C-terminal peptide 407–438 (the least abundant). IP with 77G7 was somewhat similar to those of Tau12 and HT7 but displayed a more pronounced presence of 3R + 4R, which clearly stood out as the most abundant peptides in the profile. From here, the 77G7 profile decreased towards the C-terminal end, but in a more gradual way compared with Tau12 and HT7. Finally, when immunoprecipitated with TauAB, the profile appeared to be for the most part flat, with a subtle increasing trajectory toward the C-terminal end (407–438 was the most abundant), which clearly contrasted with the other antibodies. Altogether, TBS tau in the PSP group seems to be a mixture of full-length tau and fragments containing N-terminal, mid-region, and MTBR regions.

Sarkosyl-insoluble fraction

Similar to the TBS fraction, tau peptide profiles in the SI fraction for PSP resemble those seen in the control group. All three antibodies yielded reasonably flat profiles. For all antibodies there was a moderate increase from the N-terminus towards the MTBR region followed by a decrease towards the C-terminus. The differences between the profiles are that HT7 displayed similar amounts of mid-region and MTBR peptides, while TauAB profile seemed to show higher abundance of MTBR and a small drop for 212–221, whereas IP with 77G7 resulted in an in between profile of HT7 and TauAB. Altogether, PSP SI fraction seems to mainly comprise full-length tau and some mid-region and MTBR fragments.

Corticobasal neurodegeneration and pick’s disease

TBS-soluble fraction

The tau abundance profiles of CBD and PiD in TBS fraction were virtually identical to one another. Additionally, the profiles of CBD and PiD were overall similar to controls and PSP, although the latter two groups having higher levels of all peptides (Suppl. Figure 6). CBD and PiD TBS fractions were comprised of a mix of full-length tau and fragments containing N-terminal, mid-region and MTBR. When using Tau12, CBD and PiD profiles were rather flat (i.e., had similar abundance) across N-terminus, mid-region, and deep into the MTBR (until 3R + 4R), after which they started to gradually drop, with C-terminal 407–438 being the least abundant species. When immunoprecipitated with HT7, mid-region and MTBR peptides were the most dominant, and a similar gradual decrease towards the C-terminal region was observed. With 77G7, MTBR peptides dominated the profile, closely followed by mid-region and neighbouring N-terminal peptides (0N + 1N + 2N). Least abundant peptides were C-terminal and most N-terminal (6–23 and 25–44). Finally, with TauAB, the profile remained virtually flat, showing more or less the same abundance throughout whole sequence, which may indicate that tau species containing the C-terminal region are rather intact and mostly comprised by full-length tau. Overall, CBD and PiD TBS fraction seemed to include a mixture of full-length tau and fragments comprising N-terminus, mid-region, and C-terminus.

Sarkosyl-insoluble fraction

Like in TBS, the CBD and PiD tau peptide profiles in the SI fraction displayed the same shape, but this time their peptide abundance was markedly different, with CBD having higher levels of tau. Interestingly, the CBD and PiD profiles showed a clear predominance of MTBR-containing tau species, which resemble those of AD, but comparatively less abundant. Regardless of the antibody used, the N-terminal, mid-region, and C-terminal peptides were the least abundant, whereas peptides belonging to the MTBR, i.e., 243–254, 3R + 4R, and 354–369, were the most abundant with fold changes up to 2 and 4 compared with controls for CBD and PiD, respectively (Suppl. Figure 7 h, k). There was also a small drop for the 212–221 peptide when immunoprecipitated with TauAB and 77G7. Overall, these results suggest that CBD and PiD SI fractions are comprised by a large amount of MTBR peptides and to a lesser degree, full-length tau.

Phosphorylations

Using the explorative mode of analysis, 36 sites could be identified as harbouring phosphate groups (Fig. 2a) and 16 phospho-peptides were quantified (Table 2). Suppl. Table 5 contains a list of all phospho-peptides identified when using PEAKS Studio Xpro as search engine.

Sometimes it was difficult to pinpoint the exact location of the phosphorylation, since the discriminatory information could rely on as little as two possible fragment peaks. Since peptides with the same number of phosphate groups, but at different positions, also tended to elute at (almost) the same time the fragment spectra could contain peaks from both variants. Nonetheless, with manual evaluation it was possible to confirm the identity of many of the singly, doubly, and triply phosphorylated peptides.

Quantification of the phosphorylated peptides was performed in two ways. Since no useful phosphorylated protein standard was available, to obtain the best relative quantification for comparison between the different patient groups synthetic labelled extended phospho-peptides were used as internal standards. However, since the quantitative method was optimised mainly using TBS fraction material, some phosphorylated peptides later found in SI fraction were not observed and investigated until after spectra from all samples were acquired. Here, the strength of the approach became evident; as spectra were acquired in a data dependent mode, database searches could be performed. Several phospho-epitopes were then detected that were not observed in the method development stages. A number of phospho-peptides were then added to the quantification analysis. For the quantification of these newly found peptides different surrogate internal standards were used (Table 2).

Singly phosphorylated peptides

In the TBS fraction, two of the six measured singly phosphorylated peptides were generally lower in CBD and PiD when compared to all other groups; i.e., peptide 175-190-p1 (p181, Fig. 5, Suppl. Figure 8b) and peptide 225-240-p1 (p231, Fig. 5, Suppl. Figure 8i). For HT7 there was a tendency for higher levels in the AD group for all peptides; this could also be seen for TauAB for 210-224-p1 and 225-240-p1. Peptide 212-224-p1 (mostly p217 but also p214, Fig. 5, Suppl. Figure 8 h), was the only mid-region singly phosphorylated peptide being substantially more abundant (4–20 times) in the AD group already in the TBS fraction and regardless of the antibody used for IP.

Fig. 5

Selected scatter-plots of quantified tryptic phospho-peptides after IP with HT7. Shown are abundancies of peptides carrying one (p1, upper pannels), two (p2, middle panels), or three (p3, bottom panels) phosphorylations. Data is grouped according to fractions. TBS fractions are indicated in blue and SI fractions in pink. Panel top: box/scatter plots with medians, inter-quartile intervals, and min/max intervals indicated. Fold changes of the mean values compared with the controls in respective fraction are indicated for each disease group. Panel bottom: significances according to one-way ANOVA not assuming homoscedasticity (Brown-Forsythe/Welch) followed by unpaired t-test with Welch’s correction. Calculations are only performed within each fraction, upper right (blue) for TBS and lower left (pink) for SI; ns not significant (p ≥ 0.05), * p < 0.05, ** p < 0.01, *** p < 0.001, **** p < 0.0001, – test not possible to perform due to zero variance in one of the groups compared. See Suppl. Figure 7 for all antibody-fraction combinations analysed

The C-terminal phospho-peptide 386-406-p1 (mainly p396 or p403/404, Fig. 5, Suppl. Figure 8 l) was also more abundant (21 times) in the AD group already in the TBS-soluble fraction compared with the other groups, which did not differ between themselves. For the most C-terminal phospho-peptide 407-438-p1 (p413, p416, or p422, Fig. 5, Suppl. Figure 8o) AD appeared to have higher levels than other groups but the general abundance of these species in TBS was low and the quantitative data less reliable.

In the SI fraction, all singly phosphorylated peptides were much more abundant in AD than in any of the other groups, showing fold changes of more than 1000 compared with the control group, i.e., considerably more than in the TBS fraction (Fig. 5, Suppl. Figure 8b, c, e, h, i, l, o). CBD exhibited the second highest abundance of all groups for all singly phosphorylated peptides (up to 880 times more than for controls) except 195-209-p1 (mainly p202 or p199; Fig. 5, Suppl. Figure 8b, c, e, h, i, l, o). Singly phosphorylated peptides in PSP, PiD, and controls were often comparatively similar with some variation depending on peptide and antibody.

Doubly and triply phosphorylated peptides

Several peptides with two and three phosphate groups were detected and quantified (Fig. 5, Suppl. Table 5, Suppl. Figure 8). In the TBS fraction, all peptides carrying two phospho-sites were found increased in the AD group when compared with the other groups (Fig. 5, Suppl. Figure 8d, f, j, m, p, ), with the peptides 210-224-p2 (p212 + p217; 30–500 times higher than controls and 12–70 times more than CBD, Fig. 5, Suppl. Figure 8f) and 225-240-p2 (p231 + p235; 18–66 times higher than controls and 5 times more than CBD, Fig. 5, Suppl. Figure 8j) being the most abundant (Suppl. Figure 9). Also triply phosphorylated peptides observed were markedly increased in AD compared with the other groups in the TBS fraction. Their profiles followed that of doubly phosphorylated, but generally triply phosphorylated peptides appeared to be less abundant than their doubly phosphorylated counterparts, especially in the TBS-soluble fraction where, for example, the C-terminal peptides 386-406-p3 and 407-438-p3 had a relatively low signal also for the AD group (Suppl. Figure 9). The only exception was 210-224-p3, which was similar in abundance to the corresponding doubly phosphorylated peptide (Suppl. Figure 9), but displayed higher fold changes (110–380 times more than controls and 5–7 times more than for CBD; Fig. 5, Suppl. Figure 8j, k). It is noteworthy that also the other tauopathies, in particular CBD, had increased multiple-phospho-peptide levels in TBS compared with controls, displaying moderate to high fold changes (Fig. 5, Suppl. Figure 8).

In the SI fraction, the multiple phosphorylated peptides were even more distinctly elevated and showed higher fold changes in AD than in the TBS fraction (up to more than 1000 compared with controls; Fig. 5, Suppl. Figure 8f, g, j, k, m, n, p, q and 9), but also the other tauopathies had clearly higher levels than controls (Fig. 5, Suppl. Figure 8f, g, j, k, m, n, p, q). As previously observed for singly phosphorylated peptides in the SI fraction, CBD exhibited the second highest abundance after AD for all doubly and triply phosphorylated peptides, followed by PSP and PiD, which generally displayed comparable levels (Fig. 5, Suppl. Figure 8f, g, j, k, m, n, p, q and 9).

Particular phosphorylated peptides

Since the tryptic digestion was sometimes incomplete, several sequence stretches were covered by peptides of different lengths. One example is the group 210–221, 210–224, 212–221, and 212–224, which is the result of missed cleavages in the sequence 210SRTPSLPTPPTREPK224, where boldface indicates the possible phospho-sites 212, 214, and 217, and a shift between italic and plain text indicate a possible tryptic cleavage site. The phospho-peptide that gave the highest signal was 210–224; this peptide was observed with one, two, or three phosphate groups. The 212–224 peptide was mainly observed with one phosphate group, although the same peptide carrying two phosphate groups was detected as well. When inspecting the MS/MS spectra, the most prominent phospho-site on 212–224 was the one at position Thr-217, but single phosphorylation was also observed at position Ser-214 as well, but not at Thr-212. In contrast, the longer 210–224 form was mainly phosphorylated at position Thr-217, but single phosphorylation was observed to a large extent also at Thr-212. The reason for the Thr-212-phosphorylation being observed only for 210–224 is most likely because the phosphate group at Thr-212 hinders the trypsin cleavage at that position to generate the shorter peptide.

The 195–209 peptide was not visible in TBS because of technical reasons (the peptide is relatively hydrophilic and was borderline to bind to the column), but the MS/MS data obtained from the SI fraction showed that single phosphorylations were most frequent at Ser-202 and Ser-199, followed by Thr-205. Further, Fig. 5 and Suppl. Figure 8d shows that, in the SI fraction, the peptide carrying two phosphate groups was highly increased in AD only, as compared with the same peptide phosphorylated only at one position which seemed to be ubiquitously present in other groups (Fig. 5, Suppl. Figure 8c; data from HT7 and 77G7 only).

Non-AD tauopathies

In TBS, all three non-AD tauopathies showed increased levels of multiply phosphorylated peptides compared with controls, especially for CBD. This was not the case for singly phosphorylated peptides. In SI, all non-AD tauopathies generally showed increased phospho-peptide levels compared with controls (Fig. 5, Suppl. Figure 8). This was again particularly evident for CBD, which in SI fraction showed high levels of peptides carrying two or three phospho-groups.

Distribution profiles of phosphate groups

Naturally, it was also of interest to compare the relative abundance of the different phosphorylated peptides at different positions. Most phospho-peptides could be quantified, thus relative comparisons across disease groups were made. Similar to non-phospho peptides, the relative tryptic phospho-peptide abundance along the tau protein chain is presented in Suppl. Figure 9. In the TBS fraction, the phospho-tau profile of control, PSP, CBD, and PiD were similar in both shape and peptide abundance, with 175-190-p1 (p181) and 225-240-p1 (p231) being the most prominent peptides. In addition, IP with TauAB in CBD resulted in 225-240-p2 (p231 + p235) being also noticeable and with similar abundance as 175-190-p1 (p181) and 225-240-p1 (p231). On the other hand, AD was characterized by higher levels of doubly and triply phosphorylated peptides compared with all other groups. The most abundant peptides in AD TBS fraction included 175-190-p1 (p181) and 225-240-p2 (p231 + p235), followed by 210-224-p2 (mainly p212 + p217), see Suppl. Figure 9.

In the SI fraction, the controls exhibited a similar profile to that in the TBS fraction but with lower abundance, similarly as previously observed in non-phosphorylated peptide profiles. PSP, CBD, and PiD displayed similar profiles, all characterized by the increased presence of multiply phosphorylated peptides compared with their TBS counterpart and also compared with control SI. The main difference between the groups was the overall peptide abundance, with PSP showing the lowest relative number of phospho-peptides followed by PiD and CBD. Interestingly, as was observed for non-phosphorylated tau profiles, the PSP profile was again similar to that of the control group. The higher abundance of phospho-peptides in CBD was visible in almost all investigated tau species, and generally characterized by the dominance of multiply phosphorylated peptides over their singly phosphorylated counterparts. Phospho-peptides were remarkably more abundant in AD SI than in any other disease group and were also higher than in the AD TBS fraction. The AD SI profile was characterized by the very pronounced presence of 225-240-p2 (p231 + p235) and 225-240-p3 (p231 + p235 + p237/238). They were followed by 175-190-p1 (p181), 210-224-p2 (p212 + p217), 210-224-p3 (p212 + p214 + p217), 386-406-p2 (p396 + p400/403/404), and 386-406-p3 (p396 + p400 + p403/404), see Suppl. Figure 9.

Antibody comparison

In TBS, there were clear differences in the non-phosphorylated and phosphorylated tau peptide profiles when using different antibodies for IP, suggesting that although a large part of the soluble tau species likely are full-length tau, a substantial proportion of soluble tau exists also as fragments (Figs. 4 and 5, Suppl. Figures 6–9). The singly phosphorylated peptides from the proline-rich mid-region were generally more abundant when immunoprecipitated with Tau12 or HT7. This contrasted with 386-406-p1 (mainly p396 and p403/404) which was most abundant and with 77G7 followed by TauAB. The most C-terminal, singly phosphorylated peptide, 407-438-p1 (mainly p413, p416, or p422), displayed slightly different results (Fig. 5, Suppl. Figure 8). When using Tau12 or TauAB, 407-438-p1 was observed in the control, AD, and PSP groups and when using 77G7 also in PiD. However, when using HT7 it was detected in only two AD cases (Fig. 5, Suppl. Figure 8).

The results from IP with the C-terminally directed TauAB antibody are noteworthy. In the SI fraction, the relative amounts of phosphorylated mid-region peptides were much higher than when using the mid-region directed HT7 antibody (which has epitope close to most quantified mid-region phospho-peptides) or the MTBR directed 77G7 antibody (Suppl. Figure 8). This effect did not occur in controls and was rather small in PSP, but in PiD, CBD, and AD it became increasingly clear. Even in the TBS fraction, there were noticeably high levels for multiply phosphorylated peptides in AD and CBD.

Other detected but not quantified phosphorylations

There were several phosphorylated sites that were detected and identified but not investigated in a quantitative manner (Suppl. Table 5). In the N-terminal region, phosphorylation at Ser-46, Thr-50, as well as at both residues, were detected for the 25–67 peptide (1 missed cleavage) belonging to both the 1N and 2N isoforms. The 1N specific 45–97 peptide (1 missed cleavage) was also found phosphorylated at Ser-68/Thr-69 or both, and Ser-113 was found phosphorylated at peptides from all three N isoforms. In the mid-region, 171–180 phosphorylated at Thr-175 was detected with TauAB in the SI fraction (most likely its absence at the other occasions were because of issues with the binding to the column).

There were also several phospho-epitopes detected in the MTBR that were not quantified. Ser-262/Thr-263 at peptide 258–274 was frequently detected but was excluded from the quantification due to high CV; the same was the case for Ser-305 at the 4R specific 299–317 peptide as well as Ser-324 at the 322–350 peptide. Finally, Ser-289 at the 4R specific peptide 281–290 and Ser-356 at the 354–369 peptide were also detected and identified in a few samples.

Correlations

Analysis of the correlations between the different tau peptides revealed disease-specific patterns.

Correlations of non-phospho peptides

Correlations between non-phospho peptides are shown in Suppl. Figure 10. Most peptides in the control group correlated positively with each other and there was no particular distinction between the TBS and SI fractions. PSP exhibited a similar pattern, although for the SI fraction the more C-terminal peptides in the MTBR (4R and 354–369) had a different behaviour showing lower correlation. In CBD the pattern was similar to that of PSP but C-terminal peptides in the MTBR (4R and 354–369) in SI show weak and moderate negative correlations with TBS fraction peptides. PiD also showed a similar pattern, but here all SI fraction MTBR peptides except 4R behaved differently from the general correlation pattern with a weak negative correlation. Finally, in AD there was a clear separation in correlation between the N-terminal and mid-region peptides compared with the MTBR peptides as well with the C-terminal 407-438-peptide (Suppl. Figure 10b). There was a general pattern of negative correlation between TBS N-terminal and mid-region peptides and SI fraction. N-terminal and mid-region peptides associated positively and strongly within TBS or SI. The same was observed for MTBR species. In the TBS fraction, peptide 407–438 correlated well with the other non-MTBR peptides.

Correlations of phospho-peptides

The correlations between phosphorylated peptides are shown in Suppl. Figure 11. For controls, the most common single phosphorylations 175-190-p1 (p181) and 225-240-p1 (p231) correlated best with 225-240-p2 (p231 + p235), r = 0.67–0.98 and r = 0.63–0.92, respectively. For AD, most peptides correlated well; especially within the SI fraction. For PSP, CBD, and PiD the situation was somewhere between that of controls and AD, with relatively good correlations within the SI fraction, and more variable otherwise. PSP in TBS and SI fractions generally showed positive and moderate correlations within and with each other, whereas in CBD, this only happened within each fraction but not between TBS and SI. Finally, PiD showed a stronger correlation in the SI fraction.

Correlations of phospho-peptides to non-phospho peptides

The correlations between phosphorylated and non-phosphorylated peptides are shown in Suppl. Figure 12. In controls, the only noticeable pattern was that the TBS phospho-peptides exhibited higher correlations with all non-phospho peptides than the SI phospho-peptides did. In AD, most phospho-peptides correlated best with the MTBR non-phospho peptides, especially in the SI fraction. Most phospho-peptides in TBS and SI anti-correlated with non-phospho peptides outside the MTBR in TBS (Suppl. Figure 12b). Different tau regions, N-terminal to mid-region, MTBR, a C-terminal region, associated very differently with the phosphorylated peptides.

For PSP, the pattern was more similar to that of controls, but with weaker correlations. In general, TBS phospho-peptides had low or positive correlation with most non-phospho peptides. Most SI phospho-peptides, on the other hand, had low or negative correlation with most non-phospho peptides. The main exception to this was that most SI phospho-peptides (and to a lesser degree TBS phospho-peptides) correlated positively with the SI MTBR non-phospho peptides 299–317 4R and 354–369 while the SI 275–286 3R peptide showed no correlation. CBD showed a similar pattern as PSP but with more pronounced positive correlations, where most SI phospho-peptides correlated strongly with all SI MTBR peptides except 3R, also 243–254. The SI phospho-peptides also had a weak positive correlation with SI N-terminal to mid-region non-phospho peptides. In PiD, most phospho-peptides had, with some exceptions, low or negative correlation with most non-phospho peptides in TBS. TBS 175-190-p1 (p181) and TBS 225-240-p1 (p231) in general correlated positively with non-phospho peptides. The most noticeable feature, however, was the strong positive correlation of SI phospho-peptides with all SI MTBR non-phospho peptides except 4R. This was also the case for several TBS phospho-peptides.

As an example, the correlations between selected phospho-peptides and selected non-phospho peptides for both TBS and SI fractions immunoprecipitated with 77G7 are shown in Fig. 6. Highlighted are the high positive correlations between phosphorylated peptides in SI and the disease associated MTBR peptides in SI, i.e., non-isoform-specific 354–369 for all diseases (AD r = 0.72–0.89, p < 0.03; PSP r = 0.55–0.77, p < 0.09; CBD r = 0.91–0.97, p < 0.001; PiD r = 0.95–0.96, p < 0.0001), 4R for AD (r = 0.75–0.95, p < 0.02), PSP (r = 0.63–0.87, p < 0.04), and CBD (r = 0.94–0.98, p < 0.001), and 3R for AD (r = 0.76–0.90, p < 0.02) and PiD (r = 0.97–0.99, p < 0.0001). No such correlations were observed in TBS.