Tau phosphorylation at Alzheimer's disease biomarker sites impairs its cleavage by lysosomal proteases

Abstract

INTRODUCTION

Phospho‐tau peptides from the proline‐rich domain (PRD) of tau are sensitive biomarkers for Alzheimer's disease (AD). The PRD is known to be relatively resistant to lysosomal proteolytic cleavage, but the effects of phosphorylation on cleavage are unknown.

METHODS

Using in silico modeling and in vitro protease assays, we quantified the effects of phosphorylation on lysosomal proteolysis of tau. We further assessed levels of lysosomal proteases in patient‐derived cerebrospinal fluid (CSF) relative to phosphorylated tau‐181 (p‐tau181).

RESULTS

Phosphorylation renders the PRD significantly resistant to cleavage by the lysosome, especially at less acidic pH setpoints. In Alzheimer's disease subjects, CSF levels of lysosomal proteases correlate with p‐tau181, suggesting that p‐tau peptides are released with lysosomal contents.

DISCUSSION

Loss of lysosomal acidity may contribute to the release of phospho‐tau biomarkers. This study shows that phosphorylation of tau impairs its cleavage by proteases in a pH‐dependent manner and provides a novel molecular basis for p‐tau biomarker accumulation in AD.

Highlights

• Phosphorylated tau‐181 (p‐tau181) and p‐tau217 originate from tau regions that are poorly cleaved by lysosomal proteases.

• Phosphorylation further impairs the proteolytic cleavage of AD biomarker peptides.

• Impaired proteolytic cleavage of phosphorylated tau is pH dependent.

• Levels of p‐tau181 are correlated with lysosomal proteases in Alzheimer's disease (AD) cerebrospinal fluid samples.

• AD‐associated lysosomal dysfunction may contribute to presence of disease biomarkers.

1. BACKGROUND

Alzheimer's disease (AD) is characterized by the accumulation of neuritic plaques of amyloid beta (Aβ) and neurofibrillary tangles of hyperphosphorylated tau. Although tau accumulates in many neurodegenerative disorders (collectively referred to as the tauopathies), tau has only been successfully developed as an early clinical biomarker for AD. In AD, tau that is hyperphosphorylated at specific residues in its proline‐rich domain (e.g., T181, T205, and T217) can be detected in the cerebrospinal fluid (CSF) or plasma up to two decades before symptom onset and, importantly, these biomarkers differentiate AD from other tauopathies. ¹ , ² , ³ , ⁴ It is surprising that in plasma, the phosphorylated tau biomarkers also correlate strongly with amyloid burden and can also be considered an intermediate marker between amyloid and tau pathology. ⁵

Structurally, tau comprises an N‐terminal domain containing 0, 1, or 2 N‐terminal domain repeats, the proline‐rich domain (PRD), a microtubule‐binding region (MTBR) containing three or four repeats, and a C‐terminal domain (Figure 1A). ⁶ In the CSF, most of the detected tau is a protein fragment comprising the N‐terminal domain and the PRD, the latter of which contains the biomarker sites described in AD. ⁷ , ⁸ It has been hypothesized that the high percentage of proline residues in the PRD render it relatively resistant to cleavage by proteases. The restricted secondary structure of a proline makes it difficult for many proteases to cleave the amino acid sequences containing this residue. ⁹

FIGURE 1.

The proline‐rich domain of tau is resistant to cleavage by lysosomal proteases in silico and in vitro. (A) The structure of 4N2R tau is shown, with all potential phosphorylation sites denoted by black dot. The phosphorylation sites noted in red have been studied as biofluid markers in AD. The epitopes of four anti‐tau antibodies are shown. (B) A cleavage prediction map for the amino acid sequence of tau by cathepsins B, D, L, S, and K was generated using ProsperousPlus. ¹⁴ (C) Recombinant 2N4R tau was incubated with human liver lysosomes at pH 4.0 for the times indicated. Western blots with the Tau‐1 (PRD), 4R Tau (MTBR) and Tau 46 (C‐terminal) antibodies are shown. N = 3. PRD, proline‐rich domain. AD: Alzheimer's disease, MTBR: Microtubule Binding Region, PRD: Proline‐Rich Domain.

The lysosome is a highly acidic, catabolic organelle containing proteases that are responsible for the degradation of many proteins, including tau, and numerous genetic risk factors implicate lysosomal proteins in the pathogenesis of AD. ¹⁰ , ¹¹ The lysosomal proteases vary in their preferred cleavage targets and pH optima, with most proteases operating maximally at the relatively acidic pH of the lysosome. ¹² Our group previously demonstrated that the tau PRD contains relatively few lysosomal protease cleavage sites. ¹³

Because AD biomarkers originate from the PRD, we wondered if phosphorylation renders those peptides even more resistant to lysosomal proteolytic cleavage, resulting in their persistence and availability as AD biomarkers. To test this, we first performed in silico predictions followed by in vitro protease assays with peptides from the PRD domain of tau with and without phosphorylated residues. We showed that phospho‐peptides from AD biomarker sites (p‐tau181, 199, 205, and 217) are cleaved more slowly by lysosomal proteases in a pH‐dependent manner. We confirmed these findings using full‐length recombinant phosphorylated tau and showed that this effect was not recapitulated by phosphomimetic peptides. Finally, in human CSF, levels of p‐tau181 correlated with lysosomal proteases, suggesting that they are co‐released into the extracellular space.

2. METHODS

2.1. Cleavage analysis in silico

The FASTA amino acid sequence for 2N4R tau was obtained from UniProt (UniProt ID: P10636‐8). Predictions of tau cleavage in silico were generated by the ProsperousPlus web server and the cathepsins B, L, K, S, and D using the graphing function available. ¹⁴

RESEARCH IN CONTEXT

1. Systematic review: The authors reviewed the literature establishing plasma and cerebrospinal fluid (CSF) phosphorylated tau (p‐tau) fragments as early biomarkers for Alzheimer's disease (AD). Both genetic risk factors and postmortem tissue implicate lysosomal dysfunction as a contributor to disease pathogenesis; however, impaired lysosomal processing of tau has not been studied previously relative to AD biomarker accumulation.

2. Interpretation: This study demonstrates that lysosomal proteases degrade p‐tau fragments containing AD biomarkers more slowly than other tau fragments. This difference is pH dependent. Levels of p‐tau181 in the CSF of patients with AD correlate with that of lysosomal proteases.

3. Future directions: These findings suggest a model in which disease‐associated misprocessing of cleavage‐resistant portions of tau lead to the co‐release of p‐tau biomarkers and lysosomal contents. The next steps include: (1) determining the cellular source and mechanisms of p‐tau release and (2) utilizing the principles revealed here to identify relevant biomarkers for other neurodegenerative processes.

2.2. Antibodies

The following antibodies were used for western blotting: Mouse anti‐HT7 (Thermo Scientific, MN1000, 1:1000), mouse Tau 46 (Santa Cruz, sc‐32274, 1:1000), mouse Tau‐1 (Millipore MAB342, 1:1000), rabbit 4R‐tau (1:1000, Cosmo bio CAC‐TIP‐4RT‐P01), Licor Donkey anti mouse green (Licor 926‐68072, 1:5000), and Licor Donkey anti rabbit red (Licor 926‐68073, 1:5000). 

2.3. Recombinant proteins

Full‐length recombinant tau (Sigma T0701) and full‐length tau phosphorylated by GSK3‐beta (Sigma SRP0689) were received as fluid or reconstituted per manufacturer's instructions. Cathepsin B (EMD Millipore 9047‐22‐7), Cathepsin D (R&D Systems 1014‐AS‐010), Cathepsin F (Abcam 157039), Cathepsin G (MilliporeSigma 21‐937‐3100MI), Cathepsin K (MilliporeSigma 21‐946‐125UG), Cathepsin L (MilliporeSigma 21940225UG), Cathepsin O (Abcam ab158237), Cathepsin S (R&D Systems 1183‐CY‐010), Cathepsin V (R&D Systems 1080‐CY‐010), Cathepsin X/Z/P (R&D Systems 934‐CY‐010), and Legumain/AEP (R&D Systems 2199‐CY‐010) were reconstituted or diluted to 4 uM or 1 uM per the manufacturer's instructions. Confirmation of enzyme activity was done previously. ¹³

2.4. Protease cleavage assay in vitro

Full‐length recombinant tau and tau phosphorylated by GSK3‐beta were incubated for 30 min at 30°C with or without lambda phosphatase in lambda phosphatase buffer, and then quenched with phosphatase inhibitors. They were then combined with human liver lysosomes (Xenotech X008036) at a ratio of 250 ng lysosomes to 75 ng tau in protease cleavage buffer (50 mM sodium acetate, 2 mM dithiothreitol (DTT), 1 mM ethylenediaminetetraacetic acid (EDTA), pH with HCl to pH 4.0, 4.5, 5.0, 5.5, or 6.0 as indicated) at 37°C in a water bath. Alternatively, 75 ng tau was mixed with 400 nM cathepsin B or 40 nM cathepsin L. At the indicated time points, 15 uL was removed from the reaction and quenched with Invitrogen NuPAGE LDS Buffer (4X) and Invitrogen NuPAGE Reducing Buffer (10X). The samples were boiled at 80°C for 10 min. The samples were resolved on a 4%–12% Bis‐Tris gel and transferred to nitrocellulose membrane using the Bio‐Rad Mini‐PROTEAN and Mini Trans‐Blot systems, respectively. Blots were blocked in Licor TBS‐T blocking buffer at room temperature for 1 h, and then incubated in primary antibody in blocking buffer overnight at 4°C. Blots were rinsed 3× in TBS‐T, incubated with secondary antibodies at room temp for 1‐h, rinsed 3× in TBS‐T, and then imaged on the Licor Odyssey CLx. Time courses were run as three independent replicates.

2.5. FRET peptide analysis

The FRET peptides were synthesized through GenScript. Sequences are listed in Figure S1. In addition, all peptides had two N‐terminal d‐Arg residues to promote solubility. All peptides had >95% purity, and trifluoroacetic acids (TFAs) were removed. All peptides were expected to have solubility in DMSO and thus were dissolved in DMSO to a concentration of 1 mM. The FRET peptides were mixed with protease cleavage buffer (above) to a concentration of 20 uM in 49 uL. One microliter of either human liver lysosomes (250 ng) or cathepsins (final concentration 20 nM) was added to the sample on ice. Samples were run in triplicate in a 384‐well clear bottom plate on a VANTAstar microplate reader at 37°C for 24 h. The samples were run on the FRET setting with a reading every 3 min, with excitation 320 nm and emission 405 nm for the MCA/Dnp peptides and excitation 320 nm and emission 420 nm for Abz/Dnp peptides. Using the MARS data analysis software (BMG Labtech), the data were normalized to 0%–100% of the maximum and minimum values. A maximum slope of the rising phase was calculated. GraphPad Prism was used for statistical analysis and graphing. Statistical significance was calculated by two‐way analysis of variance (ANOVA) followed by Tukey post hoc multiple comparisons. Significance is denoted by * (p < 0.05), ** (p < 0.01), *** (p < 0.001), **** (p < 0.0001), and n.s. (not significant).

2.6. Human CSF proteomics

CSF levels of lysosomal proteases were analyzed from two independent human cohorts with large‐scale proteomic data: (1) 160 patients with symptomatic, biomarker‐confirmed AD and 140 biomarker‐negative controls from the Emory ADRC and (2) 51 cognitively unimpaired older adults with and without asymptomatic elevations in AD biomarkers from the UCSF Branch cohort. ¹⁵ , ¹⁶ CSF cathepsins and legumain data were collected on the aptamer‐based SomaScan proteomics platform (v4.1, >7k proteins) in both cohorts and also via TMT mass spectrometry in the Emory cohort. ¹⁶ Details of the CSF proteomic data processing pipeline are described elsewhere. ¹⁶ In each cohort, CSF protease levels were log₂‐transformed and correlated against orthogonally‐measured CSF biomarkers of interest. In the Emory cohort, CSF p‐tau181 and Aβ1‐42 were measured using the INNO‐BIA AlzBio3 Luminex Assay. ¹⁷ In BrANCH, CSF p‐tau181 was measured using the Lumipulse platform, and neurofilament light (NfL) was measured using an in‐house enzyme‐linked immunosorbent assay (ELISA). ¹⁵ , ¹⁸ , ¹⁹ Analyses were performed using Spearman's rho correlation coefficients, and results were visualized using heatmaps and scatterplots.

2.7. Figures and statistical analysis

Figures were created in Biorender or Adobe Illustrator. FRET peptide cleavage statistical analysis was performed using two‐way ANOVA on GraphPad Prism followed by Tukey post‐hoc multiple comparisons. Significance is denoted by * (p < 0.05), ** (p < 0.01), *** (p < 0.001), **** (p < 0.0001), and n.s. (not significant).

3. RESULTS

3.1. Analysis of tau lysosomal cleavage in silico and confirmation by in vitro testing

Tau contains over 80 potential phosphorylation sites (Figure 1A). ²⁰ The biological basis for why phosphorylated residues in the PRD of tau accumulate in AD is currently unknown. In our earlier mapping of tau lysosomal protease cleavage sites, we noted that relatively fewer cathepsins could proteolyze tau in the PRD. ¹³ Similarly, when we utilized ProsperousPlus ¹⁴ to perform in silico predictions of 2N4R tau cleavage by cathepsins B, D, L, S, and K, few cleavage sites were identified between residue ≈150 and 220, which is contained within the PRD (Figure 1B). When tau is subjected to an in vitro protease cleavage assay with human liver lysosomes, two protease‐resistant fragments between 15 and 20 kDa appeared. Because the fragments were recognized only by antibodies directed against epitopes in the PRD (Tau‐1, HT7), they likely reflect the protease‐resistant properties of the PRD (Figure 1C).

3.2. Effect of phosphorylation on tau cleavage by lysosomal proteases

The PRD of tau contains the phosphorylated residues (threonines 181 and 217) that are utilized as AD biomarkers. ²¹ Thus we wondered if phosphorylation impacted tau cleavage in a way that would extend the half‐life of these phospho‐peptides. We obtained tau that had been phosphorylated in vitro by the kinase GSK3β (which has been shown to phosphorylate most tau biomarker sites and has been implicated in AD disease pathogenesis ²² ) and then compared its proteolysis by human liver lysosomes to unmodified tau.²⁴ Phosphorylation of tau appeared to both slow the overall proteolysis of tau and enhance the levels of the small PRD‐containing peptides (Figure 2A). To ensure that this was not an effect of GSK3β treatment, we performed a control experiment in which the phosphorylated tau was dephosphorylated by lambda phosphatase prior to cleavage, which restored cleavage of the PRD of tau to normal levels (Figure S2).

FIGURE 2.

Phosphorylation of tau impairs cleavage by lysosomal proteases. (A) Recombinant tau or phospho‐tau (p‐tau) was incubated with human liver lysosomes at pH 4.0 for the times indicated and then subjected to western blotting with HT7. Representative of three replicates. (B) Design of tau FRET peptides using the nine amino acid sequence for Thr181 as an example. The phosphorylated residue is indicated in bold. The N‐terminus contains a fluorescent reporter (MCA or Abz), which is quenched by C‐terminal Dnp until cleavage occurs. (C) In vitro protease assays showing cleavage of FRET peptides pairs (Thr181, Ser199, Thr205, Thr217) by human liver lysosomes across pHs ranging from 4.0–6.0. Maximum slopes were derived from traces normalized to 0%–100% of minimum–maximum fluorescence, representative traces of the pH extremes are showed in insets. If a trace did not reach asymptote by the end of the time course, a maximum slope was not calculated. N = 3. Dnp: 2,4‐dinitrophenol, FRET: Fluorescence Resonance Energy Transfer, MCA: 7‐Methoxycoumarin‐4‐acetic acid.

3.3. Phosphorylation has minimal impact at the highly acidic pH of healthy lysosomes

Lysosomal proteases recognize linear amino acid sequences that, by convention, can be labeled P4‐P4' from amino to carboxy ends. For the biomarker phosphorylation sites studied, ProsperousPlus predicted the most likely site of cleavage was when the phosphorylated residue was at either P1' or P2', depending on the protease. To further explore this phenomenon, we designed FRET peptides to evaluate the effect of a single phosphorylation site on cleavage by lysosomal proteases. We designed four FRET peptides to investigate four biomarker‐related phosphorylation sites in tau (Thr181, Ser199, Thr205, Thr217). For each pair, the peptides were identical except that the biomarker site in question was either phosphorylated or unphosphorylated. To ensure capture of both P1' and P2' cleavage events, we designed FRET peptides containing the five amino acids N‐terminal to the phosphorylated residue, the phosphorylated or unphosphorylated residue itself, and the three amino acids C‐terminal to the residue of interest (the FRET peptide sequences are available in Figure S1). The FRET peptides contained fluorescence/quencher pairs, such that fluorescence occurred only when the peptide was cleaved by a protease (Figure 2B).

We tested cleavage of these FRET pairs by human liver lysosomes at five pH levels (pH 4.0, 4.5, 5.0, 5.5, and 6.0) in conditions by which the substrate to protease ratio was ≈1000:1. We found a pH‐dependent effect of cleavage in three FRET pairs. Specifically, at lower pH values, the maximum slope of cleavage (i.e., Vmax) was similar between phosphorylated and unphosphorylated peptides, whereas at higher pH values cleavage of the phosphorylated FRET peptide was significantly slower, even when accounting for lower protease activity at higher pH (Figure 2C–F). The exception was T217, which had impaired cleavage of the phosphorylated FRET peptide at all pH levels tested.

3.4. Phosphomimetic peptides do not mimic effect of phosphorylation on tau cleavage by lysosomal proteases

Phosphorylation is the reversible addition of a phosphate (PO₄ ⁻) group to a serine, threonine, or tyrosine amino acid residue; it regulates protein interactions and signaling cascades. ²³ , ²⁴ It was demonstrated recently that the phosphorylation of serine and threonine residues introduces a hydrogen bond between the charged phosphate and amide side chains, resulting in a proline‐like “pseudocyclization.” ²⁵ The pseudocyclization depends on the ambient pH, as at a more acidic pH, the second charge on the phosphate group is lost, breaking the hydrogen bond. ²⁵

Thus we asked if phosphorylation impedes cleavage through conferring negative charge or via the structural change induced by pseudocyclization. Glutamate residues are often used to mimic phosphorylation, since they similarly confer a negative charge. However, a phosphomimetic glutamate does not induce the pH‐dependent pseudocyclization that a bona fide phospho‐serine or phospho‐threonine generates. ²⁵ We generated phosphomimetic FRET peptides containing a glutamate compared to a phosphorylated threonine. Unlike the phosphorylated FRET peptides, the phosphomimetic peptides showed a different pattern of cleavage that was not pH dependent (Figure 3A–D), suggesting that in this case, the phosphomimetic glutamate does not mimic the same effects of phosphorylation.

FIGURE 3.

Phosphomimetic residues do not recapitulate pH‐dependent impairment of phosphorylated tau cleavage by human liver lysosomes. FRET peptides were designed in which the relevant tau amino acid (Thr181, Ser199, Thr205, Thr217) is either nonphosphorylated, phosphorylated, or replaced with a glutamic acid. These peptides were incubated with human liver lysosomes across a pH range (4.0–6.0). Maximum slopes were derived from traces normalized to 0%–100% of minimum–maximum fluorescence. If a trace did not reach asymptote by end of time course, a maximum slope was not calculated. N = 3. FRET: Fluorescence Resonance Energy Transfer.

3.5. Tau phosphorylation impairs the proteolytic capacity of specific individual lysosomal proteases

Recognizing that each lysosomal protease has preferred cleavage substrates and pH optima, we then assessed how phosphorylation affects the ability of individual lysosomal cathepsins to cleave the AD biomarker FRET peptides. Again using a 1:1000 protease to substrate molar ratio, we studied the ability of cathepsins B, D, F, G, K, L, O, S, V, and X/P/Z, and legumain (AEP) to cleave the four phosphorylated and unphosphorylated FRET peptide pairs at five pH levels. We observed a striking difference between cathepsins, with some cathepsins (e.g., cathepsin B and cathepsin L) dramatically impaired by client phosphorylation (Figure 4A‐D). Indeed, in the case of cathepsin L, a V_max could not be calculated for many of the phosphorylated FRET peptides. Conversely, some cathepsins (e.g., cathepsin D) could not cleave any the FRET peptides. Finally, some cathepsins showed cleavage that was not impacted by phosphorylation (e.g., cathepsin K) or a variation between FRET peptides. These results have been summarized in a table (Figure 4E). All pH curves have been included in the supplemental data for the Thr181 (Figure S3), Ser199 (Figure S4), Thr205 (Figure S5), and Thr217 (Figure S6) FRET peptides. As a confirmatory test, we then took the two most impacted cathepsins, cathepsin B and cathepsin L, and performed a time course of cleavage of phosphorylated versus nonphosphorylated recombinant tau at pH 4.0, 5.0, and 6.0. For both cathepsins, the phosphorylated PRD took longer to cleave, particularly at higher pH values (Figure 4F).

FIGURE 4.

Impact of phosphorylation on cleavage of tau FRET peptides and tau recombinant varies by individual lysosomal protease. (A–D) Cleavage of FRET peptides pairs (Thr181, Ser199, Thr205, Thr217) across a pH range (4.0–6.0) by individual lysosomal proteases. Maximum slopes were derived from traces normalized to 0%–100% of minimum–maximum fluorescence; representative traces of the pH extremes are shown in insets. If a trace did not reach asymptote by end of time course, a maximum slope was not calculated. N = 3. (E) Summary of cleavage properties of individual lysosomal proteases and FRET peptide pairs. (F) Western blot showing time course of phosphorylated (p‐tau) and nonphosphorylated tau (tau) by cathepsin B (top) or cathepsin L (bottom) at pH 4.0 or pH 6.0. Representative of three replicates. AEP: Asparagine Endopeptidase, CTSB: cathepsin B, CTSD: cathepsin D, CTSF: cathepsin F, CTSG: cathepsin G, CTSK: cathepsin K, CTSL: cathepsin L, CTSO: cathepsin O, CTSS: cathespin S, CTSV: cathespin V, CTSZ: cathespin X/Z/P, FRET: Fluorescence Resonance Energy Transfer.

3.6. p‐tau181 levels from CSF of patients with AD correlate with lysosomal proteases

Lysosomes can fuse with the plasma membrane, releasing the entirety of their contents in a process known as lysosomal exocytosis. We wondered if lysosomal‐protease resistant p‐tau enters the extracellular space through lysosomal exocytosis, in which case other lysosomal proteins would accompany it. We examined human CSF proteomic datasets to determine whether biofluid‐detectable concentrations of lysosomal proteins correlated with tau phosphorylation in vivo. We queried CSF proteomic data from a recently published study of 160 patients with symptomatic, biomarker‐confirmed AD and 140 biomarker‐negative controls from the Emory ADRC (ages 45–90). ¹⁶ CSF levels of lysosomal proteases (from SomaScan and TMT mass spectrometry [MS] approaches) were compared with immunoassay‐based CSF pTau181 and Aβ_1‐42 for each patient. ¹⁶ , ¹⁷ , ¹⁹ Levels of 7 of 11 proteases from the SomaScan dataset and 7 of 9 from the TMT‐MS dataset were significantly correlated with p‐tau181 (p < 0.05) (Figure 5A). Notably, the four cathepsins that exhibited reduced cleavage of the tau Thr181 FRET peptide (CTSB, CTSL, CTSS, and CTSZ) and other cysteine proteases all exhibited positive correlations with CSF p‐tau181 (Spearman's rho range: 0.14 to 0.28, p range: 0.015 to < 0.0001), with strong concordance in effects across platforms. Furthermore, the majority of these proteases were not significantly correlated with CSF Aβ1‐42 (rho range: −0.09 to 0.10, p range: 0.088 to 0.984), with the exception of one SomaScan target for CTSB (rho = 0.16, p = 0.007) and one for CTSZ (rho = 0.12, p = 0.044), supporting their specificity to tau phosphorylation (Figure S7). These findings may suggest that certain lysosomal proteases are exported into the extracellular space with p‐tau181. In contrast, one SomaScan target for legumain (protein symbol: LGMN) exhibited a negative correlation with p‐tau181 that reached significance (rho = −0.23, p < 0.001), whereas two additional LGMN targets and CTSD exhibited negative correlations with p‐tau181 that approached significance (rho range: −0.10 to −0.11, p range: 0.090 to 0.053). These divergent correlations could reflect the lysosomal composition associated with p‐tau release or underlying lysosomal changes leading to disease.

FIGURE 5.

Correlation of CSF lysosomal cathepsin levels with p‐tau181 across cohorts. (A) Correlation between p‐tau181 and individual lysosomal protease levels, grouped by protease family, in CSF in patients with symptomatic AD (Emory ADRC) or healthy aging adults (BrANCH). The levels were measured by SomaScan (SOMA) and TMT:MS (TMT) for most lysosomal proteases. (B) Sample correlations (CTSB and CTSL) from the Emory ADRC dataset. Alzheimer's disease patient samples are in red; control patients are in blue. (C) Sample correlation (CTSB and CTSD) from BrANCH data set. (Emory ADRC n = 160 AD patients, n = 140 healthy controls. UCSF BrANCH. n = 51.). ADRC: Alzheimer's Disease Resource Center, BrANCH: Brain Aging Network for Cognitive Health, CSF: Cerebrospinal fluid, CTSB: cathepsin B, CTSD: cathepsin D, CTSF: cathepsin F, CTSG: cathepsin G, CTSK: cathepsin K, CTSL: cathepsin L, CTSO: cathepsin O, CTSS: cathepsin S, CTSV: cathepsin V, CTSZ: cathepsin X/Z/P, LGMN: legumain (AEP), TMT:MS, Tandem Mass Tag Mass Spectrometry.

To validate our findings in an asymptomatic cohort, we leveraged CSF SomaScan data from 51 cognitively unimpaired older adults from the UCSF Brain Aging Network for Cognitive Health (BrANCH), which included individuals with and without asymptomatic elevations in AD biomarkers. ¹⁵ In BrANCH, CTSB, CTSS, and CTSZ (CTSL was not measured) again exhibited significant positive correlations with p‐tau181 (rho range: 0.24 to 0.38, p range: 0.094 to 0.007), but not CSF NfL (rho range: −0.28 to 0.11, p range: 0.116 to 0.591), supporting their sensitivity and specificity to early tau phosphorylation (Figure 5A,C, Figure S7). CTSD, CTSG (not available in Emory), and CTSV all exhibited strong negative correlations to p‐tau181 (rho range: −0.65 to −0.38, p range: < 0.0001 to 0.006). Overall, these findings suggest that certain lysosomal proteases, particularly in the cysteine protease family, are exported to the CSF with p‐tau181. Neurofilament light (NfL), as a measure of neurodegeneration, is less correlated with lysosomal proteases, reflecting its cytosolic cellular source. ²⁶

4. DISCUSSION

In this study, we sought to determine whether phosphorylation at specific residues of tau affects its cleavage by lysosomal proteases, thereby promoting the persistence of tau phospho‐peptides, which could serve as a rationale for why they are sensitive and specific disease biomarkers. Through in silico and in vitro testing, we showed that phosphorylation at currently used AD biomarker residues (T181, S199, T205, and T217) in the PRD converts this already protease‐cleavage poor domain into a protease‐resistant region. In our data, the phosphorylation‐induced resistance is not seen at the highly acidic pH typical of healthy lysosomes. As lysosomal pH in model organisms increases with age, ²⁷ , ²⁸ our findings suggest that both aging and disease may contribute to the production of AD biomarker tau peptide. The AD biomarker phospho‐tau peptides must be released into the extracellular space in order to be detected. If this occurs through lysosomal exocytosis, then lysosomal cathepsins would accompany them. This is exactly what we found, with multiple relevant lysosomal proteases in human CSF correlating with p‐tau181 levels from two separate AD cohorts.

One of the open questions in AD biomarker development has been why plasma p‐tau biomarkers correlate more strongly with amyloid pathology than tau pathology. Historically, p‐tau accumulation was thought to be due to overproduction and hyperphosphorylation of tau; however, our data suggest that biomarker accumulation may reflect impaired p‐tau clearance. In this model, pseudocyclization of p‐tau does not occur at more acidic pHs. Thus, an acidified lysosome could clear p‐tau, whereas more alkaline lysosomes (as may occur with aging) cannot (Figure 6). In a disease state where the lysosome is slightly more alkaline, p‐tau cannot be degraded easily and preferentially accumulates in the lysosomes, ultimately getting released from the cell via exocytosis. Here, p‐tau may be a waste product of tau degradation and not necessarily a pathogenic agent by itself. Real world evidence to support this concept comes from clinical trials of humanized antibodies against the N‐terminal and PRD epitopes, which were unsuccessful in slowing AD progression. ²⁹ This does not necessarily preclude a pathogenic role for the MTBR and C‐terminal of tau. MTBR biomarkers are in development that correlate much more strongly with neurofibrillary tangles and tau PET positivity. The MTBR of tau forms the core of tau fibrils; thus its accumulation may correlate more directly with neurofibrillary tangles, and humanized antibodies directed against this domain are currently in clinical development. ⁶ , ³⁰ , ³¹

FIGURE 6.

Model of p‐tau accumulation from lysosomal dysfunction. Left panel shows that phosphorylated serine or threonine residues adopt a pseudocyclical structure due to hydrogen bonding between the amide group and phosphate group, which is structurally similar to a proline. Middle panel shows that in an acidified lysosome (top), the psuedocyclization is interrupted, allowing recognition of the amino acid sequence by the protease active site and subsequent protein degradation. In a de‐acidified lysosome (bottom), the pseudocyclization impairs binding to the protease active site, reducing protein cleavage. The right panel shows the cellular impact of de‐acidification, whereby a healthy cell can process both tau and p‐tau, whereas an unhealthy cell with de‐acidified lysosomes cannot process the PRD of p‐tau, resulting in its release from the cell as a biomarker fragment. The fragment persists in the plasma and CSF due to the neutral pH of these compartments. CSF: cerebrospinal fluid, PRD: proline‐rich domain.

Disruption of the lysosome is implicated in AD both through genetic risk factors and postmortem pathology. ³² , ³³ In vitro and, recently, in vivo studies have found that mutations in PS1 and APP both cause relative alkalinization of the lysosome. ³⁴ , ³⁵ Although this overall is expected to make lysosomal proteases function less optimally, our findings indicate that lysosomal alkalinization would disproportionately impact cleavage of p‐tau. Thus the increased levels of p‐tau could either represent a lysosomal dysfunction induced by Aβ or lysosomal dysfunction could cause the accumulation of both Aβ and p‐tau simultaneously.

One limitation to this study is that we are not able to provide in vivo or cell‐based evidence to support our findings. Typically, the effect of phosphorylation in vivo is studied by replacing residues with phosphomimetic amino acids, usually glutamate to mimic the charge of phosphorylation. However, pH‐dependent pseudocyclization is not observed in phosphomimetic residues. In addition, although we suspect that the structural effect of phosphorylation on protease cleavage is generalizable to other proteins, we focused on tau in this study due to the potential impacts on AD biomarkers and disease pathogenesis. Although outside the scope of this study, the effects of phosphorylation on the activity of other non‐lysosomal proteases, including extracellular proteases, should also be interrogated. It is important to note that if phosphorylation impacts cleavage by other proteases, it may have an even stronger effect as most proteases operate at neutral pH. Thus, once p‐tau peptides reach the CSF or plasma, it is further protected from cleavage due the neutral pH of these compartments, preserving it as a biomarker. Finally, these findings do not rule out overproduction of tau contributing to p‐tau accumulation.

To our knowledge, this study is the first to show that phosphorylation impacts the cleavage of a protein by proteases. We propose that the mechanism is due to pseudocyclization of the secondary structure of the phosphorylated threonine rather than the negative charges conferred by phosphorylation. Of interest, phosphorylation‐induced pseudocyclization is most prominent when the phosphorylated residue directly precedes a proline. ²⁵ Consistent with this, the phospho‐threonines that represent the best AD biomarkers each precede a proline. Although not studied herein, we anticipate that there is direct applicability of our findings to other phosphorylated biomarkers of disease. Within neurodegenerative diseases, both TDP‐43 (amyotrophic lateral sclerosis and frontotemporal dementia) and alpha‐synuclein (Parkinson's disease, Lewy body disease) feature the increase of phosphorylated biomarkers in domains predicted to have less cleavage by lysosomal proteases. ³⁶ , ³⁷ In contrast to tau biomarkers, which feature predominantly phosphorylated threonine residues, candidate TAR DNA‐binding protein 43 (TDP‐43) and alpha‐synuclein biomarkers feature serine residues. Both phosphorylated residues will undergo pseudocyclization; however, serines have more structural flexibility and will do so in a graded fashion, whereas threonines exhibit a step‐like structural change. ²⁵ It is unclear how these structural differences will impact pH‐dependent proteolytic cleavage. As lysosomal dysfunction is implicated in these diseases, further research in this area is indicated. Thus, we propose a template by which new phosphorylated biomarkers could be identified in regions of proteins that are poorly cleaved, either due to high levels of prolines or other poorly cleaved residues, with the highest potential biomarker amino acid being that immediately preceding a proline.

In conclusion, we describe a novel mechanism by which phosphorylation of tau reduces cleavage of tau by lysosomal proteases, particularly in cases of slightly elevated lysosomal pH, which can lead to persistence of the phospho‐peptides, allowing them to serve as disease biomarkers. Overall, this suggests a novel framework in which lysosomal dysfunction can lead to preferential accumulation of p‐tau in AD. Future work should be directed at understanding the underlying lysosomal dysfunction that contributes to disease pathogenesis.

CONFLICT OF INTEREST STATEMENT

A. W. K. serves on the Scientific Advisory Boards for Nine Square Therapeutics and Junevity, Inc., and has received research support from Eli Lilly & Co. and Ono Pharmaceuticals. B.L.M. serves on the Scientific Advisory Board of the Bluefield Project to Cure Frontotemporal Dementia (FTD), the John Douglas French Alzheimer's Foundation, Fundación Centro de Investigación Enfermedades Neurológicas, Madrid, Spain, Genworth, Inc., the Kissick Family Foundation; the Larry L. Hillblom Foundation, and the Tau Consortium of the Rainwater Charitable Foundation; serves as a scientific advisor for the Arizona Alzheimer's Consortium, Massachusetts General Hospital Alzheimer's Disease Research Center (ADRC), and the Stanford University ADRC; receives royalties from Cambridge University Press, Elsevier, Inc., Guilford Publications, Inc., Johns Hopkins Press, Oxford University Press, and the Taylor & Francis Group; serves as editor for Neurocase and section editor for Frontiers in Neurology. The other authors declare no competing interests. Author disclosures are available in the Supporting Information.

CONSENT STATEMENT

Cerebrospinal fluid samples were collected with the informed consent of the patients or through their authorized representatives.

Supporting information

Supporting Information

Supporting Information

Lane‐Donovan C, Smith AW, Saloner R, Miller BL, Casaletto KB, Kao AW. Tau phosphorylation at Alzheimer's disease biomarker sites impairs its cleavage by lysosomal proteases. Alzheimer's Dement. 2025;21:e70320. 10.1002/alz.70320