Localization, induction, and cellular effects of tau phosphorylated at threonine 217

Abstract

Introduction:

Tau phosphorylation at T217 is a promising AD biomarker, but its functional consequences were unknown.

Methods:

Human brain and cultured mouse neurons were analyzed by immunoblotting and immunofluorescence for total tau, tau_pT217, tau_pT181, tau_pT231 and tau_pS396/pS404. dSTORM super resolution microscopy was used to localize tau_pT217 in cultured neurons. EGFP-tau was expressed in fibroblasts as wild type and T217E pseudo-phosphorylated tau, and fluorescence recovery after photobleaching (FRAP) reported tau turnover rates on microtubules.

Results:

In brain, tau_pT217 appears in neurons at Braak stages I-II, becomes more prevalent later and co-localizes partially with other phospho-tau epitopes. In cultured neurons, tau_pT217, is increased by extracellular tau oligomers (xcTauOs), and is associated with developing post-synaptic sites. FRAP recovery was fastest for EGFP-tau_T217E.

Conclusion:

Tau_pT217 increases in brain as AD progresses and is induced by xcTauOs. Post-synaptic tau_pT217 suggests a role for T217 phosphorylation in synapse impairment. T217 phosphorylation reduces tau’s affinity for microtubules.

1 ∣. INTRODUCTION

The realization that Alzheimer’s disease (AD) comprises a pre-symptomatic phase that might last for more than two decades and that massive, irreversible brain damage has accumulated by the time symptoms are evident has forced a re-evaluation of how to approach the disease clinically. It is now widely accepted that accurate early diagnosis combined with effective interventions that might include drugs and lifestyle adjustments are key to delaying the onset and slowing the progression of AD symptoms, if not preventing symptoms from ever arising. It follows naturally that achieving the tightly linked goals of improved early diagnostics and therapeutics will be aided by a deeper understanding of the molecular mechanisms of AD pathogenesis. While the effort to develop effective therapeutics has not yet yielded any game changing breakthroughs, progress on the diagnostic front has been impressive during the past several years. PET imaging for plaques and tangles is now well established^1–3, and cerebrospinal fluid (CSF) biomarkers have emerged as an alternative approach^4–5. PET imaging is sporadically available and costly, though, and obtaining a CSF sample for biomarker analysis requires an invasive procedure in the form of a spinal tap.

Thanks to recent advances in assay sensitivity, it has become possible to substitute blood plasma for CSF to quantify several promising biomarkers for identification of patients at risk of developing AD symptoms years in advance of expected symptom onset. Prime examples of promising plasma biomarkers are phosphorylated forms of the neuron-specific, axon-enriched, microtubule-associated protein, tau, particularly tau_pT181, tau_pT217 and tau_pT231^6–9. Though several studies have focused on the localization and functional properties of tau_pT181^10–11 and tau_pT231^12–13, little is known about tau_pT217 other than its promise as an early fluid-based biomarker for AD, and two reports about its histological distribution in AD brain^14–15.

As a first step toward understanding functional consequences of tau phosphorylation at T217 we sought to define its distribution in vivo by surveying brains of human AD patients at various disease stages and in age-matched, cognitively normal controls, in AD model mouse brains, and in cultured mouse neurons. We also used cultured neurons to determine if extracellular oligomers of amyloid-β (xcAβOs) or tau (xcTauOs) affect the abundance or localization of tau_pT217. Finally, to determine if tau phosphorylation at T217 affects its affinity for microtubules, we expressed EGFP-tagged wild type (WT) and pseudo-phosphorylated tau in fibroblasts, and used fluorescence recovery after photobleaching (FRAP) to measure turnover rates of microtubule-bound tau.

We determined that: 1) tau_pT217 appears in human brain neurons at Braak stages I-II, and becomes more prevalent at later disease stages in axons, dendrites, neuronal cell bodies, neuropil threads, pre-tangles and tangles that are extensively co-labeled by the AT8 (tau_pS202/pT205) and PHF1 (tau_pS396/pS404) antibodies; 2) in cultured neurons, tau_pT217 is increased by xcTauOs, but not by xcAβOs, and is associated with developing post-synaptic sites; and 3) pseudo-phosphorylation by a T217E amino acid substitution reduces tau’s affinity for microtubules. These results raise the possibility that xcTauOs drive T217 phosphorylation of tau in vivo, and that tau_pT217 contributes to synaptic decline and favors formation of tau-tau interactions at the expense of tau binding to microtubules.

2 ∣. METHODS

2.1 ∣. Brain tissue sections

Paraffin-embedded, 5–6 μm thick human cortical brain autopsy sections were obtained from the archives of the Department of Pathology of the University of Virginia, and pertinent clinical information about each donor is shown in Supplementary Table 1. Institutional approval for use of archival autopsy tissue was obtained from the University of Virginia Biorepository and Tissue Research Facility. Classification of disease stages for each donor was performed by ABMS-certified Neuropathologists, James W. Mandell, M.D., Ph.D., and Maria-Beatriz Lopes, M.D., Ph.D., using the “B” (neurofibrillary tangle pathology) part of the ABC staging criteria for AD, in which a score of 0 indicates no detectable tangles, and scores of 1, 2 and 3 respectively correspond to Braak tangle stages I-II, III-IV and V-VI¹⁶.

CVN (APP_SwDI/NOS2^−/−) mice, which overexpress human APP with three amyloidogenic mutations, lack a nitric oxide 2 synthase gene, and accumulate plaques and tangles in an age-dependent manner¹⁷ were originally obtained from Drs. Michael Vitek and Carol Colton of Duke University, and were maintained as a breeding colony. Paraffin-embedded, 5–6 μm thick sagittal sections of brain tissue were cut following trans-cardiac perfusion of 4% paraformaldehyde in PBS of mice that had been deeply anesthetized intraperitoneally with ketamine/xylazine (280/80 mg/kg). Animals were maintained, bred and euthanized in compliance with all policies of the Animal Care and Use Committee of the University of Virginia.

2.2 ∣. Cultured neurons

As described in our earlier work, primary mouse cortical neuron cultures were prepared from dissected brain cortices of E16–18 WT C57/Bl6 mice and tau knockout (TKO) mice in the same background strain¹⁸, and were maintained in Neurobasal medium supplemented with B27. For immunofluorescence microscopy, the neurons were grown on 1.5 thickness glass coverslips. All experiments were completed after neurons had been in culture for 10–14 days. Animals were maintained, bred and euthanized in compliance with all policies of the Animal Care and Use Committee of the University of Virginia.

2.3 ∣. CV-1 cells

CV-1 African green monkey kidney fibroblasts were cultured in Dulbecco’s modified Eagle’s medium supplemented with fetal bovine serum to 10% by volume and 50 μg/ml gentamycin.

2.4 ∣. Immunofluorescence microscopy

Human and mouse brain sections were processed identically. First paraffin was removed with a graded series of xylenes. Next, antigen retrieval was performed by immersing brain slides in a beaker of citrate buffer, and then heating in a microwave at high power for 2.5 minutes followed by low power for 8 minutes. Then, the beaker was slowly cooled to room temperature for 20 minutes, after which it was placed in an ice slurry for 20 minutes. Brain sections were then washed with PBS for 5 minutes and incubated with blocking solution (PBS with 5% bovine serum albumin and 0.1% Triton X-100) for 1 hour, followed by sequential incubations for 2 hours each with primary and secondary antibodies. Multiple washes with PBS followed each antibody step, and immediately prior to being sealed with a coverslip using Fluoromount G (ThermoFisher catalog # 00-4958-02) and Hoechst 33342 (ThermoFisher catalog # 62249), sections were washed for 10 minutes with Autofluorescence Inhibitor Reagent (Millipore catalog # 2160).

An analogous procedure was used for immunofluorescence labeling of cultured neurons following fixation and permeabilization for 5 minutes with methanol at −20° C. In this case, though, primary and secondary antibody incubations were for 30–60 minutes each, and the Autofluorescence Inhibitor Reagent step was omitted.

Dephosphorylation of brain sections and neurons was carried out prior to application of primary antibodies by treating with bovine intestinal mucosa alkaline phosphatase (43 μg/ml, Sigma-Aldrich, Catalog # P7640) overnight at 4° C.

See Supplementary Table 2 for descriptions of all antibodies used for immunofluorescence and western blotting.

An inverted Nikon Eclipse Ti microscope with planapochromatic 10X and 20X dry, and 40X, 60X and 100X oil immersion objectives, a Yokagawa X1 spinning disk confocal head, 4 diode lasers (405, 488, 561 and 640 nm), and a Hamamatsu Flash 4.0 CMOS camera was used for confocal imaging. Quantification of fluorescence micrographs was performed using the Fiji software derivative of ImageJ, including the Coloc2 plugin for calculating Pearson correlation coefficients.

A dSTORM imaging system (Abbelight; Paris, France) connected to an inverted microscope (DMi8; Leica; Wetzlar, Germany) was used for super resolution microscopy. The microscope was configured with a 63X, NA 1.47 HC planapochromatic oil immersion Leica objective and an EM-CCD camera (Sona 4.2B-6; Andor Technology). Fixed and stained cells were embedded in STORM mounting buffer (Abbelight). Fluorophores were excited using 560 nm (200 mW), and 640 nm (500 mW) lasers. Optimal images were obtained through a combination of corrections and background removal using NEO analysis software (Abbelight).

Pearson correlation coefficients were used for analyzing colocalization and student t tests were used for comparing sample sets. For cultured neurons imaged by dSTORM (Figure 4), a minimum of 50 cells per condition from a total of at least 3 biological repeats were used. Graphs were generated using GraphPad Prism 9.0 software. Frequency distribution graphs were generated by first defining each contiguous set of tau_pT217-positive pixels and PSD95-positive pixels as a single object. Then, for each tau_pT217-positive object we plotted the shortest distance to a PSD95-positive object, and the shortest distance from each PSD95-positive object to a tau_pT217-positive object.

FIGURE 4.

Tau_pT217 is associated with developing dendritic spines marked by PSD95 in cultured mouse cortical neurons. (A) Single plane confocal images taken using a 60X objective. (B) Single plane confocal (100X objective) and dSTORM super-resolution (63X objective) images. (C) Pearson correlation coefficients for quantifying co-localization of tau_pT217 and PSD95. (D) Frequency distributions of the nearest tau _pT217-positive structure to each PSD95-positive structure and the nearest PSD95-positive structure to each tau _pT217-positive structure. At least 50 cells from each of 3 biological replicates with 12 technical replicates apiece for confocal and 3 technical replicates apiece for dSTORM were used to generate the graphs in panels C and D.

2.5 ∣. Fluorescence recovery after photobleaching (FRAP) microscopy

pRK5-EGFP-Tau, a mammalian expression vector encoding an EGFP-human 0N4R tau fluorescent fusion protein was obtained from Addgene (catalog # 46904). This plasmid was developed by the lab of Karen Ashe¹⁹, which generously deposited it with Addgene. 0N4R tau was replaced with 2N4R tau after inserting 2N4R tau between BamHI and SalI restriction sites using a cold fusion cloning kit (System Biosciences, Catalog # MC100B-1). A Q5 Site-Directed Mutagenesis Kit (New England BioLabs) was used to convert the WT threonine at position 217 of tau (relative to the largest human CNS tau isoform, 2N4R) to a pseudo-phosphorylated glutamate (T217E). Sequencing of the resulting modified plasmid confirmed the mutation. The new plasmids described here have been deposited in Addgene with plasmid numbers 187023,187024 and 187025 for EGFP-Tau _(WT), EGFP-Tau _(T217E) and EGFP-Tau _(T217A), respectively.

To measure tau-microtubule interactions in cells it is necessary to visualize individual microtubules. This can be accomplished using fibroblasts, but not neurons, because neuronal microtubules are too tightly packed to be detected as discrete filaments. WT, T217E (pseudo-phosphorylated) and T217A (non-phosphorylatable) versions of EGFP-tau, and Venus-Tau-Teal²⁰ were expressed in CV-1 African green monkey kidney fibroblasts using Lipofectamine 3000 (ThermoFisher). The cells were maintained in phenol red-free Dulbecco’s modified Eagle’s medium supplemented with fetal bovine serum to 10% by volume and gentamycin to 50 μg/ml.

FRAP microscopy was performed in three sequential steps (see Figure 6A) using a 488 nm excitation laser at the W.M. Keck Center for Cellular Imaging at the University of Virginia: 1) pre-bleach, in which a region of interest (ROI) was selected and imaged at ~1% power to visualize microtubule-bound tau; 2) photobleach the ROI at 50% power; 3) post-bleach time-lapse imaging for 100 seconds at 1 second intervals at 1% power. Images were captured in each interval on a Zeiss 980 laser scanning confocal microscope with a 63X oil immersion planapochromatic objective. Fluorescence intensity in each ROI was plotted as a function of time before and after photobleaching (see Figures 6B–E). The rate of fluorescence recovery of microtubule-associated, photobleached, EGFP-tau is an indicator of the protein’s turnover rate on microtubules, and by extension, its affinity for microtubules. Relative to a baseline fluorescence recovery rate, faster recovery indicates more rapid turnover and lower affinity.

FIGURE 6.

T217E pseudo-phosphorylation reduces tau’s affinity for microtubules. (A) Fluorescence recovery after photobleaching (FRAP) microscopy summary. FRAP images of CV-1 cells transfected to express Venus-Tau_{WT human 2N4R}-Teal (B), or WT (C) or T217E (D) human 2N4R tau-EGFP. Note in B that photobleaching of Venus did not remove tau from the microtubules, because microtubules were still visible in the Teal channel. (E) Quantification of FRAP. At least 23 cells from at least 3 biological replicates with 3 technical replicates apiece were used to generate the graphs for both Tau_WT-EGFP and Tau_T217E-EGFP.

To normalize the FRAP data (equation 1), the bleached ROI intensity (F(t)_ROI) was divided by the whole cell intensity for each time point (F(t)_cell) after subtracting background fluorescence (F_bkgd) at each time point to correct for the loss of fluorescence during the bleach step. The data were then normalized to the pre bleach intensity (F_i-cell, F_i-ROI) and multiplied by 100 to yield a percentage of initial fluorescence. The resulting normalized data were then averaged for different cells and generated the graphs.

$$F{\left(t\right)}_{\mathit{\text{norm}}}=100\times \left(F{\left(t\right)}_{ROI}-F_{\mathit{\text{bkgd}}}\right)\left(F_{i\mathit{\text{-cell}}}-F_{\mathit{\text{bkgd}}}\right)/\left(F{\left(t\right)}_{\mathit{\text{cell}}}-F_{\mathit{\text{bkgd}}}\right)\left(F_{i\text{-}ROI}-F_{\mathit{\text{bkgd}}}\right)$$ (1)

To calculate the mobile fraction of molecules (Mf; equation 2) that recover during the time course of the experiment, we used the data obtained from the normalized recovery curves (equation 1) where F_E, F₀, and F_i are the normalized fluorescence intensities at the end of the experiment, immediately following the bleach, and prior to the bleach, respectively.

$$M_f=\left(F_E-F_0\right)/\left(F_i-F_0\right)$$ (2)

We then calculated the immobile fraction of molecules (IM_f) by using equation 3.

$$I{M}_f=1-M_f$$ (3)

Halftime of recovery (t_1/2), corresponding to 50% of recovered fluorescence intensity (I_1/2), was then monitored. To compare the halftimes of a molecule under different experimental conditions, ROI size was held constant.

2.6 ∣. Extracellular Aβ_1–42 oligomers (xcAβOs)

Lyophilized, synthetic Aβ_1–42 (AnaSpec, Catalog # AS20276) was dissolved in 1,1,1,3,3,3-hexafluoro 2-propanol (Sigma-Aldrich) to ~1 mM to convert all Aβ_1–42 aggregates that may have been present initially into monomers²¹, and then was evaporated for 4 hours at room temperature. The dried powder was resuspended for 5 minutes at room temperature in 40 μl of 20mM sodium hydroxide and sonicated for 10 minutes in a water bath. To prepare xcAβOs, the monomeric peptide was diluted to 400 μl (100 μM final Aβ concentration) in Neurobasal medium (GIBCO) and incubated for 48 hours at 4° C with rocking. This was followed by centrifugation at 14,000 g for 15 minutes to remove fibrils and any other large aggregates that may have been present. Cultured neurons were exposed to 2 μM xcAβOs for 16–18 hours before being processed for immunofluorescence or western blotting.

2.7 ∣. Extracellular tau oligomers (xcTauOs)

A bacterial expression plasmid for human 2N4R tau his-tagged at its C-terminus was generously provided by Dr. Nicholas Kanaan of Michigan State University. The protein was expressed in competent BL21 E. coli cells induced with 100 mM IPTG and purified from bacterial extracts using TALON beads (Takara Bio, Catalog # 635502) according to the vendor’s instructions.

Purified tau was diluted to 8 μM in a solution of 100 mM NaCl, 5 mM dithiothreitol, 10 mM HEPES (pH 7.6), 0.1 mM EDTA and 300 μM arachidonic acid (AA; Cayman Chemicals, Catalog # 90010). The protein was then allowed to aggregate for 18 hours at room temperature in the dark. Resulting xcTauOs were diluted into neuron cultures to yield a final total tau concentration of 250 nM. 24 hours later, cultures were processed for immunofluorescence or western blotting.

2.8 ∣. SDSPAGE and Western blots

Cultured neurons, mouse brains and CV-1 fibroblasts were lysed using N-PER Neuronal Protein Extraction Reagent (ThermoFisher catalog # 87792). Samples were resolved by SDSPAGE using 10% acrylamide/bis-acrylamide gels and transferred to 0.2 μm pore size nitrocellulose (Bio-Rad). Membranes were blocked with Odyssey blocking buffer (LI-COR) and were incubated with primary antibodies and LI-COR secondary IRDye-labeled antibodies diluted into Odyssey blocking buffer. After each antibody step, the blots were washed five times in PBS containing 0.1% Tween 20. The blots were imaged using a BioRad ChemiDoc MP. See Supplementary Table 2 for descriptions of all antibodies used for immunofluorescence and western blotting.

Dephosphorylation of blots was carried out prior to application of primary antibodies by treating with bovine intestinal mucosa alkaline phosphatase (43 – 60 μg/ml, Sigma-Aldrich, Catalog # P7640) overnight.

3 ∣. RESULTS

3.1 ∣. The challenge of antibody specificity: presence of the tau_pT217 epitope in multiple proteins

Throughout the course of this study, the only commercially available antibodies against tau_pT217 were affinity purified rabbit polyclonal IgG made against a peptide surrounding pT217 in human 2N4R tau. Products fitting this description are sold by multiple vendors, including but not necessarily limited to ThermoFisher, GenScript, Medimabs, and Abcam, the last of which was the source of the anti-tau_pT217 that we used. Some vendors, including Abcam, list the immunogen as the LPpTPP peptide that is found in the proline-rich region of all human and mouse tau isoforms, and corresponds to amino acids 215–219 in human 2N4R tau, with the threonine at position 217 being phosphorylated. The other vendors do not specify the immunogen in such detail, but it is possible that they all sell the same product.

Because the peptide used to make Abcam anti-tau_pT217 was so short, we were concerned that it might be present in other proteins, in which case antibody specificity for tau_pT217 could be problematic. To determine if the immunogen peptide exists in other mouse or human proteins, we used PhosphoSitePlus to search for proteins with documented threonine phosphorylation within an identical LPpTPP peptide. Numerous perfect hits were detected by the search (see Supplementary Table 3), emphasizing the need to seek conditions in which the anti-tau_pT217 recognizes only tau phosphorylated at T217.

In light of the potential for anti-tau_pT217 to recognize multiple proteins we established the following criteria that must be met to justify using the antibody for immunolocalization of tau_pT217 in cultured neurons and brain tissue. On multiplexed western blots of cultured neuron and brain extracts the antibody must: 1) label one band or multiple closely spaced bands that co-migrate with bands labeled by the Tau5 mouse monoclonal anti-tau, which recognizes an epitope that is found in all tau isoforms and is not subject to any post-translational modifications²²; 2) the immunoreactive bands must be sensitive to treatment of blots with alkaline phosphatase prior to incubation with anti-tau_pT217; and 3) the anti-tau_pT217 must not label any bands in extracts of brain tissue or cultured neurons derived from tau knockout (TKO) mice. All of those criteria were met by diluting the antigen affinity purified anti-tau_pT217 5000-fold to 200 ng/ml (1.43 nM; Supplementary Figure 1). At concentrations of 1 μg/ml (7.14 nM) or higher, though, the Abcam anti-tau_pT217 also recognized a prominent ~100 kDa protein that is found in WT and TKO mice, and in human brain (not shown) so that band cannot represent any form of tau. Although we have not determined the identity of the ~100 kDa band, based on its size it might represent Map3k14, which is ubiquitously expressed, and like tau_pT217 contains the LPpTPP sequence. It is important to note that other anti-tau_pT217 antibodies might be similarly affected by this specificity issue. Examples of such antibodies include the aforementioned rabbit polyclonals sold by vendors that do not divulge the immunogen sequence, and Eli Lilly and Company’s IBA493, which is being used for AD biomarker research²³ and is not commercially available.

Another condition we established to inspire confidence that anti-tau_pT217 specifically and exclusively recognizes tau phosphorylated at T217 for localization is to show by immunofluorescence that staining of brain tissue and primary neuron cultures can be eliminated when the cells or tissue are incubated with alkaline phosphatase prior to the primary antibody step. Supplementary Figure 2 illustrates that when anti-tau_pT217 is used at 200 ng/ml for immunofluorescence, it preferentially labels neurons in primary mouse cortical neuron cultures that also contain glial cells, as well as in CVN mouse and human AD brain tissue. Moreover, pretreatment of the cultured cells or brain tissue sections with alkaline phosphatase abolished staining by anti-tau_pT217. We thus conclude that using anti-tau_pT217 at a concentration of 200 ng/ml allows interference-free detection of tau_pT217 in brain tissue and cultured neurons by both western blotting and immunofluorescence.

Similarly, to the anti-tau_pT217 antibody, the specificities of mouse anti-tau_pT181 and mouse anti-tau_pT231 were tested in primary WT and TKO mouse cortical neurons and brain tissue. At a concentration of 50 ng/ml both anti-tau_pT181 and anti-tau_pT231 labeled cultured WT neurons under basal conditions by immunofluorescence (Supplementary Figure 3) but not cultured TKO neurons (Supplementary Figure 4). Likewise, when antibodies to tau_pT217, tau_pT181 and tau_pT231 were used respectively at 200 ng/ml, 50 ng/ml and 50 ng/ml for immunofluorescence microscopy of CVN and TKO mouse brain, immunoreactivity for each of those antibodies was observed only in the CVN brain. When the same antibodies were used for western blotting at the aforementioned concentrations, each labeled a single band that co-migrated with a total tau band labeled by Tau5 in WT neuron extracts, but no immunoreactive bands were observed for extracts of TKO neurons (Supplementary Figure 6). We therefore conclude that the anti-tau_pT181 and anti-tau_pT231 antibodies were specific for tau phosphorylated at T181 and T231, respectively, under the conditions in which we used them. Specificity for the PHF-1 (pS396/pS404) antibody was also validated on western blots of primary WT and TKO mouse cortical neurons at concentration of 45 ng/ml (Supplementary Figure 6).

3.2 ∣. Tau_pT217 localization in human AD brain

To study the in vivo localization of signature AD phospho-tau epitopes, we performed triple immunofluorescence microscopy of human brain sections using antibodies specific for tau_pT217 and the dendrite-enriched protein, MAP2, plus total tau (detected by the Tau5) antibody), tau_pT181 (AT180 epitope, but detected by a different antibody), tau_pS202/pT205 (detected by the AT8 antibody) tau_pT231, or tau_pS396/pS404 (detected by the PHF1 antibody). The human tissue was from 14 individuals ranging in age from 64–95 years old and with Braak stages 0-VI (Supplementary Table 1). The in vivo localization results were intended to serve as a pilot survey of the cellular and intracellular distribution of tau_pT217, rather than a comprehensive cataloging of tau_pT217 localization as functions of disease status and neuroanatomical location in humans.

As shown in Figure 1, tau_pT217 was barely detectable in human brain cortex at Braak stage 0, when sparse immunoreactivity with antibodies to tau_pT181 and tau_pT231 was observed. Appreciable tau_pT217 immunoreactivity in cortex was first seen at Braak stages I-II, in cytoplasmic puncta, neuropil threads and perinuclear cytoplasm. At those Braak stages some puncta and neuropil threads were labeled by both anti-tau_pT217 and anti-tau_pT181, but staining by the latter was much more extensive. In contrast, tau_pT217 did not colocalize with tau_pT231, which was frequently confined to neuronal nuclei.

FIGURE 1.

Tau_pT217 localization in Braak stages 0-II human brain cortex. Representative fields of view show ~5 μm thick sections that were triple-labeled with antibodies to tau_pT217, MAP2, and total tau (Tau5), tau_pT181 or tau_pT231. All images illustrate single confocal planes.

At Braak stages III-VI in human brain cortex, illustrated in Figure 2, tau_pT217 was much more abundant in variably shaped aggregates and perinuclear cytoplasm, and was similar to the distribution of total tau. Depending on the field of view at those Braak stages, the tau_pT217 distribution either resembled or did not match the distributions of tau_pT181 and tau_pT231, the latter of which still was frequently found in neuronal nuclei.

FIGURE 2.

Tau_pT217 localization in Braak stages III-VI human brain cortex. Representative fields of view show ~5 μm thick sections that were triple-labeled with antibodies to tau_pT217, MAP2, and total tau (Tau5), tau_pT181 or tau_pT231. All images illustrate single confocal planes.

To assess the possible presence of tau_pT217 in pre-tangles and tangles, human brain sections were labeled with anti-tau_pT217, and AT8 or PHF1, the latter two of which are gold standard markers for those structures at advanced Braak stages. We observed extensive overlap between structures labeled by anti-tau_pT217, and AT8 (Figure 3) or PHF1 (Supplementary Figure 7) in hippocampus at all Braak stages from I-VI in hippocampus. In cortex, however, PHF1 labeling was mainly in perinuclear puncta, that infrequently co-localized with structures labeled by anti-tau_pT217 (Supplementary Figure 7). We therefore conclude that at late Braak stages tau_pT217 is commonly present in pre-tangles and tangles in hippocampus, but less frequently in cortex.

FIGURE 3.

Tau_pT217 localization in Braak stages 0-VI human brain hippocampus. Representative fields of view show ~5 μm thick sections that were triple-labeled with antibodies to tau_pT217, MAP2 and tau_pS202/pT205 (AT8 epitope). All images illustrate single confocal planes.

3.3 ∣. Tau_pT217 is associated with developing post-synaptic sites in cultured neurons

As shown in Supplementary Figures 2 and 3, under basal conditions 10–14-day old, cultured mouse cortical neurons, which reportedly complete synaptogenesis after 15 days in culture²⁴, express some tau_pT217 that appears as puncta within MAP2-positive dendrites. The presence of tau_pT217 in these young WT neurons is consistent with the Tau Phosphorylation Sites table that lists all known, validated phospho-amino acids in tau, including pT217 in the brains of normal and AD humans, and fetal rats. To test the hypothesis that at least some of the puncta correspond to dendritic spines, we double-labeled such neurons with antibodies to tau_pT217 and the dendritic spine protein, PSD95. When the cells were imaged by a spinning disc laser confocal microscopy with a resolution of ~200 nm in the x-y plane (Figure 4A, B), tau_pT217 appeared to be extensively co-localized with PSD95, with a Pearson correlation coefficient, of ~0.7 (Figure 4C). The Pearson correlation coefficient measures co-localization on a pixel-by-pixel basis, and importantly the images shown in Figure 4A were not pixel-limited for full optical resolution.

To gain deeper insight into the spatial relation of tau_pT217 to dendritic spines marked by anti-PSD95, we also examined doubly labeled cells using a dSTORM super-resolution microscope with an x-y resolution of ~15–20 nm (Figure 4B). Under those conditions, the Pearson correlation coefficient was only 0.1 (Figure 4C), although some pixels were positive for both proteins. When combined with the analogous data for confocal microscopy, the dSTORM data indicated that tau_pT217 is typically located close to dendritic spines, rather than frequently being spatially coincident with them.

Further evidence for association of tau_pT217 with dendritic spines was obtained by first defining each contiguous set of tau_pT217-positive pixels and PSD95-positive pixels as a single object. Then, for each tau_pT217-positive object we plotted the shortest distance to a PSD95-positive object, and conversely, the shortest distance from each PSD95-positive object to a tau_pT217-positive object. As can be seen in Figure 4D, both types of measurements revealed that most PSD95-positive and tau_pT217-positive objects are within 1 μm of each other, with many sharing overlapping pixels and thereby are separated by no more than 15–20 nm. Altogether, these data indicate a very close association of tau_pT217 with dendritic spines.

3.4 ∣. Exposure of cultured neurons to xcTauOs, but not to xcAβOs, increases intraneuronal tau_pT217.

xcTauOs and xcAβOs are toxic to neurons and self-propagate in AD brain by prion-like mechanisms²⁵. To seek conditions that modulate levels of intracellular tau_pT217, we exposed cultured mouse cortical neurons to xcTauOs made from recombinant 2N4R human tau and separately to xcAβOs made from synthetic Aβ_1–42 (Supplementary Figure 8). The total tau and Aβ_1–42 concentrations were 250 nM and 2 μM, respectively, and the treatment times were 24 hours for xcTauOs and 16–18 hours for xcAβOs.

As shown in Figure 5, xcTauOs induced an ~56% increase in tau_pT217 as determined by quantitative immunoblotting, and an ~13% increase as measured by quantitative immunofluorescence microscopy. Statistical significance was achieved by both methods: p = 0.0084 by immunoblotting and p = 0.0188 by immunofluorescence. In contrast, tau_pT217 levels were unaffected by xcAβO exposure as measured by either immunoblotting or immunofluorescence. These results identify xcTauOs, but not xcAβOs, as agents that can increase neuronal levels of tau_pT217. In contrast, levels of tau_pT181 and tau_pT231 in primary mouse cortical neurons tended to rise after exposure of the cells to xcTauOs, but that increase did not reach statistical significance (Supplementary Figure 9).

FIGURE 5.

Tau phosphorylation at T217 is stimulated by extracellular tau oligomers (xcTauOs) made from 2N4R human tau (A), but not by extracellular amyloid-β oligomers (xcAβOs) made from Aβ_1–42 (B). At least 50 cells from each of 3 biological replicates with 4 technical replicates apiece were used to generate the graphs.

3.5 ∣. Pseudo-phosphorylation at T217 reduces tau’s affinity for cytoplasmic microtubules

Tau was originally described as a microtubule-associated protein in brain²⁶, and T217 is found in its proline-rich region (Supplementary Figure 10), which reportedly modulates the structure of the adjacent microtubule-binding repeat domains, and by extension, the affinity of tau for microtubules²⁷. Importantly, sub-molecular domains of tau that facilitate tau-tau binding in paired helical filaments^28–31 are embedded within tau’s microtubule-binding repeat domains³². Conditions that reduce tau’s affinity for microtubules might therefore increase tau’s propensity to form toxic aggregates.

To test the hypothesis that phosphorylation at T217 affects tau’s affinity for microtubules, we expressed EGFP-tagged human 2N4R tau in CV-1 (African green monkey kidney) cells, and performed fluorescence recovery after photobleaching (FRAP) microscopy to study the kinetics of tau turnover on microtubules. CV-1 cells, rather than primary neurons, were used for these experiments, because microtubules in the neurons, but not in the CV-1 cells, are too densely packed to allow microtubule-bound tau to be distinguished from tau that is not associated with microtubules. Two forms of EGFP-tau with respect to the tau sequence were expressed for FRAP: WT and pseudo-phosphorylated T217E (Supplementary Figure 10). As shown in Figure 6A, cells with microtubule-bound EGFP-tau were photographed before and immediately after photobleaching of selective regions of cytoplasm, and at several additional post-bleach times.

Loss of fluorescence in the photobleached areas simply reflected quenching of the EGFP signal, rather than a loss of tau from microtubules. This phenomenon is demonstrated by photobleaching of Venus-Tau-Teal: 2N4R human tau tagged at its N-terminus with Venus fluorescent protein and at its C-terminus with Teal fluorescent protein²⁰. As can be seen in Figure 6B, following complete photobleaching of Venus, microtubules labeled by fluorescent tau can still be clearly seen in the blue-shifted Teal channel that was minimally excited or quenched by the photobleaching laser.

As illustrated in Figures 6C–E, the peak of fluorescence recovery were achieved within ~10 seconds for the T217E form of EGFP-tau, but the fluorescence intensity of the WT form rose continuously for 100 seconds post-bleaching. These results indicate that tau_pT217 cycles on and off of microtubules much more rapidly than WT tau. We also noticed that microtubules labeled with EGFP-tau_(T217E) appeared to be much thinner than those labeled with EGFP-tau_(WT). A possible explanation for this observation is that EGFP-tau_(T217E) was less densely packed onto microtubules than EGFP-tau_(WT).

To verify that the T217E amino acid substitution structurally mimics tau_pT217, we used western blotting with anti- tau_pT217 of CV-1 cells transfected to express EGFP-tau_WT, EGFP-tau_T217E or EGFP-tau_T217A, the latter of which cannot be phosphorylated at position 217 because of the alanine substitution at that site. Prior work has shown that substitution of naturally encoded threonine or serine with glutamate functionally mimics phospho-threonine or phospho-serine, respectively, at multiple amino acid positions in tau³³. As shown in Supplementary Figure 10, anti- tau_pT217 recognized EGFP-tau_WT and EGFP-tau_T217E, but not EGFP-tau_T217A. The immunoreactivity with EGFP-tau_WT reflects a baseline level of WT tau phosphorylation of T217. More importantly, the immunoreactivity of anti- tau_pT217 with EGFP-tau_T217E and the absence of immunoreactivity with EGFP-tau_T217A is consistent with the idea that the T217E amino acid substitution structurally mimics phosphorylation of T217. Altogether, the EGFP-tau expression experiments strongly imply that pseudo-phosphorylation at T217 substantially reduces tau’s affinity for microtubules.

4 ∣. DISCUSSION

A flurry of recent studies of potential AD biomarkers has provided evidence that assays for CSF or blood plasma levels of tau_pT217 can rival or exceed the accuracy of assays for other fluid biomarkers or PET imaging for detecting AD, especially when combined with other diagnostic assays^7–9–34. Indeed, a study that was reported at the Tau 2022 conference but has not been published yet prompted a tentative conclusion that tau_pT217 levels in CSF define a “clock” that predicts AD progression over a 30 year time span (P-Tau217 Clock). Despite these encouraging advances about the diagnostic value of tau_pT217 for AD, to the best of our knowledge only two peer-reviewed papers^14–15 have addressed the histopathology of tau_pT217, and not a single study published until now has investigated what tau_pT217 does to neurons or what provokes its intracellular accumulation.

To shed more light on the pathobiology of tau_pT217 we therefore completed a histological survey of its cellular and subcellular distribution in the brains of human AD patients, and of its cell biological properties in cultured neurons and fibroblasts. Doing so required optimization of an anti- tau_pT217 antibody that can recognize other proteins with the identical epitope, an issue that likely applies to other antibodies directed against tau_pT217. Consistent with prior immunohistological studies of human brain^14–15, we found that tau_pT217 first appears in limbic structure neurons at Braak stages I-II, and that its cellular and subcellular distribution overlaps partially with those of other phospho-tau variants associated with AD and other tauopathies. Using cultured mouse neurons, we found baseline levels of tau_pT217 closely associated with nascent dendritic spines, and stimulation of the intracellular tau_pT217 level by neuronal exposure to xcTauOs, but not to xcAβOs. Finally FRAP analysis of EGFP-tagged tau in fibroblasts implied that pseudo-phosphorylation of tau at T217 potently decreases tau’s affinity for microtubules. Altogether, these new data raise the possibility that phosphorylation at T217 provokes rapid tau turnover on microtubules, concomitant impairment of fast axonal transport³⁵ and synaptic activity, and stimulation of tau aggregation^36–37.

Because AD biomarker assays that rely on antibodies to tau_pT217 are being developed for clinical use, it is essential that such antibodies do not cross-react with other proteins. The polyclonal rabbit anti-tau_pT217 that we used for this study may be the only anti-tau_pT217 antibody that is currently available from commercial sources, although not all vendors provide sufficient information to be certain. The antibody we used was sold by Abcam and it was produced by immunizing a rabbit with an LPpTPP peptide that is found in at least 18 human and/or mouse proteins other than tau (Supplementary Table 3). The possibility that this antibody, as well as the other commercially available rabbit polyclonals and proprietary antibodies to tau_pT217, recognize one or more proteins besides tau phosphorylated at T217 is thus a serious concern. Fortunately, we found that diluting the antigen affinity purified antibody 5000-fold to a concentration of 200 ng/ml (1.43 nM) for western blotting of human brain, mouse brain and mouse neuron cultures allowed detection of tau_pT217 without detecting other bands that were recognized by the antibody when it was used at higher concentrations (Supplementary Figure 1). We therefore also diluted the antibody to 200 ng/ml for immunofluorescence detection of tau_pT217. This concentration-dependent specificity of the antibody presumably reflects protein-specific differences in amino acid sequences that flank the LPpTPP sequence and affect the affinity of the antibody for tau_pT217 versus other proteins that contain the same peptide. Moreover, antibody dilutions that are sufficient to eliminate spurious immunoreactivity on western blots are direct indicators of the dissociation constants of the antibodies for their respective antigens. When used at such dilutions for other applications, like immunofluorescence and immunohistochemistry, antibody specificities should be as valid as for western blotting because the dissociation constants do not change as long as similar or identical binding and wash buffers are used for all of the various applications. This cautionary tale about the anti- tau_pT217 antibody that we used should be taken into account for other basic and clinical research studies, as well as clinical applications aimed at detecting tau_pT217 for all antibodies directed against that tau phosphorylation site.

Our immunohistological analysis of tau_pT217 in human brain yielded results similar to those reported earlier by two other groups^14–15. In cortex at Braak stage 0, tau_pT217 was barely detectable, but occasional immunostaining of axons, neuropil threads and neuronal cell bodies was evident by Braak stages I-II (Figure 1), and more widespread by Braak stages III-VI (Figure 2). This gradual rise in cortical tau_pT217 immunoreactivity was recapitulated in hippocampus, where frequent labeling of pre-tangles and tangles was also evident at late Braak stages (Figure 3).

In contrast, Moloney and colleagues reported pre-tangles and mature tangles in hippocampus at much earlier Braak stages than we found¹⁴. Those results were obtained using what may have been a different rabbit polyclonal anti-tau_pt217 (Thermo-Fisher 44–744) than what we used (Abcam ab192665), however, and at an unspecified nanomolar concentration. The possible use of two different antibodies might therefore explain the discrepancy about how early tau_pT217-positive pre-tangles and tangles appear in AD brain, at least in part. Other potential explanations include patient-to-patient variability, post-mortem intervals and lab-specific differences in immunostaining methods.

Frequent hippocampal CA1 tangles containing tau_pT217 were also reported by Wennström and colleagues, who used a proprietary rabbit anti-tau_pT217 made by Eli Lilly and Company¹⁵. Wennström and colleagues also reported that the Lilly antibody robustly labeled granulovacuolar degeneration bodies and multi-vesicular bodies, neither of which we observed. Because further specifications about that antibody are not publicly available, we do not know if it detected granulovacuolar degeneration bodies and multi-vesicular bodies because it was used at too high a concentration, or if other factors, such as antigen retrieval methods and additional aspects of the immunohistochemical protocol might explain the discrepancies with our results.

Interestingly, we found by multi-color immunofluorescence of Braak stages I-VI brain that the distribution of tau_pT217 was generally similar, but not identical to the distributions of the other phospho-tau epitopes we examined: tau_pT181, tau_pT202/205, tau_pT231 and tau_pS396/pS404. Analogous observations were published by Wennström and colleagues¹⁵, but that group did not report the nuclear localization of tau_pT181 and tau_pT231 that we observed. A possible explanation for that discrepancy concerns the antibodies that were used. Whereas Wennström and colleagues used AT270 and Abcam EPR2488 to detect tau_pT181 and tau_pT231, respectively, we used Biolegend 846602 to label tau_pT181 and tau_pT231, and Biolegend 828901 for detecting tau_pT231. Regardless of what the explanation may be, both the Wennström group and ours agree that each of the phospho-tau epitopes discussed here has a unique cellular and subcellular distribution relative to the others in AD brain. Considering that tau has ~80 potential phosphorylation sites, of which ~50 have been validated (https://bit.ly/2JyZTbS) it should not be surprising that tau phosphorylation does not occur synchronously at its numerous sites. Deciphering how tau phosphorylation at any individual site influences phosphorylatability at other sites, as well as tau function, remain challenging goals.

To gain further insight into functions and regulation of tau_pT217 we examined its localization and induction in cultured neurons, and used FRAP microscopy to study its association with microtubules in fibroblasts expressing EGFP-tau. Using conventional confocal and dSTORM super-resolution microscopy, we found a striking correspondence between tau_pT217 puncta and nascent dendritic spines marked by anti-PSD95 in primary mouse cortical neurons cultured under basal conditions (Figure 4). In addition, exposure of such cultured neurons to xcTauOs, but not to xcAβOs, caused a substantial increase in intracellular tau_pT217 (Figure 5). Finally, we found that tau-EGFP pseudo-phosphorylated at position 217 by a T-to-E amino acid substitution had a dramatically shorter dwell time on fibroblast microtubules than WT tau coupled to EGFP (Figure 6). Altogether, these results raise the possibilities that tau_pT217 affects synapses, that soluble xcTauOs in the brain parenchyma provoke tau_pT217 accumulation in AD neurons in vivo, and that the reduced microtubule affinity of tau phosphorylated at T217 causes dysregulated fast axonal transport, which we described previously as occurring in cultured neurons derived from tau knockout mice³⁵. The presence of tau_pT217 near developing dendritic spines in cultured neurons that are still consolidating synapses²⁴ suggests a positive role for tau_pT217 in synapse formation. In contrast, the virtual absence of tau_pT217 in Braak stage 0 brain and its steady increase at later Braak stages (Figures 1–3 and Supplementary Figure 7) raises the possibility that tau_pT217 is not involved in synapse maintenance, but that its appearance and buildup during AD pathogenesis is harmful to synapses.

While much further work is required to refine understanding of what causes tau_pT217 to accumulate in neurons and how it affects neuronal functions, the evidence shown here provides the first indications of how tau_pT217 might be produced and compromise neuronal health beginning at the earliest stages of AD pathogenesis. At a more clinically relevant level, important questions that arise concern the applicability of the findings presented here to non-Alzheimer’s tauopathies, and specifically how these basic science findings might translate to patient care. For example, does tau_pT217 play pathogenic roles in other mixed tauopathies, such as Parkinson’s disease and Huntington’s disease, and in pure tauopathies, like Pick’s disease and progressive supranuclear palsy? Likewise, while numerous protein kinases can phosphorylate tau at T217 (https://bit.ly/2JyZTbS) would identifying the clinically relevant ones lead to the development of kinase inhibitors or other classes of compounds that interfere with the pathogenesis of AD and other tauopathies for which tau_pT217 may prove to be important?

Supplementary Material

supinfo

SUPPLEMENTARY FIGURE 1 Validation of anti-tau_pT217 specificity by western blotting. Anti-tau_pT217 and total tau (Tau5) western blots were performed against (A) cultured wild type (WT) and tau knockout (TKO) mouse cortical neurons, (B) 24 month old WT and CVN mouse brain homogenates, and (C) elderly, age-matched, cognitively normal and AD human brain homogenates. When used at 200 ng/ml (1.43 nM), as for these experiments, anti-tau_pT217 labeled a single band with the same electrophoretic mobility as total tau, and the band was sensitive to alkaline phosphatase treatment prior to the primary antibody step and was undetectable in TKO mouse brain.

SUPPLEMENTARY FIGURE 2 Validation of anti-tau_pT217 specificity by immunofluorescence. Braak stage V-VI human brain (A), 24 month old CVN mouse brain (B) and cultured WT mouse cortical neurons were double-labeled with anti-tau_pT217 at 200 ng/ml (1.43 nM), and to mark the neuronal somatodendritic compartment, with anti-MAP2. Note that anti-tau_pT217 was sensitive to prior treatment of the tissue sections or cells with alkaline phosphatase.

SUPPLEMENTARY FIGURE 3 Tau_pT217 is partly co-localized with tau_pT181, tau_pT231 and tau_pS396/pS404 in WT mouse cortical neurons cultured under basal conditions. Neurons were obtained from E16-E17 embryonic brain and were cultured for 10–14 days. Single plane confocal images are shown.

SUPPLEMENTARY FIGURE 4 Tau_pT217, tau_pT181 and tau_pT231 are detectable by immunofluorescence microscopy in 14 day old WT, but not TKO mouse cortical neurons. Single plane confocal images are shown.

SUPPLEMENTARY FIGURE 5 Tau_pT217, tau_pT181 and tau_pT231 are detectable by immunofluorescence microscopy in brains of 18–24 month old CVN, but not TKO mice. Single plane confocal images are shown.

SUPPLEMENTARY FIGURE 6 Western blotting validation of mouse anti-tau_pT181, mouse anti-tau_pT231 and mouse anti-tau_pS396/pS404 (PHF1) antibodies in 10–14-day old primary WT and TKO mouse cortical neurons. Tau5 was used as a total tau marker.

SUPPLEMENTARY FIGURE 7 Tau_pT217 localization in Braak stages I-VI human brain cortex versus hippocampus. Representative fields of view show ~5 μm thick sections labeled with antibodies to tau_pT217, MAP2, and tau_pS396/pS404 (PHF1 epitope). All images illustrate single confocal planes.

SUPPLEMENTARY FIGURE 8 Analysis of extracellular oligomers of 2N4R human tau and Aβ_1–42 by western blotting. Tau5 was used to probe the tau blot and 6E10 was used to probe the Aβ_1–42 blot. *Classification of the indicated immunoreactive bands as dimer and timer is based on mass spectrometry analysis of similarly sized tau oligomers prepared using arachidonic acid, the same method used in this study, in Patterson, et al. 2011. J Biol. Chem. 286: 23063–23076.

SUPPLEMENTARY FIGURE 9 Tau phosphorylation at T181 and T231 are not stimulated by extracellular tau oligomers (xcTauOs) made from 2N4R human tau. At least 2 biological replicates with 2 technical replicates apiece were used to generate the graphs.

SUPPLEMENTARY FIGURE 10 EGFP-tau variants used for fluorescence recovery after photobleaching (FRAP) microscopy. The fluorescent fusion proteins were expressed in CV-1 cells as WT, T217E pseudo-phosphorylated and T217A non-phosphorylatable versions of 2N4R human tau. Note that by western blotting anti- tau_pT217 labeled EGFP-tau_WT and EGFP-tau_T217E, but not EGFP-tau_T217A, indicating that the T217E amino acid substitution accurately mimics that structure of the tau_pT217 epitope.

SUPPLEMENTARY TABLE 1 Clinical characterization of human autopsy brain samples used in the study.

SUPPLEMENTARY TABLE 2 Primary and secondary antibodies used in this study.

SUPPLEMENTARY TABLE 3 Human and mouse proteins containing the LPpTPP sequence (from PhosphoSite Plus: https://www.phosphosite.org/homeAction).

RESEARCH IN CONTEXT.

1. Systematic review:

Tau phosphorylated at threonine-217 (tau_pT217) is a promising early biomarker for Alzheimer’s disease (AD), but little is known about its cellular and subcellular distribution, or the functional consequences of phosphorylation at that site. The purpose of this study was to begin exploring the role of tau_pT217 in AD pathogenesis, and by extension, to reveal potential new strategies for prevention or delay of AD symptom onset.

2. Interpretation:

Anti-tau_pT217 can detect other proteins, so validation of antibody specificity is of paramount importance. In human brain, tau_pT217 appears at Braak stage I-II and becomes more prevalent at later stages in axons, dendrites, neuronal cell bodies and tangles. In cultured neurons, Tau_pT217 is increased by extracellular tau oligomers and is associated with developing post-synaptic sites. Pseudo-phosphorylation by T217E mutation reduces tau’s affinity for microtubules.

3. Future directions:

As a possible stepping stone to identification of potential new targets for preventive intervention, more functional and mechanistic studies are warranted for further characterizing how tau_pT217 contributes to neuron decline in AD.