Molecular Investigation of the Unfolded Protein Response in Select Human Tauopathies

Materials

The primary antibodies used in this study were: anti-BiP (RRID: AB_2119845, Cell Signaling, #3177, 1 : 500 dilution), anti-CHOP (RRID:AB_2089254, Cell Signaling, #2895, 1 : 1000 dilution), anti-p-eIF2α (RRID:AB_2096481, Cell Signaling, #3398, 1 : 1000 dilution), anti-eIF2α (RRID: AB_836874, Cell Signaling, #2103, 1 : 1000 dilution), anti-tau (RRID:AB_10013724, DAKO, #A0024, 1 : 1000 dilution), anti-GAPDH (RRID:AB_2107448, Abcam, #ab8245, 1 : 5000 dilution), anti–actin (Sigma-Aldrich Cat# A4700, RRID:AB_476730), PHF-1 (detecting p-Ser396/p-Ser404, RRID:AB_2315150, generous gift from Prof. Peter Davies, 1 : 5000 dilution, [23]), AT8 (detecting p-Ser202/p-Thr205, RRID:AB_223647, #MN1020, ThermoScientific).

Secondary antibodies were conjugated to IRDyes (LI-COR Biosciences) and used at 1 : 10000 dilution.

The oligonucleotide primers used in the qPCR and PCR reactions were obtained from Eurofins Genomics. The sequences of the primers for each gene are shown in Table 1.

Cases

AD brain tissue was sourced from the South West Dementia Brain Bank comprising 20 AD cases and 20 controls (nAD). The inferior parietal lobule, as an area of the cerebral cortex typically affected by AD pathology [24], was investigated for all cases (Table 2). Eleven PiD and 11 controls (nPiD) as well as 10 PSP and 10 controls (nPSP) were provided by the London Neurodegenerative Diseases Brain Bank, the South West Dementia Brain Bank and the Oxford Brain Bank with the frontal cortex investigated for the FTD and control cases (Table 3 for nPiD and PiD cases and Table 4 for nPSP and PSP cases). Cases with any other significant brain pathologies such as stroke, tumor, or traumatic brain injury were excluded from the study. Controls with no history of neurological or psychiatric disease or symptoms of cognitive impairment were matched with age, gender, and postmortem delay as closely as possible. To minimize the time in formalin, which has an effect on the quality of the immunostaining, the selection was performed on the availability of formalin fixed paraffin embedded tissue, and thus on blocks processed at the time of the original postmortem examination. Fresh frozen tissue with a pH > 5.5 was selected to ensure RNA integrity [25, 26]. Brain tissue adhering to this criterion was homogenized and used as a source of control and disease associated mRNA and protein.

Ethics approval

Ethical approval was provided by the South West Dementia Brain Bank (REC approval 08/H0106/28 + 5, the London Neurodegenerative Diseases Brain Bank (REC approval 08/H0704/128), and the Oxford Brain bank (REC approval 51/SC/0639). This was a non-registered postmortem study.

Immunohistochemistry

Immunohistochemistry against phosphorylated (p)-tau protein was performed on 6μm paraffin-embedded sections using the clone AT8 (RRID:AB_223647, #MN1020, ThermoScientific). Sections were deparaffined, rehydrated, and incubated in a solution of 3%H₂O₂ in methanol to block the endogenous peroxidase. The saturation step was performed before incubation with primary antibody overnight at 4°C. Biotinylated secondary antibody goat anti-mouse from Vector Laboratories (RRID:AB_2336171, Peterborough, UK) was applied to the sections the next day followed by development of immunohistochemical reactivity using the avidin–biotin–peroxidase complex method (RRID:AB_2336819, Vectastain Elite, Vector Laboratories) with 3,3^′-diaminobenzidine as chromogen and 0.05%hydrogen peroxide as substrate (RRID:AB_2336382, Vector Laboratories). All sections were counterstained with hematoxylin, then dehydrated and mounted in Pertex (Histolab Products AB). Staining was performed in several batches with each batch containing disease and control cases to ensure comparability of immunolabeling. All experiments included a negative control slide incubated in buffer with no primary antibody.

Image quantification

Quantification was performed by observers blinded to the experimental group and identity of the cases. The slides were scanned at a magnification of x20 in an automated slides scanner microscope Olympus VS110 (Olympus America Inc.) and visualized with Olympus VS-Desktop software to allow extraction of 40 regions of interest (ROIs) in the grey matter. The ROIs were obtained in a zigzag sequence to ensure sampling of all six cortical layers as previously published [27, 28]. Quantitative image analysis was carried out with ImageJ (version 1.52p, Wayne Rasband, NIH, USA) using an automated macro to obtain an area fraction expressed as protein load (%). The mean value was calculated for each case.

Tissue homogenization

The brain samples were homogenized in 5 volumes (w/v) of sterile PBS containing cOmplete Mini EDTA-free Protease Inhibitor Cocktail tablets (Roche, #4693159001), sodium fluoride (10 mM, Fisher Scientific, #10528070) and sodium orthovanadate (2 mM, Sigma, #S6508). The samples were homogenized using Kontes pellet pestle motor and plastic pellet pestles. The homogenized samples were then used for RNA or protein extraction [29].

RNA extraction

RNA was isolated from 100μl of homogenate using the Trizol method. The RNA was purified using RNeasy kit (Qiagen, #74104). RNA amount and quality was analyzed using NanoDrop spectrophotometer (Thermo Fisher Scientific). The RNA exhibited an absorbance maximum at 260 nm (A260) and the ratio of A260/A280 was ∼2. The extracted RNA was treated with Precision DNase (Primer Design, #DNASE-50) according to manufacturer’s instructions. 200 ng of RNA was reverse transcribed using Precision nanoScript 2 reverse transcription kit (Primer Design, #RT-NanoScript2) according to manufacturer’s instructions.

XBP-1 splicing

RedTaq ReadyMix PCR Reaction Mix (Sigma Aldrich, #R2523), 1μl of cDNA (obtained as described above) and primers (0.2μM) detecting both, spliced and unspliced, forms of XBP-1 were used for end point PCR. The specificity of the primers was verified using a positive control, cDNA from HEK cells treated with tunicamycin (Supplementary Figure 1A). The samples were kept on ice before being heated at 94°C for 2 min. Subsequently, the samples were subjected to 40 cycles of denaturation at 94°C for 40 s, annealing at 60°C for 30 s, extension at 72°C for 1 min using GeneAmp PCR System 9700 (Applied Biosystems). A final extension was performed at 72°C for 10 min. The amplicons of this high cycle number PCR reaction were separated on 2.5%agarose gel. The no template control that did not contain cDNA was run alongside the samples and failed to show XBP-1 amplification. It shows the specificity of the amplification. The details of the primers used for the detection of spliced XBP1, and GAPDH as control, can be found in Table 1.

Western blotting and analysis

Proteins from PBS-homogenized samples were extracted with the equal volume of 2x extraction buffer containing HEPES-NaOH (40 mM, pH 7.4, Fisher Scientific, #10081113), NaCl (250 mM, Sigma Aldrich, #S7653), SDS (4%), cOmplete Mini EDTA-free Protease Inhibitor Cocktail tablets (Roche, #4693159001), sodium fluoride (10 mM, Fisher Scientific, #10528070), and sodium orthovanadate (2 mM, Sigma, #S6508). After extraction, protein concentration was measured using Bio-Rad DC protein assay kit (#5000112) according to manufacturer’s instructions.

The samples were mixed to give a final concentration of 2μg/μl with 5x sample buffer containing Tris-HCl (312.5 mM, pH 6.8), SDS (10%), glycerol (50%), dithiothreitol (25 mM), and bromophenol blue dye (0.005%).

These samples were boiled for 10 min at 95°C, briefly spun before loading to 12%acrylamide gel. Tauopathy cases were always run on the same gel as the relative age-matched controls. After loading, the samples were separated by SDS-PAGE and transferred to nitrocellulose membrane. The membranes were blocked in the blocking solution (2.5%BSA, TBS with 0.1%Tween) for 1 h and then incubated at 4°C overnight with the primary antibody diluted in the blocking solution. The membrane was then incubated with the secondary antibody at room temperature for 1 h. Immunoreactivity was revealed using an Odyssey Infrared Imaging System (LI-COR Biosciences). The Image Studio Scanner software was used to capture the image and Image Studio Lite software was used to quantify the intensities of the bands, using the area above and below the band for background subtraction. Cases were analyzed across multiple blots. For this they were split to ensure equal numbers of control and diseased cases to be resolved on individual gels before transfer to membranes that were processed in parallel. Measurement of GAPDH allowed for normalization to individual sample loading as well as comparison of matched membranes. Exposures that showed minimal difference in GAPDH intensity between membranes were used for quantification. The ability of the antibodies to detect UPR markers induction was verified using positive control, protein extracted from tunicamycin-treated HEK cells (Supplementary Figure 1B–D). Only antibodies that were capable of detecting increased immunoreactivity of positive control were used in this study.

Statistical analysis

The normality of distribution of the data was determined using the Shapiro-Wilk test. Data that passed the normality test and had equivalent standard deviations were compared using a T-test for parametric data, otherwise the Mann-Whitney U-test for non-parametric data was used. One-sided tests were used when comparing the level of p-tau and total tau between disease and control cases based on the prediction that p-tau is increased and tau inclusions accumulate in tauopathy cases. Total tau levels were further compared using JASP statistics software to perform a Bayesian independent samples t-test [30]. For the UPR markers, two-sided tests were assumed. Correlations between the different markers were analyzed by either Pearson’s (parametric) or Spearman’s (non-parametric) test. Fisher’s exact test was used for comparison between disease and control cases to assess the presence of XBP-1 splicing. No test for outliers was performed. All analyses were performed with SPSS software (version 25, IBM) or GraphPad Prism. p values less than 0.05 for intergroup comparisons and 0.01 for correlations were considered statistically significant. Graphs were prepared with GraphPad Prism software (version 8, La Jolla, CA).

Tau phosphorylation in neurodegeneration

Tauopathies are characterized by the increased phosphorylation and accumulation of hyperphosphorylated tau. We used immunohistochemistry to investigate the presence of tau phosphorylation in our samples. This analysis confirmed the presence of tau pathology in the AD, PiD, and PSP cases, using the commonly utilized AT8 antibody that recognizes the disease associated increase in pSer202/pThr205 of tau. In AD, dystrophic neurites, neuropil threads and NFTs were observed in the disease sections but were absent in the non-demented age matched controls (Supplementary Figure 2A). For PiD, this AT8 p-tau was localized in neurons, labelling Pick bodies, and neuropil (Supplementary Figure 2B). In the PSP cases, the pathology clearly presented NFTs in the cortex (Supplementary Figure 2C). Representative stains that highlight these features that were selectively found in the AD, PiD, and PSP disease cohorts, together with the corresponding quantitation, are shown in Supplementary Figure 2. In parallel, we homogenized and extracted mRNA and proteins from the brains of the listed cases (Methods section). To benchmark the cases investigated, we performed western blot with the PHF1 antibody directed against the tau phospho-epitopes Ser396/Ser404. As expected, the AD, PiD, and PSP cases showed increased tau phosphorylation when compared to the control cases (Fig. 1). The increase in tau phosphorylation in AD cases was accompanied by an increase in total tau when compared to age-matched non-demented controls in face of equivalent GAPDH signal (Fig. 1A). In contrast, total tau levels in PiD cases were similar to those in control cases, despite showing an 80-fold increase of p-tau level compared to controls (Fig. 1B). We did not find evidence for an increase in total tau level in PSP cases despite the significantly increased p-tau compared to control cases (Fig. 1C).

OPEN IN VIEWER

Fig. 1

Relative levels of phospho-tau in the brain samples from distinct tauopathy cohorts. Brain homogenates from age-matched controls and (A) AD, (B) PiD, and (C) PSP brains were probed for p-tau, total tau and GAPDH. The PHF1 and total tau immunoreactivity is shown alongside the corresponding GAPDH. Immunoreactivity was expressed as ratios of PHF1/total tau, PHF1/GAPDH and total tau/GAPDH. The normalized data were analyzed using a one-tailed unpaired t-test (total tau/ GAPDH) or Mann-Whitney test (PHF1/ total tau, PHF1/ GAPDH). Shown is the mean, error bars are SEM. ^****p≤0.0001; ^***p≤0.001; ^**p≤0.01, ns- not significant. p-tau/GAPDH: AD: p < 0.0001 (U = 11), PiD: p = 0.0012 (U = 16), PSP: p = 0.0045 (U = 16); p-tau/total tau: AD: p < 0.0001 (U = 16), PiD: p = 0.0004 (U = 12), PSP: p = 0.001 (U = 11). Total tau/ GAPDH: AD: (p = 0.0002, t = 3.811, BF ₊₀ = 69.957); PiD: (p = 0.1027, t = 1.309, BF ₊₀ = 0.165); PSP: (p = 0.286, t = 0.5759, BF ₊₀ = 0.356). Number of cases: ND = 20, AD = 20, nPiD = 11, PiD = 11, nPSP = 10, PSP = 10. Each lane corresponds to one individual case. Colored symbols in (A) correspond to individuals with LBD co-morbidity (red), APOE 4.4 genotype (blue), and CVD co-morbidity (magenta).

Interestingly, the phosphorylation level in the cases within the diseased groups (AD, PiD, and PSP) was heterogenous showing up to an 80-fold variation in p-tau between the highest and lowest levels of PHF-1 immunoreactivity detected by Western blotting. In contrast, the relative levels of total tau showed a much more constrained variability of immunoreactivity.

Finally, this side-by-side comparison of homogenates allowed us to score the relative levels of p-tau in the investigated tauopathies. Although increased p-tau was detected for all the diseases, the degree of tau phosphorylation differed amongst the tauopathies, with PSP cases having the lowest fold increase in comparison to non-demented cohort (Fig. 1C). The data confirm the increased phosphorylation of tau in the brains of patients affected with dementia. In addition, the results highlight the variability of p-tau level that can be observed within the diseased groups. The data thus can be used to analyze whether the expression of UPR markers correlates with p-tau levels.

XBP-1 splicing in tauopathy

One of the commonly used and sensitive markers of UPR activation is splicing of XBP-1 mRNA. This arises from the activation of the endoribonuclease IRE1, leading to the removal of an intron from mRNA. This generates an active transcription factor that induces expression of genes involved in ER homeostasis [31]. We have previously shown that chemical perturbation of cultured rodent neurons with tunicamycin triggers a UPR that is readily detected by primers designed to pick up XBP-1 splicing [22]. We designed human primers that span the spliced intron and enable detection of both the unspliced mRNA and the shorter spliced form that is an indicator of an active and ongoing UPR. These primers were first verified on the mRNA extracted from tunicamycin-treated HEK293 cells. This identified that they detect both the unspliced and spliced forms of XBP-1 and the UPR activated splicing is selectively found in the mRNA extracted from the tunicamycin treated cells (Supplementary Figure 1A).

The primers amplified the products of expected sizes in cDNA from human brain samples (377 bp for the unspliced and 351 bp for the spliced transcript). We then used the distribution of the spliced and unspliced amplicons to probe if XBP-1 splicing was selectively occurring in the investigated tauopathies when compared to the respective controls (Fig. 2). We observed that splicing occurs independently of the presence of the disease diagnosis. In AD cases, 40%of the samples were found to be positive for the splicing compared to 75%in controls. There was no significant difference between both groups (Fig. 2A). In PiD, 55%of the cases presented the XBP-1 splicing but this was not statistically different from the control cases (Fig. 2B). Similarly, in PSP, 40%of the cases were positive for XBP-1 splicing with no significant difference when compared to respective control cases (Fig. 2C). For each group, some cases had undetectable level of XBP-1 despite readily amplifying the housekeeping gene, GAPDH. There was no association between the XBP-1 splicing and the level of p-tau (data not shown). Altogether, the data suggest that XBP-1 splicing occurs to comparable levels in both control and disease cohorts. The distribution of samples that detect a splicing event shows that XBP-1 is a widely observed event in the ageing brain based on its presence in the non-demented control samples (Supplementary Figure 3).

OPEN IN VIEWER

Fig. 2

XBP-1 splicing in tauopathies and age-matched control brains. cDNA from brain samples from (A) AD, (B) PiD, and (C) PSP together with respective non-demented controls were subjected to PCR with primers designed to amplify the unspliced and spliced variants of XBP-1. The products were resolved on a 2.5%agarose gel. The PCR products were classified as undetectable, spliced, or unspliced, and the distinct PCR outcomes are expressed as percentage of total samples and analyzed using Fisher’s exact test for (A) AD (p = 0.1560), (B) PiD (p = 0.9999), and (C) PSP samples (p = 0.6372). Number of cases: ND = 20, AD = 20, nPiD = 11, PiD = 11, nPSP = 10, PSP = 10. Each lane corresponds to one individual case.

p-eIF2α level in tauopathy

The PERK arm of the UPR has emerged as a candidate stress response in AD and other tauopathies [14, 32]. Here we used the phosphorylation of eIF2α that can occur upon PERK activation as a surrogate for the activation of the PERK branch of the UPR. Importantly, in the preclinical studies an increase in phosphorylation of eIF2α has been discussed as mediating UPR in face of no XBP-1 signaling. An antibody capable of detecting a tunicamycin-induced eIF2α phosphorylation in human cells by western blot (Supplementary Figure 1B) was used to probe whether PERK arm of UPR was activated in humans. Single or double bands of the correct size of 38 kDa were detected for all the samples investigated. In tauopathies, no increase in p-eIF2α levels was detected relative to their respective controls (Fig. 3), irrespectively of whether phosphorylation was normalized to GAPDH (Fig. 3A–C), total eIF2α (Fig. 3B, C) or actin (Fig. 3B, C). Further, we did not detect a difference in eIF2α levels between PiD cases, PSP cases and their respective controls (Fig. 3B, C). The absence of increased p-eIF2α selective to the diseased cases argues against a disease-associated engagement of the PERK arm of the UPR.

OPEN IN VIEWER

Fig. 3

Detection of eIF2α phosphorylation in tauopathies and age-matched controls. p-eIF2α and GAPDH immunoreactivity for (A) AD, (B) PiD, and (C) PSP together with relative age-matched non-demented controls. Quantification of p-eIF2α immunoreactivity from individual samples was normalized to the immunoreactivity of a loading control, GAPDH. Data were analyzed using a two-tailed unpaired t-test revealing no significant differences (p = 0.4892, t = 0.6983; p = 0.3604, t = 0.9362; p = 0.8813, t = 0.1514 for AD, PiD, and PSP, respectively). (B, C) p-eIF2α, eIF2α, and actin immunoreactivity for (B) PiD and (C) PSP. Quantification revealed no significant difference when normalizing p-eIF2α over eIF2α (p = 0.3653, U = 46; p = 0.5288, U = 41 for PiD, PSP), p-eIF2α over actin (p = 0.8977, U = 58; p = 0.5607, t = 0.5927 for PiD, PSP) or eIF2α over actin (p = 0.6522, U = 53 p = 0.9705, U = 49 for PiD, PSP). Shown are means, error bars show SEM. ns, not significant. Number of cases: ND = 20, AD = 20, nPiD = 11, PiD = 11, nPSP = 10, PSP = 10. Each lane corresponds to one individual case. Colored symbols in (A) correspond to individuals with LBD co-morbidity (red), APOE 4.4 genotype (blue), and CVD co-morbidity (magenta).

Expression levels of the common sensor of the UPR- BiP

The ER chaperone BiP is a pivotal sensor of the UPR and the level of BiP expression is increased upon UPR activation. In contrast to the selective markers described above, p-eIF2α and XBP-1, BiP is implicated in all three branches of the UPR. The investigation of BiP expression was performed using previously tested reagents (Supplementary Figure 1C). The expression of BiP protein was not increased in the disease versus control cases (Fig. 4), suggesting that the UPR is not activated in the tauopathies.

OPEN IN VIEWER

Fig. 4

BiP protein expression in tauopathies and age-matched non-demented controls. A) BiP and GAPDH immunoreactivity for AD, PiD, and PSP together with relative age-matched non-demented controls. B) Quantification of BiP immunoreactivity from individual samples was normalized to the immunoreactivity of a loading control, GAPDH. Data were analyzed using a Mann-Whitney test (AD, PiD) or two-tailed unpaired t-test (PSP) revealing no significant differences (p = 0.2888 and U = 160, p = 0.4779 and U = 49, p = 0.6099 and t = 0.5193 for AD, PiD disease and PSP, respectively). Shown are means, error bars are SEM. ns, not significant. Number of cases: ND = 20, AD = 20, nPiD = 11, PiD = 11, nPSP = 10, PSP = 10. Each lane corresponds to one individual case. Colored symbols for nAD, AD cases correspond to individuals with LBD co-morbidity (red), APOE 4.4 genotype (blue), and CVD co-morbidity (magenta).

CHOP expression in tauopathies

Finally, CHOP, another sensitive marker that is robustly induced by each of the three branches of the UPR, was investigated. Similar to BiP, protein level was determined using previously tested reagents that detected an induced UPR in tunicamycin-treated human cells (Supplementary Figure 1D). The antibody raised against CHOP protein showed the clear induction of a protein of expected size in the tunicamycin-treated cells. However, we noted a confounding co-migration of non-specific bands next to the protein of interest that were enhanced in the untreated cells. To quantify the relative CHOP expression in the brain samples, the immunoreactivity that migrates at the same size of the induced protein in HEK293 cells was used.

No difference in the CHOP protein expression was observed between disease and control cases (Fig. 5). These findings suggest that this downstream target of the UPR branches is not induced in the diseased cohorts, further supporting the notion that UPR is not activated in the investigated tauopathies.

OPEN IN VIEWER

Fig. 5

CHOP protein expression in tauopathies and age-matched non-demented controls. A) CHOP and GAPDH immunoreactivity for AD, PiD, and PSP together with relative age-matched non-demented controls. B) Quantification of CHOP immunoreactivity from individual samples was normalized to the immunoreactivity of a loading control, GAPDH. Data were analyzed using Mann-Whitney test (AD) or two-tailed unpaired t-test (PiD and PSP) revealing no significant differences (p = 0.6749 and U = 184, p = 0.7539 and t = 0.3178, p = 0.7649 and t = 0.3036 for AD, PiD, and PSP, respectively). Shown are means, error bars are SEM. ns, not significant. Number of cases: ND = 20, AD = 20, nPiD = 11, PiD = 11, nPSP = 10, PSP = 10. Each lane corresponds to one individual case. Colored symbols for nAD, AD cases correspond to individuals with LBD co-morbidity (red), APOE 4.4 genotype (blue), and CVD co-morbidity (magenta).

Correlations of p-tau with the UPR markers

Given the heterogenous levels of p-tau within the disease groups, we explored whether the p-tau load correlated with the expression of UPR markers within the different groups (disease and controls) of the different tauopathies (AD, PiD, and PSP). No significant association was detected between the p-tau markers AT8 and PHF1 with the UPR markers p-elF2α, BIP and CHOP in any of the control or disease groups (Tables 5–7). This again supports the suggestion that late stage human tauopathies are not associated with an activated UPR.