Tau reduction in the presence of amyloid-β prevents tau pathology and neuronal death in vivo

Abstract

Several studies have now supported the use of a tau lowering agent as a possible therapy in the treatment of tauopathy disorders, including Alzheimer’s disease. In human Alzheimer’s disease, however, concurrent amyloid-β deposition appears to synergize and accelerate tau pathological changes. Thus far, tau reduction strategies that have been tested in vivo have been examined in the setting of tau pathology without confounding amyloid-β deposition. To determine whether reducing total human tau expression in a transgenic model where there is concurrent amyloid-β plaque formation can still reduce tau pathology and protect against neuronal loss, we have taken advantage of the regulatable tau transgene in APP/PS1 × rTg4510 mice. These mice develop both neurofibrillary tangles as well as amyloid-β plaques throughout the cortex and hippocampus. By suppressing human tau expression for 6 months in the APP/PS1 × rTg4510 mice using doxycycline, AT8 tau pathology, bioactivity, and astrogliosis were reduced, though importantly to a lesser extent than lowering tau in the rTg4510 alone mice. Based on non-denaturing gels and proteinase K digestions, the remaining tau aggregates in the presence of amyloid-β exhibit a longer-lived aggregate conformation. Nonetheless, lowering the expression of the human tau transgene was sufficient to equally ameliorate thioflavin-S positive tangles and prevent neuronal loss equally well in both the APP/PS1 × rTg4510 mice and the rTg4510 cohort. Together, these results suggest that, although amyloid-β stabilizes tau aggregates, lowering total tau levels is still an effective strategy for the treatment of tau pathology and neuronal loss even in the presence of amyloid-β deposition.

Introduction

Alzheimer’s disease is the most common neurodegenerative disease, pathologically defined by two hallmarks: intraneuronal neurofibrillary tangles, composed of hyperphosphorylated tau, and extracellular plaques, which result from the accumulation of the amyloid-β peptide (Hyman et al., 1984; Braak and Braak, 1991). While both pathological hallmarks are required for the development of Alzheimer’s disease, the close correlation between tau deposition with cognitive decline (Arriagada et al., 1992; Nelson et al., 2012), in addition to genetic evidence that mutations in the tau gene (MAPT) alone can result in widespread neurodegeneration (Buee and Delacourte, 1999), highlights the importance of tau pathology in the pathogenesis of the disease.

One therapy proposed for Alzheimer’s disease is reducing total tau levels. Lowering endogenous tau in mice either genetically (Ittner et al., 2010; Vossel et al., 2010; Roberson et al., 2011; Li et al., 2014) or with a drug post-development (DeVos et al., 2013) is well tolerated with a minor Parkinson’s phenotype (Lei et al., 2012), though this appears to vary based on the background strain of the mouse (van Hummel et al., 2016). This endogenous mouse tau reduction is also protective against several amyloid-β induced deficits, such as hyperexcitability and cognitive decline (Roberson et al., 2007, 2011; Ittner et al., 2010; Vossel et al., 2010; Leroy et al., 2012). The lowering of human tau has been performed using a genetic model as well as with exogenously applied tools, such as siRNA and antisense oligonucleotides. In all scenarios, neuronal loss was prevented while tau pathology was able to be both prevented and reversed (Santacruz et al., 2005; Xu et al., 2014; DeVos et al., 2017).

However, all reports of human tau reduction have been in pure tauopathy models. While this mimics what may happen in diseases where tau is the lone pathological species, such as FTDP-17, these models do not reflect the micro-environment in the brain of Alzheimer’s disease. To address this gap, we used a newly characterized mouse model (Bennett et al., 2017) generated from a cross between the regulatable rTg4510 tauopathy line, which overexpresses a mutant form of human tau (Santacruz et al., 2005), and the APP/PS1 amyloid-β depositing mice (Jankowsky et al., 2004). This APP × rTg4510 line develops tau pathology and amyloid-β plaques in an age-dependent manner that has an accelerated tau pathology phenotype as compared to tau-alone expressing rTg4510 mice (Bennett et al., 2017). Using this model, which has both pathological hallmarks of Alzheimer’s disease, we sought to evaluate a tau reducing strategy by taking advantage of the regulatable mutant human tau transgene. By genetically lowering human tau in the context of amyloid-β plaques, we assessed (i) whether lowering human tau in an Alzheimer’s disease murine model is capable of reducing tau pathology and protecting against neuronal loss; and (ii) whether a tau reduction therapy is equally efficacious with and without concurrent amyloid-β deposition.

Materials and methods

Animals

All animal experiments were performed in accordance with the Massachusetts General Hospital’s (MGH) and McLaughlin Research Institute’s Institutional Animal Care and Use Committees. The mice were housed under a 12-h light/dark cycle and were given food and water ad libitum. B6.Cg-Tg(APPswe,PSEN1dE9)85Dbo/Mmjax mice (APP/PS1) obtained from Jackson Laboratory were crossed to the B6.Cg-Tg(Camk2a-tTA)1/Mmay tet transactivator (tTA) strain, which expresses tTA from the CamKIIα-tTA transgene in the murine forebrain (Mayford et al., 1996). B6.CK-tTA,APP/PS1 double transgenic males were crossed with dams from the tetracycline-responsive element line FVB-Tg(tetO-MAPT*P301L)4510/Kha/Jlws (rTg4510). The resulting cross produced mice with the experimental genotype (termed APP × rTg4510). Genotyping was performed pre-weaning using previously described protocols (Jankowsky et al., 2001, 2004; Santacruz et al., 2005). For the study herein, the following genotypes were studied: (i) APP × rTg4510 mice that carry the APP/PS1 array, the CamkIIα-tTA transactivator, and the rTg4510 tau responder transgene; (ii) rTg4510 mice that have both the CamkIIα-tTA and the rTg4510 tau transgenes; (iii) APP/PS1 only mice; and (iv) non-transgenic CamkIIα-tTA only mice.

At 6 months, half of the mice were placed on a doxycycline (DOX) diet, whereby mice received chow containing 200 mg/kg DOX (Fisher Scientific) ad libitum for 6 months to suppress the CamkIIα-tTA-driven human tau transgene expression, until the point of collection (Santacruz et al., 2005). The remaining mice received standard chow.

Immediately prior to euthanasia, mice were anaesthetized with 3–5% isoflurane for CSF collection. CSF was drawn through the cisterna magna and immediately frozen on dry ice (Barten et al., 2011). Five mice did not give adequate CSF volumes for enzyme-linked immunosorbent assay (ELISA) analysis (n = 2 naïve non-transgenic, n = 1 naïve rTg4510, n = 1 DOX rTg4510, n = 1 DOX APP × rTg4510). For tissue collection, mice were anaesthetized with isoflurane and perfused using chilled phosphate-buffered saline (PBS). The brain was removed and weighed. The left hemisphere was post-fixed in 4% paraformaldehyde (Electron Microscopy Sciences) and transferred to 30% sucrose 24 h later. The right hemisphere was microdissected into cortex and hippocampus and snap frozen on dry ice. All frozen tissue was stored at −80°C.

Human tissue selection

Human tissue samples from parahippocampal gyrus or frontal cortex were obtained from the Neuropathology Core of the Massachusetts Alzheimer’s Disease Research Center at MassGeneral Institute for Neurodegenerative Disease (Table 1). Brains were coronally sliced at the time of autopsy, flash frozen between metal plates on dry ice, and stored at −80°C. A 1-cm square of grey matter was dissected out and immediately homogenized. Cases had been formerly assessed by an MGH neuropathologist to generate a neuropathological diagnosis in addition to a neuritic plaque scoring system, developed by the Consortium to Establish a Registry for Alzheimer’s disease (CERAD; Morris et al., 1989). All cases also received a tau Braak rating based on the location of neurofibrillary tau tangles as seen with a total tau immunostain.

Quantitative real-time PCR

All mRNA analyses were done using quantitative real-time RT-PCR (qRT-PCR). Total RNA was extracted from the frontal cortex using the QIAGEN RNeasy® Kit. Messenger RNA was reverse transcribed and amplified using the EXPRESS One-Step SuperScript® qRT-PCR Universal Kit (Invitrogen) with TaqMan™ probe technology. All qRT-PCRs were run and analysed on the CFX96 Touch Real-Time PCR Detection System (Bio-Rad). Total human and mouse tau expression levels were normalized to mouse glyceraldehyde 3-phosphate dehydrogenase (Gapdh) mRNA levels. Relative expression levels were generated using the ΔΔCt method. Primer/probe sequences: Human total tau (MAPT): Forward 5'-AGAAGCAGGCATTGGAGAC-3'; Reverse 5'-TCTTCGTTTTACCATCAGCC-3'; Probe 5'-/56-FAM/ACGGGACTGGAAGCGATGACAAAA/MGBNFQ/-3′, Mouse total tau (Mapt): Forward 5′-GAACCACCAAAATCCGGAGA-3′; Reverse 5′-CTCTTACTAGCTGATGGTGAC-3’; Probe 5'-/56-FAM/CCAAGAAGG TGG CAG TGG TCC/MGBNFQ/-3′, GAPDH: Forward 5′–TGCCCCCATGTTGTGATG-3′; Reverse 3′-TGTGGTCATGAGCCCTTCC-3′; Probe 5′/56-FAM/AATGCATCCTGCACCACCAACTGCTT/MGBNFQ/3′ (Thermo Fisher Scientific).

Preparation of brain homogenate

Thawed mouse or human cortex was placed in 500 µl of PBS + protease inhibitor (Roche) in a 2 ml glass dounce homogenizer and dounce homogenized with 30 up/down strokes on ice by hand. The lysate was centrifuged at 3000g for 10 min at 4°C and supernatant collected and aliquoted so that no sample was frozen/thawed more than three times. A bicinchoninic acid assay (Thermo Scientific Pierce) was performed to determine protein concentration.

SDS-PAGE and western blot

For total tau analysis, 5–10 µg of total protein per well was loaded on 4–12% Bis-Tris SDS-PAGE gels (Invitrogen) and run in MES buffer (Invitrogen). Proteins were transferred to Immobilon PVDF membrane (EMD Millipore) and incubated overnight at 4°C with anti-mouse HT7 total tau antibody (amino acid 159-163) (1:1000, Thermo Fisher Scientific) and anti-rabbit GAPDH antibody (1:2000, Abcam) in 1:1 Odyssey blocking buffer:distilled water. Blots were washed three times for 10 min in Tris-buffered saline + 0.25% Tween (TBS-T), incubated with infrared secondary anti-mouse 800 and anti-rabbit 680 (1:2000, Licor) antibodies in blocking buffer, washed three times for 10 min TBS-T, and imaged on an Odyssey Infrared Imaging System (Licor). Three carry-over samples were used for normalization across blots. Blots were converted to greyscale and densitometry analysis was performed in ImageJ (NIH v1.51 n).

ELISAs

ELISAs were performed using 3000g PBS soluble cortex homogenates or CSF. The Invitrogen total tau and pS396 tau ELISAs (Invitrogen) were run following the manufacturer’s instructions. Plates were developed using the Wallac plate reader (Perkin Elmer) at 450 nm. For the Meso Scale Diagnostics total tau and pT231 tau multiplex ELISA (Meso Scale Diagnostics), the manufacturer’s protocol was followed. Plates were developed using the MESO QuickPlex SQ 120 Plate Reader (Meso Scale Diagnostics). To analyse amyloid-β peptide 1-42 (amyloid-β_1-42) in cortical homogenates, the human/rat amyloid-β42 ELISA kit (Wako) was used that measures amyloid-β(x-42) peptides. Plates were developed using the Wallac plate reader (Perkin Elmer). For all ELISAs, samples were fit to an eight-point standard curve. All ELISA values were adjusted for protein concentration values to determine the amount of analyte per weight of tissue (volume for CSF).

Semi-denaturing detergent agarose gel electrophoresis

Semi-denaturing detergent agarose gel electrophoresis (SDD-AGE) blots were carried out as previously described with minor modifications (Halfmann and Lindquist, 2008; Sanders et al., 2014). Agarose (1.5%) was dissolved in buffer G (20 mM Tris-Base, 200 mM glycine) and 0.02% SDS added. Lysate (15 µg) was incubated with 0.02% SDS sample buffer for 7 min at room temperature prior to loading. The SDD-AGE was run using Laemmli buffer (Buffer G + 0.1% SDS) at 30 V for 16 h at 4°C. The transfer was performed using capillary action using 20 pieces of thick Whatman® paper (GB 005) and eight pieces of medium high absorbent Whatman® paper (GB 003) to Immobilon PVDF (Millipore) membrane at 4°C for 24 h using TBS. Following transfer, the membrane was blocked in 5% non-fat dry milk TBS-T for 1 h and probed for total tau (anti-rabbit polyclonal; 1:4000; Abcam) overnight at 4°C in 5% non-fat dry milk TBS-T. The membrane was washed three times with TBS-T, probed with a HRP-conjugated goat anti-rabbit IgG secondary antibody (1:4000; Thermo Fisher) for 1.5 h at room temperature in 5% non-fat dry milk TBS-T, washed three times in TBS-T, and detected using chemiluminescent HRP substrate (Thermo Fisher) and film (GE Healthcare). To quantify, the low and high molecular weight portions of the blot were quantified separately using ImageJ. Previous reports of high and low molecular weight tau run on SDD-AGE show an accumulation of low molecular weight tau at the bottom of the gel with smears of high molecular weight tau oligomers at the top (Yanamandra et al., 2013; Sanders et al., 2014; Takeda et al., 2016; DeVos et al., 2017). For full lane analysis, 13 equal size bins were generated to span all samples and densitometry analysis for each bin of each sample was performed using the ImageJ gel quantification tool. For per cent tau quantification, the densitometry value for each bin was divided by the sum of all 13 densitometry values for each sample.

Proteinase K digestion

Proteinase K digestion was carried out as previously described with minor modifications (Falcon et al., 2015). Lysate was diluted in proteinase K buffer (50 mM Tris-HCL pH 8.0, 1 mM CaCl₂, 3 mM DTT, and 2 M urea) and proteinase K (Thermo Fisher Scientific) added to the appropriate concentration (0–100 µg/ml). Samples were incubated in proteinase K/proteinase K buffer for 30 min at room temperature. The digestion was halted with 5 mM phenylmethylsulfonyl fluoride (Sigma-Aldrich). NuPAGE® LDS Sample Buffer (4× concentration) (Invitrogen) and 10× NuPAGE® Sample Reducing Agent (Invitrogen) was added to 5 µg of digested lysate and boiled for 10 min at 95°C and run on 4–12% Bis-Tris SDS-PAGE gels (Invitrogen) in MES buffer (Invitrogen) and transferred to PVDF membrane. Blots were incubated with total tau antibody (anti-rabbit polyclonal, 1:2000, Abcam) overnight at 4°C. Blots were washed three times for 10 min with TBS-T and incubated with secondary anti-rabbit 680 (1:1000) for 1.5 h at room temperature and imaged on the Licor imaging system (Odyssey). For quantification, images were converted to greyscale in ImageJ. Using the gel quantification densitometry tool, the full lane of tau was quantified as well as tau fragments <14 kDa for mouse and >28 kDa for human lysate, both of which had been identified a priori based on the proteinase K dose escalation blots. The <14 kDa and >28 kDa band intensity was divided by the total tau intensity.

Size-exclusion chromatography

PBS-soluble brain lysates from the frontal cortex of two control and two Alzheimer’s disease cases (Table 1) were separated by size-exclusion chromatography (SEC) on single Superdex200 10/300GL columns (GE Healthcare) in PBS at a flow rate of 0.5 ml/min, with an AKTA purifier 10 (GE Healthcare). Each brain lysate was diluted to the same concentration and filtered through a 0.2 µm membrane filter before loading onto the SEC column. Each fraction encompassed 500 µl and based on our previous work (Takeda et al., 2015) and molecular weight markers run through the SEC, Fraction 2 contains high molecular weight tau and Fraction 14 contains low molecular weight tau.

Immunohistochemistry

Half mouse brains were cut coronally into 40 µm sections with a freezing sliding microtome. All sections were stored in glycerol cryoprotectant solution (30% glycerol in PBS) at −20°C. All immunohistochemistry staining—for tau and amyloid-β—for all mice was done at the same time and imaged together to reduce staining batch variability. Six coronal sections 400 -µm apart were washed three times for 15 min with TBS and incubated with 0.3% hydrogen peroxide for 10 min at room temperature. Sections were blocked in 3% non-fat dry milk in TBS and 0.25% Triton™ X-100 (TBS-X), followed by incubation at 4°C overnight in 1% non-fat dry milk TBS-X with either biotinylated anti-mouse AT8 antibody (1:500; Thermo Fisher Scientific), which recognizes phosphorylated tau at Serine 202 and Threonine 205 (Goedert et al., 1995) or anti-rabbit total human amyloid-β (1:1000; #18584 Immuno-Biological Laboratories) that identifies human amyloid-β 40, 42, and 43 fragments. The next day, sections were washed three times for 15 min with TBS. For the amyloid-β stain, sections were then incubated with goat anti-rabbit biotinylated secondary antibody (1:1000; Thermo Fisher Scientific) in TBS-X for 1 h at room temperature. Amyloid-β incubated sections were washed three times for 15 min with TBS. ABC Elite solution (1:400) was made (Vector) and incubated for 30 min at room temperature. All sections were incubated in the ABC Elite solution for 2 h at room temperature and washed three times for 15 min with TBS. Sections were transferred to a 12-well plate with nets to ensure all sections were exposed to 3,3′diaminobenzidine (DAB) equally. This DAB solution with the Nickel addition was made (Vector) and sections developed. Sections were transferred to distilled water for 5 min and then washed three times for 15 min with TBS before mounting, dehydrating in ascending ethanol/xylene, and coverslipping with Cytoseal™ XYL (Thermo Fisher). All sections were imaged using brightfield at ×20 on the Olympus NanoZoomer 2.0-HT slide scanner (Hamamatsu).

For the human brain sections, the contralateral hemisphere of all cases was post-fixed in formalin and the parahippocampal gyrus was paraffin-embedded and cut into 7–8 -µm thick sections and mounted on slides. Sections were deparaffinized with xylene followed by descending ethanol series and incubated in 3% hydrogen peroxide prior to incubation with 90% formic acid for 5 min. Sections were then rinsed with TBS and incubated with a mouse anti-amyloid-β antibody (1:200, Clone 6F/3D, Dako) for 1 h at room temperature. Staining was then visualized with the Vectastain rabbit IgG horse-radish peroxidase kit (VectorLabs) and Vector peroxidase substrate kit DAB (VectorLabs). Slides were finally dehydrated and coverslipped with Cytoseal™ XL. Slides were imaged using brightfield at ×10 on the Olympus BX51.

Immunofluorescence

Sections were washed three times for 15 min in TBS and blocked in 3% normal goat serum (Vector Laboratories) in TBS-X for 1 h at room temperature. Sections were then incubated with primary antibodies mouse anti-GFAP (1:1000; Abcam), rabbit anti-NeuN (1:1000; Cell Signaling), and/or rabbit anti-Iba-1 (1:1000; Wako) in blocking solution overnight at 4°C. The next day, sections were washed in TBS and incubated with the appropriate secondary antibodies conjugated to Alexa 488, Cy3, or Cy5 (1:400) for 1.5 h at room temperature. In a set of sections following secondary incubation, slices were then incubated in freshly made 0.005% thioflavin-S (50% ethanol/50% TBS) for 5 min, then rinsed three times for 30 s in 80% ethanol, 20% TBS, and eventually three times for 10 min in TBS. All sections were mounted on slides, coverslipped using DAPI containing Fluoromount-G® (SouthernBiotech) or, in thioflavin-S-stained sections, Fluoromount-G®, and sealed with clear nail polish.

Tau and amyloid-β pathology quantification

To quantify thioflavin-S positive pathology, we performed stereological quantifications using the Computer Assisted Stereological Toolbox version 2.3.1.5 (Olympus). Thioflavin-S positive neurofibrillary tangles and amyloid-β plaques were simultaneously counted in three serial brain sections containing hippocampus 400 µm apart. Neurofibrillary tangles and amyloid-β plaques are morphologically very distinct and easily discernible from each other. Ten per cent of cortex was counted using a random meander sampling algorithm allowing for per mm² count. For AT8 and amyloid-β immunohistochemistry pathology, three to four brain sections containing hippocampus, starting between bregma −1.5 mm and −2.0 mm and 400 µm apart, were used from each mouse. Using the NDP viewer software (Hamamatsu), each section was exported and saved as .tiff image file. For AT8 pathology analysis, image files converted to 8-bit greyscale and re-saved in ImageJ. Using the freehand selection tool on the greyscale images, the outline of the cortex and hippocampus was drawn. A uniform threshold value was then applied followed by the ‘analyse particles’ task to determine the per cent covered by positive signal. The average per cent coverage of all three to four sections was used as the final per cent AT8 value for that mouse. For amyloid-β pathology, all three brain sections were captured in the same field of view using the NDP viewer software and exported at ×10 as a single .tiff image per mouse. Images were processed through ImageJ with the biovoxxel plugin using the following batch process script in beanshell:

• run(“8-bit”);

• run(“Subtract Background…”, “rolling=40 light sliding”);

• run(“Invert”);

• run(“OtsuThresholding 8Bit”);

• //setThreshold(33, 255);

• setOption(“BlackBackground”, false);

• run(“Make Binary”, “thresholded remaining black”);

• run(“Extended Particle Analzyer”, “extent=0.00-.99 circularity=0.17-1.00 compactness=0-1.00 aspect=0-10000max_feret=0-2000 show=Nothing redirect=None keeP = None display summarize add exclude include”);

• run(“Measure”);

• dir = getDirectory(“image”);

• name = getTitle;

• index = lastIndexOf(name, “.”);

• if (index!=-1) name = substring(name, 0, index);

• name = name + “.xls”;

• saveAs(“Measurements”, dir+name);

• print(dir+name)

To account for total section area, the same three sections used for plaque quantification were outlined in the NDP viewer software to generate a single section area (mm²) value. The number of plaques determined by the ImageJ Fiji analysis was divided by the section area to obtain a final plaques/area (mm²) value for each mouse.

GFAP and Iba1 quantification

For GFAP quantification, the cortex of three brain sections spaced 800 µm apart starting at bregma −1.5 mm were imaged for GFAP on the Olympus BX51 for all mice at the same exposure. The images were uploaded into ImageJ, the cortex outlined, and images converted to 8-bit greyscale. The same threshold value was applied to all images and then the ‘analyse particles’ function was applied to measure the per cent area covered by positive GFAP signal. The average of the three sections was determined for one value per mouse. For Iba1 quantification, three brain sections spaced 800 µm apart, starting at bregma −2.0 mm, were imaged at ×20 for fluorescence on the Olympus Nanozoomer 2.0-HT (Hamamatsu). Using the NDP viewer software, each brain section was exported at ×10. Images were uploaded into ImageJ and background subtracted using a rolling ball radius of 40 pixels. Images were then converted to 8-bit greyscale. Using the freehand selection tool, the outline of the cortex was drawn. A uniform threshold value was then applied followed by the ‘analyse particles’ task to determine the per cent covered by positive Iba1 signal. The average per cent coverage of all three sections was used as the final Iba1 per cent coverage for that mouse.

Hippocampal and cortical volume analysis

Beginning at the anterior tip of the hippocampus (bregma −1.0 mm), six sections were taken 400 -µm apart, mounted and imaged at ×20 for brightfield on the Olympus Nanozoomer 2.0-HT (Hamamatsu). Using the NDP viewer software, the hippocampus and cortex in each section was outlined to obtain the total hippocampal and cortical area for each section. Each of the hippocampal and cortical areas was then multiplied by 400 µm to obtain final hippocampal and cortical volumes in mm³ for each mouse.

NeuN count in CA1

Two full hippocampi images per mouse spaced 800 -µm apart were imaged for NeuN and DAPI on the Olympus BX51. Corresponding NeuN and DAPI images were merged and a rectangle spanning 200 µm was drawn over CA1 in ImageJ. The same size rectangle was used for each section. The numbers of NeuN/DAPI positive nuclei were counted within the defined CA1 region. The number of NeuN positive DAPI nuclei for the two sections was averaged to generate a single value per mouse.

Tau seeding assay

In vitro tau seeding activity was measured as previously described (Holmes et al., 2014). For the longitudinal assessment, HEK293 cells stably expressing the repeat domain of tau with the P301S mutation (TauRD^P301S) fused with cyan fluorescent protein (CFP) and yellow fluorescent protein (YFP) were plated in poly-d-lysine coated 96-well plates (Corning) at 40 000 cells/well. The next day, cells were transduced with 0.1 µg/well brain lysate plus 1% Lipofectamine® 2000 (Invitrogen) diluted in Opti-MEM^™. Every 3–6 h for 24 h, one well per sample was collected. At collection, cells were trypsinized and transferred to a 96-well U-bottom plate (Corning) with chilled Dulbecco’s modified Eagle medium 10% foetal bovine serum media. Cells were pelleted at 1000g, resuspended in 2% paraformaldehyde (Electron Microscopy Services) for 10 min and pelleted again. Cells were resuspended in chilled PBS and immediately run on the MACSQuant® VYB (Miltenyi) flow cytometer. CFP and Forster resonance energy transfer (FRET) were measured by exciting the cells with the 405-nm laser and reading fluorescence at the 405/50 and 525/50 nm filters, respectively. To quantify the FRET signal, a bivariate plot of FRET versus the CFP donor was created. Using cells that received Lipofectamine® alone as a marker of the FRET-negative population, the integrated FRET density was calculated (percentage of FRET-positive cells multiplied by median fluorescence intensity of FRET-positive population). Forty thousand cells per well were analysed. Data analyses were performed using the MACSQuantify software (Miltenyi). Representative images of tau aggregates along the time course were captured using the ZOE Fluorescent Cell Imager (Bio-Rad) using the GFP excitation/emission filters.

For the per cell analysis, the cells were treated with 0.1 µg lysate/well with 1% Lipofectamine^®. Cells were incubated with lysate/Lipofectamine complexes in Opti-MEM™ for 24 h and collected as above. Cells were resuspended in flow cytometry buffer containing HEPES-buffered RPMI medium (without phenol red). Image flow cytometry was performed using the ImageStream®X Mark II imaging flow cytometer (Amnis Corporation) equipped with a 40× objective. CFP:YFP FRET was measured from images obtained by 405 nm laser excitation on the ImageStream® using the FRET (YFP:CFP) Intensity Ratio feature. To quantify FRET, CFP bleed-through into the YFP and FRET channel was compensated using IDEAS software (version 6.2). For each experiment, 5 × 10⁵ events per sample were collected and analysed. To quantify the number of FRET-positive tau inclusions/spots in each FRET-positive cell, single cells that contained high and low tau spot counts were identified and gated first. Then, to determine the number of FRET-positive tau inclusions/spots in each cell, masks were created based on the previous high and low tau spot count gates to identify the spot region of interest. The number of individual masks in a cell was enumerated using the Spot Count feature in the IDEAS software per sample.

Experimental design

For selection of mice in the study, treatment groups were gender-, age-, and litter-matched. For sample size calculations, a main readout of treatment efficacy was identified a priori as tau pathology and seeding. Thus, we derived the mean and standard deviations from the APP × rTg4510 12-month cortical AT8 pathology and tau seeding values from Bennett et al. (2017). We planned a study of continuous response variable from independent control (APP × rTg4510 naïve) and experimental (APP × rTg4510 DOX) subjects with one control per experimental subject. Using the mean and standard deviations (SDs) from Bennett et al. (2017) and an effect size of 0.5 based on previous literature of the effect of tau suppression in rTg4510 (Polydoro et al., 2013; Holmes et al., 2017) and PS19 (DeVos et al., 2017) mice on tau pathology and seeding, we performed a power analysis that determined we would need to study at least six experimental subjects and six control subjects to be able to reject the null hypothesis that the tau pathology and seeding means of the APP × rTg4510 DOX and APP × rTg4510 naïve groups are equal with power 0.8. The error probability associated with this test of the null hypothesis is 0.05 (alpha). To ensure objective quantification, the following experiments were blinded by a third-party concealment of image file names and microscope slides: AT8 tau pathology quantification (hippocampus and cortex), thioflavin-S pathology stereology counts (neurofibrillary tangles and amyloid-β plaques), amyloid-β plaque pathology quantification (plaque number and plaque size), GFAP per cent coverage, Iba1 per cent coverage, hippocampal and cortical volume, and NeuN counts in the hippocampus.

Statistical analysis

The data were analysed for statistical significance using the graphing program GraphPad Prism 6. The number of mice used throughout the manuscript are depicted in Fig. 1A. The D’Agostino and Pearson omnibus normality test was applied to all datasets to test for normal distribution where possible. For the proteinase K digest and amyloid-β_1-42 ELISA quantification, a two-tailed Student’s t-test was used. A two-way repeated measures ANOVA with Sidak post hoc analysis was used to analyse the SDD-AGE per cent tau per fraction, the longitudinal tau seeding analysis, and the amyloid-β plaque size distribution. Pearson’s correlation was used for the CSF versus brain tau direct comparison, fitted with the line of best fit and 95% confidence interval (CI) bands. For all other experiments, two-way ANOVA with Sidak post hoc analyses were used. Graphical data are represented as individual points for each mouse with mean ± standard error of the mean (SEM) overlapping or, in the case of the SDD-AGE total and per cent tau graphs, longitudinal seeding data and amyloid-β plaque size distribution, a single point representing the mean with error bars representing ± SEM.

Results

Doxycycline treatment effectively reduces human MAPT mRNA expression levels

To test whether lowering human tau in the presence of amyloid-β plaques is protective, we crossed the amyloid-β depositing APP/PS1 mice (Jankowsky et al., 2003) with the regulatable rTg4510 tauopathy mice (Santacruz et al., 2005), termed APP × rTg4510. In the present study, human tau was lowered for 6 months, starting at 6 months of age, in half of the study mice by feeding the mice DOX containing chow (Fig. 1A). Half of the non-transgenic and APP/PS1 (APP) mice were treated with DOX to control for any off-target effects due to 6-month exposure to DOX. At sacrifice, the frontal cortex was collected and total RNA extracted to determine if the DOX treatment successfully reduced human tau transgene expression. Human MAPT (tau) mRNA levels, not mouse Mapt (tau) mRNA levels, were significantly lowered in both rTg4510 and APP × rTg4510 DOX mice compared to their naïve genotype controls (Fig. 1B and C), though it is worth noting that the human MAPT/tau transgene was not 100% repressed with the DOX treatment.

Lowering of human tau expression reduces total human tau protein in the brain and CSF

The cortex was homogenized in PBS and the supernatant was run on western blot and probed for total human tau and GAPDH (Fig. 2A). ImageJ densitometry analyses shows a significant decrease in total human tau in both rTg4510 and APP × rTg4510 DOX mice when compared to naïve genotype controls with no difference in total human tau protein between the two DOX-exposed genotypes (Fig. 2B). Similarly, when total, pS396, and pT231 human tau protein levels were measured using ELISA (from both Invitrogen and Meso Scale Diagnostics) a significant decrease in tau expression can be seen in DOX-treated rTg4510 and APP × rTg4510 mice when compared to their naïve genotype controls (Supplementary Fig. 1A–D). Immediately prior to euthanasia, CSF was collected and human tau levels measured on ELISA (Invitrogen). Notably, CSF tau levels measured in the rTg4510 and APP × rTg4510 naïve mice are equal to previously analysed 4-month naïve rTg4510 and APP × rTg4510 mice (Bennett et al., 2017), suggesting CSF tau release is in part an active process and not just a result of widespread neuronal loss. Similar to brain, total human tau levels in the CSF were significantly lowered in both rTg4510 and APP × rTg4510 DOX mice compared to their respective 12-month naïve treatment cohorts (Fig. 2C). This results in a significant direct correlation between total human brain tau and CSF tau (Supplementary Fig. 2). This effect has been seen previously with lowering endogenous mouse tau (DeVos et al., 2013) and endogenous non-human primate tau (DeVos et al., 2017), suggesting CSF tau measurement is a potential biomarker for a tau-lowering therapy.

Tau aggregation in the presence of amyloid-β adopts a unique conformation

Previous reports have suggested that amyloid-β is capable of altering tau biology in mice, rats, non-human primates and humans (Sigurdsson et al., 1997; Price and Morris, 1999; Gotz et al., 2001; Lewis et al., 2001; Delacourte et al., 2002; Ribé et al., 2005; Bolmont et al., 2007; Hurtado et al., 2010; Seino et al., 2010; Mairet-Coello et al., 2013; Forny-Germano et al., 2014; Héraud et al., 2014; Barthélemy et al., 2016; Manassero et al., 2016; Bennett et al., 2017; He et al., 2018). To analyse tau aggregates biochemically, brain lysate was first run on an SDD-AGE gel to assess the level of high and low molecular weight tau protein (Fig. 2D). We have previously demonstrated a mild increase in high molecular weight tau as early as 4 months of age in APP × rTg4510 mice (Bennett et al., 2017), though the amount of high molecular weight tau is substantially lower at this age than at the 12 month time point used in the present study. Despite the increase in the amount of high molecular weight tau in the older mice, the SDD-AGE shows a significant decrease in both high and low molecular weight tau protein in the DOX treatment groups (Supplementary Fig. 3). The absolute amount of tau for each sample was then quantified in a series of bins down the length of the blot and plotted (Fig. 2E), highlighting not only the decrease in total tau in the DOX cohorts, but also the different patterns in total tau. On the blot, there was a noticeable high molecular weight band in the APP × rTg4510 DOX samples, which was reduced or absent in the rTg4510 DOX group. To further analyse this, the total amount of tau per sample was set to 100% and the per cent per bin along the gel was plotted for both the DOX cohorts (Fig. 2F) and the naïve cohorts (Supplementary Fig. 4A). This revealed a relative increase in the high molecular weight segments with an accompanying decrease in low molecular weight tau in the APP × rTg4510 mice when compared to rTg4510 mice in both the naïve and DOX conditions, though the effect was more prominent in the DOX cohort. These data suggest that amyloid-β enhances soluble tau aggregation, either through direct (tau aggregates could be in complex with amyloid-β plaques) or indirect interaction. We speculate that this is a long-lived species, given that the human tau transgene had been suppressed for 6 months.

Additionally, the two total human tau ELISAs that were used recognize different regions of tau, with the Invitrogen ELISA measuring a very N-terminal portion of tau (Barten et al., 2011) that has been reported to not recognize aggregated tau as efficiently (Acker et al., 2013) and the Meso Scale Diagnostics human tau ELISA detecting tau between amino acids 150 and 200 (unpublished observations). In the naïve treated mice, while there is a noticeable decrease in Invitrogen total tau in the APP × rTg4510 naïve mice compared to rTg4510 naïve mice, that difference does not exist in the Meso Scale Diagnostics total tau ELISA, perhaps reflecting a difference in conformation and exposed tau epitopes in the two genotypes (Supplementary Fig. 1A–D). To explore whether the Invitrogen and Meso Scale Diagnostics total human tau ELISAs recognize the same fractions of tau equally, we isolated high and low molecular weight tau fractions using size-exclusion chromatography from human control and Alzheimer’s disease frontal cortex lysates (Takeda et al., 2015) (Table 1) and measured each fraction on both the Invitrogen and Meso Scale Diagnostics total human tau ELISAs (Supplementary Fig. 1E–H). While both ELISAs were able to similarly detect human tau in the low molecular weight fraction, the Meso Scale Diagnostics ELISA was capable of recognizing high molecular weight tau with much higher sensitivity, suggesting that the N-terminus tau epitope targeted by the Invitrogen ELISA is either masked by the conformation of the tau molecule or removed following a tau cleavage event. If the Invitrogen ELISA recognizes low molecular weight over high molecular weight tau while the Meso Scale Diagnostics ELISA is capable of recognizing both low and high molecular weight tau, as shown by the size-exclusion chromatography data, it would suggest that the APP × rTg4510 mice have a higher high:low molecular weight tau ratio than rTg4510 mice. This corroborates the SDD-AGE data (Fig. 2F and Supplementary Fig. 4A).

To analyse the possibility of a unique confirmation of tau in the presence of amyloid-β further, we used a protease digestion that others have reported to differentiate tau conformations (Falcon et al., 2015). As a pilot, a rTg4510 and APP × rTg4510 sample were both digested with varying concentrations of proteinase K and the resulting tau banding patterns observed on western blot. Interestingly, there were more fragmented bands in the rTg4510 sample at 1 µg/ml of proteinase K than in the APP × rTg4510 sample (Fig. 2G). Using this a priori identified difference in tau fragmentation, all human tau containing samples were then incubated with 1 µg/ml proteinase K and probed for tau on western blot (Fig. 2H, I and Supplementary Fig. 4). The ratio of fragmented tau to total tau was calculated and plotted for each sample. In the DOX cohort (Fig. 2I) and the naïve cohort (Supplementary Fig. 4), there were significantly more tau fragments generated by proteinase K-digestion at 1 µg/ml in the rTg4510 mice than in amyloid-β expressing APP × rTg4510 mice, suggesting that the conformation of tau in aggregates in the presence of amyloid-β may be more resistant to protease degradation.

Human brain derived tau aggregates are more protease resistant in the presence of amyloid-β

We have previously compared human neuropathological samples that either had plaques and tangles, or tangles only, to explore the analogies between the experimental mouse models and human phenotypes (Bennett et al., 2017). To test the idea of amyloid-β associated tau protease resistance in the human brain, brain lysate was generated from the parahippocampal gyrus from human cases with Braak III/IV tau staging that did (amyloid-β+) or did not (amyloid-β−) have co-occurring amyloid-β neuritic plaques (Table 1 and Supplementary Fig. 5). An amyloid-β− and amyloid-β+ case were both digested with varying concentrations of proteinase K and run on western blot (Fig. 2J). While there were not the same <14 kDa tau digest banding patterns in the human lysate, there was a visible increase in the persistence of >28 kDa tau bands in the amyloid-β+ case at 2.5 µg/ml proteinase K. All human lysates were then incubated with 2.5 µg/ml proteinase K and probed for tau on a western blot (Fig. 2K and L). The ratio of >28 kDa tau to total tau was calculated and plotted for each sample, revealing that amyloid-β+ cases had a significantly higher ratio of >28 kDa protease resistant tau when compared to amyloid-β− cases. These data from human neuropathological material, combined with the data from mouse models, suggest that tau aggregates in the presence of amyloid-β adopt a more proteinase K resistant conformation, though reducing total human tau in the APP × rTg4510 mice is still capable of significantly lowering these proteinase K resistant tau aggregates.

Human tau repression decreases tau pathology throughout the hippocampus and cortex

A major pathological hallmark in rTg4510 mice is the presence of AT8 positive tau accumulations. At 12 months of age, we have previously shown that both rTg4510 and APP × rTg4510 mice have extensive AT8 tau pathology throughout the cortex and hippocampus (Bennett et al., 2017). When total human tau is lowered, there is less AT8 positivity in both the rTg4510 and APP × rTg4510 DOX cohorts when compared directly to their naïve genotype controls (Fig. 3A–C). Despite a reduction in tau pathology in the APP × rTg4510 DOX mice as compared to APP × rTg4510 naïve mice, more AT8 tau pathology remains in the APP × rTg4510 DOX mice when compared to the human APP negative rTg4510 DOX mice, perhaps because of a more stable conformation of tau aggregate in the presence of amyloid-β (Fig. 2 and Supplementary Fig. 4). When tau neurofibrillary tangles were assessed by thioflavin-S—a compound that recognizes beta-sheet structure in protein aggregates—there was a reduction in the number of thioflavin-S neurofibrillary tangles in both the rTg4510 DOX and APP × rTg4510 DOX cohorts (Fig. 3D and E). Unlike with AT8 pathology where the per cent area covered by pathology was assessed, the neurofibrillary tangle burden was measured by counting the number of neurofibrillary tangles using a stereology setup since thioflavin-S also binds to the beta sheets in amyloid-β plaques. Interestingly, unlike AT8 pathology analyses, there was an equal level of neurofibrillary tangle reduction in both the rTg4510 DOX and APP × rTg4510 DOX mice, bringing the neurofibrillary tangle numbers down close to what we have previously reported for rTg4510 and APP × rTg4510 4–6-month-old mice (Bennett et al., 2017).

Human tau reduction decreases tau bioactivity in the presence of amyloid-β

To measure tau bioactivity levels, we employed the stably expressing HEK293 cells that express the mutant P301S repeat domain of tau (Tau^RD) fused with CFP and YFP. In the presence of tau aggregates, Tau^RD-CFP and Tau^RD-YFP are recruited to the aggregate where they are then in close enough proximity to exhibit FRET, which can be detected on a flow cytometer for quantification.

Because low levels of rTg4510 and APP × rTg4510 lysate yield high bioactivity, we collected treated cells every 3 h to ensure we were not reading out bioactivity at a saturated point. Lysates were applied to the stably expressing cells and collected every 3–6 h for 24 h. Similar to our previous experiments at 12 months of age, naïve rTg4510 and APP × rTg4510 mice showed similar levels of tau bioactivity (Bennett et al., 2017) (Fig. 4). Both the rTg4510 DOX and APP × rTg4510 DOX cohorts showed a significant reduction in bioactivity as compared to their respective naïve genotype controls, starting at 15 h post-treatment for rTg4510 DOX and 18 h post-treatment for APP × rTg4510 DOX. At the final 24-h time point, there was a significant increase in bioactivity in the APP × rTg4510 DOX mice when compared to the non-amyloid-β expressing rTg4510 DOX mice, echoing the phenomenon we have seen previously with younger rTg4510 and APP × rTg4510 mice (Bennett et al., 2017). We also assessed the number and size of FRET-positive inclusions that were induced in the HEK293 biosensor cells. Cells treated with rTg4510 lysate exhibited an increased number of smaller FRET-positive inclusions per cell as measured by total FRET mean fluorescent intensity and spot count as compared to APP × rTg4510 lysate treated cells (Supplementary Fig. 6). This may serve as a proxy as to the types of tau aggregates in the lysate that subsequently recruit Tau^RD-CFP and -YFP monomer, with rTg4510 lysate containing smaller and less stable aggregates and APP × rTg4510 lysate possessing larger aggregates with increased stability. These bioactivity data combined with the previous biochemical (Fig. 2) and pathological (Fig. 3) data together suggest that it may be the more stable, AT8-aggregated, high molecular weight tau that is the more bioactive tau species as compared to the thioflavin-S positive neurofibrillary tangles.

The presence of aggregated tau decreases the number of amyloid-β plaques

While tau-containing neurofibrillary tangles are one of the major pathologies in Alzheimer’s disease, the other is the formation of extracellular amyloid-β plaques (Masters et al., 1985; Braak and Braak, 1991). In the previous characterization of the APP × rTg4510 mice, a decrease in the number of cortical amyloid-β plaques was observed as compared to APP/PS1 mice (Bennett et al., 2017). To determine the effect of tau lowering on amyloid-β pathology, we first measured the amount of PBS soluble amyloid-β_1-42 peptide in the brain homogenate (Iwatsubo et al., 1996). We did not observe any changes in the amount of soluble amyloid-β_1-42 peptide levels per gram of tissue between either the APP × rTg4510 naïve and DOX cohorts, nor in non-tau containing APP/PS1 mice (Fig. 5A). Next, brain sections were stained with a pan anti-human amyloid-β antibody and both the number of amyloid-β antibody-positive plaques per mm² as well as the size distribution of plaques counted was assessed (Fig. 5B–D). No plaques were detected in non-transgenic or rTg4510 mice (Fig. 5B and Supplementary Fig. 7). Both the plaque number and size were reduced in the APP × rTg4510 mice when compared to non-tau containing APP/PS1 mice. When human tau was reduced, both the number and size of plaques were partially restored. As a second measure of amyloid-β plaque burden, the number of cortical thioflavin-S-positive plaques were counted. Plaque number was reduced in the APP × rTg4510 naïve mice when compared to APP/PS1 mice, as previously reported, and was partially restored when human tau was lowered (Fig. 5E and F). This partial restoration in plaque number as measured by both amyloid-β antibody and Thioflavin-S in the APP × rTg4510 DOX cohort is still significantly lower than the APP/PS1 mice. This phenomenon of decreased plaque number in the APP × rTg4510 naïve mice compared to APP/PS1 mice may be due to a combination of an increase in astrogliosis and/or microgliosis in the presence of tau pathology, in addition to substantial neuronal loss, which would decrease the amount of APP overexpressing cells in the brain, arguably diminishing amyloid-β plaque deposition.

Lowering human tau in the presence of amyloid-β significantly reduces astrogliosis but not microgliosis

In Alzheimer’s disease, inflammation plays a major role in the disease pathogenesis (Beach et al., 1989; Serrano-Pozo et al., 2013; Heppner et al., 2015). To test the effect of lowering total human tau on astrogliosis and microgliosis, sections were stained for GFAP and ionized calcium-binding adapter molecule 1 (Iba1), respectively. Astrogliosis, as measured by an increase in GFAP per cent coverage, is increased throughout the cortex in both the presence of amyloid-β plaques in APP/PS1 mice and tau pathology in rTg4510 mice (Fig. 6A and B). APP/PS1 mice treated with DOX for 6 months did not have significantly different levels of astrogliosis than naïve APP/PS1 mice, suggesting that DOX alone did not have an impact on amyloid-β-driven astrogliosis. APP × rTg4510 naïve mice showed an additive effect for astrogliosis when compared to APP/PS1 naïve and rTg4510 naïve groups, similar to our previous report (Bennett et al., 2017). Reducing human tau alone with DOX in the rTg4510 mice significantly reduced astrogliosis. Further, compared to the naïve APP × rTg4510 cohort, the APP × rTg4510 DOX mice showed a significant reduction in astrogliosis. However, when compared to the rTg4510 DOX cohort, there was still a significant increase in astrogliosis in the APP × rTg4510 DOX mice. This is likely due, at least in part, to the remaining amyloid-β plaques in the APP × rTg4510 DOX mice.

In addition to astrogliosis, we assessed microgliosis by staining for Iba1 and looking at per cent coverage of Iba1-positive glia. As microglia become activated, they extend out processes, causing an increase in the per cent area occupied (Blackbeard et al., 2007; Trapp et al., 2007; Chen et al., 2012). Similar to the astrogliosis effect, both amyloid-β and tau pathology alone were capable of inducing an increase in microgliosis (Fig. 6C and D). Having both amyloid-β plaques and tau accumulation also resulted in an increase in microgliosis, though it was not significantly higher than either of the individual Alzheimer’s disease pathologies. While there was a reduction in microgliosis in rTg4510 DOX-treated mice, when human tau was lowered in APP × rTg4510 DOX mice, there was no significant reduction in microgliosis, likely because of the continued presence of amyloid-β plaques that also cause an increase in microgliosis.

Lowering tau in the presence of amyloid-β is sufficient to rescue neuronal loss

A major causative factor in the loss of memory and function in Alzheimer’s disease is the loss of neurons. Because amyloid-β deposition does not appear to result in neuronal death—APP/PS1 mice do not experience neuronal loss—and tau deposition is directly correlated to cognitive decline in human patients (Arriagada et al., 1992; Nelson et al., 2012), we next sought to test in vivo whether lowering human tau protects against neuronal loss when amyloid-β plaques are present by measuring wet brain weight, cortical volume, and neuronal counts. In the rTg4510 and APP × rTg4510 DOX cohorts, there was an increase in wet brain weight when compared to their genotype naïve controls, though only the APP × rTg4510 DOX group reached statistical significance (Fig. 7A). Next, cortical and hippocampal volume was analysed (Fig. 7B–D). Reducing total human tau was sufficient to protect against volume loss in the brain in rTg4510 and APP × rTg4510 DOX mice, though again only the APP × rTg4510 DOX cohort was statistically significant. The lack of significance in the rTg4510 genotype is likely due to a lower number of mice in rTg4510 naïve cohort since there are strong trends and previous studies have reported a robust neuronal protection phenotype in rTg4510 DOX mice (Santacruz et al., 2005; Spires et al., 2006; Holmes et al., 2016; Blackmore et al., 2017). To directly assess neuronal loss, the number of NeuN positive cells—a pan-neuronal marker—in the CA1 portion of the hippocampus were counted (Fig. 7E and F). This revealed a significant, and equal, protective effect of lowering human tau on neuronal loss in both rTg4510 and APP × rTg4510 DOX mice.

Discussion

By definition, primary tauopathies include human neurodegenerative disorders where there are neural or glial accumulations of the protein tau, namely Alzheimer’s disease, progressive supranuclear palsy, corticobasal degeneration, Pick’s disease, and ∼40% of frontotemporal dementia cases (Goedert et al., 2012). Of these tauopathies, only Alzheimer’s disease has the additional pathological hallmark of extracellular amyloid-β plaques. All previous studies that have analysed a human tau lowering therapy have been done so solely in the presence of tau pathology. However, given the observation of synergy between tau and amyloid-β across multiple systems (Ribé et al., 2005; Bolmont et al., 2007; Hurtado et al., 2010; Seino et al., 2010; Forny-Germano et al., 2014; Héraud et al., 2014; Bennett et al., 2017), we asked whether a human tau reduction approach, which has been shown to be highly effective in human tau transgenic mice (Santacruz et al., 2005; Polydoro et al., 2013; Xu et al., 2014; DeVos et al., 2017), would still be as effective in an environment combining tau and amyloid-β.

To test this idea, we genetically lowered the human tau transgene in APP × rTg4510 mice that develop both amyloid-β plaques and neurofibrillary tangles in an age-dependent manner (Bennett et al., 2017). DOX treatment was equally effective at lowering human MAPT/tau mRNA and protein in both the rTg4510 and APP × rTg4510 lines (Figs 1 and 2). Subsequently, a significant reduction in tau pathology—both AT8 and thioflavin-S—and tau seeding activity was observed in the rTg4510 DOX and APP × rTg4510 DOX mice when compared to their naïve genotype controls (Figs 3 and 4). While tau aggregation is one of the major pathological hallmarks in Alzheimer’s disease, the other is the presence of amyloid-β plaques. We have previously reported that there are fewer plaques in the context of tau pathology in the APP × rTg4510 mice compared to APP/PS1 mice (Bennett et al., 2017). In the study herein, we found a similar plaque number phenotype and observed a partial restoration in the number and size of plaques in APP × rTg510 DOX mice when compared to APP/PS1 mice (Fig. 5). Astrogliosis was also significantly lowered in both rTg4510 and APP × rTg4510 DOX cohorts, though microgliosis was only reduced in the rTg4510 DOX group (Fig. 6). As a final measure, neuronal integrity—as measured by brain weight, hippocampal and cortical volume, and number of neurons—was equally protected in both the rTg4510 and APP × rTg4510 DOX cohorts (Fig. 7). Together, the data presented here support a tau lowering therapy for primary tauopathies, both with and without concurring amyloid-β plaque deposition.

While there was a reduction in tau aggregation, tau pathology, and tau bioactivity in APP × rTg4510 DOX mice, differences in tau aggregates emerged when amyloid-β plaques were present—notably an increased stability as measured by the persistence of high molecular weight after 6 months of transgene repression and increased bioactivity. Additionally, we noted that tau aggregates in the presence of amyloid-β demonstrated an increased resistance to protease digestion. The use of proteinase K to better understand conformations of aggregated proteins has long been used in the prion field to biochemically discriminate prion aggregate conformations, i.e. strains (Parchi et al., 1996; Aguzzi et al., 2007). Proteinase K digestion has also recently been used to study different conformations of alpha-synuclein (Miake et al., 2002; Guo et al., 2013; Monsellier et al., 2016; Jung et al., 2017), huntingtin (Monsellier et al., 2016), and tau (Falcon et al., 2015; Narasimhan et al., 2017) aggregate conformations. When rTg4510 and APP × rTg4510 lysate was digested with proteinase K, fewer tau fragments are generated in human amyloid-β containing lysate, suggesting that the presence of amyloid-β has altered the conformation of tau to be more protease resistant. This phenotype of amyloid-β-induced tau aggregate resistance to proteinase K was recapitulated in human brain lysate containing neurofibrillary tangles with or without concurrent amyloid-β pathology (Fig. 2). Together, we believe these data support the notion that amyloid-β, either directly or indirectly, alters the conformation of soluble tau aggregates, resulting in a more stable conformation of tau aggregate that is longer lived in vivo. We do note that the timing of pathological tau and amyloid-β deposition in this new double transgenic model may not be exactly as seen in human Alzheimer’s disease patients and that this may impact the degree of effect that amyloid-β has on tau aggregation. Future studies of transgenic mouse models that develop amyloid-β deposition well before the onset of tau pathology would be informative and important in better understanding the range of impacts that amyloid-β can have on tau.

We have suggested that Alzheimer’s-induced neurodegeneration can be conceptually divided into an ‘amyloid dependent’ and ‘amyloid independent’ phase, whereby the deposition of plaques leads ultimately to tangles and neurodegeneration, which become independent to the initiating factor. One evocation of this hypothesis is that amyloid-β deposits lead to a synergistic effect on tau, which adopts a more pathological phenotype which, due to enhanced bioactivity, can lead to enhanced Alzheimer’s disease-related phenotypes. The current studies are consistent with this model: tau in the presence of amyloid-β adopts different biochemical characteristics, as measured by size on SDD-AGE blots, protease resistance, and, importantly, both a presumed longer half-life and increased bioactivity as measured on an aggregation assay. Perhaps, though, neurons are capable of withstanding a lower level of tau aggregation, such that a reduction of tau expression is sufficient to be profoundly neuroprotective, either in the presence or absence of amyloid-β deposits.

In addition to neurofibrillary tangle and amyloid-β pathologies impacting disease, both microgliosis and astrogliosis are thought to play an important role in the pathogenesis of Alzheimer’s disease (Serrano-Pozo et al., 2013). In the tau only rTg4510 mice, when tau is decreased by DOX, both microgliosis and astrogliosis are significantly reduced. However, in the presence of amyloid-β deposition, tau reduction is only associated with a reduction in the astrogliosis phenotype. The significant reduction in astrogliosis in the APP × rTg4510 DOX mice is likely due to an additive effect of astrocyte activation when both tau and amyloid-β pathologies are present, such that with decreased tau expression, the astrogliosis is reduced to a similar level as amyloid-β alone APP/PS1 mice. In the human brain, astrogliosis is directly correlated with neurofibrillary tangle deposition, synapse loss, and cognitive decline (Ingelsson et al., 2004; Kashon et al., 2004; Serrano-Pozo et al., 2013) and attenuating the activation of astrocytes in vivo rescues synaptic dysfunction and cognitive decline (Furman et al., 2012), suggesting that lowering of astrogliosis is therapeutically beneficial.

Understanding the consequences of endogenous tau repression is also of importance in thinking of a human tau lowering therapy for the clinic. Tau deletion can protect against a growing number of amyloid-β-mediated toxicities in vivo, including cognition (Roberson et al., 2007; Leroy et al., 2012), seizures (Roberson et al., 2007, 2011; Ittner et al., 2010; Suberbielle et al., 2013; Li et al., 2014), survival (Roberson et al., 2007, 2011; Ittner et al., 2010), axonal transport deficits (Vossel et al., 2010), and double-stranded breaks in DNA (Suberbielle et al., 2013). This complete deletion of tau has also been shown to be phenotypically normal in regard to learning/memory, cognition, and neuroanatomy (Roberson et al., 2007; Morris et al., 2013; Li et al., 2014; van Hummel et al., 2016) with a minor motor phenotype developing later in life (Lei et al., 2012; Morris et al., 2013; Li et al., 2014; van Hummel et al., 2016). Additionally, when murine tau was lowered in adult mice, no deficits in any sensory, motor, or cognitive tasks were detected (DeVos et al., 2013). Endogenous tau has also been lowered in non-human primates with no alterations from baseline reported (Olson et al., 2016; DeVos et al., 2017), further supporting the safety of reducing tau levels in the adult brain.

Alzheimer’s disease is a complex disease with amyloid-β plaque deposition, tau accumulation, and neuroinflammation all contributing to the overall neurodegeneration. In our studies here, amyloid-β causes an increase in glial activation while also having an impact on the pathogenicity of tau aggregation. Whether amyloid-β interacts directly or indirectly with tau aggregates to induce a unique, more pathogenic conformation will be an important subject for further study. We believe these current data, while providing pre-clinical support for a tau lowering therapy in human Alzheimer’s disease patients, provide a basis to begin dissecting the contributions of each of these different pathologies—tau, amyloid-β, and gliosis—in the mature phase of Alzheimer’s disease.

Acknowledgements

We thank Dr Marc Diamond for the HEK293 CFP/YFP-Tau^RD cell line and Dr Kathleen Schoch for providing manuscript edits. Acknowledgement is also made to the donors of Alzheimer’s Disease Research, a program of BrightFocus Foundation, for support of this research.

Funding

This research was funded by NIA training grants to the Division of Medical Sciences at Harvard University (NIH/NIA T32 AG000222, S.L.D. and R.E.B.), Brightfocus Foundation (S.L.D. and R.E.B.), the Massachusetts Alzheimer’s Disease Center grant P50AG05134 (B.T.H.), the JPB Foundation (B.T.H.) and by NIH Grants P30 NS045776 (M.J.) and AG026249 (B.T.H.).

Supplementary material

Supplementary material is available at Brain online.

Abbreviations

Abbreviations
 
• CFP

  cyan fluorescence protein

 
• DOX

  doxycycline

 
• ELISA

  enzyme-linked immunosorbent assay

 
• FRET

  Forster resonance energy transfer

 
• SDD-AGE

  semi-denaturing detergent agarose gel electrophoresis

References