p-tau Ser356 is associated with Alzheimer’s disease pathology and is lowered in brain slice cultures using the NUAK inhibitor WZ4003

All postmortem brain tissue used in this study was obtained with ethical approval from the Edinburgh Sudden Death Brain Bank. This study was reviewed and approved by the Edinburgh Brain Bank ethics committee, a joint office for NHS Lothian and the University of Edinburgh, under ethical approval number 15-HV-016. The Edinburgh Brain Bank is a Medical Research Council funded facility with research ethics committee (REC) approval (16/ES/0084). Individual demographics/ case numbers for human postmortem studies are listed in Table 1, with the specific experiment they were included in referenced by the figure numbers in the table. Different cases were used for different experiments in line with tissue availability and type required (e.g. frozen, paraffin sections). Types of tissue used and grouped demographic information are detailed in the methods sections below for each experiment. All cases were screened by Edinburgh MRC Brain Bank for tau Braak stage, Aβ Thal phase, presence of Lewy bodies and p-TDP43. p-TDP43 was assessed according to a condensed protocol to assess limbic-predominant age-related TDP-43 encephalopathy neuropathologic change (LATE-NC), which can detect LATE-NC stages 2–5. All staining for the included cases was negative for p-TDP43 using this method, but we cannot rule out the presence of stage 1 LATE-NC in these samples [40]. Inclusion criteria for AD cases were: clinical dementia diagnosis, Braak stage V–VI and a post-mortem neuropathological diagnosis of AD. Control subjects were chosen to be age/sex matched as much as possible, with inclusion criteria of no diagnosed neurological or psychiatric condition. Where possible, exclusion criteria for both AD and control samples were substantial non-AD neuropathological findings (such as diagnosis of Lewy body dementia (with substantial cortical Lewy bodies in the region of interest), or substantial haemorrhage).

At the MRC Edinburgh Brain Bank, fresh postmortem brain tissue was sectioned into small blocks of around 500 mg each, flash frozen in on dry ice then stored at − 80 °C until requested by the research group. Ten different frozen postmortem human brain case samples were obtained per Braak stage for the Brodmann area 20/21 region (summary details listed in Table 2). Total protein homogenate and synaptoneurosomes were generated as previously described [29]. For each case sample, 300–500 mg of frozen tissue was thawed and immediately homogenised, on ice, in a glass-teflon dounce homogeniser in 1 ml of homogenisation buffer (25 mM HEPES (pH 7.5), 120 mM NaCl, 5 mM KCl, 1 mM MgCl₂, 2 mM CaCl₂, protease and phosphatase inhibitors (Roche complete mini: 11836153001)). The homogenate was then flushed through an 80 μm-pore nylon filter (Millipore NY8002500), with 300 μl of the resulting crude total homogenate set aside and frozen at − 80 °C. To generate synaptoneurosome preps, total homogenate was further filtered through a 5 μm filter (Millipore SLSV025NB). Synaptoneurosomes were centrifuged at 1000xg for 5 min and the pellet was collected. Proteins (either total homogenate or synaptoneurosome) were extracted in an SDS buffer (100 mM Tris–HCl, 4% SDS) and the protein concentration determined by commercial BCA assay before equal protein amounts were diluted in 2 × Laemmli buffer, boiled then loaded for Western blot (below).

(Summary patient details in Table 3) 4 μm sections of formalin-fixed, paraffin-embedded tissue from Broadmann area 20/21 (inferior temporal gyrus) were de-waxed in xylene, then decreasing concentrations of ethanol, incubated with autofluorescence inhibitor reagent (Millipore 2160), and non-specific antigens blocked by incubation in a solution of 0.1 M PBS containing 0.3% Triton X-100 and 10% normal donkey serum for 1 h. Primary antibodies to p-tau Ser356 (Abcam: ab75603, diluted 1:1000) and GFAP (Abcam: ab4674, diluted 1:3000) were diluted in blocking buffer and incubated on sections overnight at 4 °C. Sections were washed with 0.1 M PBS containing 0.3% triton X-100, and then incubated in secondary antibodies (donkey anti-rabbit Alexa-fluor 594, Invitrogen A21207 and donkey anti-chicken Alexa-fluor 647, Invitrogen A78952) diluted 1:200 in block buffer at room temperature for 1 h. Sections were washed with PBS, then incubated in 0.5% Thioflavin S in 50% ethanol, 50% H₂O solution for 8 min in the dark at room temperature. Slides were dipped in 80% ethanol to reduce Thioflavin S background, and then rinsed thoroughly in distilled water before coverslipping with Immu-Mount™ (Epredia™: 9990402). Sections were imaged with a 20 × 0.8NA objective on a Zeiss AxioImager Z2 microscope equipped with a CoolSnap camera. Ten regions of interest were imaged in the cortex of each slide sampling in a systematic random fashion to sample throughout all six cortical layers. Images were examined in Fiji (ImageJ) by a blinded experimenter. Neurofibrillary tangles were identified by Thioflavin S staining in a flame shape with a hole in the centre (where the nucleus resides) at the approximate size of a neuronal cell body (~ 10 μm in diameter). Cells with Thioflavin S and GFAP staining were excluded to exclude astrocytic tau pathology. Dystrophic neurites were identified as p-tau Ser356 containing swellings of > 5 μm in diameter associated with Thioflavin-S-positive Aβ plaques. Neuropil threads were identified as Thioflavin S and/or p-tau Ser356 staining in linear thread-like patterns consistent with axonal or dendritic staining. The presence or absence of p-tau Ser356 was noted for each identified neurofibrillary tangle and quantified as a percentage of total detected tangles in the ten regions of interest for each case.

Tissue from non-AD control patients, with Braak stages I–II and Thal phase 1–2, and AD patients with Braak stage VI and Thal phase 5 were acquired (summary patient details in Table 4 and Supplementary Table 2). As described previously [9], fresh tissue was fixed in 4% paraformaldehyde for 3 h, dehydrated in ethanol and embedded in LR White Resin. A diamond knife (Diatome) mounted onto an Ultracut microtome (Leica) was used to cut the embedded tissue into 70 nm serial sections. For Fig. 3, 15–30 serial section ribbons were collected onto gelatin-coated coverslips and immunostained with the following primary antibodies for 1 h: 1:500 Rabbit p-tau Ser356 (Abcam: ab75603), 1:100 goat synaptophysin (R&D Systems: AF555), 1:200 chicken GFAP (Abcam: ab4674), 1:200 guinea pig PSD95 (Synaptic Systems: 124 014) and 1:100 mouse p-tau Ser202/Thr205 (AT8) (Thermo Fisher: MN1020). Secondary antibodies, donkey anti-rabbit Alexa-fluor 405 (Abcam: ab175651), donkey anti-goat Alexa-fluor 594 (Abcam: ab150136), goat anti-chicken Alexa-fluor 405 (Invitrogen: A48260), goat anti-guinea pig Alexa-fluor 488 (Abcam: ab150185) and donkey anti-mouse Alexa-fluor 647 (Thermo Fisher: A32787) were then applied at 1:50 concentration to ribbons for 45 min. Images were obtained with a 63 × 1.4 NA objective on a Leica TCS confocal microscope. Details of staining and imaging for Supp. Fig. 2 can be found in supplemental information. Images from the same locus in each serial section along a ribbon were then aligned, thresholded, and parameters quantified using in-house scripts in Fiji (ImageJ) and MATLAB. Any signals that do not appear in the same x–y location in a least two adjacent (z) brain sections are discarded as noise. Resulting parameter data were statistically analysed using custom R Studio scripts. Analysis code is available on GitHub (https://​github.​com/​Spires-Jones-Lab).

APP/PS1 (APPswe, PSEN1dE9 [31]) and wildtype litter mate male and female mouse pups, aged 6–9 days old (P6-9), were obtained from a breeding colony at the Bioresearch and Veterinary Services Animal Facility at the University of Edinburgh. Animals were culled via cervical dislocation, performed by a trained individual, who was assessed by the Named Training and Competency Officer (NTCO). All animal work was conducted according to the Animals (Scientific Procedures) Act 1986 under the project licence PCB113BFD and PP8710936. All animals were bred and maintained under standard housing conditions with a 12/12 h light–dark cycle.

Mouse organotypic brain slice cultures (MOBSCs) were generated and maintained as described previously [19, 20, 28, 56], with minor modifications. Mouse pups aged postnatal day (P)6–9 were culled by cervical dislocation. Brains were rapidly transferred to ice-cold 0.22 μm-filtered dissection medium composed of 87 mM NaCl, 2.5 mM KCl, 25 mM NaHCO₃, 1.25 mM NaH₂PO₄, 25 mM glucose, 75 mM sucrose, 7 mM MgCl₂, 0.5 mM CaCl₂, 1 mM Na-Pyruvate, 1 mM Na-Ascorbate, 1 mM kynurenic acid and 1× penicillin/streptomycin (Thermo Fisher: 15140122) (340 mOsm, pH 7.4), bubbled with 95% O₂, 5% CO₂. All salts and chemicals were purchased from Merck. Brains were then mounted on and glued (cyanoacrylate, Loctite) to a vibratome stage. A Leica VT1200S vibratome was used to cut 350 µm-thick horizontal slices, from which the hippocampus was dissected with fine needles. Hippocampal slices were plated on membranes (Millipore: PICM0RG50) sitting on top of 1 ml of maintenance medium, placed inside 35 mm culture dishes. The maintenance medium was 0.22 µm-filtered and composed of MEM with Glutamax-1 (50%) (Invitrogen: 42360032), heat-inactivated horse serum (25%) (Thermo Fisher: 26050070), EBSS (18%) (Thermo Fisher: 24010043), d-glucose (5%) (Sigma: G8270), 1× penicillin/streptomycin (Thermo Fisher: 15140122), nystatin (3 units/ml) (Merck: N1638) and ascorbic acid (500 μM) (Sigma-Aldrich: A4034). Slice cultures were then immediately placed into an incubator and maintained at 37 °C with 5% CO₂ thereafter. The slice culture medium was changed fully within 24 h of plating, then at 4 days in vitro (div), 7 div and weekly thereafter. Four culture dishes were made per pup, with 1–2 slices plated per dish. Cultures were kept for either 2 weeks or 4 weeks, with 5 μM WZ4003 (APExBIO: B1374) or DMSO (Sigma-Aldrich: D2438) control applied in culture medium at every feed during the treatment period (either 0–2 weeks or 2–4 weeks).

Human brain slice cultures (HBSCs) were generated from surplus neocortical access tissue from patients undergoing tumour resection surgery, with ethical approval from the Lothian NRS Bioresource (REC number: 15/ES/0094, IRAS number: 165488) under approval number SR1319. Additional approval was obtained for receiving data on patient sex, age, reason for surgery and brain region provided (NHS Lothian Caldicott Guardian Approval Number: CRD19080). Patient details are listed in Table 5. The informed consent of patients was obtained using the Lothian NRS Bioresource Consent Form. Dissection and culture methods have been adapted from published studies [41, 51, 53, 54]. Access, non-tumour, neocortical tissue was excised from patients (would normally be disposed of during surgery) and immediately placed in sterile ice-cold oxygenated 0.22 μm-filtered artificial cerebrospinal fluid (aCSF) containing 87 mM NaCl, 2.5 mM KCl, 10 mM HEPES, 1.62 mM NaH₂PO₄, 25 mM d-glucose, 129.3 mM sucrose, 1 mM Na-Pyruvate, 1 mM ascorbic acid, 7 mM MgCl₂, and 0.5 mM CaCl₂. The tissue was then sub-dissected and mounted in 2% agar, before being glued (cyanoacrylate, Loctite) to a vibratome stage. 300 µm-thick slices were then cut in ice-cold and oxygenated aCSF, before being sub-dissected into smaller slices. Slices were placed into 0.22 μm-filtered wash buffer composed of oxygenated Hanks Balanced Salt Solution (HBSS, Thermo Fisher: 14025092), HEPES (20 mM) and 1X penicillin–streptomycin (Thermo Fisher: 15140122) (305 mOsm, pH 7.3) for 15 min at room temperature. Slices were then plated on membranes (Millipore: PICM0RG50) sitting on top of 750 μl of a second wash medium, placed inside 35 mm culture dishes. The second wash medium was 0.22 µm-filtered and composed of BrainPhys Neuronal Medium (StemCell Technologies: 5790) (96%), N2 (Thermo Fisher: 17502001) (1×), B27 (Thermo Fisher: 17504044) (1×), hBDNF (StemCell Technologies: 78005) (40 ng/ml), hGDNF (StemCell Technologies: 78058) (30 ng/ml), Wnt7a (Abcam: ab116171) (30 ng/ml), ascorbic acid (2 μM), dibutyryl cAMP (APExBIO: B9001) (1 mM), laminin (APExBIO: A1023) (1 ug/ml), penicillin/streptomycin (Thermo Fisher: 15140122) (1×), nystatin (Merck: N1638) (3 units/ml) and HEPES (20 mM). Slice cultures were kept in the second wash medium in an incubator at 37 °C with 5% CO₂ for 1 h, after which the medium was aspirated and replaced with maintenance medium. The maintenance medium composition was identical to that of the second wash medium, but without HEPES. For all human cases, the time taken from patient tissue removal to the brain slices being placed in the incubator (post-operation interval) was kept under 2 h. 100% medium exchanges occurred twice weekly thereafter. Cultures were kept for 2 weeks and treated with either 10 μM WZ4003 (APExBIO: B1374) or DMSO (Sigma-Aldrich: D2438) control applied in culture medium, and with every feed thereafter from 0 days in vitro (div). To assess neuronal integrity in HBSCs, 14 div slices were fixed overnight in 4% PFA, washed 3 × in PBS, blocked for 1 h in PBS + 3% normal goat serum + 0.5% Triton X-100 and then incubated overnight at 4 °C in 1:500 guinea pig anti-MAP2 (Synaptic systems: 188004) in blocking solution. The slices were then washed 3 × in PBS, before incubation for 2 h in 1:500 secondary antibody (goat anti-guinea pig-488 (Thermo Fisher: A11073)) in block buffer at room temperature. Slices were washed 3 × in PBS, counterstained with DAPI for 15 min, washed 3 × in PBS then mounted on slides in Vectashield Antifade mounting medium (2B Scientific: H-1900). Images were taken using a 63 × 1.4 NA objective on a Leica TCS confocal microscope.

MOBSCs or HBSCs were removed from the culture membrane with a scalpel blade into RIPA buffer (Thermo Fisher Scientific: 89901) with protease inhibitor cocktail (1×) and EDTA (1×) (Thermo Fisher Scientific: 78429). Slices were thoroughly homogenised via trituration through 20 pipette fill and empty cycles (using a 100 μl pipette tip). 50 μl of RIPA buffer was used per MOBSC slice and 100 μl of RIPA per HBSC slice. RIPA-buffered MOBSCs, HBSC samples, or human postmortem total protein/synaptoneurosome samples were then mixed into equal volumes of 2X Laemmli buffer (Merck: S3401-10VL) and boiled for 10 min at 98 °C. 12 µl of each sample was loaded into 4–12% NuPage Bis–Tris gels (Invitrogen: NP0336BOX), before proteins were separated by electrophoresis using MES SDS running buffer (Invitrogen: NP0002). Proteins were then transferred onto PVDF transfer membranes (Invitrogen: IB24002), before a total protein stain (Li-Cor Biosciences: 926-11016) image was acquired using a Li-Cor Odyssey Fc machine. Membranes were de-stained and then blocked for 1 h using PBS Intercept Blocking Buffer (Li-Cor Biosciences: 927-70001). Primary antibodies were diluted in PBS Intercept Blocking Buffer with 0.1% Tween-20 and incubated with membranes overnight at room temperature, with shaking. Membranes were washed three times for 5 min with PBS-Tween, then incubated in darkness for 2 h with IRDye anti-rabbit (Li-Cor Biosciences: 680RD) and anti-mouse (Li-Cor Biosciences: 800CW) secondary antibodies, each at 1:10,000 concentration. Membranes were washed 3 × in PBS-Tween, 1 × in PBS and then imaged using a Li-Cor Odyssey Fc machine. The following primary antibodies were used: 1:500 mouse Tau-5 (Abcam: AB80597), 1:1000 rabbit ps356 tau (Abcam: AB75603), 1:500 rabbit PSD-95 (Abcam: AB18258), 1:2500 rabbit Tuj-1 (Sigma: T2200), 1:500 rabbit p-tau Ser202/Thr205 (Cell Signalling Technology: 30505S), 1:500 rabbit NUAK1 (ProteinTech 22723-1-AP) and 1:2000 rabbit cyclophilin-B (Abcam: AB16045). Western blot images were analysed using Empiria Studio (Version 2.3).

All data were analysed using R (v 4.2.2) and R Studio (v 2023.03.1, Build 446). Statistical tests were chosen according to the experimental design and dataset type. Unpaired T tests, ratio paired T tests and N-way repeated-measures ANOVA tests (car package) were conducted using linear mixed effects models (LMEM) using the lme4 package (each model is listed in the relevant results section). Human or mouse case was included as a random effect in linear mixed effects models to avoid pseudoreplication of the data [57]. T tests used Satterthwaite’s method, whilst Type-III F-Wald ANOVA tests used the Kenward–Roger method, to compute degrees of freedom. For N-way ANOVA analyses resulting in significant interaction effects or significant main effects where more than two within-factor levels existed, multiple comparison tests using the Tukey method for adjusting p values were computed with the emmeans package. Model assumptions were tested using Shapiro–Wilk and F tests, along with plotting model residuals. To assess equality of variance, standardised model residual scatter was plotted against fitted values. Residual distribution was assessed for normality by plotting a histogram and kernel density estimate of standardised model residuals over a standard normal distribution, and by generating a quantile–quantile plot of dataset quantiles against normal quantiles. Datasets in Fig. 3 were Tukey-transformed prior to performing statistical tests if assumptions of normality were violated. Statistical tests performed on datasets represented in Figs. 4, 5, 6 were conducted on absolute data that was log-transformed, to compute ratio differences between differentially treated samples from the same animal/human and to satisfy assumptions of normality. Figures 4, 5 and 6 represent control-normalised data to display within-animal and within-human case effects. For the Thal phase correlation graph data represented in Supp. Fig. 1, a Spearman’s rank correlation test was performed for each preparation group (TH or SN) due to data being non-normally distributed. Significance values are reported as p < 0.05 *, p < 0.01 *, p < 0.001 *** and error bars represent the mean ± SEM.

To investigate how the levels of p-tau Ser356 change with Braak stage in postmortem human temporal cortex (BA20/21), soluble fractions of total homogenate and synaptoneurosome preparations were generated from frozen postmortem brain from ten Braak 0–I individuals, ten Braak III–IV individuals and ten Braak VI individuals (with clinically diagnosed AD). Individual patient details are listed in Table 1 with summary demographic information listed in Table 2. Levels of p-tau Ser356 and total tau (tau5) were quantified by Western blot (Fig. 1a). Total tau (tau5) was detected in all cases and was normalised to total protein (REVERT protein stain). For statistical analysis, the following LMEM was applied: Protein level ~ Braak Stage * Preparation + (1|CaseID). The levels of total tau did not increase with Braak stage (Fig. 1b, effects of Braak stage: F_(2,46.01) = 1.32, p = 0.28) and there were no detectable differences in the levels of tau in total homogenate compared to synaptoneurosome preparations (Fig. 1b, effect of preparation: F_(1,27.00) = 0.15, p = 0.71). By contrast, the levels of p-tau Ser356, normalised to total tau levels, showed a significant increase with Braak stage (Fig. 1c effect of Braak stage: **F_(2,42.28) = 5.72, p = 0.006) and a detectable effect of preparation (Fig. 1c, effect of preparation: *F_(1,27.00) = 4.94, p = 0.035) driven by an increase in p-tau Ser356 in total homogenate, compared to synaptoneurosome, in Braak VI cases (Fig. 1c, multiple comparisons test: *p = 0.034). This is possibly reflective of increased tau in the cell bodies at a stage where tangle prevalence will be high. There was a significant increase in p-tau Ser356 in Braak VI AD brains compared to Braak 0–I control brains (Fig. 1c, multiple comparisons test: ***p = 0.0009), and a strong trend for increase between Braak III–IV and Braak VI (Fig. 1c, multiple comparisons test: p = 0.053). A similar pattern was seen when examining the levels of p-tau Ser202/Thr205 (a well characterised phosphoepitope known to be upregulated in later Braak stages [45]) (Supp. Fig. 1a,c), with a significant increase in the ratio of p-tau Ser202/Thr205/total tau with Braak stage (Supp. Fig. 1c, effect of Braak stage: ***F_(2,32.12) = 10.7, p < 0.001). Interestingly, in contrast to previous studies which found increased levels of NUAK1 in AD versus control samples[37], we did not see an overall effect of Braak stage on soluble NUAK1 levels (Supp. Fig. 1b,d) (effect of Braak stage: F_(2,36.83) = 0.64, p = 0.531). We did, however, find an accumulation of NUAK1 in the synaptoneurosome compartment, relative to the total homogenate in Braak stages III–IV or VI brains (but not in Braak 0–I brains), indicating that there may be a shift in location of NUAK1 throughout disease progression (Supp. Fig. 1d, effect of Braak stage*preparation: *F_(2,27.00) = 3.48, p = 0.045). When analysing correlation of protein levels with Aβ Thal phase, we found a significant correlation between increasing Thal phase and increased levels of both p-tau Ser356 (Supp. Fig. 1e) and p-tau Ser202/Thr205 (Supp. Fig. 1f). For NUAK1, we did not see a significant correlation with Thal phase, but there appeared to be divergence between synaptoneurosome and total homogenates in the later stages (Supp. Fig. 1 g).

Having established a significant increase in p-tau Ser356 in AD brains, we sought to establish whether this phosphoepitope of tau is found in pathological lesions including neurofibrillary tangles (NFTs), dystrophic neurites around plaques and reactive astrocytes. Paraffin sections were obtained from five control (Braak 0–II, Thal 0–2) and five confirmed AD (Braak VI, Thal 5) cases (individual patient details listed in Table 1, summary demographic details listed in Table 3). Staining in paraffin sections demonstrated that p-tau Ser356 is readily detectable in tangles, dystrophic neurites, neuropil threads and some reactive astrocytes in AD brain (Fig. 2). Of the 85 Thioflavin-S-positive tangles, we identified in the AD cases, 79 were positive for p-tau Ser356 (93%) (Fig. 2f). Interestingly, whilst control brains showed, as expected, significantly lower levels of neuropathology (Fig. 2a), we also found evidence of p-tau Ser356 staining in some control brains, in areas co-localising with Thioflavin S, most notably in dystrophic neurites (Fig. 2b). Together, these results suggest that p-tau Ser356 is a common component of dystrophic neurites and Thioflavin-S-positive NFTs, and may, therefore, be involved early in the tau aggregation cascade.

Given recent work highlighting the role of synaptic tau in the early stages of AD pathology [9], we next sought to examine the presence of p-tau Ser356 at synaptic terminals. Here we used array tomography, which physically overcomes the diffraction limit of light in the axial plane to enable imaging of the protein composition of individual synapses [33]. Effective detection of proteins located within synapses using this method has been validated against super-resolution and electron microscopy imaging [9, 33, 48, 52]. We imaged five confirmed AD (Braak V–VI) and five age/sex-matched control cases (Braak 0–III) for p-tau Ser356, AT8 (tau phosphorylated at Ser202 and Thr205) and the pre-synaptic marker synaptophysin (Fig. 3a–e) or post-synaptic marker PSD95 (Fig. 3f–j). Individual patient demographic data are listed in Table 1 and summary demographic data are listed in Table 4. For statistical analysis, the following LMEM was applied: Co-localisation % ~ Diagnosis + Sex + (1|CaseID). Whilst largely undetectable in control cases (Fig. 3a), both p-tau Ser356 (Fig. 3c) and AT8 (Fig. 3d) were found to co-localise with synaptophysin in AD cases (**F_(1,7.14) = 18.7, p = 0.0033 and **F_(1,7.13) = 14.6, p = 0.0063, respectively). Both p-tau Ser356 (Fig. 3h) and AT8 (Fig. 3i) were also found to co-localise with PSD95 in AD brains (**F_(1,7.38) = 12.5, p = 0.0088, *F_(1,7.14) = 9.08, p = 0.019, respectively). To provide additional confirmation that p-tau Ser356 is located within synaptic compartments, we conducted Förster resonance energy transfer (FRET) imaging of AD brain samples (see Supplementary methods) [69]. We find that p-tau Ser356 is located within 10 nm of both synaptophysin (Supp. Fig. 2c: *t₍₄₎ = 4.18, p = 0.014) and PSD95 (Supp. Fig. 2f: ***t₍₅₎ = 7.56, p = 0.0006).

Interestingly, whilst p-tau Ser356 and AT8 were found to co-localise with a similar percentage of pre-synaptic terminals (median of 1.38% (Fig. 3c) and 1.60% (Fig. 3d), respectively), the percentage of synaptophysin puncta that co-localised with both tau epitopes, (whilst still significantly higher in AD than controls (*F_(1,7.08) = 10.5, p = 0.014)), was considerably smaller (median of 0.36% (Fig. 3e)). For post-synapses, there was no significant increase in the number of puncta that co-localised with both epitopes in AD compared to control brain (Fig. 3j F_(1,7.27) = 1.28, p = 0.294). This could suggest that different synapses have unique signatures of tau phosphorylation patterns in AD, or that there are technical considerations (e.g. masked antibody binding sites) that make it harder to resolve both epitopes when they co-localise.

In light of recent findings that astrocytes ingest synapses in post-mortem AD brain, we examined the levels of pre- (Supp. Fig. 2a,b) and post-synapses (Supp. Fig. 2d,e) contained within astrocytes [59, 62], and whether synapses containing p-tau Ser356 were more likely to be ingested. We found that pre-synapses containing p-tau Ser356 were around 5 × more likely to be found within astrocytes (Supp. Fig. 2b: **F_(1,14) = 9.94, p = 0.00705), than the general pre-synapse population, whilst there was no effect on post-synapse ingestion with tau status (Supp. Fig. 2e: F_(1,17) = 0.0341, p = 0.856).

We next sought to evaluate pharmacological tools to lower p-tau Ser356 in live brain tissue. To determine whether response to NUAK inhibition differs under physiological conditions versus conditions of early Aβ dysregulation (which may initiate downstream tau changes), we generated mouse organotypic brain slice cultures (MOBSCs) from wildtype and APP/PS1 littermates. Previous work in MOBSCs has found that cultures undergo significant changes in the first 2 weeks in vitro, (including resolution of inflammation, increased neuronal outgrowth, developmental maturation and increased production of synaptic proteins [14, 24, 56]), before entering a relatively stable period. We, therefore, designed our experiments to capture both this initial period, and the following stable period.

WT or APP/PS1 MOBSCs were split into two culture phase groups (0–2 weeks in vitro, versus 2–4 weeks in vitro) and two treatment groups (DMSO or WZ4003) resulting in four experimental conditions represented in tissue from the same animal (Fig. 4a). In accordance with prior literature [37], we applied 5 µM WZ4003 to cultures and analysed the impact on proteins using Western blot (Fig. 4b). We explored whether 1) p-tau Ser356 levels change over time in culture (effect of phase), 2) whether early Aβ dysregulation results in changes to tau (effect of genotype), 3) whether WZ4003 treatment impacts levels of tau and 4) whether either genotype or age of the cultures impacts response to WZ4003 treatment. All conditions are graphically displayed normalised to the 0–2 week DMSO-treated condition from the same animal, whilst analysis was performed using ratio repeated measured analysis with the absolute data (Graphs of absolute data are in Supp. Fig. 3). For statistical analysis, the following LMEM was applied: Protein level ~ Genotype * Phase * Treatment + (1|Litter/Animal).

Regardless of genotype, we found that the levels of both total tau (Fig. 4c, effect of phase: *F_(1,51.00) = 6.28, p = 0.015) and p-tau Ser356 (Fig. 4d, effect of phase: **F_(1,51.00) = 8.73, p = 0.005) protein rise over time in culture. This likely reflects the regrowth or maturation of neurites in MOBSCs following the initial slicing procedure. The ratio of p-tauSer356/total tau remained stable over time in culture (Fig. 4e, effect of phase: F_(1,51.00) = 1.25, p = 0.27). In line with previous studies in neuroblastoma cells [37], application of WZ4003 resulted in a significant, and largely proportional, reduction in both total tau (Fig. 4c, effect of treatment: *F_(1,51.00) = 7.01, p = 0.011) and p-tau Ser356 (Fig. 4d, effect of treatment: **F_(1,51.00) = 10.7, p = 0.002) (Fig. 4e, effect of treatment on p-tau Ser356/total tau ratio: F_(1,51.00) = 2.10, p = 0.15). Interestingly, there was a trend interaction between culture phase and treatment for p-tau Ser356 (Fig. 4d, effect of treatment*phase: F_(1,51.00) = 2.83, p = 0.098), where the mean percentage loss of p-tau Ser356 between treated and control samples trended to be larger in the 0–2 week phase than the 2–4 week phase. The reduction in p-tau Ser356 averaged at 26.4% for WT and 26.6% for APP/PS1 0–2 week cultures, compared with 9.1% for WT and 6.8% for APP/PS1 2–4 week cultures.

In this study, we found no differences between genotypes in either baseline levels of tau, p-tau Ser356/total tau ratio, or in the degree of response to treatment (Fig. 4c–e). This shows that, in MOBSCs up to 4 weeks in vitro, the presence of mutations in the amyloid processing pathway does not result in perturbations to p-tau Ser356.

As we saw no differences between genotype, we performed an additional experiment in MOBSCs from WT mice to characterise WZ4003 impact on potential downstream phosphorylation sites of tau (Supp. Fig. 4). p-tau Ser202/Thr205 has previously been shown to be strongly associated with AD Braak stage [45] and phosphorylation of this site is sufficient to induce tau aggregation in vitro [15]. Preventing tau phosphorylation at Ser356 has also previously been found to prevent downstream phosphorylation of Ser202/Thr205 [46]. When examining the levels of p-tau Ser202/Thr205 in MOBSCs, we find a significant interaction between phase and treatment (*F_(1,18.00) = 5.09, p = 0.037), indicating that p-tau Ser202/Thr205 is lowered to a greater degree in 0–2 week cultures by WZ4003 than 2–4 week cultures (Supp. Fig. 4d, e). NUAK1 levels are not impacted by WZ4003 treatment (F_(1,18.00) = 0.82, p = 0.378) (Supp. Fig. 4 g, h). In this independent experiment, we also confirm the same effect of both phase (***F_(1,18) = 27.9, p < 0.001) and WZ4003 treatment (**F_(1,18.00) = 11.7, p = 0.003) on the levels of p-tau Ser356 (Supp. Fig. 4a, b), with a trend for a treatment*phase interaction (F_(1,18.00) = 3.45, p = 0.08).

We next sought to examine the impact of WZ4003 treatment on both synaptic and neuronal protein levels in MOBSCs (Fig. 5). Here we tested whether the levels of PSD95 and Tuj1 are altered by: 1) age of the culture (effect of phase), 2) presence of APP/PS1 mutations (effect of genotype), 3) WZ4003 treatment and 4) whether genotype or culture age impacts response to treatment. For statistical analysis, the following LMEM was applied: Protein level ~ Genotype * Phase * Treatment + (1|Litter/Animal). Western blot analysis (Fig. 5a) revealed that the expression of the synaptic protein PSD95 increased over time in culture (Fig. 5b, effect of phase: ***F_(1,51.00) = 26.7, p < 0.001), but, interestingly, PSD95 was reduced in response to WZ4003 (Fig. 5b, effect of treatment: *F_(1,51.00) = 5.60, p = 0.022). The neuronal tubulin marker Tuj1 is similarly upregulated in 2–4 week cultures (Fig. 5c, effect of phase: ***F_(1,51.00) = 13.4, p < 0.001), but was not significantly impacted by WZ4003 treatment (Fig. 5c, effect of treatment: F_(1,51.00) = 1.99, p = 0.164). To examine the impact of synaptic proteins in the context of changes to neuronal proteins, we next normalised the levels of PSD95 to Tuj1 levels. Interestingly, the effect of phase remains (Fig. 5d, effect of phase: *F_(1,51.00) = 4.39, p = 0.041), indicating that synaptic proteins rise in excess of the rise in neuronal proteins over time, but the effect of treatment disappears (Fig. 5d, effect of treatment: F_(1,51.00) = 1.63, p = 0.207). This suggests that, despite the loss of Tuj1 in response to WZ4003 treatment not reaching significance alone, a WZ4003-induced loss of neuronal protein may partly contribute to the loss of PSD95. There were no differences between genotypes for either PSD95 or Tuj1 at baseline or in response to treatment (Fig. 5, Supp. Fig. 3).

When normalising total tau (Fig. 5e) or p-tau Ser356 (Fig. 5f), to Tuj1, we see that, relative to neuronal tubulin, there is a reduction in the levels of total tau (Fig, 5e, effect of phase: **F_(1,51.00) = 7.38, p = 0.009) and p-tau Ser356 (Fig. 5f, effect of phase: *F_(1,51.00) = 5.40, p = 0.024) in the 2–4 week cultures. This indicates, whilst there may be increased neuronal (Fig. 5c) and tau (Fig. 4c, d) protein over time, the proportion of tau relative to neuronal protein declines as the cultures age. Of particular note, is that the effect of WZ4003 treatment is abolished when normalising total tau (Fig. 5e, effect of treatment: F_(1,51.00) = 0.00, p = 0.964), or p-tau Ser356 (Fig. 5f, effect of treatment: F_(1,51.00) = 0.22, p = 0.642) to Tuj1, demonstrating that the lowering of tau, inhibition of NUAK, or other impacts of WZ4003 results in a proportional reduction in both tau and neuronal protein in MOBSCs up to 4 weeks in vitro.

Finally, we sought to assess the impact of WZ4003 treatment in live adult human brain tissue. Healthy, peri-tumoral access tissue from five patients (demographics details listed in Table 5), was processed into 300 μm slices and cultured for 2 weeks in vitro (Fig. 6a). Cultures showed intact MAP2 positive neuronal cell bodies and processes at 14 div (Fig. 6b) and tau, neuronal and synaptic proteins were readily detectable by Western blot (Fig. 6c). p-tau Ser356 was detectable by Western blot in live human brain tissue prior to culturing (Supp. Fig. 5i, j). Slices from the same individuals were divided into both control (DMSO) and treatment (10 μM WZ4003) conditions, and protein levels were compared to the untreated control from the same patient. Graphs showing control-normalised data are used for display purposes (Fig. 6), with statistics run on ratio paired T tests using absolute data. Graphs of absolute data are shown in Supp. Fig. 5. For statistical analysis, the following LMEM was applied: Protein level ~ Treatment + (1|CaseID).

We found that WZ4003 treatment did not impact the level of total tau (Fig. 6d, mean % decrease = 2.2%, t₍₄₎ = 0.43, p = 0.69), but there was a trend for reduced levels of p-tau Ser356 (Fig. 6e, mean % decrease = 35.0%, t₍₄₎ = 2.43, p = 0.072) and a trend for reduced p-tau Ser356/tau ratio (Fig. 6f, mean % decrease = 31.6%, t₍₄₎ = 2.17, p = 0.096). Interestingly, in contrast to the response in MOBSCs, we found a significant increase in the levels of the neuronal protein Tuj1 (Fig. 6g, mean % increase = 89.2%, *t₍₄₎ = 3.39, p = 0.028). When normalising tau levels to Tuj1, we saw no significant loss of total tau (Fig. 6h, mean % decrease = 29.0%, t₍₄₎ = 2.04, p = 0.11), but a significant loss of p-tau Ser356 (Fig. 6i, mean % decrease = 55.7%, **t₍₄₎ = 4.81, p = 0.0086), indicating that WZ4003 treatment results in preferential lowering of p-tau Ser356 without incurring a loss of neuronal protein in HBSCs. We saw a trend for increased PSD95 in the WZ4003 treated cultures (Fig. 6j, mean % increase = 105%, t₍₄₎ = 2.37, p = 0.077), that was likely proportional to the rise in Tuj1 levels (Fig. 6k, t₍₄₎ = 0.739, p = 0.50). Overall, this demonstrates that WZ4003 treatment specifically reduces p-tau Ser356 in adult human brain tissue, relative to increases in neuronal and synaptic proteins.