Glymphatic inhibition exacerbates tau propagation in an Alzheimer’s disease model

Generation of homozygous Thy1-hTau.P301S transgenic mice (line 2541, henceforth referred to as P301S mice) has been reported previously [2]. Mice were bred on a mixed C57BL/6JxCBA/ca background, and breeders transferred from the MRC Laboratory of Molecular Biology (Cambridge, UK) for colony establishment at UCL (London, UK). For the non-transgenic (wildtype) experiments, C57BL/6J mice were acquired from Charles River, UK. Animals were housed on a 12-hr light/dark cycle in groups of two to five in individually ventilated cages with ad libitum access to food and water and provided with environmental enrichment (cardboard houses and tubes, paper nesting). Equal numbers of both male and female animals were used in experimental groups. All animal work was performed in accordance with the UK’s Animals (Scientific Procedures) Act of 1986 and approved by UCL’s internal Animal Welfare and Ethical Review Board.

Glymphatic function was assessed in the mouse brain by infusion of a fluorescent tracer via the cisterna magna, followed by whole brain and brain slices imaging for quantification of CSF-brain influx, similar to that previously published [22]. Briefly, mice were anesthetized with 2% isoflurane in O₂ at a delivery rate of 1 l/min and positioned in a stereotaxic frame with the head flexed to ~ 50°. A midline incision was made at a midpoint between the skull base and the occipital margin to the first vertebrae. The underlying muscles were parted to expose the atlanto-occipital membrane and dura mater overlaying the cisterna magna, and a durotomy was performed using a 23-gauge needle. An intrathecal catheter (35–40 mm port length, 80–90 mm intravascular tippet length, Braintree Scientific, US) extended with polyethylene tubing (0.4 mm × 0.8 mm, Portex™, Smiths Medical, UK) and attached to a 100 µl glass Hamilton syringe driven by a microinfusion pump (sp210iw syringe pump, World Precision Instruments, US) was filled with low molecular weight fluorescent tracer, Texas Red™-conjugated dextran (TxR-d3, 3 kDa; ThermoFisher Scientific, UK) at 0.05% in filtered artificial CSF (Tocris, UK), advanced 1 mm into the cisternal space, and sealed and anchored in place with cyanoacrylate. Once dry, TxR-d3 was infused via the implanted cannula (2 µl/min over 5 min). 30 min after the start of the infusion, mice were killed by overdose with sodium pentobarbital (10 ml/kg, given intraperitoneally) and the brain was quickly removed.

Whole brain fluorescence was imaged immediately in excised brains (Ex. 570 nm, Em. 600 nm) using an IVIS Spectrum imaging system (PerkinElmer, UK). Brains were then drop-fixed in 4% paraformaldehyde (PFA) for 48 h, prior to cryoprotection (30% sucrose in phosphate-buffered saline (PBS) for 72 h), freeze embedding in Optimal Cutting Temperature (OCT) compound, and sectioning (100 μm sagittal sections) onto SuperFrost Plus™ microscope slides (VWR, UK) using a cryostat (CM3050 S, Leica, UK). Slices were mounted using Fluoromount-G™ with DAPI (Invitrogen, UK), and single plane images were captured on a Leica microscope (Leica DMi8) using a 10 × 1.3 NA objective with Texas Red and DAPI filter sets. Whole brain fluorescence in three aspects (dorsal, ventral and lateral) was measured using the Living Image Analysis Software (PerkinElmer, UK), and area coverage of Texas Red fluorescence in brain slice images was measured using ImageJ (v1.51j8).

Brain extracts for subsequent intracerebral infusion were prepared from end-stage (5-5.5 months-of-age) P301S and age-matched wildtype mice, as previously described [1]. Briefly, mice were killed by overdose with sodium pentobarbital (10 ml/kg, given intraperitoneally), brainstems were rapidly dissected on wet ice and tissue extracts snap frozen on dry ice for storage at -80 °C. Frozen brainstems from five mice were combined, weighed, and homogenised in 10% (w/v) sterile PBS (containing protease inhibitor cocktail, phosphatase inhibitor cocktails I and II (Sigma, UK), at a final dilution of 1:100, and 1 mM phenylmethylsulphonyl fluoride) by sonication (Fisherbrand™ Model Sonic Dismembrator, Fisher Scientific, UK). The resultant brain extract homogenates (henceforth referred to as either ^P301SBE or ^WTBE, originated from P301S or wildtype animals, respectively) were centrifuged at 3,000 g at 4 °C for 5 min, and the supernatants stored at -80 °C until characterisation and subsequent intracerebral infusion.

Total tau content was quantified in brain extracts by ELISA (Human Tau (total) ELISA kit (#KHB0041, Invitrogen, UK), see Supplementary methods) (31.2(± 2.8) µg/ml in ^P301SBE, 0.1(± 2.4) µg/ml in ^WTBE), and dot blots performed (see Supplementary methods) to confirm immunoreactivity of ^P301SBE (and not ^WTBE) with MC1 (abnormal AD tau conformation) and TG5 (total tau) antibodies (both a kind gift from the late Prof Peter Davies), and with the 4R tau specific RD4 (1E/A6) antibody (a kind gift from Prof Rohan de Silva [10]) (Suppl. Fig. 1). Additionally, presence of amyloid fibrils in ^P301SBE was verified by its fluorescence upon incubation with Thioflavin T (see Supplementary methods) (signal intensity (a.u.) 1.8(± 0.1) for ^P301SBE, 0.9(± 0.1) for ^WTBE). Prior to intracerebral infusion, ^P301SBE was adjusted to 20 µg/ml total tau (based on ELISA determined concentration) with homogenization buffer (^WTBE diluted by the same factor), briefly sonicated, and kept on ice until infusion.

Mice (2 months-of-age) were anesthetized with 2% isoflurane in O₂ at a delivery rate of 1 l/min and positioned in a stereotaxic frame in the horizontal skull position. Buprenorphine (Vetergesic, 0.1 mg/kg, subcutaneously) was administered for peri/post-operative analgesia, prior to making a midline incision on the top of the head to expose the skull. A burr hole was then made with a microdrill above the injection location (–2.5 mm anteroposterior, and + 2 mm mediolateral to bregma). 2.5 µl/site of either ^P301SBE or ^WTBE was infused at a rate of 0.5 µl/min into the hippocampus and overlying cortex (-1.8 and − 0.8 mm ventrodorsal to the brain surface respectively) using a 5 µl Hamilton syringe. After each infusion, the needle was left in situ for 5 min to prevent injectate reflux. The scalp was then sutured closed, and the animal left to recover in a heated recovery chamber until it had regained consciousness. It was then returned to its home cage, and its weight and features of the mouse grimace scale [29] monitored daily for 7 days post-surgery to ensure complete recovery.

Pharmacological inhibition of AQP4 was achieved using TGN-020 (N-1,3,4- thiadiazol-2-yl-3-pyridinecarboxamide, Tocris Bioscience, UK). Unless otherwise stated, mice were treated intraperitoneally with either TGN-020 (50 mg/kg in 20 ml/kg 0.9% NaCl (saline)) or vehicle (20 ml/kg saline). For chronic treatment studies, animals received injections of either TGN-020 or vehicle, 3 times per week during a 10-week period, starting ~ 3 days post-surgery, unless otherwise stated.

Mice were injected with an overdose of sodium pentobarbital (10 ml/kg, given intraperitoneally) and transcardially perfused with approximately 10 ml PBS immediately followed by the same volume of 4% PFA in PBS. Brains were extracted from the skull and drop-fixed in 4% PFA for 24 h at 4 °C. Brains were then cryopreserved (30% sucrose in PBS for 72 h at 4 °C), freeze embedded in OCT, and serial coronal sections collected (50 μm thick) onto SuperFrost Plus™ microscope slides (VWR, UK) using a cryostat (CM3050 S, Leica, UK). Slides were stored at -20 °C until further processing. On the day of staining, sections were brought to room temperature (RT) before blocking in 10% donkey serum (in 0.02% Triton X-100 PBS) for ~ 4 h at RT. Sections were then incubated with primary antibody (mouse anti-phospho-Tau (AT8, 1:500, Invitrogen, UK), rabbit anti-NeuN (D3S3I, 1:600, Cell Signalling, UK) or rabbit anti-AQP4 (1:500, Millipore, UK)), overnight at RT in blocking buffer. Slides were washed three times and incubated with secondary antibody (anti-mouse or anti-rabbit (depending on primary antibody host) conjugated to Alexa Fluor™ 568 (1:800, donkey, Invitrogen, UK)) for ~ 4 h at RT in the dark. Slides were then washed again three times and mounted under coverslips using Fluoromount-G™ with DAPI (Invitrogen, UK). Slides were kept in the dark until imaging.

Single plane, tiled images were captured on a fluorescence microscope (Leica DMi8) using a 10 × 1.3 NA objective. For analysis, using ImageJ (v1.51j8), a region-of-interest (ROI) was drawn on images encompassing the target region (e.g. hippocampus) in at least 6 sections from each brain, with the experimenter blind to animal experimental/treatment group. For tau, measures of signal intensity and % area coverage (after thresholding for immunoreactivity) were extracted from each and averaged across slides analysed from each brain. The interhemispheric intensity ratio (contralateral divided by ipsilateral intensity) was additionally calculated, as an indicator of transhemispheric propagation. For AQP4, polarisation was measured in images in addition to signal intensity, as has been previously described [16]. For this, the median immunofluorescence intensity of perivascular regions was measured. A threshold analysis was then used to measure the percentage of the region exhibiting AQP4 immunofluorescence greater than or equal to perivascular AQP4 immunofluorescence. Polarization was expressed as the percentage of the region that exhibited lower AQP4 immunoreactivity than the perivascular endfeet.

At the end of drug treatment studies (10-weeks post-inoculation), structural 3D MR images of mouse brains were acquired. For this, mice were anesthetized with 2% isoflurane in a mixture of air (0.4 l/min) and O₂ (0.1 l/min), and positioned into an MR-compatible stereotaxic frame, with 1.5% isoflurane delivered (in the same gas mixture) via a nose cone. A horizontal-bore 9.4T Bruker preclinical system (BioSpec 94/20 USR, Bruker, Germany) was used for acquisition of 3D structural T2-weighted MR images of the entirety of the mouse brain, using a volume coil and 4-channel surface coil for RF transmission and signal detection, respectively. A fast spin echo sequence with the following parameters was used: FOV = 16 × 16 × 16 mm, matrix = 128 × 128 × 128, repetition time = 700ms, echo time = 40ms, 2 averages.

For analysis, a multi-atlas brain structure parcellation framework was used [35] to extract the whole brain structural volumes. The raw brain MR images went through the pipeline of reorientation, intensity non-uniformity correction through the N4 bias correction algorithm, skull stripping, and whole-brain segmentation. The template images of an in vivo 3D digital atlas of the adult C57BL/6J mouse brain [36] were first registered affinely to the original MR image data using a block-matching algorithm [43]. The brain masks in the atlas were then transformed and resampled to create a consensus brain mask. A further non-rigid registration based on fast free-form deformation was then performed to correct any remaining local misalignment of the affinely registered atlas to the brain volumes [41]. The brain structure labels from the atlas were then transformed and resampled to the individual image space transformation and fused using the STEPS label fusion algorithm [26] to create the final parcellation. Hemispheric volumes were extracted for interrogation of interhemispheric differences, calculating the degree of interhemispheric volume ratio (ipsilateral/contralateral volume) change from 1 (hemispheric equality), and comparing this across groups/regions to determine if the degree of interhemispheric volume difference was altered with drug treatment in any brain region.

All data is presented as group means ± standard error of the mean (SEM), with individual animal datapoints shown where possible. Unpaired t-tests were used for comparisons made between two groups. For comparison of three or more groups, either 1-way, 2-way or 3-way Analysis of Variance (ANOVA) were used, depending on the dataset (ANOVA type denoted on individual graphs), with either Tukey’s multiple comparisons tests for individual comparisons, or Sidak’s multiple comparisons tests for grouped data. Differences in MRI data were analysed by multiple t-tests, correcting for multiple comparisons using the Holm-Sidak method, plotting -log(Adjusted P Value) against the magnitude of change of differences in volcano plots. For XY plots, the lines of best fit were smoothed using 2 neighbour datapoints. All graphs were constructed, and statistical comparisons made using GraphPad Prism (v8.1.2).

Studies have demonstrated that knockout model systems for AQP4 deletion reproducibly lead to significant decreases in glymphatic function and protein clearance [22, 38, 49]. In addition, we have previously demonstrated that acute selective pharmacological blockage of AQP4 channels, using TGN-020 [19, 20, 50] leads to a > 80% reduction in CSF-ISF exchange across different regions of the brain, and a corresponding decrease in solute clearance [16]. Our initial objective was therefore to investigate whether TGN-020 would have a similar effect when administered chronically. Adult wildtype mice (2 months-of-age) were treated with either TGN-020 or vehicle over a period of 3-weeks, with 3 doses/week (50 mg/kg TGN-020 in 20 mg/kg saline or 20 mg/ml saline alone, i.p.); concentration and dose based on our drug efficacy findings in relation to brain tau clearance (see Supplementary methods, Suppl. Fig. 2). Severe disruption of CSF-ISF exchange was observed in animals chronically treated with TGN-020 following cisterna magna infusion with TxR-d3, with only half of the fluorescent signal detected compared to vehicle-treated animals (p < 0.001, Fig. 1A, B). Further, when sectioned, a dramatic reduction in the percentage area of brain tissue covered by the CSF tracer was observed in drug treated mice (Fig. 1C, D): area under the curve of distance vs. signal plots indicated a 93(± 13) % decrease in tracer influx (Fig. 1E), with regions closest to the medial part of the brain (midline) being the most affected (Fig. 1D). These results demonstrate that chronic pharmacological blockade of AQP4 via TGN-020 treatment of adult mice significantly reduces CSF-ISF exchange and supresses brain-wide glymphatic function.

Recent studies suggest a relationship between tau accumulation, AQP4 polarisation and distribution, and glymphatic function [16, 23], highlighting the possible impact the latter can have on propagation of tau directly. Given this evidence, we set out to investigate the consequences of chronic glymphatic suppression, via pharmacological AQP4 blockade, on tau propagation, in an in vivo model setting. For this, we induced tau propagation in P301S mice, as has been previously described [1], and chronically treated animals with the AQP4 inhibitor, TGN-020, as above (3 doses/week, 50 mg/kg TGN-020 in 20 mg/kg saline or 20 mg/ml saline alone, i.p.), for 10 weeks, to determine the resultant effects on the propagation of tau in the brain.

Due to their endogenous expression of mutant human tau, ‘seeding’ of young P301S mice leads to robust and rapid development of tau pathology within days of inoculation, free from the confounds of regional transgene driven deposition [1]. Of note, this phenomenon is specific to transgenic P301S mice, as when wildtype mice are injected with ^P301SBE and left to age for 10-weeks (either treated with vehicle or TGN-020), brain tau pathology remains barely detectable by histology (Suppl. Fig. 3). AQP4 expression (see Supplementary methods) also appears to be unaffected by disease onset in ^P301SBE infused transgenic mice (Suppl. Fig. 4), supporting its targeting in this model for mechanistic study, as we have done here. Furthermore, in control (wildtype) animals, chronic treatment with TGN-020 has no negative or confounding effects on animal health, body weight (Suppl. Fig. 5A), cognitive performance (Suppl. Fig. 5B), or locomotor ability (Suppl. Fig. 5C) throughout treatment. And neither expression nor polarisation of AQP4 are affected by TGN-020 treatment in control animals (Suppl. Fig. 6).

Despite a noticeable increase in the amount of tau (AT8 immunopositive) pathology in the molecular layer of the hippocampus upon inoculation with tau (^P301SBE, containing pathogenic tau) (Fig. 2A), no difference in tau area coverage was observed between TGN-020 treated and vehicle treated ^P301SBE injected mice (Fig. 2B). When assessing AT8 immunopositive signal intensity however, TGN-020 treated ^P301SBE injected mice exhibited a greater signal than the vehicle treated ^P301SBE injected controls (p < 0.01, Fig. 2C). Furthermore, contralateral tau (propagation) in these TGN-020 treated mice was significantly greater than that in the contralateral hippocampus of vehicle treated ^P301SBE injected mice (p < 0.05, Fig. 2C), and was significantly greater, relative to the ipsilateral side, in drug treated compared to vehicle treated animals (p < 0.05, Fig. 2D). The discrepancy between tau pathology measured by area coverage compared to signal intensity would suggest that TGN-020 does not affect the number of cells burdened by tau, but rather results in greater intracellular burden, consistent with exacerbated propagation.

A similar trend towards increased tau immunopositive area with inoculation with tau (^P301SBE) was observed in the cortex (encompassing the somatosensory, visual and retrosplenial cortices, Fig. 2E). However, in this region of the brain there appeared to be no increase in AT8 immunopositive signal intensity with TGN-020 treatment, or any significant interhemispheric differences in tau pathology (Fig. 2F, G).

Together, these results show that chronic suppression of glymphatic function through AQP4 inhibition in a mouse model of tau propagation, exacerbates pathological spread of tau in the hippocampus of the brain.

Given that chronic pharmacological suppression of AQP4 in vivo leads to glymphatic impairment and increased propagation of tau, we next tested whether this led to any change in the behaviour of the animals. After only 4 weeks of TGN-020 treatment, we observed a marked decline of preference index in ^P301SBE injected mice compared to baseline in a NOR paradigm (p < 0.001 compared to baseline, Fig. 3A); a behavioural task dependent on the hippocampal formation and its cortical connections [3, 7, 48]. This substantial decline in object preference was greater than that observed in ^P301SBE injected vehicle treated mice (p < 0.05 compared to baseline, Fig. 3A), resulting in a significantly lower mean preference index in drug treated compared to vehicle treated animals at week 4 post-inoculation (p < 0.05, Fig. 3A), suggesting that the exacerbated effect of TGN-020 treatment on tau burden and spread has a negative impact on function of recognition memory in these animals.

In an open field arena 10 weeks post-inoculation, TGN-020 treated mice were on average less mobile in comparison to the other groups, with almost half of the drug treated mice remaining immobile for extended periods during the task (Fig. 3B), despite no significant difference in total path length, or time spent in the centre of the test arena, relative to the other groups (Suppl. Fig. 7A, B). Further, at this same timepoint, TGN-020 treated mice exhibited the highest overall limb clasping score amongst the groups (p < 0.01 compared to ^WTBE injected animals, Fig. 3C, D), indicative of a substantial decline in corticospinal function in these animals. We also attempted to perform the NOR task at this 10-week timepoint following drug treatment. However, we noticed that animals treated with TGN-020 spent insufficient time exploring either of the objects (Suppl. Fig. 7C), and as such this timepoint could not be reliably considered, as lack of interaction and interest with the objects resulted in confounding of the test. This behaviour could be attributed to animals’ immobility and inactivity, as evident from the above datasets (Fig. 3B-D).

It is proposed that tau propagation is related to synaptic connectivity rather that spatial proximity, and indeed, it has been shown that in the mouse model used in this study, tau pathology spreads to synaptically-connected extra-hippocampal/cortical areas [1]. In order to assess the extent of tau induced volume changes outside the hippocampus, we acquired structural magnetic resonance images of the whole brain and automatically parcellated them to extract volumes of gross neuroanatomical structures. For each brain region in each group, interhemispheric volume differences were examined by calculating the interhemispheric volume ratio (ipsi/contralateral volume), allowing appreciation of tau/drug induced changes both between hemispheres of a region of the brain and between brain regions.

Exacerbated interhemispheric volume differences were observed in TGN-020 treated ^P301SBE injected mice, as is evident from volcano plotting of the differences in interhemispheric volume ratio (Fig. 4). Significant ^P301SBE injected TGN-020 treated mice datapoints exhibit both an upper-left (accentuated ipsi < contralateral difference) and upper-right (accentuated ipsi > contralateral difference) shift away from the centre of the plot compared to vehicle treated ^P301SBE injected group points (Fig. 4). Not only are the degrees of significance of hemispheric differences exaggerated in TGN-020 treated mice compared to vehicle, but the magnitudes of effect also appear greater upon treatment with the drug. For example, in TGN-020 treated compared to vehicle treated ^P301SBE injected mice, we observed a 49% and 29% exacerbation of ipsilateral atrophy in the striatum and amygdala respectively (Suppl. Figs. 8, 9). Likewise, the hemispheric differences in anterior commissure and midbrain volumes (ipsi > contralateral) were magnified by 36% and 11% respectively in drug treated mice compared to vehicle (Suppl. Figs. 8, 9). These data suggest that TGN-020 aggravates tau induced volume changes outside the site of initial tau seeding, indicating intensified propagation of tauopathy as a result of glymphatic inhibition in the brain.

The above data suggests that chronic suppression of glymphatic function leads to a clear increase in the propagation of tau, resulting in discernible cerebral atrophic changes and cognitive behavioural outcomes in a mouse model of tau propagation. To investigate this phenomenon further and to gain insights into the role of the glymphatic system in disease progression, we conducted additional animal experiments in which we initiated drug treatment at either 2-weeks (early/mild stages of tauopathy) or 4-weeks (moderate stages of tauopathy) after tau inoculation [1] (Fig. 5A).

The intensity of hippocampal tau immunoreactivity in mice receiving TGN-020 immediately after tau inoculation was significantly greater than both the 2-week and 4-week delayed treatment groups (p < 0.05 in both comparisons, Fig. 5B), suggesting a critical influence of the glymphatic system on tau burden in the initial 2-weeks following ‘seeding’. Moreover, a significant linear trend was observed between the decline in tau immunoreactivity and the delay in TGN-020 treatment, indicating clear influence of the glymphatic system on the pathological accumulation of tau in the brain.

Following 10-weeks of TGN-020 treatment, ^P301SBE injected mice spend significantly less time exploring objects in a NOR task compared to earlier timepoints (Fig. 5C, Suppl. Fig. 7C). However, it appeared that motivation to perform the task was spared when treatment initiation was delayed (linear trend, R² = 0.5015, p < 0.001, Fig. 5C); ratio of time spent exploring the novel object at week 10 compared to baseline was significantly greater in both the 2-week and 4-week delayed groups (p < 0.05 and p < 0.01 respectively) compared to mice which received drug treatment immediately post-inoculation. Limb clasping behaviour was also significantly alleviated by delaying drug treatment (linear trend, R² = 0.3437, p < 0.05, Fig. 5D) with the 4-week delay treatment group exhibiting a significantly lower limb clasping score compared to the mice treated immediately (p < 0.05). Consistent with histological measures, these behavioural outcomes similarly suggest that the period immediately followed tau inoculation in this model is pivotal for glymphatic influence on tau pathology.

Upon examination of structural MR images of mouse brains at week 10, as per our previous findings, significant tau induced interhemispheric differences were observed in the striatum and amygdala (ipsi < contra), and in the midbrain, anterior commissure, fimbria and hippocampus (ipsi > contra) (Suppl. Figs. 8, 9). However, with each 2-week delay in treatment initiation, fewer brain regions appeared significantly altered between hemispheres (Suppl. Figs. 8, 9). This time dependent mitigation is particularly evident in the hippocampus (Fig. 5E) and amygdala (Fig. 5F), illustrating the significant trends towards hemispheric equality with treatment delay in these regions (linear trends, R² = 0.3561 and 0.3797 respectively, p < 0.01 in both).

Together these data indicate that glymphatic inhibition in the initial period following seeding results in severe impact on tau propagation and pathogenesis, and that the impact of glymphatic alteration diminishes with pathological progression.