The radiosynthesis of [18F]AV1451 was achieved in two steps following the previously published protocol by the Scott group [20] (see supplementary information).

Male SHR24 transgenic tau rats (SHR24) (n = 13) and their age-matched wildtype SHR littermates (SHRwt) (n = 15) were imported from Axon Neuroscience (Bratislava, Slovak Republic) at 13–14 months old (average weights, SHRwt = 392 g, SHR24 = 261 g) and allowed to acclimatise for a minimum of seven days before the scans. They were housed in Techniplast 2000P IVC cages with play tunnels and aspen bricks as enrichment, on a layer of Aspen bedding at a constant temperature (21 ± 2 °C) and fixed 12 h light–dark regime (lights on at 7:00 am). Food and water were available ad libitum. Further details of the health, number of rats used for each of the studies, the weights and injected doses are provided in Supplementary Table 2.

This research was regulated under the Animals (Scientific Procedures) Act 1986 Amendment Regulations 2012 following ethical review by the University of Cambridge Animal Welfare and Ethical Review Body (AWERB).

Rats were anaesthetised using isoflurane and femoral vein and artery of each rat was cannulated, as previously described [21] for the injection of the radiotracer and blood sampling respectively. Individual rats were placed prone on the warmed bed of the PET-CT scanner (Mediso nanoScan PET-CT, Hungary) and physiologically monitored during the scans.

A CT scan (helical acquisition, 360 projections, 50 kVp/980uA tube voltage, 300 ms exposure time and 1:4 binning) was performed and images were reconstructed with Cosine Filter Back Projection method. [18F]AV1451 was injected over ~ 20 s via the femoral vein cannula and a 60 min PET scan (packet timestamp list-mode, single field-of-view, 1–5 coincidence mode, 5 ns coincidence time window) of the head started simultaneously. The list mode data were re-binned in 22 time frames (6 × 10 s, 4 × 30 s, 4 × 60 s, 1 × 180 s, 4 × 300 s, 3 × 600 s) and dynamically reconstructed (Tera-Tomo 3-dimensional reconstruction [22] with 4 iterations and 6 subsets; 400–600 keV; 0.4 mm3 voxel size). Data were corrected for decay, attenuation, scatter and random coincidences.

Sixteen arterial blood samples were collected and placed on ice at approximately 5, 10, 15, 30, 45, 60, 75, 90, 105 s (~ 70 uL) and 2, 3, 5, 10, 15, 30 and 60 min (~ 150 uL) post injection. A 25 μL aliquot of whole blood was reserved, and remainder centrifuged (Eppendorf 5430-R centrifuge, at 5 °C, 5 min, 30,130 RCF) to obtain plasma. Radioactivity in plasma and whole blood (25 μL each) was measured using a gamma counter (Hidex AMG 425–601). Whole blood and plasma time-activity curves (TACs) were plotted and the results expressed as %ID/g. Remaining plasma samples (5, 10, 15, 30 and 60 min) were further processed (see supplementary information) for metabolite analysis. The parent radioligand concentration was expressed as a percentage of the total plasma radioactivity.

At the end of the scan, the animals were culled whilst under deep anaesthesia, brain dissected out, bisected along the sagittal plane, frozen over dry ice and stored at −80 °C. A biodistribution study was performed in six rats from each group (see supplementary information).

Post processing and kinetic modelling of the PET data was performed using PMOD software (v3.8; PMOD technologies, Zurich). CT images from individual rats were co-registered (rigid match) with an MR template (Schiffer T2 rat). The transformations were applied to individual PET scans and manually adjusted where required. 3D volumes of interest (VOIs) available on PMOD for the MR template (Px Rat W.Schiffer) were modified (see supplementary information) to obtain cortex, midbrain, brainstem, cerebellar grey, striatum and hippocampus VOIs (Fig. 1a and b). TACs were obtained for each of the VOIs and the results expressed as %ID/g. Area under the TACs were obtained (Graphpad Prism, version 9.4.1). Additionally, the ratio between the region and cerebellar grey was calculated and differences between groups analysed using 2-way ANOVA followed by post hoc Bonferroni test.

Representative PET scan images (%ID/g, 0–60 min summed) of [18F]AV1451 uptake in the brains of (a) SHRwt and (b) SHR24 rats co-registered with an MRI template and ROIs visualised: cortex (blue), midbrain (pink), brainstem (red), striatum (orange), hippocampus (purple), cerebellum grey (yellow) and cerebellum white (green). Average TACs (without error bars for visual clarity) of the brain regions of (c) SHRwt and (d) SHR24 show higher variability in uptake between regions in SHR24 rats. (e) Area under the curves of the groups show higher overall uptake in the brains of SHR24 rats. (f) Region to cerebellum grey ratio of 50–60 min post injection reveals that only the cortex has significantly higher uptake

Two-tissue compartment model (2-TCM), with the parameter for cerebral blood volume (VB) fixed at 3.6% [23], was fitted to the dynamic PET data of brain regions using PMOD software. This model was chosen following comparison of various models (1-TCM with VB free, 1-TCM with VB fixed at 3.6%, 2-TCM with VB free, 2-TCM with VB fixed at 3.6% and 2-TCM with VB fixed at 3.6% with added blood delay) via visual comparison and using Akaike information criterion values during preliminary analysis of the whole brain VOI. Blood and plasma TACs were used with linear interpolation. Group population average metabolite curve fractions were used for correcting individual plasma curves. The rate constants K1, k2, k3 and k4 were estimated from the curve fit and macro parameters, non-displaceable volume of distribution (VND = K1/k2), total volume of distribution (VT = K1/k2 * (1 + k3/k4)) and non-displaceable binding potential (BPND = k3/k4) were obtained.

Additionally, Logan graphical analysis (t* = 20, VB = 3.6%) was also used to estimate VT (VT Logan). To evaluate the possibility of using this model in longitudinal studies without blood sampling, Simplified reference tissue model (SRTM) and Logan reference tissue model (LRTM) were fitted using cerebellum grey as the pseudo-reference region. The cerebellum was chosen because it does not exhibit hyperphosphorylated tau or tau pathology at any age. The constant \(\mathrm\) from SRTM was used for LRTM. The correlation between VT Logan and VT and between k3/k4 and BPND from the reference tissue models were quantified using Pearson correlation coefficient r. The predictive value of VT Logan for VT and BPND from reference tissue models for k3/k4 from 2-TCM were evaluated using regression analysis.

The primary outcome measures were differences in VT and BPND from the cortex. Statistical analyses (t-test and 2-way ANOVA with Bonferroni correction for multiple comparison) were carried out using GraphPad Prism (v 9.1.2 and 9.4.1). Significance levels are denoted as *P < 0.05, ** P < 0.01, *** P < 0.001 and **** P < 0.0001.

Fresh frozen sagittal sections (30 μm) of the brain stored at −80 °C was used for autoradiography [24]. The sections were fixed in ice cold methanol (20 min) and incubated with 0.85 MBq/ml [18F]AV1451 in phosphate buffered saline (PBS) at 37 °C (1 h). Adjacent slides were similarly incubated with [18F]AV1451 in the presence of 10 μM [19F]AV1451. Slides were washed as follows: ice-cold PBS (1 min), 70% ethanol/30% PBS (2 min), 30% ethanol/70% PBS (1 min), ice-cold PBS (1 min) and dipped into ice-cold de-ionised water (1 s). The slides were dried in a stream of air and exposed to a phosphor-imaging plate (GE Healthcare) overnight. The plates were read on a phosphorimager (Dürr Medical CR35 Bio Digital Imaging System) and analysed using AIDA Image Analyser (v.4.27). Regions of interest were drawn on cortex, midbrain, brainstem, cerebellum (pseudo-reference region) and parietal/retrosplenial cortex (low uptake region) (Fig. 6c). [18F]AV1451 binding data from each region is expressed as a ratio to the binding on individual cerebellum. Statistical analyses of all autoradiography results were carried out with two-way repeated measure ANOVA with Bonferroni multiple-comparison correction. Statistical levels are denoted as *P < 0.05, ** P < 0.01, *** P < 0.001 and **** P < 0.0001. Five representative slides from the autoradiography study were stored at −80 °C until immunofluorescence and fluorescence staining experiments (see supplementary information).