Seven female domestic pigs (crossbreed of Landrace × Yorkshire × Duroc) weighing on average 27 ± 1 kg (ranging from 25 to 31 kg) and approximately 10–11 weeks old were used in the present study (Table 1). Animals were sourced from a local farm and prior to any experiments they were acclimatized for 7–9 days in an enriched environment. All animal procedures were performed following the European Commission’s Directive 2010/63/EU, approved by the Danish Council of Animal Ethics (Journal no. 2017-15-0201-01375), and complied with the ARRIVE guidelines. The overall design of the study is shown in Figure 1.

Pigs were injected in the medial prefrontal cortex (mPFC) with 25 μL of α-synuclein preformed fibrils (6.4 mg/mL), AD human brain homogenate (10% homogenate in saline), DLB human brain homogenate (10% homogenate in saline), or saline (Table 1). The details and characteristics of the preformed fibrils and human brains are provided in Supplementary Table 1. The substrates were injected in both hemispheres, as detailed for each pig in Table 1, and in accordance with our procedure for targeting mPFC (Jørgensen et al., 2017, 2018).

A detailed description of preparation, anesthesia and transport has previously been described by us (Jørgensen et al., 2021). Briefly, anesthesia was induced by intramuscular (IM) injection of Zoletil mixture and maintained with 10–15 mg/kg/h intravenous (IV) propofol infusion. Analgesia was achieved with 5 μg/kg/h fentanyl IV infusion. Endotracheal intubation allowed for ventilation with 34% oxygen in normal air at 10–12 mL/kg. The left and right femoral arteries were catheterized with Seldinger Arterial Catheter (Arrow International, Inc., Reading, PA, United States). The left and right superficial mammary veins and ear veins were also catheterized. A urinary catheter was placed to avoid discomfort and stress throughout the surgery and scanning schedule. The animals were monitored for heart rate, blood pressure, peripheral oxygen saturation (SpO₂), end-tidal CO₂ (EtCO₂), blood glucose, and temperature throughout the scan, except while undergoing MRI scans.

Intracerebral injections were performed using a modified stereotactic approach: An in-house instrument for modified stereotactic procedures containing a head-rest plate, a flexible arm attached with a micro-manipulator (World Precision Instruments, Sarasota, FL, United States), and a micro-syringe infusion pump system (World Precision Instruments, Sarasota, FL, United States) (Supplementary Figure 1). The flexible arm allowed the micro-manipulator to be positioned and fixed relative to the target entry point with a trajectory perpendicular to the skull, as illustrated in Supplementary Figure 1. For the first two experiments (Pig 1 and 2), we used a prototype of the device with slightly less arm flexibility and a different micro-manipulator brand and syringe-type, although the capacity, needle size, length, and tip shape were the same. However, the prototype did provide injections comparable to the remaining ones, as validated with immunohistochemistry.

After installation of local anesthesia, midline incision, and skull exposure, two burr holes were placed bilaterally, 25 mm anterior and 8 mm lateral to bregma, followed by hemostasis and dura puncture. We have previously validated this target point: 8, 25, 14 mm in the X, Y, Z coordinate relative to bregma, to center on gray matter in the mPFC (Jørgensen et al., 2017, 2018). The syringe and the needle were then positioned and fixed in a trajectory perpendicular to the skull and with the needle tip adjusted to the skull entry point. The syringes [250 μL SGE Gas-tight Teflon Luer Lock Syringes (World Precision Instruments, Sarasota, FL, United States) (different syringes for the different injectates)] were attached with SilFlex tubing (World Precision Instruments, Sarasota, FL, United States), NanoFil Injection Holder (World Precision Instruments, Sarasota, FL, United States) and 28 G Hamilton Kel-F hub blunt tip needle (Hamilton Central Europe, Giarmata, Romania). The SilFlex tubing and NanoFil Injection Holder were removed during homogenate injection because of the viscous content.

Using the micromanipulator, the needle was slowly advanced to the mPFC target point (perpendicular to the skull). The injection was performed over two steps with 10 and 15 μL injected 1 mm apart (centered at the mPFC target point). The infusion was delivered at 450 nL/min using the micro-syringe infusion pump followed by a 7-min pause before a slow withdrawal of the needle to avoid backflow. After the procedure, both burr holes were packed with an absorbable hemostatic gelatin sponge (Curaspon^®, CuraMedical BV, Assendelft, Netherlands), and the incision was sutured shut. The animals were then transported to the scanner facilities and connected to the respirator.

All pigs were PET-scanned with a Siemens high-resolution research tomograph (HRRT) scanner (Innovations/Siemens, Malvern, PA, United States). [¹¹C]PIB was prepared at the Copenhagen University Hospital, Rigshospitalet, as per routine clinical preparation. The complete method of preparation is explained in Supplementary Figure 2. Data acquisition lasted 90 min after bolus injection (over ∼20 s) of [¹¹C]PIB through one of the superficial mammary veins (IV). The injected dose was 448 ± 41 MBq, while injected mass was 4.82 ± 4.3 μg (mean ± SD).

Manual arterial blood samples were drawn at 2.5, 5, 10, 20, 30, 40, 50, 70, and 90 min after injection, while an ABSS autosampler (Allogg Technology, Strängnäs, Sweden) continuously measured arterial whole blood radioactivity during the first 20 min. The manual blood samples were measured for total radioactivity in whole blood and plasma using an automated gamma counter (Cobra 5003; Packard Instruments, Downers Grove, CT, United States) cross-calibrated against the HRRT. Radiolabeled parent and metabolite fractions were determined in plasma using an automatic column-switching radio-high performance liquid chromatography (HPLC) as previously described (Gillings, 2009), equipped with an extraction column Shim-pack MAYI-ODS (50 μm, 30 × 4.6 mm; Shimadzu Corporation, Kyoto, Japan) eluting with 50 mM HNa₂PO₄ pH 7.0 and 2% isopropanol (v/v) at a flow rate of 3 mL/min, and an Onyx Monolithic C18 analytical column (50 × 4.6 mm, Phenomenex, Torrance, CA, United States) eluting with 26% acetonitrile and 74% 100 mM HNa₂PO₄ pH 7.0 (v/v) at a flow rate of 3 mL/min. Before analysis by radio-HPLC, the plasma samples were filtered through a syringe filter (Whatman GD/X 13 mm, PVDF membrane, 0.45 m pore size; Frisenette ApS, Knebel, Denmark). Plasma was diluted 1:1 with the extraction buffer, and up to 4 mL of plasma sample was used. The eluent from the HPLC system was passed through the radiochemical detector (Posi-RAM Model 4; LabLogic, Sheffield, United Kingdom) for online detection of radioactive parent and metabolites. Eluents from the HPLC were collected with a fraction collector (Foxy Jr FC144; Teledyne, Thousand Oaks, CA, United States), and fractions were counted offline in a gamma well counter (2480 Wizard2 Automatic Gamma Counter, PerkinElmer, Turku, Finland). The parent fraction was determined as the percentage of the radioactivity of the parent to the total radioactivity collected. Examples of radio-HPLC chromatograms from a pig are shown in Supplementary Figure 3. The parent fraction curve from the last three HPLC scans could not be estimated because of a technical problem with the HPLC. For the final three scans, we used a population-based parent fraction curve derived from the first four scans (Table 1 and Supplementary Figure 6).

The integrity of the BBB post-intracerebral injection was assessed by determining the %-difference ΔT₁-map of the pre-gadolinium and the post-gadolinium scans. The MRI data were acquired on a 3T Prisma scanner (Siemens, Erlangen, Germany) using a human 64-channel head coil (active coil elements were HC3-7 and NC1). Three pigs were scanned in the MRI scanner as previously described by us (Jørgensen et al., 2021). The pigs underwent two T₁-map scans: pre-and post- gadolinium contrast injection. The protocol for T₁-weighted 3D magnetization-prepared rapid gradient-echo (MP-RAGE) MRI was: frequency direction = anterior to posterior; dimension = 144 × 256 × 256; slice thickness = 0.9 mm; repetition time = 2,000 ms; echo time = 2.58 ms; inversion time = 972 ms; flip angle = 8°; base resolution = 256 × 256, and acquisition time = 192 s. After the pre-gadolinium T₁-map scan, pigs received gadolinium IV [0.1 mmol/kg, Gadovist^® (gadobutrol), Bayer A/S, Copenhagen, Denmark] through a superficial mammary vein and were rescanned 5 min later with another T₁-map scan. Data were processed using a custom code in MATLAB 9.5.0 (R2018b) (The MathWorks Inc., Natick, MA, United States). DICOM files were converted to NIfTI-1 using dcm2niix (Li et al., 2016). The post-gadolinium T₁-map was co-registered and resliced to the pre-gadolinium T₁-map using SPM12. A%-difference map (ΔT1-map) was created from the resliced post-gadolinium and pre-gadolinium T₁-maps (Equation 1). Three regions in the ΔT₁-map were measured: left injection site, right injection site, and occipital cortex

At the end of the scanning, the animals were euthanized by IV injection of 20 mL pentobarbital/lidocaine. After euthanasia, the brains were removed, snap-frozen with powdered dry-ice, and stored at −80°C until further use. 20 μm coronal cryosections were sectioned on a cryostat (Thermo Scientific/Epredia™ CryoStar™ NX70 Cryostat, Shandon Diagnostics Limited, Runcorn, United Kingdom) and mounted on Superfrost Plus™ adhesion microscope slides (Thermo Fisher Scientific, Waltham, MA, United States). Sections were stored at −80°C for the remaining period of the study.

We performed [³H]PIB (Novandi Chemistry AB, Södertälje, Sweden, Molar activity: 78 Ci/mmol) autoradiography to calculate the total available binding sites (B_max) and equilibrium dissociation constant (K_D) in the injected pig brain, and compared this to the B_max and K_D of human brain regions that were used to create the homogenates. Saturation assays were performed using increasing concentrations of [³H]PIB for total binding and [³H]PIB + thioflavin S (100 μM) for non-specific binding on AD-homogenate-injected pig brain slices (n = 1, 0–5 nM of [³H]PIB), α-synuclein-preformed-fibril-injected pig brain slices (n = 1, 0 to 40 nM of [³H]PIB), and AD post-mortem human brain slices (n = 2 × 2 regions, 0–5 nM of [³H]PIB). Since there was no specific binding in DLB-homogenate-injected pig slices or DLB human brain slices, we could not perform saturation assays on these sections. Human brain slices were prepared in the same fashion as pig brain slices, including section thickness and storage. Detailed procedure for autoradiography is available in Supplementary Material.

The data were analyzed using GraphPad Prism (v. 9.2.0; GraphPad Software, San Diego, CA, United States). Non-linear regression analysis (Function: One site—total and non-specific binding) was used to calculate B_max and K_D values for all three assays. The fitting method used was the least squared regression with no weighting. In vitro binding potential (BP) was calculated with Equation 2.

PET list-mode emission files were reconstructed using the OP-3D-OSEM algorithm, including modeling the point-spread function, with 16 subsets, ten iterations, and standard corrections (Sureau et al., 2008). During reconstruction, attenuation correction was performed using the MAP-TR μ-map (Keller et al., 2013). Emission data were binned into time frames of increasing lengths: 6 × 10 s, 6 × 20 s, 4 × 30 s, 9 × 60 s, 2 × 180 s, 8 × 300 s, 3 × 600 s. Each time frame consisted of 207 planes of 256 × 256 voxels of 1.22 × 1.22 × 1.22 mm in size.

Brain parcellation was done with our previously published automatic PET-MR pig brain atlas method (Villadsen et al., 2017). The neocortex, occipital cortex, and cerebellum non-vermis (henceforth denoted as the cerebellum) were extracted from the Saikali atlas (Saikali et al., 2010) for the present study. Two additional regions for the injection site were hand-drawn on the atlas from an approximate injection site that was initially characterized around the site of needle penetration as visualized by the MRI scans and postmortem extracted brain. This was also further confirmed and optimized by positive immunohistochemistry slices from the region (Supplementary Figures 4, 5). Regions approximately 0.32–0.35 cm³ (∼ 250 voxels) in size were placed symmetrically in the left and right hemispheres. This region is slightly larger than the injection site itself, but this gives leeway for any potential mechanical error during the stereotactic operation. Wherever possible (not possible in, e.g., saline-injected regions), the automatic region was visually inspected with late-scan frames averaged.

Regional radioactivity concentration (kBq/mL) was normalized to injected dose (MBq) and corrected for the animal weight (kg) to provide standardized uptake values (SUV, g/mL) used to make average plots as in Figure 2. PMOD 3.7 (PMOD Technologies, Zurich, Switzerland) was used to visualize and create the representative PET and MR images.

Kinetic modeling was performed using kinfitr (v. 0.6.0) (Matheson, 2019; Tjerkaski et al., 2020) in R (v. 4.0.2; “Taking Off Again,” R core team, Vienna, Austria). The Logan graphical Analysis was applied to estimate the total volume of distribution (V_T) values (Logan et al., 1990), using a metabolite corrected input function derived from radioactivity measurements of arterial blood samples. Reference tissue modeling was performed with simplified reference tissue model 2 (SRTM2), with an average k’₂ calculated with SRTM (Lammertsma and Hume, 1996), to calculated non-displaceable binding potential (BP_ND) using the occipital cortex as the reference region (Yaqub et al., 2008). For more details on the kinetic modeling (see Supplementary Material).

Graph-Pad Prism (v. 9.2.0; GraphPad Software, San Diego, CA, United States) was used for data visualization and statistical analysis. All data are presented as mean values ± standard deviation. The difference in PET outcomes (Logan V_T and SRTM2 BP_ND) between the injected regions and reference tissues was calculated using the non-parametric Kruskal–Wallis one-way analysis of variance (ANOVA). For assessment of change in gadolinium-contrast MR, we used the Friedman non-parametric ANOVA test with paired testing. Post hoc ANOVA tests were corrected for multiple comparisons by Dunn’s multiple comparison test (Dunn, 1964).

After [¹¹C]PIB injection, we observed high brain uptake and rapid tracer wash-out. The blood and brain kinetics of [¹¹C]PIB were very fast, with less than 10% of the parent radioligand remaining in plasma after 2.5 min (Figure 2 and Supplementary Figure 6). We found higher radioactivity retention in the AD-homogenate- and α-synuclein-preformed-fibrils injected region (Figure 2A). Compared to the cerebellum and the occipital cortex, almost no retention was seen in DLB-homogenate and saline-injected regions (Figures 2A,B).

[¹¹C]PIB binding parameters from Logan graphical analysis and SRTM2 are summarized in Table 2. We found fourfold higher V_T values in the AD-homogenate injected region compared to the occipital cortex (p = 0.006) and twofold higher V_T values in the α-synuclein-preformed fibrils region (p = 0.034) (Figure 3A). We found no difference between the saline- and DLB-injected regions.

Compared to the cerebellum, the average BP_ND of 2.34 was higher (p = 0.004) in the AD-homogenate injected region, and the average BP_ND of 0.65 was also higher (p = 0.016) in the α-synuclein-preformed-fibrils injected region. There was no difference in BP_ND in the saline- or DLB-homogenate injected regions compared to the cerebellum (Figure 3B).

Using our minimally invasive method, we successfully injected all animals in the same symmetrical brain region. Prefrontal cortical immunostaining (α-synuclein and amyloid-β) and thioflavin S staining at the injection site confirmed the presence of α-synuclein preformed fibrils, AD homogenates, and DLB homogenates, respectively (Supplementary Material). To evaluate the appropriateness of our pig model, we compared B_max and K_D in both the pig and human brains. This was done for the α-synuclein-preformed fibrils and AD-homogenate injected pig brain regions as well as for the AD postmortem human brain slices using [³H]PIB autoradiography saturation assays (Figure 4 and Table 3). We determined B_max to be 477.2 fmol/mg TE and K_D of 12.07 nM in the α-synuclein-preformed fibrils region in pig brain slices (n = 1). We found a higher B_max on the AD postmortem human brain slices (366.7 fmol/mg TE, n = 3) compared to AD-homogenate-injected pig brain slices (233.4 fmol/mg TE, n = 1). However, the K_D is similar at 2.46 nM in AD-homogenate-injected pig brain slices vs. 2.54 nM in AD postmortem human brain slices. We also found 2.4 times higher binding potential in AD-homogenate-injected pig brain slices compared to α-synuclein-preformed fibrils-injected pig brain slices.

We found no statistically significant difference in the T₁-maps before and after gadolinium injection (Figure 5). Still, in cases, with local hemorrhage near the site of needle penetration (Figure 5A, red ROI), some regions had higher gadolinium uptake than the occipital cortex. Also, the amplitude of the [¹¹C]PIB time-activity curves (Figures 2B, 3) did not suggest that the injected regions had higher uptake compared to non-injured brain tissue. Finally, the uptake in saline-injected regions did not differ from that of the occipital cortex, supporting that the injection itself does not hamper the integrity of the BBB (Figures 5C,D).