Synthesis and pharmacokinetic characterisation of a fluorine-18 labelled brain shuttle peptide fusion dimeric affibody

Brain positron emission tomography (PET) imaging with radiolabelled proteins is an emerging concept that potentially enables visualization of unique molecular targets in the brain. However, the pharmacokinetics and protein radiolabelling methods remain challenging. Here, we report the performance of an engineered, blood–brain barrier (BBB)-permeable affibody molecule that exhibits rapid clearance from the brain, which was radiolabelled using a unique fluorine-18 labelling method, a cell-free protein radiosynthesis (CFPRS) system. AS69, a small (14 kDa) dimeric affibody molecule that binds to the monomeric and oligomeric states of α-synuclein, was newly designed for brain delivery with an apolipoprotein E (ApoE)-derived brain shuttle peptide as AS69-ApoE (22 kDa). The radiolabelled products 18F-AS69 and 18F-AS69-ApoE were successfully synthesised using the CFPRS system. Notably, 18F-AS69-ApoE showed higher BBB permeability than 18F-AS69 in an ex vivo study at 10 and 30 min post injection and was partially cleared from the brain at 120 min post injection. These results suggest that small, a brain shuttle peptide-fused fluorine-18 labelled protein binders can potentially be utilised for brain molecular imaging.

peptide via low-density lipoprotein receptor-related protein 1 (LRP1). ApoE (159-167) 2 peptide recognises LRP1 expressed on blood endothelial cells, and the fusion protein is transported into the brain via RMT. (C) Schematic diagram of [ 18 F]FET incorporation by genetic expansion in the cell-free protein radiosynthesis system 30 . pCNF-RS = p-cyanophenylalanine aminoacyl-tRNA synthetase; ATP = adenosine triphosphate; AMP = adenosine monophosphate; PPi = pyrophosphoric acid. (D) The design of (a) non-radiolabelled and (b) radiolabelled protein in this study. 6*His = polyhistidine-tag. These sequences were inserted into a pET-21a plasmid vector for bacterial expression (a) or cell-free protein radiosynthesis (b).
Although several methods for protein radiolabelling with fluorine-18 have been reported, the development of mild, site-specific, and robust radiolabelling methods has been challenging 13,16 . As an option to solve these issues, we have developed a cell-free protein radiosynthesis (CFPRS) system for protein fluorine-18 labelling with a genetic expansion strategy using p-cyanophenylalanine aminoacyl-tRNA synthetase (pCNF-RS), tRNA optimised for amber codon recoding (tRNA CUA opt ), and O-2-[ 18 F]fluoroethyl-L-tyrosine ([ 18 F]FET). The CFPRS system enables site-specific fluorine-18 radiolabelling of various proteins with a high molar activity under mild conditions and with short reaction times through a simple addition of the template DNA 30 . In the biological translation process of the CFPRS system, the amber codon inserted after the start codon of the original sequence encodes the [ 18 F]FET, resulting in the facile radiolabelling of proteins (Fig. 1C,D). Using this CFPRS system, we successfully labelled an affibody molecule that binds to human epidermal growth factor receptor type 2 with fluorine-18 and demonstrated the utility of the construct in tumour PET imaging 30 .
In the present study, we determined the binding characteristics of AS69 and the newly designed AS69-ApoE construct in vitro and radiolabelled AS69 and AS69-ApoE with fluorine-18 (obtaining 18 F-AS69 and 18 F-AS69-ApoE) using the CFPRS system. We used 18 F-AS69 and 18 F-AS69-ApoE in wild-type mice to evaluate and compare their pharmacokinetics and to assess the brain penetration of the labelled proteins.

Results
In vitro characterisation of AS69 and AS69-ApoE. First, we investigated the binding selectivity of AS69 and AS69-ApoE against α-synuclein and Aβ 1-42 monomers through dot blot ( Fig. 2A). Both AS69 and AS69-ApoE showed selective binding to α-synuclein monomers but not to Aβ 1-42 monomers, suggesting that AS69-ApoE retained the binding properties of AS69. Subsequently, we evaluated the binding affinity of AS69 and AS69-ApoE against α-synuclein monomers (Fig. 2B). AS69-ApoE retained the binding affinity of AS69; however, its binding affinity was slightly improved. The EC 50 values were 7.05 µM for AS69 (95% CI 6.25-7.95 µM) and 4.01 µM for AS69-ApoE (95% CI 2.58-6.22 µM), which indicated the effect of ApoE peptide fusion to AS69 on its binding forms. Furthermore, we evaluated the binding properties of AS69 and AS69-ApoE against different conformations of α-synuclein (Fig. 2C). The morphological state of α-synuclein was monitored via ThT fluorescence, which reflected the amount of fibril structure present 31 . The signals of AS69 and AS69-ApoE binding declined in a time-dependent manner, suggesting that AS69 and AS69-ApoE bound to ThT-negative α-synuclein conformations but showed less binding to ThT-positive aggregated α-synuclein. 18 F-AS69-ApoE. 18 F-AS69 and 18 F-AS69-ApoE were successfully synthesised by the CFPRS system via the proposed mechanism described in Fig. 1C within 60 min, which included the purification steps. Analysis of purification integrity using SDS-PAGE is shown in Fig. 3A and Fig S1, which shows a single band in lane 1 and 2 derived from the purified, radioactive proteins, indicating the successful preparation of 18 F-AS69 and 18 F-AS69-ApoE. The radiochemical purity (radioactivity from main products/ radioactivity from total products) of each construct was estimated as > 95% by image analysis (Fig. 3B), and the decay-corrected radiochemical yields (RCY, radioactivity from main products/total radioactivity used) of 18 F-AS69 and 18 F-AS69-ApoE were 17.0% ± 1.98% and 2.79% ± 0.84%, immediately following His-tag purification. The estimated molar activity of the radiolabelled proteins was 498 ± 151 GBq/μmol after a 30-min reaction, based on the molar activity of [ 18 F]FET at the reaction start time (549 ± 174 GBq/μmol). The molar activity of each experiment is shown in Table S1.

Radiosynthesis of 18 F-AS69 and
Biodistribution study of 18 F-AS69 and 18 F-AS69-ApoE. Next, we compared the ex vivo biodistribution of 18     ]. The black and orange bars indicate the %ID/g value ± SEM in mice administered 18 F-AS69 and 18 F-AS69-ApoE, respectively. The injected dose in all experiments ranged from 185-370 kBq. Statistical significance was calculated using two-way ANOVA and Sidak's multiple comparisons test (*p < 0.05, **p < 0.01, and ***p < 0.001). (B) Comparison of time-activity curves in the brain and blood following the administration of 18 F-AS69 and 18 F-AS69-ApoE. The values of %ID/g in the blood (red dot) were fitted to a one-phase decay model, and the curve (red) was drawn. Black bars indicate the %ID/g value ± SEM. Statistical analysis was performed using two-way ANOVA and Sidak's multiple comparisons test (**p < 0.01). (C) PET images with 18 F-AS69 and 18 F-AS69-ApoE in the mouse brain (25-30 min post-injection, sagittal views). White arrows indicate the position of the mouse brain. www.nature.com/scientificreports/ Table S2. The brain accumulation of 18 F-AS69-ApoE was significantly higher than that of 18 F-AS69 at 10 min and 30 min post-injection (p = 0.045 and p = 0.0003, respectively). The ratios of %ID/g for the brain ( 18 F-AS69/ 18 F-AS69-ApoE) at each time point were 2.38 and 1.64, respectively. In addition, 18 F-AS69 was retained in the brain at 120 min post-injection; however, 18 F-AS69-ApoE levels were significantly reduced ( Fig. 4B, p = 0.0022). The estimated clearance rate of 18 F-AS69-ApoE from 60 to 120 min was 0.013 (%ID/g)/min (the difference between the %ID/g values at the 60 and 120 min/60 min), which indicates faster clearance of 18 F-AS69-ApoE from the brain. The tracer accumulation in the liver increased at the early time points, as predicted in the literature 26 . The blood concentration curve and tracer accumulation in the brain are presented in Fig. 4B and show that both 18 F-AS69 and 18 F-AS69-ApoE were rapidly cleared from the blood circulation within 10 min. The fitted biological half-lives of 18 F-AS69 and 18 F-AS69-ApoE were 2.67 min (95% CI 1.76-6.13 min) and 4.41 min (95% CI 3.29-6.71 min), respectively. Additionally, PET imaging was performed using 18 F-AS69 and 18 F-AS69-ApoE in wild-type mice (Fig. 4C). The images from the mice at 25-30 min after the injection of 18 F-AS69-ApoE showed increased SUV values in the brain compared with those from the 18 F-AS69-administered mice, which was consistent with the results from the ex vivo biodistribution study. PET images of other timeframes are shown in Fig  S2. In vitro stability of 18 F-AS69 and 18 F-AS69-ApoE. To support the idea that 18 F-AS69-ApoE, not its metabolites, crossed the BBB and showed radioactivity in the brain, we evaluated the stability of 18 F-AS69 and 18 F-AS69-ApoE in mouse plasma in vitro. Following a 60-min incubation of 18 F-AS69 and 18 F-AS69-ApoE in the mouse plasma, most radioactivity remained detectable in intact forms (Fig. 5A,B), indicating that both radiolabelled proteins were mostly stable in the mouse plasma as well as the 10-, 30-, and 60-min incubation results (Fig. S3). On the other hand, some weak signals were also observed at the top of the gels.
In vivo stability of 18 F-AS69 and 18 F-AS69-ApoE. To confirm the in vivo stability of 18 F-AS69-ApoE, plasma and urine samples were analysed via SDS-PAGE autoradiography at 30 min after intravenous administration of 18 F-AS69 or 18 F-AS69-ApoE into mice (Fig. 6A,B, Fig S4). We found that both 18 F-AS69 and 18 F-AS69-ApoE were predominantly detected in the plasma, suggesting that protein tracers can exist as their intact forms www.nature.com/scientificreports/ in the body. Furthermore, we also investigated excreted mouse urine, because the ex vivo results (Fig. 4A) and PET images (Fig. S2) showed quick accumulation in the kidney. No parent radiolabelled proteins were observed in the urine, while the dominant unknown radioactivity at the bottom of the gel was observed. It was not [ 18 F] FET because it did not migrate on the NuPAGE gel as described in Fig. S1. This result indicates that 18 F-AS69 and 18 F-AS69-ApoE were completely metabolised or degraded before urinary excretion.

Discussion
While brain PET imaging of senile plaques and neurofibrillary tangles with 18 F-labelled small molecules has been successfully utilised for the diagnosis and drug development of AD, there remain challenges to visualise other misfolded proteins found in neurodegenerative diseases. Typical small molecule PET tracers recognise the rigid β-sheet structure of amyloid fibrils; however, this can be a limitation for visualising other misfolded proteins (e.g., α-synuclein and TAR DNA-binding protein 43). Hence, alternative strategies using BBB-permeable protein radiotracers have the potential to overcome limitations of the conventional approach with their unique binding mechanisms and would provide new insights into the field of neuroscience and neuroimaging.
In this study, we investigated the possibility of brain PET imaging with fluorine-18-labelled proteins as the first attempt. Fluorine-18 is the most widely used radionuclide because its decay half-life (t 1/2 = 109.8 min) is appropriate for clinical applications 16 . On the other hand, we needed to design a protein molecule that reaches the brain and is cleared within the optimal time window of fluorine-18 (approximately 2 h). For this purpose, we focussed on affibody molecules that would be quickly eliminated from the bloodstream, and AS69 was selected as a potential candidate protein as an α-synuclein imaging tracer. We then designed a BBB-permeable affibody molecule, AS69-ApoE, by fusing the ApoE (159-167) 2 peptide (Fig. 1A). This molecular design did not affect the binding properties of AS69, as previously reported 27,28 , suggesting that this strategy might be applicable to other proteins. Furthermore, we found that AS69-ApoE exhibited a slightly stronger binding affinity against α-synuclein monomers than AS69, while maintaining binding selectivity (Fig. 2B). This may be due to the electrostatic interactions between the positively charged ApoE (159-167) 2 peptide (pI 12.70) and the negatively charged α-synuclein (pI 4.67). Our binding assay additionally showed that both AS69 and AS69-ApoE exhibited poor binding against α-synuclein samples when ThT fluorescence was detected (Fig. 2C), suggesting that AS69 www.nature.com/scientificreports/ and AS69-ApoE have low affinity against α-synuclein fibrils and limited utility for the visualisation of abnormal α-synuclein fibril accumulation in diseases 32 . The production of 18 F-AS69 and 18 F-AS69-ApoE using the CFPRS system was successfully confirmed via gel autoradiography (Fig. 3A). However, the RCY remained relatively low compared with that obtained using other methods in previous studies 30 , suggesting that further system optimisation is required to improve RCY. Furthermore, the RCY of 18 F-AS69-ApoE (2.79%) was less than that of 18 F-AS69 (17.0%), probably because of the difference between AS69 and AS69-ApoE in terms of certain biological factors (e.g., translation efficacy and stability of product proteins). This indicates that the efficiency of the CFPRS system depends on the protein of interest. Although we could not directly measure the molar activity of 18 F-AS69 and 18 F-AS69-ApoE, this value can be estimated from the molar activity of [ 18 F]FET because [ 18 F]FET is theoretically incorporated in the protein in a 1:1 stoichiometry. Hence, we expected to synthesise both 18 F-AS69 and 18 F-AS69-ApoE with high molar activity comparable to that of [ 18 F]FET (454 ± 144 GBq/μmol, immediately following 30 min of synthesis). This molar activity was remarkably higher than that from other methods of protein fluorine-18 labelling using N-succinimidyl 4-[ 18 F]fluorobenzoate or aluminium 18 F-fluoride-1,4,7-triazanonane-1,4,7-triacetate conjugate (typically 1-100 GBq/mol) [33][34][35][36] . This advantage allows for more sensitive target detection and high-contrast images. On the other hand, we cannot demonstrate in vitro binding experiments with both 18 F-AS69 and 18 F-AS69-ApoE. Based on the estimated molar activity, we used 50 nM of radiolabelled products, but this concentration was not sufficient to evaluate their binding against α-synuclein monomers (K D = 250 nM). For further investigations, we need a protein binder that possess low nanomolar affinity.
The BBB permeability of AS69 was increased by the addition of a partial peptide of ApoE (Fig. 4A,C), which was consistent with previous findings 25,26 . We confirmed more than 2%ID/g brain uptake of 18 F-AS69-ApoE at 30 min and 60 min post-injection. This value seems low compared with successful small radioligands for neuroimaging that showed over 4%ID/g in the brain within a few minutes of intravenous injection 7 . However, the PET images (Fig. S1) and ex vivo results (Fig. 4B) indicate that 18 F-AS69-ApoE gradually entered the brain, in contrast with the dynamic pharmacokinetics of small compounds. Hence, small protein tracers have unique pharmacokinetic characteristics and may be difficult to apply to the principles of small-molecule PET tracers. In support of this consideration, amyloid brain imaging with antibody-based protein binders has been achieved with < 0.5%ID/g brain radioactivity 10,13 . Furthermore, we can exclude the effect of the blood activity on the brain radioactivity because a higher brain uptake was observed in 18 F-AS69-ApoE despite having no significant difference between 18 F-AS69 and 18 F-AS69-ApoE in blood radioactivity at 30 min post-injection. Moreover, at earlier time points, blood activity was rapidly decreased from 2 to 10 min post-injection, but there was no obvious difference in brain uptake between 2 and 10 min post-injection, indicating that the effect of the blood radioactivity does not seem to be critical for the evaluation of radiotracer uptake into the brain. Additionally, the accumulation of AS69 in the liver was significantly increased at early time points in the biodistribution study, as predicted from the literature 26 , perhaps because of the cationic properties of the fused ApoE (159-167) 2 peptide. Since cell-permeating peptides are typically positively charged 37 , the ApoE (159-167) 2 peptide may also exhibit the propensity of cell-permeating peptides, accelerating accumulation in the liver.
Brain-targeting PET tracers must show rapid clearance from the brain to evaluate its retention at specific targets in the brain. 18 F-AS69 was retained in the brain from 60 to 120 min post injection, while the amount of 18 F-AS69-ApoE in the brain was reduced (Fig. 4B), suggesting that the clearance of 18 F-AS69-ApoE from the brain was initiated between 60 and 120 min after administration. To our knowledge, this is the first report to demonstrate enhanced clearance from the brain following the addition of a brain shuttle peptide. One hypothesis is that 18 F-AS69-ApoE in the brain tissue interacts with LRP1, which is expressed in the brain, and moves into the blood via the BBB according to its concentration (using the reverse pathway described in Fig. 1B), following the clearance of most of the 18 F-AS69-ApoE in the blood. This hypothesis implies that the BBB permeability of 18 F-AS69-ApoE is directly related to its clearance from the brain. This finding is based on the short biological half-lives of affibody molecules, unlike antibodies that exhibit longer biological half-lives. Hence, further optimisation of fused brain shuttle peptides will facilitate greater BBB permeation and subsequent rapid clearance.
For the quantitative measurement of target proteins in the brain, radiotracers must be stable, and their radioactive metabolites should not enter the brain 38 . Our biodistribution study showed no significant defluorination of the labelled proteins over 120 min following administration (Fig. 4A). However because both 18 F-AS69 and 18 F-AS69-ApoE were labelled with incorporated [ 18 F]FET located adjacent to the first methionine, [ 18 F]FET might be released from radiolabelled proteins by endogenous protease hydrolysis in vivo. [ 18 F]FET was shown to cross the BBB probably via a specific amino acid transport system 39 ; thus, [ 18 F]FET potentially appears to be a representative radiolabelled metabolite that merits consideration. As presented in Fig. 5, the in vitro stability assay of 18 F-AS69 and 18 F-AS69-ApoE showed that they were stable in the mouse plasma for at least 60 min. The signals at the top of the gel might be [ 18 F]FET that has the property to stay in the wells, or fluorine-18 labelled proteins that did not migrate into the gels by forming aggregates with other proteins. Notably, the in vivo stability assay demonstrated that both protein radiotracers are likely intact in mouse plasma at 30 min post-injection, supporting the idea that intact protein tracers entered the brain. However, no intact tracers were observed in mouse urine, indicating that protein tracers were thoroughly metabolised before urinary excretion. Most of the radioactivity in urine migrated into the gels was not [ 18 F]FET with a reason above, but could be small molecular weight peptides or molecules. On the other hand, the limitation of this study is its inability to directly evaluate metabolites in the brain with two issues in analysing the radioactive fraction from the brain after intravenous administration of radiotracers. First, it was difficult to separate all radioactive protein fractions from whole brain tissues. Second, the radioactivity in the brain was relatively low compared to the blood, liver, and kidney (Fig. 4), to perform further analysis with SDS-PAGE autoradiography or column chromatography. Further metabolism analyses are required to fully characterise the in vivo stability of protein radiotracers. www.nature.com/scientificreports/ Our goal in this study was to demonstrate in vivo brain PET imaging using mouse models and 18 F-labelled proteins within several hours. However, there remain challenges that need to be addressed. First, to our knowledge, there are no protein binders available that meet the following requirements: (1) sufficiently small size to enable rapid clearance from the body, (2) ability to be produced in the CFPRS system, and (3) high affinity (K D < 1 nM) and selectivity against target molecules in the brain. Second, although the ApoE (159-167) 2 peptide fusion improved the brain penetration of AS69, we cannot conclude that the fusion of an ApoE shuttle peptide to protein binders is the most effective approach for visualisation of the brain within several hours. Hence, our future study will focus on the identification of protein binders suitable for protein brain PET imaging and the consideration of another shuttle peptide fusion.
In conclusion, a newly designed protein fused with a brain shuttle peptide, AS69-ApoE, retained the binding affinity and selectivity of AS69. We successfully labelled AS69 and AS69-ApoE with fluorine-18 using our CFPRS methods. The results additionally suggest that the CFPRS system may be applicable to general protein molecules. Notably, 18 F-AS69-ApoE showed higher BBB permeability and faster clearance from the brain than 18 F-AS69. This indicates that the strategy of shuttle peptide fusion to small proteins such as affibody molecules is potentially applicable for neuroimaging. Further optimisation of shuttle peptides and identification of small protein binders will enable molecular brain PET imaging with fluorine-18-labelled proteins.
Preparation of α-synuclein, AS69, and AS69-ApoE. The pET-28a plasmid encoding human α-synuclein was a gift from Prof. Koji Sode of North Carolina State University (USA). The AS69 and AS69-ApoE genes were custom-synthesised (GeneScript, USA) and cloned into the expression vector pET-21a(+) into the NdeI/XhoI restriction site. Protein expression was induced by the addition of 0.5 mM isopropyl β-D-1thiogalactopyranoside when the optical density of the bacterial solution at 600 nm reached 0.5-0.7. Subsequently, the cells were collected via centrifugation and lysed via sonication. From the crude solution, α-synuclein was purified by heating, followed by anion-change chromatography and size-exclusion chromatography (SEC) using HiTrap Q HP and HiLoad Superdex 75 16/60 preparatory-grade columns (Cytiva, USA), respectively. AS69 and AS69-ApoE were purified using immobilized metal ion-adsorption chromatography and SEC with HisTrap FF and HiLoad Superdex 75 16/60 prep grade columns, respectively. Its purity was confirmed by sodium dodecyl sulphate polyacrylamide gel electrophoresis (SDS-PAGE).
In vitro binding characterisation of AS69 and AS69-ApoE. Monomeric Aβ 1-42 was prepared by dissolving the peptides in 100% anhydrous dimethyl sulfoxide at 5 mM, followed by dilution of this solution to 100 µM with phosphate-buffered saline (PBS). Monomeric α-synuclein was prepared in PBS following gel filtration. BSA was used as a control. The samples were blotted on a cellulose membrane (1 µL/dot). The membrane was blocked with 10% skim milk in Tris-buffered saline containing 0.1% Tween 20 (TBST). To evaluate the difference in binding affinity between AS69 and AS69-ApoE, diluted α-synuclein samples in PBS (100 µM, 50 µM, 10 µM, 5 µM, 1 µM, 500 nM, 100 nM, 50 nM, and 10 nM) were blotted onto a cellulose membrane (1 µL each). For the monitoring of α-synuclein fibrillation, α-synuclein (140 µM) was incubated in PBS at 37 °C for 0-7 days with 10 mM sodium azide; sampling was performed every day. Aliquots of the collected samples (5 µL) were added to 95 µL of 25 µM thioflavin T (ThT), and fluorescence was measured using an FP-6300 spectrofluorometer (JASCO, Japan) using excitation at 440 nm and emission at 480 nm. The membranes were incubated for 60 min with 280 nM AS69 or AS69-ApoE for the assessment of their binding. Membrane-bound AS69 and AS69-ApoE were probed using anti-His-tag antibody (Proteintech, Japan, 1:1000 in TBST, incubated for 1 h at 20 °C) and HRP-conjugated secondary antibodies (1:1000 in TBST, incubated for 30 min at room temperature). Three rounds of washing for 5 min were performed between each procedure. Anti-Aβ antibody (4G8, 1:1000) and anti-α-synuclein antibody (4D6, 1:1000) were used as positive controls. HRP-driven chemical luminescence from the membrane was imaged with an Ez-Capture MG luminescent image analyser (ATTO). For evaluation of binding, grey values of detected dots on the membrane were measured using ImageJ software and analysed using GraphPad Prism 5 (GraphPad, USA). www.nature.com/scientificreports/ Radiosynthesis of 18 F-AS69 and 18 F-AS69-ApoE. The CFPRS system was constructed using the RTS 100 E. coli HY Kit as previously described 30 . The additional components per 375 µL of the reaction mixture include pCNF-RS (300-375 µg), tRNA CUA opt (50.3 µg), RNase inhibitor (300 U), template pET-21a plasmids (4.5 µg, amber codon inserted as described in Fig. 1D by site-directed mutagenesis), and [ 18 F]FET. The prepared solution was allowed to react at 30 °C for 30 min. The synthesised proteins were purified using a His SpinTrap column (Cytiva) according to the manufacturer's instructions. For the experiments, the solvent was replaced with PBS using a NAP-5 column (Cytiva).

Radiosynthesis of [
Gel autoradiography. The radiochemical purity of the product was confirmed by SDS-PAGE based on the NuPAGE system (Invitrogen). To estimate the molecular weight of the products, Novex Sharp Pre-stained Protein Standard (Invitrogen) was loaded onto the first lane of each gel, and radioactive liquid was spotted on a fixed thin layer chromatography plate at the position corresponding to 10, 20, and 40 kDa standard bands. The gel was placed in contact with a BAS-IP TR 2025 imaging plate (Cytiva) overnight, and autoradiographic images were acquired using a Typhoon FLA 9500 laser scanner (Cytiva). The images were analysed using ImageQuant TL (Cytiva).

Ex vivo biodistribution experiments in mice.
All protocols using mice were approved by the Laboratory Animal Care Committee of Tohoku University and all animal experiments were performed in accordance with relevant guidelines and regulations including the ARRIVE guidelines. The mice (slc:ICR, male, 6-7 weeks) were anaesthetised, cervically dislocated, and dissected at defined time points (2, 10, 30, 60, and 120 min) after the intravenous injection of radiolabelled proteins (185-370 kBq/0.2 mL). The tissues (blood, brain, kidney, liver, and bone) were collected in vials (AS69-ApoE: 2 and 10 min (n = 3); AS69: 10, 30, and 120 min; AS69-ApoE: 30 min (n = 4); AS69-ApoE: 120 min (n = 5); AS69: 2 and 60 min; AS69-ApoE: 60 min (n = 7)]. The radioactivity of the respective tissues was measured using the gamma-ray counter AccFLEX γ7000 (Hitachi, Japan), and the counts were divided by the counts for injected radioactivity (decay-corrected) and the tissue weight for standardisation. The tissue weights were measured simultaneously using a gamma counter. In the figures, the standardised values are shown as the percentage of injected dose per gram (%ID/g). The values of %ID/g in the blood (red dot) were fitted to a one-phase decay model using GraphPad Prism 5. Results are shown as the mean ± standard error of the mean (SEM).

Small-animal PET imaging.
A PET study was performed using Clairvivo PET scanner (Shimadzu, Kyoto, Japan). Before the PET scans, the mice (slc:ICR, male, 6 w, n = 1) were anaesthetised with 1.5% (v/v) isoflurane. Emission scans were acquired for 120 min in three-dimensional (3D) list mode following intravenous administration of 18 F-AS69 (4.56 MBq) or 18 F-AS69-ApoE (3.34 MBq) dissolved in PBS via tail vein catheters. The resulting sinograms were reconstituted with the 3D-DRAMA algorithm into 25 frames (1 min × 5, 2 min × 5, 5 min × 9, and 10 min × 6). Standardised uptake value (SUV) images were obtained by normalising the tissue radioactivity concentrations according to the injected dose and body weight using the AMIDE software 41 . In vitro stability of protein tracers in mouse plasma. Mouse blood was collected via cardiac puncture using a syringe-connected needle previously wetted with heparin sodium (Mochida Pharmaceutical, Japan), followed by anaesthetisia of the mice and cervical dislocation for euthanasia, and centrifugation (1200×g) for 30 min at 4 °C. The cleared supernatants were carefully collected from the samples. The radiolabelled proteins (5 µL, 37-74 kBq) were added to 95 µL of PBS or mouse plasma and incubated for 60 min. The incubated samples (10 µL) were evaluated using gel-autoradiography as described above.
Statistical analysis. All statistical analyses were performed using GraphPad Prism 7. Significant differences were analysed with two-way ANOVA and Sidak's multiple comparisons test (*p < 0.05, **p < 0.01, and ***p < 0.001) for comparison between the radiolabelled proteins with respect to their brain uptake and pharmacokinetics, respectively.