AAV serotype 8 vectors for expression of the murine D377Y-PCSK9 cDNA (AAV-PCSK⁹) were produced as previously described (13) by using the two-plasmid method and cotransfecting AAV/D377Y-mPCSK9 (gift from Jacob Bentzon; Addgene plasmid # 58376) (10) together with the helper plasmid pDP8 (14) in HEK293T cells using polyethylenimine (Sigma Aldrich). AAV vectors were purified using iodixanol step gradients and titrated as previously described (15). Here, this vector is denoted as AAV-PCSK9^DY.

All animal care and experimental procedures were conducted in accordance with the NIH guidelines for care and use of laboratory animals and were approved by the local animal ethics committee (Ministerium für Landwirtschaft, ländliche Räume, Europa und Verbraucherschutz des Landes Schleswig-Holstein, Germany) under the application number 44-6/21. The results of all studies involving animals are reported in accordance with the ARRIVE guidelines (16). The group sizes of n=10 per condition were assessed by a power analysis (corrected α = 0.01, power 80%) by considering cerebral blood flow (CBF) (17). In total, 20 10- to 11-week-old male C57BL/6N mice (Janvier Labs, Germany) were used for this study. All mice were kept in groups of 2-3 individuals randomized by weight. Mice had ad libitum access to standard chow diet for a 2-week habituation period before starting the study and receiving the AAV-PCSK9^DY injections.

Mice received AAV-PCSK9^DY injections (2×10¹¹ VG in 100 µl) or the same volume of saline via the tail vein at an age of 10 to 11 weeks. Afterwards, AAV-PCSK9^DY-injected mice were fed a WD (WD, EF TD88137 mod. +1.25% cholesterol, Sniff Spezialdiäten GmbH, Metabolizable Energy 19.1 MJ/kg, 42kJ% fat, 15kJ% protein, 43kJ% carbohydrates; Crude Nutrients [%], crude protein (N x 6.25) 17.3, crude fat 21.1, crude fibre 5.0, crude ash 4.2, starch 13.4, sugar 34.1, N free extracts 48.7, cholesterol 12,950 mg/kg) for 3 months, while the saline-treated controls received standard maintenance chow (12). Body weight was monitored and body composition was determined at week 15 using the nuclear magnetic resonance (NMR) method (Minispec BCA analyzer, LF-110, Bruker) 2 h after the mice were transferred to the NMR room to acclimatize to the environment. For the measurements, mice were put into a restrainer, which was then placed in the analyzer (12, 17). Cognition of mice was determined from week 11 to week 14 by open-field (OF), object place recognition (OPR), elevated plus maze (EPM), as well as Barnes maze (BM) tests as previously described (12, 17). At the end of the study, CBF was determined by ASL-MRI at week 16 (17). Immediately after ASL-MRI measurements, blood samples (non-fasted) were collected by cardiac puncture. Mice were sacrificed and transcardially perfused with Ringer-heparin solution. Aortic arch, descending aorta, and brain were snap frozen in liquid nitrogen and stored at -80°C for further analyses. The protocol is summarized in Supplementary Figure S1.

The OF test was performed at week 11, the OF apparatus being 370 mm x 370 mm and light set to 19 lux. Mice were placed in the center of apparatus and allowed to freely explore for 10 min; the test was repeated for 3 days. Time spent in the center and immobile time were measured here.

EPM was performed at week 12 with a plus sign-shaped apparatus. The maze consisted of a plastic cross, placed 40 cm above the floor. The two opposing, closed arms of the EPM apparatus (30 cm x 5 cm x 16 cm) were dimly illuminated by a light bulb at 50 lux. The two opposing open arms of the EPM (30 cm x 5 cm x 0.5 cm) were brightly illuminated by a light bulb at 250 lux. The four arms were connected by a central platform (5 cm x 5 cm, illuminated at 100 lux). At the beginning of the test, the mouse was placed on the central platform, facing one of the closed arms and allowed to explore for 5 min. Immobile time, exploration events in open arm, and time spent in open or closed arm were measured.

OPR test was performed at week 11 by using an open box (370 mm x 370 mm, light set to 19 lux). Mice were habituated for 10 min on 3 consecutive days and, on the testing day, also placed in the center and allowed to explore two identical objects placed in two of the corners for 10 min. After 1 h, mice were allowed to explore the apparatus, in which one of the objects was relocated into another corner, for 5 min. The interaction with the displaced object was evaluated as preference index.

For BM testing at week 13-14, the mice were placed in the center to explore the brightly lit (300 lux) platform (90-cm diameter with 20 holes) for a maximum of 180 s to find the dark escape hole. Mice were gently led to the escape hole if they could not find it. During the training phases (for 5 consecutive days), the test was repeated 4 times for each mouse each day with an inter-trial interval of 15 min. approximately. Long-term and remote memory tests were performed at day 6 and 17, respectively; on these days the escape hole was closed and the maze orientation was turned 180°, then mice were allowed to explore for 90s. During all trials, the video analysis was made by semi-automatic measurements of primary latency (time until the escape hole was reached the first time), time mobile (the time the mice were mobile in the given 90sec during the experiment) and manual measurements of primary errors (number of head dips into wrong holes until first contact with the escape hole). Search strategy were categorized as defined by Harrison et al., 2006 as direct, serial or random and we determined the search strategy preference to find the escape hole (18). Example of search strategy category can be found in Supplementary Figure S2. In addition, the following secondary parameters were measured: mean speed distance travelled (distance that the mice has moved until the end of the test), secondary latency (during the learning phase, this is the time the mice required to solve the test, from second 0 until they enter into the escape hole), exit interaction^time (this is the total time the mice had been exploring the escape hole, regardless of whether the mice enters or not), and exit interaction^events (this is the total number of interactions with the escape hole, regardless of whether the mice enters or not). All procedures were recorded and analyzed with ANY-maze Video Tracking System ver. 4.114 (©1999-2013 Stoelting Co).

According to previous reports (17), the mice were scanned at a 7-Tesla small-animal MRI scanner (BioSpec, Bruker, Germany) using an 86-mm quadrature volume transmit-receive coil and 2x2 channel receive-only surface coil. Prior to the ASL sequence, a T2-weighted spin-echo sequence (TR=2500ms, TE=33ms, FOV=20x20mm², matrix=256x256, slice thickness=0.7mm) and 3D time-of-flight (TOF) (TR=12ms, TE=3ms, FOV=20x20mm², matrix=256x256, slab thickness=15mm) were performed. Perfusion was measured by continuous arterial spin labeling (CASL) with echo-planar imaging (EPI) readout (TR=4000 ms, TE=16 ms, FOV=25x25 mm², matrix=96x96, slice thickness=1mm). The mean inversion efficiency value was calculated in the carotids; thus, the result of the CASL-EPI is an absolute CBF map. T2-weighted image analysis for tissue perfusion was performed with ImageJ (U. S. National Institutes of Health, Bethesda, Maryland, USA). TOF angiography was performed in the brain with slices in transverse orientation and images were acquired in cross sections of the internal carotid artery (ICA). All images were analyzed with Fiji ImageJ.

For all atherosclerotic plaque analyses, investigators were blinded to the viral load. As described previously (12, 19, 20), serial cross sections (8- to 10-μm thick) were obtained, starting below the aortic root to the proximal aorta, below the aortic arch. The sections were stained using Oil Red-O (ORO) and the mean atherosclerotic lesion area was calculated from 8-10 sections at 40-μm intervals, starting at the appearance of at least two aortic valves until the aortic valves disappeared. Images were acquired using a Keyence microscope (BZ-X800). Sections were manually cropped using GIMP software, version 2.6 (The GIMP Development Team) to yield aortic root areas (12). Areas of lesions and ORO-positive regions were determined using an in-house Python script (available on request). In brief, the Python package OpenCV (https://pypi.org/project/opencv-python/) was utilized to process images and to determine lesions based on color thresholds (e.g., reddish pixel for ORO). The ratio of ORO-positive lesions in each animal was determined as the percentage lesion area, normalized to the total area of the aorta.

RNA sequencing (RNAseq) analysis was performed in the descending aorta. Aortic tissue was isolated, trimmed from fat, and snap frozen in liquid nitrogen. For the RNAseq analyses, the aortic segments of 2 mice were pooled and a total of 4 samples were analyzed. This means that the results represent 8 of the 10 mice used each group. Total RNA was isolated with the RNA Mini Kit from Bio& SELL (Nuremberg, Germany) combined with on-column DNase digestion (DNase-Free DNase Set, Qiagen) to avoid genomic DNA contamination. RNA and library preparation integrity were verified with LabChip Gx Touch 24 (Perkin Elmer). As input for VAHTS Stranded mRNA-seq Library preparation, 500 µg of total RNA was used following the manufacturer’s protocol (Vazyme). Sequencing was performed on a NextSeq2000 instrument (Illumina) with 1x72-bp single-end setup. The resulting raw reads were assessed for quality, adapter content, and duplication rates with FastQC (RRID : SCR_014583) (Andrews Simon: Andrews Simon: FastQC: a quality control tool for high-throughput sequence data. Available online at http://www.bioinformatics.babraham.ac.uk/projects/fastqc).

Trimmomatic version 0.39 was employed to trim reads after a quality drop below a mean of Q15 in a window of 5 nucleotides and keeping only filtered reads longer than 15 nucleotides (21). Reads were aligned versus Ensembl mouse genome version mm10 (Ensembl release 101) with STAR 2.7.10a (22). Aligned reads were filtered to remove duplicates with Picard 2.27.4 (Picard: A set of tools in Java for working with next-generation sequencing data in the BAM format), multi-mapping, ribosomal, or mitochondrial reads. Gene counts were established with featureCounts 2.0.4 by aggregating reads overlapping exons on the correct strand, excluding those overlapping multiple genes (23). The raw count matrix was normalized with DESeq2 version 1.36.0 (24). Contrasts were created with DESeq2 based on the raw count matrix. Genes were classified as significantly differentially expressed at average count > 5, multiple testing adjusted p-value < 0.05, and -0.585 < log2FC > 0.585. The Ensemble annotation was enriched with UniProt data (Activities at the Universal Protein Resource (UniProt)).

Brain immunofluorescence was performed on 20-μm-thick frozen sections. Brain sections were fixed with methanol and blocked in 1% bovine serum albumin (BSA) in phosphate-buffered saline (PBS) for 1 h. For CD31-, Col IV-, CD68- and PDFGrß-staining, sections were incubated in 1% BSA primary antibodies overnight (Supplementary Table S1). Then, sections were washed in PBS and incubated in DAPI (1ug/ml), and secondary antibodies (Supplementary Table S2) for 1 h. Immunofluorescent images were analyzed by considering 10 randomly selected different regions of interest in a blinded manner in hippocampus and amygdala. String vessels and pericytes were counted manually; the string vessels were considered as thin Col IV-positive and CD31-negative vessels, while pericytes include stuby Col IV-positive and CD31-negative vessels. In this analysis, stuby structures around DAPI-positive nuclei were not counted as pericytes. Pericyte analysis was confirmed by PDGFrβ/Col IV immunostaining, the number of pericytes was measured manually. The CD31/Col IV ratio and IgG/Col IV length ratio, was calculated with ImageJ as the area ratio of CD31-positive area over Col IV-positive area. Patchy blood-brain barrier leaks were detected with the IgG-positive areas outside of Col IV-positive vessels. These patchy blood-brain barrier leaks were counted manually and the area of lesion was measured with ImageJ. IgG deposition in the posterior cerebral artery was calculated as ratio of the length of IgG-positive length over the Col IV-positive length. CD68-positive perivascular macrophages were counted manually in the posterior cerebral artery defined by DAPI staining.

Plasma was prepared by centrifugation. Plasma concentrations of total cholesterol (TC) and triglycerides (TG) were analyzed in the mice as described previously (19) to determine whether an increasing viral load plus WD affected lipid profiles.

Plasma concentrations of diverse adipocytokines were measured in 10 µL of sample by immunosorbent assays according to the manufacturer’s instructions using the Bio-Plex 200 platform and the Milliplex mouse metabolic magnetic bead panel kit MMHMAG-44K.mouse (amylin active, C-peptide 2, ghrelin, GIP, GLP-1 active or GLP-1 total, glucagon, IL-6, insulin, leptin, MCP-1, PP, PYY, resistin, secretin, and TNF), and MHSTCMAG-70K (GM-CSF, IFNγ, IL-1α, IL-1β, IL-2, IL-4, IL-5, IL-6, IL-7, IL-10, IL-12 (p70), IL-13, IL-17A, KC/CXCL1, LIX, MCP-1, MIP-2, TNF).

GraphPad Prism 8.0 (La Jolla, USA) was used for statistical analysis. All data were checked for outliers with Rout (Q=1%) test and tested for Gaussian distribution and variance homogeneity with D’Agostino & Pearson test. A two-tailed Student’s t-test was used for comparing different groups, assuming a Gaussian distribution and variance homogeneity. In case of no variance homogeneity, a Welch correction was used. Alternatively, if Gaussian distribution was not given, we used the Mann-Whitney test. A p-value <0.05 was considered statistically significant. Strategy use per day was expressed as percentage of the three strategies (direct, serial, mixed) used within the 5 daily trials. Comparisons between conditions were conducted using Fisher’s exact chi quadrat test (SPSS Statistics 27, IBM, USA). To analyze the change in strategy used across days1-5, Friedman tests were applied (SPSS Statistics 27, IBM, USA). To analyze whether different search strategies differs at d5 between both groups we used Mann-Whitney testing (SPSS Statistics 27, IBM, USA). In box graphs (representing the 25th to 75th percentiles) with whiskers (representing the maximum and minimum values), both the individual data and the medians are presented. In the line graphs, means ± SDs are depicted. Correlation analyses were performed by a one-sided Pearson test.

We first characterized the development of obesity and atherosclerosis in the PCSK9^DY+WD mouse model. The final body weight of PCSK9^DY-treated mice increased by 21% on WD compared to mice fed standard chow (37.7 ± 2.5 vs. 31.2 ± 1.8g, p<0.0001, Figure 1A). The NMR analysis indicated that body weight increase was associated with an increase in fat mass (+96%) with no change in lean mass compared to controls. In addition, NMR scans revealed increased free body fluid (+26%, Figure 1B). The increased calorie intake (+23%) in the PCSK9^DY+WD mice caused obesity (Figure 1C). PCSK9^DY+WD mice developed leptin resistance as indicated by the positive correlation between plasma leptin and body weight (Figure 1D). The development of obesity in mice fed WD was also reflected by the significantly increased plasma levels of the hormones leptin, peptide YY (PYY), resistin, and glucose-dependent insulinotropic polypeptide (GIP) (Figure 2). However, it is unclear as to whether the mice also developed a glucose tolerance disorder as we could not detect any changes in plasma insulin or plasma C-peptide (Figure 2).

Plasma concentrations of TC and TG were increased 10- and 6-fold, respectively, in PCSK9^DY+WD mice as compared to controls (Figure 3A). Moreover, plaque burden and lipid content in the plaques were clearly visible in aortas of PCSK9^DY+WD mice, but not in controls (Figures 3C, D). These findings clearly show that atherosclerosis had developed in PCSK9^DY+WD mice. In addition, elevated plasma levels of TNF (5-fold) and IL-6 (2-fold) revealed systemic inflammation (Figure 3B). Other cytokines (IL-10, IL-12, GMCSF, and LIX) were not significantly different between the two treatment groups (Supplementary Figure S3). Remarkably, the genes expressed in aortic tissue in the two groups differed significantly (Supplementary Figure S4) as shown in the principal component analysis (PCA, Figure 4A) and a Volcano blot (Figure 4B). Importantly, RNAseq analyses in thoracic aortic segments point to an inflammatory response. KEGG analysis annotate the differentially expressed genes in PCSK9^DY+WD mice to various, inflammatory pathways (e.g., natural killer cell-mediated cytotoxicity, cell adhesion molecules, hematopoietic cell lineage, B-cell receptor signaling pathway, cytokine-cytokine receptor interaction, and NF-κB signaling pathway) (Figure 4C). The most prominent genes upregulated in the PCSK9^DY+WD mice were Cd5L, Stra6l, MMP12, H2-M2, Mcoln3, Il1a, Elovl3, SPP1, Itgad, and Clec1b (Figure 4D), the most significantly down-regulated genes Hamp, Lmod2, Myh6, Csrp3, Mb, Kcnj3, MYBPHL, Sin, Nppa, and Ckmt2 (Figure 4D) (25). Moreover, we detected upregulation of genes that are members of the matrix metalloproteinase family (MMP), vascular cell adhesion molecule (V-CAM), intercellular adhesion molecule (I-CAM), and the monocyte chemoattractant protein (MCP) families in aortas of PCSK9^DY-positive mice, while sirtuins (Sirt1-Sirt7) were rather not regulated (Figure 4E).

We did not detect anxiety-like behavior in PCSK9^DY+WD mice with the OF test (Figure 5A). The number of entries to the center area or the time spent in the center area or in the distance traveled in the test did not differ between controls and PCSK9^DY+WD mice. Only the time and number of immobility events was higher in PCSK9^DY+WD mice, which might be associated with obesity and less with anxious behavior. Furthermore, the EPM test did not show signs of anxiety-like behavior in the mice (Figure 5B): entry into the closed and open arms was similar and no differences were found in the time spent in the closed or open arms. When the animals were in the open arms, both groups demonstrated the same exploratory behavior in looking outside of the maze and those events lasted the same time.

The BM test was conducted 13-14 weeks after administering AAV-PCSK9^DY. For this test, the primary errors (number of errors before first contact with the escape hole), secondary errors (number of errors after first contact with the escape hole), primary latency (time needed for first contact with the escape hole), and, finally, the time mobile (total time until entry into the exit) were analyzed. During the training phase, number of primary errors was slightly lower in the PCSK9^DY+WD mice, while primary latency did not differ (Figure 6A). Furthermore, in the PCSK9^DY+WD mice, the secondary errors were increased, and the exit interaction^time as well as the exit interaction^events tended to be increased during the training period, respectively, while the mean speed was decreased. No difference was observed in distance travelled and secondary latency (Supplementary Figure S6). Friedman testing indicated that the search strategy was time-dependently changing during the 5 days of habituation in the controls (random p<0.001; serial p=0.01, direct p=0.001) as well as in the PCSK9^DY+WD mice (random p=0.02; serial p=0.516, direct p=0.002). Chi Quadrat testing indicated that the search strategy differed between the two groups only on d5 (p=0.004). Finally, Mann-Whitney-U-test, considering search strategies at d5, further indicated that the direct strategy was higher in the controls than in the PCSK9^DY+WD mice (p=0.014), while serial was higher in the PCSK9^DY+WD than in the controls (p=0.004) (Figure 6A). Long term memory in the PCSK9^DY+WD group was not affected as test performance at d6 was similar in both groups of mice. We detected no difference in primary errors, primary latency and mobile time. The search strategy at d6 seems to be different between the two groups, although a statistical analysis (Chi square testing) was not possible due to the too small group size (Figure 6B). Other parameters from the BM evaluation regarding long term memory (d6) and remote memory (d17) were not different between the two groups (Supplementary Figure S7). The finding that long-term memory was not affected is confirmed by the OPR test as we did not detect any differences in the index preference (Supplementary Figure S5). In contrast, significant functional impairments were observed at d17, representing remote memory. Number of primary errors (3.2-fold) and primary latency (2.5-fold) were higher in PCSK9^DY+WD mice. In addition, the search strategy changed: the direct strategy was reduced, while the serial and random strategy increased, although it must be mentioned here that a statistical evaluation was not possible due to the small group size (Figure 6C).

After acquiring MRI scans, we analyzed perfusion in the areas of interest. Those areas were defined according to signal intensity of the T2-weighted MRI images. When PCSK9^DY+WD and control mice were compared, no difference was observed total anterior detected, while the total posterior blood flow was slightly lower in PCSK9^DY+WD mice (-10%, Figures 7A, E). When investigating the blood flow in specific brain areas, we also found no differences in perfusion between the PCSK9^DY+WD mice and the controls for amygdala or motor prefrontal cortex (Figures 7B, F), consistent with the lack of anxiety-like behavior. In contrast, we found decreased perfusion in the cingulate cortex (-17%), caudate putamen (-15%), hippocampus (-10%), and thalamus (-13%) of PCSK9^DY+WD mice compared to controls (Figures 7C, D, G–I). According to TOF angiography of vessel cross sections of the ICA no differences in the lumen area (0.408 ± 0,024 mm² vs. 0.397 ± 0.015 mm²; p=0.690) were observed between the two experimental groups.

Brain slices were immunostained particularly to evaluate changes in the microvascular system of the hippocampus and amygdala. The CD31/Col IV ratio was selectively reduced in hippocampus (-40%, Figure 8A) of PCSK9^DY+WD mice while it was not in the amygdala (Supplementary Figure S8A). This indicates changes in microvascular structure. In addition, we found no differences in the CD31/Col IV ratio in cingulate cortex or caudate putamen of PCSK9^DY+WD or control mice (Supplementary Figure S9). Using higher magnifications of the microscopic CD31/Col IV images, we investigated the morphology of the vascular structures in more detail. Whereas there were no changes in the number of string vessels in the hippocampus of the PCSK9^DY+WD mice, the number of pericytes was indeed increased (+56%) in the hippocampus of the PCSK9^DY+WD (Figure 9A). Those vessels developing Col IV+/CD31 protuberances and dilated vessels with a large Col IV-positive area and small or absent CD31 were classified as pericytes. Again, an increase in pericytes was not seen in the amygdala (Supplementary Figure S10). The hippocampus-specific increase in pericytes was further confirmed by PDFGrß staining, as the number of positive cells was increased in PCSK9^DY+WD mice by 53% (Figure 9B). The size, but not the number of patchy blood-brain barrier leaks was higher (2.1-fold) in the PCSK9^DY+WD mice as detected by infiltration of IgG into the brain parenchyma (Figure 8B), but, again, not in the amygdala of PCSK9^DY+WD-positive mice compared to controls (Supplementary Figure S8B). Similarly, CD68 staining (+68%) selectively increased in the hippocampus (Figure 8C).