Mice used in these experiments were described in our previous study (López et al., 2018) and were housed and bred in the animal facilities of Universidad Francisco de Vitoria (Pozuelo de Alarcón, Madrid, Spain). Experimental protocols met the European and Spanish regulations for protection of experimental animals (86/609/EEC and RD 1201/2005 and 53/2013) and were approved by the committee of Ethics for Animal Welfare of the Universidad Francisco de Vitoria and University of the Basque Country (M20/2015/093). Efforts were made to minimize the number and suffering of animals.

Mice co-expressing five familial Alzheimer’s disease mutations (5xFAD) were purchased from Jackson Laboratories (Bar Harbor, ME, United States; Oakley et al., 2006) on the C57BL/6J background and were mated with CB₂ ^EGFP/f/f and CB₂ ^−/− mice and backcrossed for at least ten generations to generate 5xFAD/CB₂ ^EGFP/f/f and 5xFAD/CB₂ ^−/− mice.

Prior to the experiment, mice were homogenously distributed per group according to bodyweight. A stock solution of 90 mg/ml RO6866945 (Roche Pharma Research and Early Development, Roche Innovation Center Basel, Basel, Switzerland) in ethanol was conserved at −20°C, and diluted in vehicle solution [5% ethanol, 5% kolliphor (Sigma, C5135), 90% NaCl 0.9% (Braun, 857367)] the day of use. 6 months old 5xFAD/CB₂ ^EGFP/f/f and 5xFAD/CB₂ ^−/− male mice were treated (i.p.) with RO6866945 10 mg/kg, or vehicle (VEH) daily for 28 days. RO6866945 ((3S)-1-[5-tert-butyl-3-[(4-methyl-1,2,5-oxadiazol-3-yl)methyl]triazolo[4,5-d]pyrimidin-7-yl]pyrrolidin-3-ol; CAS Registry Number 1433360-72-5) was synthesized as described in US20130116236 A1 (Example 136) (Adam et al. (2013). Preparation of [1,2,3]triazolo[4,5-d]pyrimidine derivatives useful as cannabinoid receptor 2 agonists, US20130116236 A1). It is a highly potent CB₂ agonist across species (human CB₂ cAMP EC₅₀ 0.2 nM, 104% efficacy; mouse CB₂ cAMP EC₅₀ 0.2 nM, 101% efficacy) which does neither interact with the CB₁ receptor in the cAMP (human CB₁ cAMP EC₅₀ > 10′000 nM) nor in the radioligand binding assay (human CB₁ Ki > 10′000 nM; Ouali Alami et al., 2018). RO6866945 exhibits an excellent early ADME profile including an oral bioavailability of 44% in mice and penetrates through the blood brain barrier.

Twenty-four hours before the end of the treatment, mice were intraperitoneally injected with 10 mg/kg methoxy-X04 (Tocris, 4920) in 15% DMSO, 15% kolliphor and 70% NaCl 0.9%. Then, mice were anaesthetised with 170 mg/kg ketamine (Richter Pharma, 580393.7) and 10.7 mg/kg xylazine (Calier, 572599.4) in NaCl 0.9%, and transcardially perfused with cold PBS pH 7.4. From each mouse, right cortex, hippocampi and cerebellum were dissected and stored at −80°C. Left cortex and the rest of the brain were immediately processed to isolate microglia for analysis by flow cytometry.

Three male CB₂ ^EGFP/f/f and three 5xFAD/CB₂ ^EGFP/f/f mice were anaesthetized with ketamine/xylazine (100mg/10 mg/kg body weight, intraperitoneal injection) and subsequently perfused transcardially at room temperature (RT) with 4% formaldehyde (freshly depolymerized from paraformaldehyde), 0.2% picric acid and 0.1% glutaraldehyde in PBS 0.1 M (pH 7.4) for 10–15 min. The brains were then removed from the skull, post-fixed in the fixative solution for 1 week at 4°C and cut into 50 μm thick coronal sections using a vibratome.

Our protocol previously published was used (Puente et al., 2019). Brain sections containing the subiculum were pre-incubated in a blocking solution of 10% bovine serum albumin (BSA), 0.02% saponin and 0.1% sodium azide in Tris-hydrogen chloride buffered saline (TBS 1X), for 30 min on a shaker at RT. Tissue was then incubated for 2 days at 4°C with both a rat monoclonal anti-GFP antibody (1:500, GF090R, Nacalai) and a rabbit polyclonal anti-Iba1 antibody (1:500, 019-19741, FUJIFILM Wako Pure Chemical Corporation) prepared in 10% BSA, 0.1% sodium azide and 0.004% saponin. After washes in 1% BSA/TBS, sections were incubated with 1.4 nm gold-conjugated goat anti-rat IgG antibody (Fab’ fragment, 1:100; Nanoprobes Inc., Yaphank, NY, United States) and with biotinylated anti-rabbit IgG antibody (1:200; Biotin-SP-AffiniPure donkey anti-rabbit IgG) diluted in 1% BSA/TBS with 0.004% saponin on a shaker for 4 h at RT. They were washed in 1% BSA/TBS and then incubated with the avidin-biotin peroxidase complex (1:50; Elite, Vector Laboratories, Burlingame, CA, United States) for 1.5 h at RT. Sections were then washed in 1% BSA/TBS and kept in the same washing solution overnight at 4°C, postfixed with 1% glutaraldehyde in TBS for 12 min at RT and washed in double distilled water. Gold particles were silver-intensified with the HQ Silver kit (Nanoprobes Inc., Yaphank, NY, United States) in the dark for 12 min at RT. The biotinylated antibody was exposed to 0.05% diaminobenzidine (pH 7.4) with 0.01% hydrogen peroxide for 3.5 min at RT. Sections were incubated with 1% osmium tetroxide, pH 7.4, in the dark for 20 min, washed in PB 0.1 M, dehydrated and embedded in Epon 812 resin. 50 nm-thick sections were cut with an ultra-diamond knife (Diatome United States) and collected on nickel mesh grids. They were counterstained with 2.5% lead citrate for 20 min and examined with a transmission electron microscope (JEOL JEM 1400 Plus, Canada). Tissue was photographed using a Hamamatsu FLASH digital camera inserted in the electron microscope. Anatomical landmarks were taken to locate the subiculum region.

To ensure homogeneous labelling between all samples, only the first 1.5 µm from the section surface of each specimen was collected. Random electron micrographs were taken of the subicula. Areas of 3,524 μm² in CB₂ ^EGFP/f/f and 4,078 μm² in 5xFAD/CB₂ ^EGFP/f/f mice were examined to assess CB₂ receptors in Iba1-positive microglia. GFP gold particles were counted and differentiated between their localization in membrane (between 0 and 30 nm of the membrane) or cytosol (more than 30 nm). Minor contrast and brightness adjustments were made to the figures using ImageJ software (NIH; RRID: SCR_003070), Adobe Photoshop and Gimp.

Flow cytometry was employed to determine the ability of microglial cells to phagocytize Aβ (stained with methoxy-X04), and the levels of CB₂ with RO7246360 probe (compound 3b in Sarott et al., 2020). 6-month-old animals were injected i. p. with Methoxy-X04 (Tocris Bioscience) at 10 mg/kg body weight. 24 h after injection, animals were deeply anesthetized by i. p. administration of a mixture of ketamine (170 mg/kg) and xylazine (10.7 mg/kg) and transcardially perfused with cold PBS 1X, pH 7.4. Brains were dissected and enzymatically digested to facilitate microglia separation. The cell suspension was mechanically dissociated and filtered through a 70 µm-cell strainer. Microglial cells, isolated by percoll gradient (GE Healthcare), were washed with PBS 1X and blocked with 1% BSA/PBS 1X for 20 min. Cells were stained with CD11b-PE and CD45-APC antibodies and with RO7246360 fluorescent probe for 40 min. Samples were read on a MACSQuant Flow Cytometer and analysed with MACS Quantify software (Miltenyi Biotec).

Debris and aggregates were eliminated from analysis by forward and side scatter characteristics. Then microglia were identified as CD11b⁺ CD45^lo. The CB₂ receptor expression was determined by the fluorescent signal of RO7246360 probe. Fluorescence signals were corrected by fluorescence minus one (FMO) control. For each hemisphere, approximately ten thousand CD11b + singlets were analysed.

Extracts from frozen brain cortices were obtained by homogenization in magnesium lysis buffer (MLB: 25 mM HEPES, pH 7.5, 150 mM NaCl, 1% Igepal CA-630, 10 mM MgCl₂, 1 mM EDTA) containing 10% glycerol, and protease and phosphatase inhibitors (1 mM Na₃VO_4, 25 mM NaF and protease inhibitor cocktail; Roche) and were maintained at 4°C. Homogenates were centrifuged at 12000 g for 20 min at 4°C and supernatants were collected to determine their protein content by BCA protein assay (Pierce™ BCA protein assay kit, Thermo Scientific). Homogenates were used to measure cAMP levels using an ELISA kit (cat.no. ab65355, Abcam) following the manufacturer’s instructions. Standards and samples were plated in duplicate, and the absorbance was measured at 450 nm using a Varioskan Flash multifunction plate reader (Sunrise, Tecan).

Frozen mouse brain cortices were homogenized in four volumes (weight: volume) of TBS extracting buffer (140 mM NaCl, 3 mM KCl, 25 mM Tris, pH 7.4, 5 mM EDTA and protease inhibitor cocktail; Roche). Homogenates were centrifuged at 16,000 g for 20 min at 4°C. The supernatants were saved to quantify the soluble Aβ_1-42 peptide fraction and the pellets were again homogenized in four volumes (weight: volume) of 5M guanidine 50 mM Tris-HCl pH 8. The supernatants obtained after the centrifugation step were collected to quantify the insoluble Aβ_1-42 peptide fraction. An equal volume of PBS containing 1 mM serine protease inhibitor AEBSF (Sigma) was added to all samples and their protein content was determined by micro-BCA protein assay (Micro BCA™ protein assay kit, Thermo Scientific). Human Aβ_1-42 Ultrasensitive ELISA kit (cat.no. KHB3544 Invitrogen) was used for the quantification of soluble and insoluble fractions of Aβ_1-42 peptide following the instructions provided by the manufacturer. Standards and samples were plated in duplicate, and the absorbance was measured at 450 nm using a Varioskan Flash multifunction plate reader (Sunrise, Tecan).

Extracts from frozen brain cortices were obtained following the procedure previously described for cAMP assay. Lysates (60 μg/lane) were separated by SDS-PAGE and transferred onto nitrocellulose membranes (BioRad) and PVDF membranes (BioRad, used for the transference of phosphorylated proteins). After blocking in 5% bovine serum albumin in TTBS (10 mM Tris pH 7.5, 150 mM NaCl, 0.1% Tween 20) membranes were incubated overnight at 4°C, as appropriate, with primary antibodies: anti-phospho-p38 MAPK (1:1000; Cell Signaling Technology, 4511T), anti-p38 MAPK (1:1000; Cell Signaling Technology, 8690T), anti-phospho-CREB (1:1000; Cell Signaling Technology, 9198S), anti-CREB (1:1000; Cell Signaling Technology, 9197T), anti-phospho-ERK1/2 MAPK (1:1000; Cell Signaling Technology, 9101S), anti-Erk1/2 MAPK (1:1000; Santa Cruz Biotechnology, sc514302), Anti-Iba1 (1:1000, FUJIFILM Wako Pure Chemical, 016-20001), Anti-APP N-terminus (1:1000, EMD Millipore, MAB348), Anti-APP C-terminus (1:2000, Sigma, A8717), anti-BACE1 (1:500, Abcam, ab 2077) and anti-GAPDH (1:1000; Abcam, ab8245). Membranes were incubated with corresponding horseradish peroxidase (HRP)-conjugated secondary antibody anti-mouse IgG-HRP (1:10000; Abcam, ab97046), anti-rabbit IgG-HRP (1:5000; Cell Signaling Technology, 7074S) and were developed using a chemoluminiscent reagent (Western Lighting ECL Plus, PerkinElmer, NEL103001EA) in the appropriate equipment (ChemiDoc, Bio-Rad). GAPDH was used as an internal control. The relative quantity of protein levels in western blot was measured using ImageJ software (ImageJ; NIH).

All statistical analyses were performed, and graphs were generated using GraphPad Prism v 9.0 (GraphPad). Graphs represent average values ± standard error of the mean. Normality of data distribution was determined with the Shapiro-Wilk or the D’Agostino-Pearson tests. For GFP labeling, data were analyzed by means of the Mann-Whitney U test. For the rest of determinations, data were analysed by means of two-way ANOVA, followed by Tukey’s post-hoc tests. A p-value < 0.05 was considered as statistically significant. Only male animals were used in the experiments. The number of animals used for each experiment is reported in the figure legends.

The GFP/CB₂ labelling was localized in Iba-1 immunopositive microglial processes in both CB₂ ^EGFP/f/f and 5xFAD/CB₂ ^EGFP/f/f mice (Figure 1). GFP-positive microglial processes increased significantly in 5xFAD/CB₂ ^EGFP/f/f (0.7126 ± 0.2311) relative to CB₂ ^EGFP/f/f (0.1648 ± 0.07686, *p: 0.0176; Figure 2). Likewise, a significant increase in the proportion of GFP-positive microglial ramifications was seen in 5xFAD/CB₂ ^EGFP/f/f (16.71 ± 3.664%) with respect to CB₂ ^EGFP/f/f (5.430 ± 2.631%; p = 0.0191; Figure 2). Also, the total number of GFP particles per area of microglial ramifications was significantly greater in 5xFAD/CB₂ ^EGFP/f/f (1.238 ± 0.2534) than in CB₂ ^EGFP/f/f mice (0.6962 ± 0.4138; p = 0.0467; Figure 2), and the number of GFP particles in microglial branches per 100 μm² was statistically higher in 5xFAD/CB₂ ^EGFP/f/f (0.8343 ± 0.2962) than in CB₂ ^EGFP/f/f (0.1648 ± 0.07686; p = 0.0176; Figure 2). Noticeably in 5xFAD/CB₂ ^EGFP/f/f, the percentage of GFP immunoparticles localized in microglial membranes (77.22 ± 11.40%) was significantly higher than the proportion distributed in the cytosol (22.78 ± 11.40%; p = 0.0106; Figure 2). As to CB₂ ^EGFP/f/f, 100% of the GFP particles were found in microglial membranes.

We then studied whether the chronic treatment with RO6866945 had an impact on the expression levels of cannabinoid CB₂ receptors. We used two different approaches: first, RT-PCR revealed no significant effects of the 28-days treatment with the agonist on CB₂ mRNA levels (F(1,23) = 0.6509, p = 0.4280) and confirmed the absence of CB₂ expression in samples from 5xFAD/CB₂ ^-/- mice (Figure 3A; F(1,23) = 437.6, p < 0.0001). Second, we employed flow cytometry to quantify the binding of the selective fluorescent probe RO7246360 to the CB₂ receptor; we found no changes induced by the chronic exposure to the agonist [F(1,22) = 0.02066, p = 0.8870] and confirmed the absence of CB₂ protein in isolated microglia from 5xFAD/CB₂ ^-/- mice (Figure 3B; F(1,22) = 55.62, p < 0.0001).

We next analyzed the signaling cascades affected by CB₂ activation or deletion (Figure 4). We found that the CB₂ agonist had a significant impact on cAMP levels [Figure 4A; F(1,19) = 8.851, p = 0.0078]. Post-hoc analysis revealed a decrease in cAMP in 5xFAD/CB₂ ^EGFP/f/f mice as a consequence of the treatment (p < 0.0001) that was absent in 5xFAD/CB₂ ^-/- mice (p = 0.9802). No differences due to the genotype were observed in vehicle-treated mice (p = 0.2537), although were significant between RO6866945-treated 5xFAD/CB₂ ^EGFP/f/f vs. 5xFAD/CB₂ ^-/- mice (p < 0.0001).

Regarding pCREB levels (Figure 4B), genotype had a significant effect [F(1,20) = 9.370, p = 0.0062]. Post-hoc analysis revealed that the agonist significantly decreased pCREB levels in 5xFAD/CB₂ ^EGFP/f/f mice (p = 0.0492) but not in 5xFAD/CB₂ ^-/- mice (p = 0.7765). No differences due to the genotype were observed in vehicle-treated mice (p = 0.9917), although were significant between RO6866945-treated 5xFAD/CB₂ ^EGFP/f/f vs. 5xFAD/CB₂ ^−/− mice (p = 0.0033).

p-p38MAPK levels (Figure 4C) were modified by the treatment with RO6866945 [F(1,20) = 21.77, p = 0.0001], by genotype [F(1,20) = 60.28, p < 0.0001] and by the interaction of both factors [F(1,20) = 6.088, p = 0.0228]. p-p38MAPK was increased in 5xFAD/CB₂ ^EGFP/f/f mice as a consequence of CB₂ activation by the agonist (p = 0.0003) and exhibited significantly lower levels in samples from both vehicle- and RO6866945-treated CB₂-lacking mice (p = 0.0064 and p < 0.0001, respectively). These observations highlight the selectivity of RO6866945 as a CB₂-selective agonist and suggest a putative constitutive activation of p-38MAPK signaling cascade by CB₂ receptors in the context of AD.

Finally, p-ERK levels remained unaltered after treatment with the agonist [F(1,18) = 2.377, p = 0.1405] as well as in 5xFAD/CB₂ ^−/− mice (Figure 4D; F(1,18) = 4.339, p = 0.0518).

As microglia are the main source of cannabinoid CB₂ receptors in the brain of 5xFAD/CB₂ ^EGFP/f/f mice, we analyzed the putative changes triggered in these cells by the activation of the receptor and by its genetic deletion (Figure 5). We found no changes in Iba1+ microglia (Figure 5A; F(1,20) = 0,7931, p = 0.3837) nor in its phagocytic activity (measured by its ability to internalize methoxy-X04-stained amyloid; Figures 5B,C; F(1,22) = 3.602, p = 0.0709) derived from CB₂ activation by the agonist. However, significant differences were evident between 5xFAD/CB₂ ^EGFP/f/f and 5xFAD/CB₂ ^-/- microglia; thus, we found a decrease in Iba1+ microglia abundance [Figure 5A; F(1,20) = 34.95, p < 0.0001] as well as an impairment in its phagocytic activity (Figure 5B; F(1,22) = 69.96, p < 0.0001).

We next measured the impact of CB₂ modulation on Aβ levels. To that end, we quantified several amyloid-related peptides (APP, C83 and BACE1) as well as the soluble and insoluble forms of Aβ_1-42, the main component of neuritic plaques (Figure 6). Our data showed no changes in APP [F(1,20) = 0.08911, p = 0.7684], C83 [F(1,20) = 0.1794, p = 0.6764] or BACE1 [F(1,20) = 3.026, p = 0.0973] after treatment with RO6866945. CB₂ deletion induced significant differences in protein levels of BACE1 [F(1,20) = 10.34, p = 0.0043], but not in C83 (F(1,20) = 1.705, p = 0.2065) and APP [F(1,20) = 2.468, p = 0.1319].

Levels of soluble amyloid were unaltered after treatment with the CB₂ agonist [F(1,14) = 0.2681, p = 0.6127] or genetic inactivation of the receptor [F(1,14) = 0.1408, p = 0.1731; Figure 6D]. Cortical amounts of insoluble amyloid, however, were significantly modified by both (Figure 6E). Thus, the treatment with RO6866945 led to a significant increase in insoluble amyloid levels [F(1,17) = 17.88, p = 0.0006] in 5xFAD/CB₂ ^EGFP/f/f mice, while 5xFAD/CB₂ ^−/− mice exhibited decreased levels of this peptide [F(1,17) = 209.3, p < 0.0001].