In a previous work we assessed the effects of cannabinoids following intraventricular administration concomitant to Aβ injection for 7 days to rats 12 . This type of administration is not feasible in the Clinic, and is restricted to very few disease conditions. Other authors have used oral administration of 0.1 mg/kg/day tetra-hydrocannabinol (THC) effectively 30 . Therefore, in this work we have chosen to administer cannabinoids in the drinking water to investigate their effects in the genetic model of AD. Given that we aimed a preventive treatment, it was initiated at 7 months of age when they do not exhibit Aβ deposition in plaques 28 .

First we examined whether continuous oral treatment with a low dose (0.2 mg/kg/day for 4 months) of cannabinoids was able to rescue the cognitive impairment of Tg APP eleven months old mice in the novel object recognition test. Exploration activity was monitored during the training trial and all the animals displayed similar attention to the two identical objects. Wild type mice treated with vehicle were able to discriminate between the familiar object (Figure 1) and a novel object, exploring the latter for a longer time. WIN and JWH did not affect this behavior in wild type mice, even after prolonged treatment. As expected Tg APP mice did not distinguish both objects, showing cognitive impairment. In contrast Tg APP mice treated with the CB₂ selective cannabinoid JWH spent more time exploring the novel object, although WIN was ineffective at that respect.

¹⁸FDG uptake is significantly decreased in areas involved in learning and memory both in AD patients and in subjects showing mild cognitive impairment 31 . Moreover it is considered to be an early marker of the neurologic disease. Prolonged WIN treatment reduced the uptake in wild type mice (Figure 2 B, C, D), while JWH was without effect. As expected 12 month Tg APP vehicle treated mice had reduced ¹⁸FDG uptake (25-30% depending on the region of interest; Figure 2 A, B, C, D). JWH continuous administration fully reversed the reduction in glucose uptake in Tg APP mice (Figure 2 A, B, C, D), which was slightly higher than that of wild type mice (approximately 15%), while WIN was ineffective on that respect. Therefore JWH, the CB₂ selective agonist, was capable of improving cognitive impairment and of reversing the reduction in ¹⁸FDG uptake in cortical areas and hippocampus of Tg APP.

Neuroinflammation is reflected in AD and its transgenic models brain as astrogliosis and microglial activation, particularly surrounding senile plaques 26 28 . GFAP immunohistochemistry showed similar staining patterns in hippocampus in the different groups studied (Figure 3 A and 3B), irrespective of the genotype or the treatment received. Similarly GFAP protein levels, as assessed by Western blotting, were maintained across the different groups (Figure 3 C and 3D). Microglial cell staining and density was assessed by using ionized Ca²⁺-binding adaptor molecule-1 (Iba-1) antibody. The antibody labeled resting microglial cells in wild type mice, characterized by the fine tortuous branching, which was unaffected by either of the cannabinoids used in the oral chronic treatment (Figure 4, A). Tg APP vehicle treated mice showed an increase (53%) in Iba-1 positive microglial cells compared with wild type mice (Figure 4, A and 4B). While continuous WIN treatment did not alter this increased density, the CB₂ selective agonist JWH decreased cell density so that cell number accounted 83% of that of wild type mice (Figure 4 B). These results indicate that at this age Tg APP mice do not show astrogliosis, but they had microglial activation, which was effectively counteracted by JWH oral treatment.

Recent works have shown increased CB₂ microglial expression following a lesion or in a neurodegeneration context 32 , including AD brain 13 33 . We also addressed this issue by means of double immuno-fluorescence. We revealed CB₂ expression in Iba-1 microglial cells (additional file 1 A1-3) in quinolinic acid injected striatum (additional file 2) and not in the contralateral striatum (additional file 1, B1-3), as expected. However there was no evident CB₂ staining in any region examined of Tg APP mice (additional file 1, C1-3). Regarding CB₂ protein expression Tg APP mice had similar levels compared to wild type mice which also received vehicle (Figure 5 A and 5B). JWH administration, but not WIN, to wild type mice reduced (45%) CB₂ protein levels (Figure 5, B). Both WIN and JWH significantly decreased CB₂ protein (Figure 5, B) in the AD model, which attained very low levels (around a 75%-80% reduction).

We used Western analysis or qRT-PCR to assess possible changes in inflammatory parameters in cerebral cortex (Figure 5). JWH, but not WIN, decreased COX-2 expression in wild type mice (Figure 5 C and 5D). This parameter was markedly increased a 49% in Tg APP mice (Figure 5 D). Noteworthy, both cannabinoids effectively reversed COX-2 protein enhancement, reaching even lower levels than in wild type mice (40% and 60% respectively vs. wild type vehicle treated mice).

Different cytokines have been shown to be increased in AD brain and in its genetic model. IL-6 mRNA expression was similar across the different groups (Figure 5, E). As expected cortical TNF-α mRNA expression was dramatically increased (around 6 fold vs wild type vehicle treated mice; Figure 5 F). The cannabinoid agonists used in this study prevented this increase, although this cytokine expression continued to be higher than that of wild type mice (Figure 5, F).

Taken together these results demonstrate that continuous cannabinoid administration has anti-inflammatory activity given that different mediators were effectively counteracted in Tg APP.

The Tg APP model parallels the amyloidosis of the neurologic condition. Eleven month old Tg APP mice have no detectable insoluble (formic acid extracted) Aβ levels by ELISA. As expected no soluble human transgenic Aβ levels were detected in wild type mice, while Tg APP mice had high levels of Aβ_1-40 (Figure 6, A) and detectable levels of the more amyloidogenic Aβ_1-42 fragment (Figure 6, B). JWH decreased by 27% the levels of Aβ_1-40, and a 30% similar reduction in Aβ_1-42 was found following the administration of either WIN or JWH (Figure 6, A and 6B).

To determine the mechanism by which both cannabinoids reduce Aβ levels first we examined the effect of both compounds on Aβ release from cultured APP/PS1 glioma cells. There was no difference in the presence or the absence of drugs (data not shown) suggesting that neither the release nor the synthesis of Aβ was altered. Next we wondered whether the cannabinoids would reduce Aβ levels by enhancing its clearance, a therapeutic strategy intensively studied for AD treatment. Therefore we assessed Aβ transport across choroid plexus monolayers in vitro. Since there are contradictory reports on the presence of CB₁ receptors in choroid plexus 34 35 first we performed stainings with CB₁ and CB₂ antibodies. We revealed the presence of both receptors in rat choroid plexus (Figure 6, C). Then we examined the possible time course of the transport of Aβ_1-40 (5 μg/ml) added to the upper compartment, by measuring the amount of peptide in the lower compartment by Western blotting at different time points (Figure 6 D). In the absence of any drug the peptide was recovered at 12 h, and at longer times (24 and 48 h), after its addition to the upper compartment. However, the transport was much quicker in the presence of WIN or JWH (200 nM) and Aβ was mainly transported at shorter times (1-3 h; Figure 6 D). To quantify this transport we took advantage of ELISA assays and assessed peptide levels at 1 and 3 h after the addition of Aβ to the cell cultures (Figure 6 E). WIN and JWH promoted the same peptide transport both at 1 and 3 h, while almost no detectable Aβ crossed the choroid plexus cell monolayer in their absence (Figure 6 E). In summary, prolonged oral administration of cannabinoids effectively reduced Aβ levels in Tg APP brain, likely due to increased peptide clearance through the blood- or the CSF-barrier.

AD is characterized by the existence of increased tau phosphorylation, mainly by the action of GSK3-β, which is deregulated. We assessed by Western blotting total and pSer9-GSK3-β (inactive form) levels in cerebral cortex, from Wt and Tg APP mice following the different treatments. Cannabinoid agonist treatment did not change p-Ser9 GSK3-β protein levels in wild type mice. In contrast, there was a reduction (around 40%) in p-GSK3-β in the AD model which was normalized by WIN administration, but not JWH (28% decrease vs wild type vehicle treated mice). Total GSK3-β protein levels remained largely unchanged. Similarly no changes in the levels of either p-Ser21 GSK3-α or total GSK3-α protein levels were observed (Figure 7). Therefore, the pathological pSer9-GSK3-β activity was normalized by the CB₁/CB₂ mixed agonist WIN.

Tg APP transgenic mice were obtained via heterozygous breeding of mice expressing the 695 aa long isoform of the human APP containing a double mutation Lys 670-Asn, Met 671-Leu 52 (swedish mutation) under transcriptional control of the hamster prion promoter on a C57/BL6 breeding background. Male Tg APP and wild type littermates, used as controls, were 7 months age at the beginning of the experiments. Mice were group-housed (4-5 animals per cage) with a 12:12 h light/dark cycle and with ad libitum access to food and water. All of the experiments were performed according to ethical regulations on the use and welfare of experimental animals of the European Union and the Spanish Ministry of Agriculture, and the procedures were approved by the bioethical committee of the CSIC.

WIN 55,212-2 (WIN) and JWH-133 (JWH) were administered in the drinking water at a dose of 0.2 mg/kg/day using ethanol (0.1%) as vehicle. The amount of water drank by the animals was assessed every other day and the treatment was adjusted to their weight. There was no difference in the body weight or the ingested water between groups, all along the experiment, discarding a possible reinforcing effect of cannabinoids.

Animals were sacrificed by cervical dislocation followed by decapitation at 11 months of age. The brain was saggitally divided. One hemisphere was rapidly dissected on a cold plate, frozen on dry ice and stored at -80°C until assayed. The other hemisphere was immersion fixed in PF 4% in PB 0.1 M for 24 h, cryoprotected in sucrose 15% (24 h) and 30% (24 h) in PB, snap frozen in hexane (-60°C) and stored at -20°C until cut with a sliding microtome.

The arena measured 40 × 40 cm, surrounded by 30-cm-high perimeter black walls, that was located in an isolated room that was novel to the animals. The floor of the arena was covered with used sawdust. The arena was monitored by a video-camera located above the arena. The procedure consisted in three visits to the arena in subsequent days. The first day mice were placed into the empty arena for 15 min (habituation). The second day they were allowed to explore two identical objects during two 10 min trials 5 min apart (training). On the third day (test), one of the objects was changed by a novel one, with different shape and color, and the mice explored the arena for 10 min. Data collection was carried out by the Ethovision software (Noldus, The Netherlands) and exploring was defined as "directing the nose at a distance equal to or less than 3 cm from the object or touching it with the nose". The time spent exploring the familiar object and the new object was expressed as a percentage. Objects were cleaned before every exposure with acetic acid 0.1% to prevent any olfactory clues. Experiments were performed at the same time of the day (9.00-14.00 h).

Fasted mice were anesthetized with isofluorane (2%) and injected (ip) with ¹⁸FDG (11,1 MBq or 300 μCi/200 μl saline, PET Technologic Institute, Madrid). Thirty min later ¹⁸FDG uptake images of each mouse were acquired for 30 min by PET imaging (Albira PET, 8 detectors, Gem-Imaging, Spain; 71 ). The regions of interest were previously delineated in magnetic resonance (MR) T2-weighted images (Bruker Biospin, Germany) of each animal. Quantification of the metabolic activity was performed by co-registering the PET images of the brains to their own MR image as described by 72 . In our case, the field of view (FOV) of the PET scanner is 80 × 80 × 40 mm and the number of pixels of the reconstructed tomographic image is 160 × 160 × 80 pixels, being the voxel size 0.5 mm³. Co-registration of the PET image to the MRI, leads to a reduction of the pixel size to 0.2 mm³, given the trilinear interpolation done by the PMOD software (PMOD Technologies, version 2.9, Switzerland). The ¹⁸FDG uptake of the different brain areas were normalized to the ¹⁸FDG uptake in the cerebellum (considered as reference region).

Immunostaining was performed on floating sections (35 μm) as described previously 12 . In brief, following several washes with PBS, the endogenous peroxidase was blocked (3% hydrogen peroxide in methanol), washed again, and incubated for 90 min in PB containing 0.2% Triton × 100 and 10% normal goat serum. Sections were incubated with the different antibodies in PB containing 0.2% Triton × 100 and 1% normal goat serum overnight at 4°C. Dilutions of antibodies were as follows: anti-GFAP (1:1500, Sigma); anti-Iba-1 (1:1000, WAKO) (additional file 2). Development was conducted by the ABC method (Pierce, Rockford, IL), and immunoreactivity was visualized by 3,3-diaminobenzidine oxidation as chromogen, with or without nickel enhancement. Images were acquired with Zeiss Axiocam high resolution digital color camera, using the same settings and the segmentation parameters (MCID software, InterFocus Imaging, UK) were constant for a given marker and experiment. The mean value for each animal per region results from the analysis of 5-6 sections.

Aβ measurements were performed by two ELISA kits, one for each fragment (Aβ1-40 and Aβ1-42), from Biosource following the manufacturer instructions. The samples were sonicated (5 sec) in 10 vol of protein lysis buffer containing protease inhibitors (see Western blotting for details). The lysate was centrifuged (18,000 × g, for 10 min at 4°C). The supernatant was considered soluble and the pellets were further extracted by sonicating with formic acid and centrifuged. Prior to the Elisa the unsoluble samples were diluted 3 times with Tris 1 M, pH 10, while the soluble samples were diluted 5 times in Elisa buffer. The results were expressed as pg/mg of protein measured by the Bradford method, using BSA as standard.

A double-chamber choroid plexus epithelial cell culture system mimicking the blood-cerebrospinal fluid (CSF) interface was used for in vitro studies, as previously described 59 . After seeding, the cells were incubated for 24 h and thereafter Aβ_1-40 (5 μg/ml) was added to the lower chamber in the absence or presence of the cannabinoid agonists (500 nM). At different time points the medium was collected from the upper chamber, and the Aβ_1-40 content was determined by immunoblotting. In other experiments Aβ_1-40 levels were assessed by selective Elisa kits (Biosource) as described above.

Western blot was performed as described previously 61 . In brief, tissues were sonicated in lysis buffer, samples were centrifuged at high speed for 10 min, and supernatants were collected. Total protein was assessed by the Bio-Rad (Hercules, CA) protein assay. An aliquot of each sample (40 μg of protein) was separated by SDS-PAGE (10%), and proteins were transferred from the gels onto nitrocellulose membranes. The blots were blocked with 1% defatted dry milk for 1 h at room temperature and incubated overnight at 4°C with the following antibodies: anti-GFAP (1:10.000, DAKO), anti-CB₂ (1:10.000, Affinity Bioreagents); anti-COX2 (1:100, Abcam); anti-p-GSK3β (1:5000, BD); anti-GSK3β (1:1.500, Cell Signaling) (additional file 2). Finally, samples were subjected to enhanced chemiluminescence and densitometric analysis. Band densitometric analysis was performed by Quantity One quantitation software (version 5.0; Bio-Rad) from film exposures; the background was subtracted, and the optical density percentage was obtained considering 100% the control samples within the same film. Tubulin was used as loading control. Every membrane contained samples from each treatment group, for comparison purposes.

The specificity of the CB₂ antibodies used (additional file 2) was assessed by preincubating with the human antigenic peptide (2 μ/ml overnight at 4°C under agitation), as stated in additional file 2, before its use in Western blotting, which resulted in blockade of the immunoreactive signal (see additional file 3).

Total RNA from cortex was extracted using TRIzol reagent according to the manufacturer's instructions (Invitrogen). To avoid interference with potential genomic DNA amplification, we treated 1 μg of total RNA with 1 μl DNAse I (Invitrogen) plus 1 μl of 10× Buffer (Invitrogen). The samples were incubated at RT for 15 min. EDTA (25 mM) was added to the mixture and the samples were incubated at 65°C for 15 min to heat inactivate the DNAse I. For cDNA synthesis a total of 1 μg of RNA from the different samples were reverse-transcribed for 75 min at 42°C using 5 U of avian myeloblastosis virus reverse transcriptase (Promega) in the presence of 20 U of RNasin (Promega). The real-time PCR reaction was performed in 25 μl using the fluorescent dye SYBR Green Master mix (Applied Biosystems) and a mixture of 5 pmol of reverse and forward primers. The primers used were for TNF-α forward primer 5' CATCTTCTCAAAATTCGAGTGACAA 3' and reverse primer 5' TGGGAGTAGACAAGGTACAACCC 3' and for IL-6 forward primer 5' GAGGATACCACTCCCAACAGACC 3' and reverse primer 5' AAGTGCATCATCGTTGTTCATACA 3'. Quantification was performed on an ABI PRISM 7900 sequence detection system (Applied Biosystems). PCR cycles proceeded as follows: initial denaturation for 10 min at 95°C, then 40 cycles of denaturation (15 sec, 95°C), annealing (30 sec, 60°C), and extension (30 sec, 60°C). The melting-curve analysis showed the specificity of the amplifications. Threshold cycle, which inversely correlates with the target mRNA level, was measured as the cycle number at which the reporter fluorescent emission appears above the background threshold (data not shown). Data analysis is based on the ΔCT method with normalization of raw data to a house-keeping gene (β-actin). All of the PCRs were performed in triplicate.

Statistical significance analysis was assessed by using one-way or two-way analysis of variance (ANOVA) followed by unpaired Student's t test (version 5.0, Prism software, GraphPad, USA). A value of p < 0.05 was considered significant.