Mice were generated at Genoway facilities (Lyon, France). A targeting strategy was designed consisting in the insertion of an enhanced Green Fluorescent Protein (EGFP) reporter gene, preceded by an Internal Ribosomal Entry Site (IRES) sequence in the 3′ untranslated region (UTR) of the cnr2 mouse gene. This approach results in the expression of the reporter gene under the control of the endogenous mouse cnr2 promoter and transcript from the same bicistronic mRNA as the CB₂R protein. Further, the entire exon 3, including the 3′ UTR and knocked-in reporter, is flanked by loxP sites, allowing the conditional inactivation of the cnr2 gene in cells expressing Cre recombinase (Fig. 1a).Fig. 1

Generation of a novel mouse model. a Genomic structure of the construct employed for the generation of CB₂^EGFP/f/f and CB₂^−/− mice. Hatched rectangles represent cnr2 coding sequences, gray rectangles indicate non-coding exon portions, solid lines represent chromosome sequences. The neomycin-positive selection cassette (Neo) IRES sequence and reporter gene (EGFP) are indicated. loxP sites are represented by blue triangles and FRT sites by double red triangles. The initiation (ATG) and Stop (Stop) codons are indicated. For the generation of knock-in mice (CB₂^EGFP/f/f), recombined mice were bred with ubiquitous FLP-recombinase expressing mice, enabling the deletion of the FRT-flanked region. b For the generation of knock-out mice (CB₂^−/−), recombined mice were bred with ubiquitous Cre-recombinase expressing mice, resulting in the deletion of the loxP-flanked region. c Representative Southern blot showing the expected wild-type (WT) and recombined (Rec) hybridization signals at 9302 and 5794 bp, respectively, in embryonic stem cells from non-transfected (right column) and successfully transfected clones (left column). d Representative Western blot showing EGFP expression in spleen tissue homogenates from CB₂^EGFP/f/f (left lane) and CB₂^−/− mice (right lane). e The basal expression level of CB₂ receptor mRNA is not modified as a consequence of the insertion of the genetic construct used for the generation of the knock-in mice (CB₂^EGFP/f/f), as revealed by qRT-PCR in spleen samples and comparison with WT mice

Three isolated sequences encompassing the murine cnr2 gene regions surrounding the targeted exon 3 were used for the construction of the targeting vector. These sequences included (i) a 3462 bp-sized fragment containing exon 2 and downstream intronic sequences, (ii) a 2980 bp-sized fragment containing the coding part of exon 3 and upstream intronic sequences, and (iii) a 3657 bp-sized fragment containing the non-coding part of exon 3 and downstream sequences. The linearized targeting construct was transfected into C57BL/6J embryonic stem cells. Homologous recombinant cells were identified by Southern analysis and five clones were used to generate chimeric mice. Chimeras were bred with C57BL/6J Flp- and Cre-deleter females, in order to generate Neo-excised EGFP reporter knock-in (CB₂^EGFP/f/f) mice (Fig. 1a) and constitutive knock-out (CB₂^−/−) mice (Fig. 1b), respectively.

Homozygous mice identified by PCR were further verified by Southern blot analysis (Fig. 1c). All mice used in this study were fourth- or fifth-generation offspring from intercrosses of C57BL/6J mice. Mice were housed and bred in the animal facilities of Universidad Rey Juan Carlos (Alcorcón, Madrid, Spain) or the Medical College of Wisconsin (Milwaukee, WI, USA). Experimental protocols met the European and Spanish regulations for protection of experimental animals (86/609/EEC and RD 1201/2005 and 53/2013) or were approved by the Institutional Animal Care and Use Committee of the Medical College of Wisconsin. Male mice were used in all experiments included in the present report with the exception of flow cytometry experiments (see below).

Mice co-expressing five familial Alzheimer’s disease mutations (5xFAD) were purchased from Jackson Laboratories (Bar Harbor, ME, USA; [25]) on the C57BL/6J background and were mated with CB₂^EGFP/f/f and CB₂^−/− mice and backcrossed for at least five generations to generate CB₂^EGFP/f/f/5xFAD and CB₂^−/−/5xFAD mice. Animals employed in the present experiments were 3 to 6 months old; this period was chosen based on previously published data [25, 36] in order to allow for the appearance of amyloid deposits.

Single cell suspensions were prepared from the spleens of wild type, CB₂^EGFP/f/f, and CB₂^EGFP/f/+ mice of both sexes as described previously [27]. Cells were incubated with combinations of anti-mouse fluorescently-conjugated antibodies as follows: anti-B220 PE, anti-CD4 APC-eFluor780, anti-CD8 eFluor450, anti-CD11b eFluor450, anti-CD11c PE, anti-Ly6C APC, anti-Ly6G APC/Cy7, and anti-NK1.1 APC. Flow cytometry was used to identify B cells (B220⁺CD4⁻CD8⁻), CD4 T cells (CD4⁺NK1.1⁻), CD8 T cells (CD8⁺ NK1.1⁻), NKT cells (CD4⁺NK1.1⁺), NK cells (CD4⁻NK1.1⁺), macrophages (CD11b⁺Ly6C^+/−Ly6G⁻), dendritic cells (CD11b⁺CD11c^hi), and granulocytes (CD11b⁺Ly6C⁺Ly6G⁺). Sample acquisition was performed on a BD Biosciences LSR II, and data was analyzed using FlowJo software to generate the geometric mean of eGFP expression in each immune cell population.

Mice (N = 4–6 mice per group) were deeply anesthetized and transcardially perfused with cold PBS (pH 7.4) followed by freshly prepared cold 4% paraformaldehyde in PBS (pH 7.4). Tissue samples were collected and post-fixed in the same fixative overnight. Afterwards, tissues were dehydrated by sequential transfer to 10 and 30% sucrose solutions. Finally, tissues were cryoprotected with Tissue-Tek and frozen in dry ice. Thirty-micrometer-thick sections were obtained in a cryostat and preserved in cryoprotectant solution until use.

Floating tissue sections were washed with Tris Buffer Saline (TBS) before overnight incubation at 4 °C with the primary antibodies used for identification of the cellular types. For EGFP identification, overnight incubation with an anti-GFP antibody (1:1500; Abcam) was followed by incubation with an Alexa 488 anti-chicken antibody conjugate (Invitrogen) carried out at 37 °C for 2 h, rendering green fluorescence. Afterwards, sections were incubated with a rabbit polyclonal anti-ionized calcium-binding adaptor molecule 1 (Iba1) (1:1000 dilution, Wako, Osaka, Japan), diluted in TBS containing 1% bovine serum albumin (BSA; Sigma, St. Louis, USA) and 1% Triton x-100 (Sigma). After the incubation, sections were washed in TBS followed by incubation with an Alexa 546 anti-rabbit antibody conjugate (Invitrogen, Eugene, OR, USA) at 37 °C for 2 h, rendering red fluorescence. Additional tissue sections were incubated with mouse monoclonal anti-GFAP-Cy3 antibody (1:1500 dilution, Sigma) in the same buffer for 2 h at 37 °C or with mouse monoclonal anti-neuron-specific nuclear protein (NeuN) antibody (1:1000 dilution, Merck Millipore, Darmstadt, Germany) followed by incubation with Alexa 594 anti-mouse antibody conjugate (Invitrogen) as described above.

In order to study amyloid plaque deposits, a subset of CB₂^EGFP/f/f/5xFAD mice received an i.p. dose of 10 mg/kg of methoxy-XO4 (a Congo Red derivative known to selectively stain amyloid plaques; Tocris Bioscience; [4]) 24 h prior to sacrifice. Brains were processed and sections were obtained and preserved for immunostaining as described above.

Sections were mounted in aqueous solution (Vectashield, Vector Laboratories, Burlingame, CA, USA), coverslipped, and sealed. Slides were studied and photographed with upright microscopes (Nikon 90i, Nikon, Tokyo, Japan; and Axioimager M2, Zeiss, Oberkochen, Germany) and using a DXM1200F camera and C1 and LSM710 confocal systems [36]. Image analysis was carried out as described [36] with Metamorph (Molecular Devices, Sunnyvale, CA, USA) and ImageJ software (Research Services Branch, National Institute of Mental Health, Bethesda, MD, USA).

Protein fractions were collected from a Tri-pure extraction of hippocampal and spleen tissues, according to the manufacturer’s instructions (Roche). Lysates (20 μg/lane or 10 μg/lane for hippocampal and spleen protein samples respectively) were separated by SDS-PAGE and transferred onto PVDF membranes (BioRad). After blocking in Tris-Tween buffered saline (TTBS; 10 mM Tris pH 7.5, 150 mM NaCl, 0.1% Tween 20 plus 5% nonfat dried milk), they were incubated overnight at 4 °C, as appropriate, with anti-GFP (1:500, Abcam, Cambridge, UK). Membranes were incubated with corresponding horseradish peroxidase-conjugated secondary antibody (1:8000) and were developed using a chemiluminescent reagent (ECL detection reagent GE Healthcare, Buckinghamshire, UK). Developed signals were recorded on X-ray film (Agfa) for densitometric analysis (ImageJ, NIH, MD, USA). N = 4–6 mice per group were used for protein quantification by Western blot.

Human ELISA kits (Invitrogen, Camarillo, CA, USA) were used for the quantification of Aβ_1-42 in the brain soluble fractions, following the instructions provided by the manufacturer. Levels were normalized to the total amount of protein.

Total RNA was isolated using Tripure Isolation Reagent (Roche, Mannheim, Germany) according to the protocol of the supplier. RNA was dissolved in RNase-free water and quantified by absorption at 260 nm. Aliquots were subjected to 1% denaturing agarose gel electrophoresis and GelRed Nucleic Acid Gel Stain (Biotium, Fremont, CA, USA) staining to verify the quantity and quality of RNA. Single-stranded complementary DNA (cDNA) was synthesized from 1 mg of total RNA using LightCycler Taqman Master (Roche Diagnostics). PCR primers and TaqMan probes were designed by Tib Molbiol (Berlin, Germany) (see Additional file 2: Figure S1). For normalization, 18S primers and probe number 55 from Universal ProbeLibrary (Roche) were utilized. Gene expression was quantified using LightCycler FastStart DNA Master HybProbe and LightCycler Taqman Master (Roche) and Quantimix Easy Probes kit (Biotools, Madrid, Spain) in a LightCycler thermocycler (Roche). The concentration of primers and probes were 0.5 and 0.2 μM, respectively. PCR assays were performed using 2 μl of the cDNA reaction. All assays were carried out twice as independent PCR runs for each cDNA sample. Mean values were used for further calculation. A negative (no template) control was measured in each of the PCR runs. Standard curves were calculated for quantification purposes using fivefold serial dilutions of cDNA from mouse brain. The transcript amounts were calculated using the second derivate maximum mode of the LC-software version 4.0. The specific transcript quantities were normalized to the transcript amounts of the reference gene 18S. All further calculations and statistical analyses were carried out with these values referred to as relative expression ratios.

Results are expressed as mean ± SEM. Statistical analysis were made using student’s t test for comparisons between two groups, analysis of variance (ANOVA), and two-way ANOVA with Tukey’s post-test for multiple comparisons. A p value < 0.05 was considered as statistically significant (see Additional file 3: Table S1). Data were analyzed with Graph Pad Prism software version 6.0 (San Diego, CA, USA).

To characterize the newly generated CB₂^EGFP/f/f mice, we performed Western blotting on spleen samples. A single band corresponding to the EGFP molecular weight was evident in CB₂^EGFP/f/f mice and was undetectable in spleen samples from CB₂^−/− mice (Fig. 1d). We determined whether the strategy for the generation of the knock-in mice modified the expression levels of CB₂R gene. Our results show that no changes were evident in CB₂R mRNA expression levels between WT and CB₂^EGFP/f/f mice in spleen (Fig. 1e; p = 0.474), thus ruling out a putative impact of the transgene on basal expression of the receptor.

We used flow cytometry to identify and quantify the EGFP expression of splenocyte cell populations from wild type, CB₂^EGFP/f/+, and CB₂^EGFP/f/f mice (Fig. 2a–c). Using wild type mice, we found that background EGFP immunofluorescence was low in all immune cell populations examined (Fig. 2a). EGFP expression levels in splenic immune cells were compared in heterozygous (Fig. 2b) and homozygous (Fig. 2c) CB₂^EGFP/f/f mice. In all immune cell populations investigated, the homozygous CB₂^EGFP/f/f mice exhibited approximately double the mean fluorescence intensity (MFI) of the heterozygous mice. EGFP expression was highest in the B cell population, which is consistent with reports that B cells have the highest CB₂ receptor expression among these cell types [12]. Among the T cell populations, CD4 T cells and NK T cells expressed a similar low level of EGFP expression, while CD8 T cells expressed ~ threefold higher levels (Fig. 2). NK cells expressed negligible levels of EGFP (Fig. 2). Monocytes/macrophages and dendritic cells expressed EGFP in a broader expression pattern than the lymphocytes (Fig. 2). Given that the spleen contains numerous macrophage and dendritic cell populations, it is likely that CB₂R, and thus EGFP, will be differentially expressed among them [7, 15]. Finally, of the myeloid subset, granulocytes exhibited the highest amount of EGFP expression. These data are consistent with the published reports of CB₂R distribution among these cell types [12] indicating that the CB₂^EGFP/f/f mouse is an excellent tool by which to determine steady state CB₂R expression in various spleen cell populations using EGFP fluorescence.Fig. 2

EGFP presence and distribution in CB₂^EGFP/f/ spleens determined using flow cytometry (a–c) and immunofluorescence (d–f). a–c Determination of cell-specific expression of EGFP in splenocytes harvested from WT (a), CB₂^EGFP/f/+ (b), and CB₂^EGFP/f/f (c) mice. The numbers next to the histograms are the mean fluorescence intensity (MFI). d–f CB₂-dependent-EGFP expression in B-lymphocytes in the spleen of CB₂^EGFP/f/f. EGFP expression was evident in follicles of the white pulp. Scale bars, 100μm (d–g)

We analyzed the expression of EGFP in spleens of CB₂^EGFP/f/f mice by immunofluorescence and found discrete cell populations showing detectable signal. EGFP⁺ B cells were detected, limited to the marginal zone of the white pulp follicles, mostly located in the follicular corona (Fig. 2d–f).

In the CNS, microscopic analysis of the brain and spinal cord of 3-, 4-, or 6-month-old CB₂^EGFP/f/f mice showed no detectable EGFP immunoreactivity above background in glial or neuronal elements of any region examined, which included hippocampus (Fig. 3a), cortex, cerebellum, thalamus, brain stem, and spinal cord (not shown). In contrast, intense EGFP signal could be seen in brain regions of CB₂^EGFP/f/f/5xFAD mice known to be rich in beta-amyloid neuritic plaques, such as hippocampus (Fig. 3b). Other regions such as cortex, thalamus, and brain stem also exhibited EGFP signal (data not shown), in concordance with the previously reported distribution of neuritic plaques [25]. EGFP⁺ cells exhibited an ameboid shape and were mostly found in clusters (Fig. 3c), suggesting they could be activated microglial cells. No signal could be observed in the hippocampus of CB₂^−/−/5xFAD mice (Fig. 3d) or in any other brain region examined (data not shown).Fig. 3

In vivo EGFP induction in the context of AD, as revealed by immunofluorescence. a Un-manipulated, healthy CB₂^EGFP/f/f mice showed no significant EGFP signal in hippocampus (a). b, c EGFP signal could be noticed in the brain of CB₂^EGFP/f/f/5xFAD mice; these cells showed ameboid shape. d No EGFP signal could be observed in any brain region of CB₂^−/−/5xFAD mice, including those enriched in amyloid plaques such as hippocampus. e–l EGFP signal in CB₂^EGFP/f/f/5xFAD mice increased with age and paralleled that of amyloid deposits. EGFP was evident in brain samples of CB₂^EGFP/f/f/5xFAD mice starting at 3 months of age (e) and progressively increasing with age (4 months, f, and 6 months, g) and matched with the pattern of distribution of amyloid-enriched plaques stained with methoxy-XO4 in those same samples (figure i–k). Note that neighboring brain regions devoid of amyloid deposits exhibited a complete absence of EGFP signal (asterisks in g, k). h, l Densitometric quantification of EGFP⁺ cells (h) and amyloid plaques (k) shows a parallel increase in neuritic plaques and EGFP expression. Data are expressed as mean ± SEM. *p < 0.05 vs CB₂^EGFP/f/f mice, # vs CB₂^EGFP/f/f/5xFAD-3mo mice, and “a” vs CB₂^EGFP/f/f/5xFAD-4mo mice. N = 5 for immunofluorescence labeling. Scale bars, 10 μm (c) and 50 μm. DG (dentate gyrus); SBC (subiculum); CA1 (CA1 region of Ammon’s horn)

As shown in Fig. 3e, EGFP immunoreactivity above background could be observed as early as 3 months of age in the CB₂^EGFP/f/f/5xFAD mice, and EGFP-labeled cells increased in density with age in these mice (Fig. 3e–h). EGFP⁺ were found in clusters throughout the brain parenchyma and their distribution and increased density with age paralleled that of neuritic plaques, identified using methoxy-XO4, a dye for amyloid deposits (Fig. 3i–l). Interestingly, no EGFP signal could be observed in regions not exhibiting neuritic plaques (asterisks in Fig. 3g–k). The number of EGFP⁺ cells was dramatically increased at 4 and 6 months of age, which also paralleled the increase in the appearance of amyloid deposits (Fig. 3h: F_3,18 = 58.46, p < 0.0001; Fig. 3l: F_3,23 = 64.70, p < 0.0001).

EGFP⁺ cells were located in association with neuritic plaques (as revealed by staining with methoxy-XO4) and exhibited morphological features of microglia (Fig. 4). Co-localization studies with Iba-1, a commonly used marker of cells of myeloid lineage, were carried out. Low magnification (a to d) images showed a match in the pattern of distribution among EGFP⁺ and Iba1⁺ cells in the subiculum of 6-month-old CB₂^EGFP/f/f/5xFAD mice; in addition, our data show that CB₂-dependent EGFP expression takes place selectively in Iba1⁺ cells located in the vicinity of neuritic plaques (Fig. 4e–l). Microglial cells not associated with these pathological structures showed no EGFP staining (see Fig. 5a–d). For example, note the microglial cell at the arrow in Fig. 5b is neither EGFP positive nor associated with a plaque. Differences in the morphological features of EGFP⁺ and EGFP⁻ microglial cells were evident, with EGFP⁺ cells exhibiting an ameboid-like shape (Fig. 5a and b), typical of activated microglia, while EGFP⁻ cells showed a highly ramified morphology, characteristic of quiescent, non-activated, microglia (arrow in Fig. 5b).Fig. 4

Restricted EGFP expression in microglial cells located in peri-plaque areas of the CB₂^EGFP/f/f/5xFAD mouse hippocampus. a–l Low-magnification photographs of EGFP⁺ microglial cells (a, b) in close association to beta-amyloid neuritic plaques (c, d). Medium- (e–h) and high-magnification (i–l) photographs of EGFP⁺ microglial cells. Detailed co-localization immunofluorescent analysis reveals a complete overlap between EGFP⁺ cells and Iba1⁺ cells, indicative of their macrophage/microglia nature, and a selective association to amyloid-enriched plaques (stained with methoxy-XO4). Scale bars, 100 μm (a–d), 25 μm (e–h), and 25 μm (i–l). SBC (subiculum)

Furthermore, we also studied whether other cell types in the CNS, such as neurons or astrocytes expressed EGFP in CB₂^EGFP/f/f/5xFAD mice. To that end, co-localization studies with a neuronal marker (NeuN; Fig. 5e–h) or with a marker of astrocytes (GFAP; Fig. 5i–l) were carried out. Our data indicate that neither of these cell types express EGFP; thus, cnr2-dependent EGFP expression is limited to microglial cells in CB₂^EGFP/f/f/5xFAD mice.

We analyzed the impact of cnr2 gene deletion on plaque formation, soluble amyloid levels and neuroinflammation (Fig. 6). We found a small but significant decrease in hippocampal neuritic plaque density (measured by staining with methoxy-XO4; Fig. 6a: p < 0.0338) in the CB₂^−/− mice that was not paralleled by changes in soluble levels of Aβ_1-42 in the hippocampus (measured by ELISA; Fig. 6b: p < 0.6413). Hippocampal microgliosis was assessed by counting Iba1⁺ cells in tissue sections. As expected, the 5xFAD mice exhibited a significant increase in Iba1⁺ cells (Fig. 6c: F_1,23 = 85.84, p < 0.0001); however, there was no difference in this measure between the wild type and CB₂^−/− mice (Fig. 6c: F_1,23 = 0.03775, p = 0.8476). Finally, a significant increase in interleukin-1 beta (IL1β) was observed as a consequence of the amyloid pathology (Fig. 6d: F_1,23 = 49.12, p < 0.0001) but CB₂R genotype had no effect (F_1,33 = 0.2229, p = 0.6400).Fig. 6

Consequences of CB₂ deletion in the context of AD. The genetic inactivation of CB₂ in CB₂^−/−/5xFAD mice led to a significant decrease in plaque density in hippocampus (a), without any changes in soluble amyloid production (b), or microgliosis (c). No changes were observed in the expression of IL1β as a consequence of gene deletion (d) in CB₂^−/−/5xFAD mice as compared to those in CB₂^EGFP/f/f/5xFAD mice. Data are expressed as mean ± SEM. *p < 0.05. Student’s t-test (a, b) and two-way ANOVA followed by Tukey’s post hoc test (c, d). N = 5 for immunofluorescence labeling measurements and N = 6 for experiments on soluble amyloid quantification and IL1β expression

Additional file 1:

CB2 Western blots. Test of different CB2 primary antibodies in spleen samples (with high CB2 expression levels in normal conditions) harvested from CB2EGFP mice (lines 1, 2, and 3) and CB2KO mice (lines 4, 5, and 6). GFP and beta-actin immunodetection was employed as internal controls. (PPTX 6471 kb)