Antagonization of OX1 Receptor Potentiates CB2 Receptor Function in Microglia from APPSw/Ind Mice Model

Microdialysis assays demonstrated a possible role of orexin in the regulation of amyloid beta peptide (Aß) levels in the hippocampal interstitial fluid in the APP transgenic model. CB2R is overexpressed in activated microglia, showing a neuroprotective effect. These two receptors may interact, forming CB2-OX1-Hets and becoming a new target to combat Alzheimer’s disease. Aims: Demonstrate the potential role of CB2-OX1-Hets expression and function in microglia from animal models of Alzheimer’s disease. Receptor heteromer expression was detected by immunocytochemistry, bioluminescence resonance energy transfer (BRET) and proximity ligation assay (PLA) in transfected HEK-293T cells and microglia primary cultures. Quantitation of signal transduction events in a heterologous system and in microglia cells was performed using the AlphaScreen® SureFire® kit, western blot, the GCaMP6 calcium sensor and the Lance Ultra cAMP kit (PerkinElmer). The formation of CB2-OX1 receptor complexes in transfected HEK-293T cells has been demonstrated. The tetrameric complex is constituted by one CB2R homodimer, one OX1R homodimer and two G proteins, a Gi and a Gq. The use of TAT interfering peptides showed that the CB2-OX1 receptor complex interface is TM4-TM5. At the functional level it has been observed that the OX1R antagonist, SB334867, potentiates the action induced by CB2R agonist JWH133. This effect is observed in transfected HEK-293T cells and microglia, and it is stronger in the Alzheimer’s disease (AD) animal model APPSw/Ind where the expression of the complex assessed by the proximity ligation assay indicates an increase in the number of complexes compared to resting microglia. The CB2-OX1 receptor complex is overexpressed in microglia from AD animal models where OX1R antagonists potentiate the neuroprotective actions of CB2R activation. Taken together, these results point to OX1R antagonists as drugs with therapeutic potential to combat AD. Data access statement: Raw data will be provided by the corresponding author upon reasonable requirement.


Introduction
Orexin (hypocretin) receptor-mediated signaling has been mainly studied from the point of view of endocrinology and in relationship to eating disorders. However, orexinactions are also important from the point of view of neurodegenerative diseases. Regarding Alzheimer's disease (AD), microdialysis assays showed that orexin treatment resulted in the regulation of amyloid beta peptide (Aß) levels in the hippocampal interstitial fluid of a transgenic model that overexpresses a mutant form of the amyloid precursor protein

Direct Interaction of Cannabinoid CB 2 R and Orexin OX 1 R Receptors in a Heterologous Expression System
We first aimed to assess whether the cannabinoid CB 2 (CB 2 R) and orexin OX 1 (OX 1 R) receptors might interact. Immunocytochemistry assays were performed in HEK-293T cells expressing the CB 2 R fused to YFP, the OX 1 R fused to Rluc, or both ( Figure 1A). CB 2 R-YFP was detected by the YFP green fluorescence and OX 1 R-Rluc was detected by a mouse monoclonal anti-Rluc antibody and a secondary Cy3 conjugated anti-mouse IgG antibody. Results in Figure 1A show that when independently transfected, both receptors are expressed at the plasma membrane level and also at the cytosol. Moreover, in cells expressing both fusion proteins, receptors colocalize at the plasma membrane and intracellularly ( Figure 1A). As colocalization does not demonstrate direct interaction, bioluminescence resonance energy transfer (BRET) assays were carried out in HEK-293T cells expressing a constant amount of OX 1 R-Rluc and increasing amounts of CB 2 R-YFP. A hyperbolic BRET saturation curve indicated a specific interaction between the receptors and the formation of CB 2 R-OX 1 R heteromers with the following parameters: BRET max = 127 ± 7 mBU and BRET 50 = 110 ± 20 mBU ( Figure 1B). BRET max results from donor/acceptor proximity and the number of CB 2 R-OX 1 R heteromers while BRET 50 gives an idea of the affinity of the interaction. Accordingly, these parameters indicate that the interaction of the two receptors is strong. As negative control, GABA B -Rluc was used instead of OX 1 R-Rluc, and a linear signal was obtained, indicating the lack of interaction between CB 2 R and GABA B receptors ( Figure 1C). between the signaling mediated by these two receptors and their potential as therapeutic targets to combat AD, with special emphasis on evaluating their role in activated microglia.

Direct Interaction of Cannabinoid CB2R and Orexin OX1R Receptors in a Heterologous Expression System
We first aimed to assess whether the cannabinoid CB2 (CB2R) and orexin OX1 (OX1R) receptors might interact. Immunocytochemistry assays were performed in HEK-293T cells expressing the CB2R fused to YFP, the OX1R fused to Rluc, or both ( Figure 1A). CB2R-YFP was detected by the YFP green fluorescence and OX1R-Rluc was detected by a mouse monoclonal anti-Rluc antibody and a secondary Cy3 conjugated anti-mouse IgG antibody. Results in Figure 1A show that when independently transfected, both receptors are expressed at the plasma membrane level and also at the cytosol. Moreover, in cells expressing both fusion proteins, receptors colocalize at the plasma membrane and intracellularly ( Figure 1A). As colocalization does not demonstrate direct interaction, bioluminescence resonance energy transfer (BRET) assays were carried out in HEK-293T cells expressing a constant amount of OX1R-Rluc and increasing amounts of CB2R-YFP. A hyperbolic BRET saturation curve indicated a specific interaction between the receptors and the formation of CB2R-OX1R heteromers with the following parameters: BRETmax = 127 ± 7 mBU and BRET50 = 110 ± 20 mBU ( Figure 1B). BRETmax results from donor/acceptor proximity and the number of CB2R-OX1R heteromers while BRET50 gives an idea of the affinity of the interaction. Accordingly, these parameters indicate that the interaction of the two receptors is strong. As negative control, GABAB-Rluc was used instead of OX1R-Rluc, and a linear signal was obtained, indicating the lack of interaction between CB2R and GABAB receptors ( Figure 1C). The OX1R and CB2R interact in a heterologous expression system. (A) Immunocytochemistry assays were performed in HEK-293T cells expressing CB2R-YFP (1 μg cDNA), which was detected by its own fluorescence (green) and/or OX1R-Rluc (1 μg cDNA) that was detected by a mouse monoclonal anti-Rluc antibody and a secondary Cy3-conjugated anti-mouse IgG antibody (red). The OX 1 R and CB 2 R interact in a heterologous expression system. (A) Immunocytochemistry assays were performed in HEK-293T cells expressing CB 2 R-YFP (1 µg cDNA), which was detected by its own fluorescence (green) and/or OX 1 R-Rluc (1 µg cDNA) that was detected by a mouse monoclonal anti-Rluc antibody and a secondary Cy3-conjugated anti-mouse IgG antibody (red). Colocalization is shown in yellow. Cell nuclei were stained with Hoechst (blue). Scale bar: 15 µm. (B) BRET assays were performed in HEK-293T cells transfected with a constant amount of cDNA for OX 1 R-Rluc (0.4 µg) and increasing amounts of cDNA for CB 2 R-YFP (0.4 to 1.6 µg) or, as negative control (C), a constant amount of GABA B -Rluc (0.4 µg cDNA) and increasing amounts of CB 2 R-YFP (0.4 to 1.6 µg cDNA). Values correspond to six independent experiments.

OX 1 R Antagonists Potentiate CB 2 R-Mediated Signalling in Transfected HEK-293T Cells
Signaling studies were first performed in single transfected cells. CB 2 R couples to G i protein, thus leading to the inhibition of adenylate cyclase and the decrease of intracellular cAMP levels. The concentration of this second messenger was measured in cells expressing CB 2 Rs and pretreated with forskolin to activate adenylate cyclase and increase the cAMP levels. In CB 2 R expressing cells, the selective agonist JWH133 induced a 70% decrease in forskolin-induced cAMP levels ( Figure 2A). This effect was specific as it was completely blocked by pretreatment with the selective CB 2 R antagonist, SR144528. OX 1 R can couple to different G proteins depending on the heteromeric complexes in which the receptor is involved, hence we first performed assays assuming that it could couple to the G i protein.
Determination of cAMP levels in HEK-293T cells expressing the OX 1 R showed that orexin-A induced a 70% decrease of forskolin-induced cAMP levels and that SB334867, a selective antagonist, completely blocked this effect ( Figure 2B). When similar assays were performed in cells coexpressing both CB 2 R and OX 1 R, the agonists of the two receptors exerted a significant effect that was reverted by the corresponding antagonists ( Figure 2E). However, the CB 2 R antagonist not only blocked the JWH133-induced effect but also the effect of the OX 1 R agonist. This phenomenon is known as cross-antagonism and can be used as a print to detect CB 2 R-OX 1 R heteromers in homologous systems. In contrast, the OX 1 R antagonist blocked the effect triggered by orexin-A, but potentiated the effect of the CB 2 R agonist ( Figure 2E). This result is relevant, as it suggests that OX 1 R antagonists might potentiate the neuroprotective effects mediated by CB 2 R. On the other hand, when activating the same cells with both agonists, orexin-A and JWH133, a similar effect was observed to that induced by the activation of only one of the receptors ( Figure 2E).
Due to the capability of OX 1 R to couple to the G q protein and subsequently increase the cytosolic concentration of calcium ion (Ca 2+ ), assays were performed to measure this second messenger in cells expressing OX 1 R or both receptors. First, experiments were performed in cells expressing the CB 2 R and the calcium sensor GCaMP6; JWH133 stimulation did not lead to any effect, thus confirming that CB 2 R is not coupled to G q ( Figure 2C). Orexin-A in cells expressing the OX 1 receptor or OX 1 and CB 2 receptor produced a transient increase of Ca 2+ concentration that was blocked by the OX 1 R antagonist ( Figure 2F,G). As expected, JWH133 in these cells did not trigger any significant variation of calcium ion levels. The CB 2 R antagonist did not significantly modify the effect of orexin-A ( Figure 2F).
Finally, the mitogen-activated protein kinase (MAPK) pathway was analyzed in single transfected and in cotransfected cells. Data of ERK1/2 phosphorylation obtained by Western blotting and normalized by total ERKs are shown in Figure 2F. JWH133 and orexin-A induced a, respectively, 70% and 50% increase in ERK1/2 phosphorylation over basal phosphorylation levels. The effect was blocked when the two receptors were simultaneously activated. This phenomenon where the agonist of one receptor blocks the activation of the other protomer of the receptor complex is named negative cross-talk, becoming a print to demonstrate the existence of this complex in native tissue. Moreover, pretreatment with the CB 2 R antagonist not only blocked the JWH133-induced effect but also the effect of orexin-A. This cross-antagonism was also found when the OX 1 R antagonist was used. and were pretreated with selective receptor antagonists (1 μM SB334867 for OX1R or 1 μM SR144528 for CB2R) or vehicle and treated with selective agonists (100 μM Orexin-A for OX1R and/or 100 nM JWH133 for CB2R) followed by 0.5 μM forskolin stimulation (15 min). Values are the mean ± S.E.M. of six independent experiments performed in triplicates. One-way ANOVA followed by Bonferroni's multiple comparison post hoc test were used for statistical analysis (** p < 0.01, *** p < 0.001) versus FK condition. (C,D,F) HEK-293T cells were transfected with the cDNA for the GCaMP6 calcium sensor (1 μg) and for the CB2 receptor (1 μg) (C), for the OX1 receptor (1 μg) (D) or both receptors (F). Receptors were activated using selective agonists (100 nM orexin-A for OX1R and/or 100 nM JWH133 for CB2R). When indicated, cells were pretreated with selective antagonists (1 μM SB334867 for OX1R or 1 μM SR144528 for CB2R). Cytosolic calcium readings were collected and data are the mean ± S.E.M. from six independent experiments. (G) ERK1/2 phosphorylation was determined in HEK-293T cells expressing the CB2 receptor (1 μg cDNA) and the OX1 receptor (1 μg cDNA). Results are expressed in percentage with respect to basal condition. Values are the mean ± S.E.M. of five independent experiments performed in triplicates. A resentative Western blot is shown (bottom). p-ERK1/2: phosphorylated ERKs; t-ERK1/2: total ERKs. * p < 0.05, *** p < 0.001. One-way ANOVA followed by Bonferroni's multiple comparison post hoc test were used for statistical analysis * p < 0.05 versus basal condition in ERK1/2 phospohorylation assays or versus FK condition in cAMP determination assays (dashed line).

CB2R and OX1R form Tetrameric Complexes via a TM4-TM5 Interface
To characterize the CB2R-OX1R complex structure, a complementation approach was first used. CB2R was fused to the non-fluorescent C-terminal part of the YFP (cYFP hemiprotein) while OX1R was fused to the non-fluorescent N-terminal part of the YFP (nYFP hemiprotein). Fluorescence in cells coexpressing these hemiproteins showed bimolecular fluorescence complementation (BiFC), i.e., YFP was reconstituted due to interaction of the receptors. To identify the interacting domains, cells were treated with interfering peptides consisting of the sequence of transmembrane domains (TM) of the receptors fused to a and were pretreated with selective receptor antagonists (1 µM SB334867 for OX 1 R or 1 µM SR144528 for CB 2 R) or vehicle and treated with selective agonists (100 µM Orexin-A for OX 1 R and/or 100 nM JWH133 for CB 2 R) followed by 0.5 µM forskolin stimulation (15 min). Values are the mean ± S.E.M. of six independent experiments performed in triplicates. One-way ANOVA followed by Bonferroni's multiple comparison post hoc test were used for statistical analysis (** p < 0.01, *** p < 0.001) versus FK condition. (C,D,F) HEK-293T cells were transfected with the cDNA for the GCaMP6 calcium sensor (1 µg) and for the CB 2 receptor (1 µg) (C), for the OX 1 receptor (1 µg) (D) or both receptors (F). Receptors were activated using selective agonists (100 nM orexin-A for OX 1 R and/or 100 nM JWH133 for CB 2 R). When indicated, cells were pretreated with selective antagonists (1 µM SB334867 for OX 1 R or 1 µM SR144528 for CB 2 R). Cytosolic calcium readings were collected and data are the mean ± S.E.M. from six independent experiments. (G) ERK1/2 phosphorylation was determined in HEK-293T cells expressing the CB 2 receptor (1 µg cDNA) and the OX 1 receptor (1 µg cDNA). Results are expressed in percentage with respect to basal condition. Values are the mean ± S.E.M. of five independent experiments performed in triplicates. A resentative Western blot is shown (bottom). p-ERK1/2: phosphorylated ERKs; t-ERK1/2: total ERKs. * p < 0.05, *** p < 0.001. One-way ANOVA followed by Bonferroni's multiple comparison post hoc test were used for statistical analysis * p < 0.05 versus basal condition in ERK1/2 phospohorylation assays or versus FK condition in cAMP determination assays (dashed line).

CB 2 R and OX 1 R form Tetrameric Complexes via a TM4-TM5 Interface
To characterize the CB 2 R-OX 1 R complex structure, a complementation approach was first used. CB 2 R was fused to the non-fluorescent C-terminal part of the YFP (cYFP hemiprotein) while OX 1 R was fused to the non-fluorescent N-terminal part of the YFP (nYFP hemiprotein). Fluorescence in cells coexpressing these hemiproteins showed bimolecular fluorescence complementation (BiFC), i.e., YFP was reconstituted due to interaction of the receptors. To identify the interacting domains, cells were treated with interfering peptides consisting of the sequence of transmembrane domains (TM) of the receptors fused to a sequence, the TAT cell-penetrating peptide. It has been previously shown that these TAT-derived peptides can be inserted into the plasma membrane and disrupt the interaction of cell surface GPCRs [20]. We found that sequences of the OX 1 R, TAT-TM4 (400 nM), and TAT-TM5 (400 nM) led to a significant decrease in the YFP reconstitution. Similar results were obtained with TAT-TM4 (400 nM) and TAT-TM5 (400 nM) when using CB 2 R TM sequences. These results indicate that the CB 2 R-OX 1 R complex has a TM4-TM5 interface ( Figure 3A). To gain insight into the structure of the complex, we explored the possibility that CB 2 R and OX 1 R could form tetrameric complexes. We attempted to obtain BRET in cells expressing a constant amount of the "hemiproteins" OX 1 R-nRLuc and OX 1 R-cRluc and increasing amounts of the "hemiproteins" CB 2 R-nYFP and CB 2 R-cYFP. A hyperbolic BRET signal saturation curve was obtained (BRET max 50 ± 2 mBU; BRET 50 50 ± 10 mBU), indicating the formation of tetramers formed by two CB 2 Rs and two OX 1 Rs ( Figure 3B). Values represent the mean ± SEM of eight independent experiments. A statistical analysis was performed using a one-way ANOVA followed by Bonferroni's post hoc test (* p < 0.05 versus basal condition). (B) BRET assays were determined in cells expressing a constant amount of cDNA for OX 1 R-cRluc (1.5 µg) and OX 1 R-nRluc (1.5 µg) and increasing amounts of cDNA for CB 2 R-cYFP (0.3-2.5 µg) and CB 2 R-nYFP (0.3-2.5 µg). (C) HEK-293T cells expressing CB 2 R, OX 1 R and the engineered calcium sensor, GCaMP6 (1 µg) were incubated overnight with vehicle or pertussis toxin (PTX; 10 ng/mL), or for 2 h with cholera toxin (CTX; 100 ng/mL) or YM254890 (YM; 100 ng/mL) and exposed to orexin-A (100 nM). Cytosolic calcium readings were collected, and data from a representative experiment are shown. (D) HEK-293T cells expressing the receptors were incubated overnight with vehicle or pertussis toxin (PTX; 10 ng/mL), or for 2 h with cholera toxin (CTX; 100 ng/mL) or YM254890 (YM; 100 ng/mL) and exposed to JWH133 (100 nM), SB334867 (1 µM) or both in the presence of forskolin (0.5 µM). (E) Cells coexpressing cDNAs for OX 1 R (1.5 µg) and CB 2 R (1.5 µg) were preincubated for 4 h with interfering peptides (0.4 µM), pretreated with CB 2 R antagonist (SB334867 1 µM) for 15 min and then treated for 15 min with the selective CB 2 R agonist, JWH133 (100 nM), in the presence of forskolin before determining cAMP levels. A statistical analysis was performed using a one-way ANOVA followed by Bonferroni's post hoc test. * p < 0.05 versus forskolin treatment # p < 0.05 versus JWH133 + FK condition. Values are the mean ± S.E.M. of five different experiments performed in triplicates. Values are expressed as percentage of cAMP accumulation provoked by forskolin (FK) (n = 3, in triplicates). One-way ANOVA followed by Bonferroni's multiple comparison tests was used for statistical analysis * p < 0.05 # p < 0.05 versus JWH133 + FK condition.
It was further investigated whether the heteromer print disappeared when disrupting the heteromeric complex using TAT-TM peptides. Assays of cAMP level determination were performed in HEK-293T cells coexpressing the two receptors and treated with TAT-TM peptides. Interestingly, it was observed that the heteromer print disappeared in the presence of TAT-TM4 (400 nM) or TAT-TM5 (400 nM) (CB 2 R sequence) but not when other TAT-TM peptides were used. These two peptides prevented the potentiation of the effect of the CB 2 R agonist-induced by the OX 1 R antagonist ( Figure 3E). The action of these peptides was specific, as the TAT-TM2 (400 nM) peptide or vehicle were ineffective.
Finally, we investigated which G alpha proteins were coupled to the macromolecular CB 2 R-OX 1 R complex. Experiments were performed in HEK-293T cells coexpressing the two receptors and pretreated with either cholera toxin (CTX; 100 ng/mL), which alters G smediated signaling, pertussis toxin (PTX; 10 ng/mL), which alters G-mediated signaling, or the G q inhibitor, YM254890 (1 µM), which blocks G q -mediated signaling and was also used as a further control. In calcium mobilization assays, the effect of orexin-A was partially blocked by PTX and, also, by YM254890 ( Figure 3C), suggesting the occurrence of one G i and one G q protein in the heteromeric complex. In cAMP determination assays, PTX treatment completely blocked the decrease in cAMP upon treatment with orexin-A or JWH133 ( Figure 3D). Thus, it may be assumed that the tetramer contains at least one G i protein. Furthermore, the fact that CTX did not affect any of the agonistic effects indicates that the complex does not couple to G s proteins.

Blockade of OX 1 R Potentiates CB 2 R Function in the Microglia from APP Sw/Ind Mice
Different studies have shown an important increase in CB 2 R expression in activated microglia compared to resting cells [21]. Accordingly, the expression and function of the CB 2 -OX 1 receptor complex in the microglia of the Alzheimer's model and in control mice was determined. To characterize the CB 2 -OX 1 receptor complex function in microglial cells, cAMP level determination assays were carried out. In resting microglia from control mice, we observed that both JWH133 and orexin-A induced a significant effect that was non-additive upon receptor coactivation. Interestingly, in the presence of the OX 1 R antagonist, JWH133 induced a more marked effect ( Figure 4A). When analyzing microglia from APP Sw/Ind mice, both JWH133 and orexin-A provoked a significant decrease over forskolin-induced cAMP levels ( Figure 4B) that was non additive in coactivation; such a negative crosstalk was not observed in control microglia. However, pretreatment with the OX 1 R antagonist strongly potentiated CB 2 R signaling, as observed in transfected HEK-293T cells, indicating again that the blockade of OX 1 R potentiates CB 2 R-induced signaling. A cross-antagonism was also found when the CB 2 R antagonist was used; SR144528 blocked the effect triggered by both JWH133 and orexin-A.
1 Figure 4. Functionality of CB 2 R and OX 1 R heteromer in the microglia from the APP Sw,Ind mouse. Primary microglia from APP Sw,ind (A,B) or age-matched control animals (B,D) were pre-treated with selective receptor antagonists (1 µM SR144528 for CB 2 or 1 µM SB334867 for OX 1 receptors) for 15 min and subsequently treated with selective agonists (100 nM JWH133 for CB 2 or 100 nM orexin-A for OX 1 receptors) in single or combined treatments. cAMP levels (A,B) were detected 15 min after forskolin addition. Values are the mean ± S.E.M. of seven different experiments performed in triplicates. One-way ANOVA followed by Bonferroni's multiple comparison post hoc test was used for statistical analysis (* p < 0.05 versus forskolin treatment). (C,D) ERK1/2 phosphorylation was analyzed using an AlphaScreen ® SureFire ® kit (Perkin Elmer). Values are the mean ± S.E.M. of five different experiments performed in triplicates. One-way ANOVA followed by Bonferroni's multiple comparison post hoc test were used for statistical analysis (** p < 0.01, *** p < 0.001ersus untreated cells). (E) A proximity ligation assay (PLA) was performed in primary microglia from APP Sw,Ind transgenic mice or age-matched controls, using specific primary antibodies against CB 2 R or against OX 1 R (1/100). Representative images corresponding to stacks of four sequential planes are shown. Cell nuclei were stained with Hoechst (blue) and receptor complexes appear as red dots. The number of red dots/cell was quantified using Andy's algorithm Fiji's plug-in (see Section 4). Scale bar: 15 µM. Values are the mean ± S.E.M. of five different experiments performed in duplicates. Oneway ANOVA and Bonferroni's multiple comparison post hoc test were used for statistical analysis (* p < 0.05 versus control).
Analysis of the MAPK signaling pathway led to the finding that coactivation with orexin-A and JWH133 induced a lower effect than that observed by activation of one of the receptors, thus, negative crosstalk was stronger in the microglia from control mice but also observed in APP Sw/Ind mice ( Figure 4C,D). Interestingly, in APP Sw/Ind mice microglia the CB 2 R antagonist potentiated the OX 1 R-induced function. This phenomenon was not observed in control miceor in transfected HEK-293T cells and could be explained due to a differential expression of CB 2 R in activated microglia.
Finally, by PLA the formation of CB 2 -OX 1 receptor complexes in control and APP Sw/Ind mice microglia was demonstrated. Interestingly, the expression of the heteromer was two-fold higher in the microglia from the AD model than in the microglia from control animals ( Figure 4E). The important increase in the CB 2 -OX 1 complexes in APP Sw/Ind mice microglia could explain the functional differences observed in cAMP levels and MAPK phosphorylation assays.

Discussion
The interaction of two receptors that are expressed in microglia opens new perspectives for the regulation of the activation of these cells. In addition, the interaction between CB 2 R and OX 1 R is relevant for understanding the mechanisms underlying orexigenic and cannabinoid functional interactions at the CNS level. A recent study has shown by ligand-binding assays that the phytocannabinoid CBD can bind OX 1 R at low micromolar range where it acts as an antagonist by decreasing calcium mobilization [22]. This result does not become clear evidence of a direct relationship between the cannabinoid and the orexinergic systems because this compound also binds to other receptors such as serotonin receptors [23]. Another study in the amygdala shows that 2-arachidonoylglycerol acting on the CB 2 R reverts the fear extinction deficits induced by orexin-A. This finding seems a consequence of a higher 2-arachidonoylglycerol production by the action of orexin-A. Importantly, the effect of the peptide correlated with an increase in the expression of the CB 2 R in microglial cells [24]. The cannabinoid receptor has been proposed as a target for neuroprotection by regulating microglial activation and polarization. The mechanisms favoring microglia polarization to a neuroprotectivephenotype, also known as M2, are not known, although there are GPCRs that may regulate their production. On the one hand, cannabinoids acting on the CB 2 R may afford neuroprotection and prevent the progression of neurodegenerative diseases (see [25,26] and references therein). On the other hand, the CB 2 R is upregulated in activated microglia. The results afforded by this manuscript demonstrate that the OX 1 R antagonist, SB334867, not only blocks the orexin-A-induced effect, but also potentiates cannabinoid CB 2 R function in cAMP and MAPK signaling pathways. Moreover, this effect is potentiated in activated microglia compared to resting cells. In this sense, OX 1 R antagonists could become a new therapeutic target to consider in AD and other neurodegenerative diseases that trigger with microglia activation.
In the periphery, at the level of the intestinal barrier, orexin-A regulates neuroinflammation by acting on enterocytes and/or microglia [27]. In another of the few papers on orexin-A actions on activated microglia it was reported that the peptide favors the production and release of neuroprotective factors [28]. In animal models, orexin-A also affords protection against neuroinflammation occurring after brain ischemic insult [29]. The hypothesis underlying the research described in this paper was based on the possible co-occurrence and potential interaction of CB 2 R and OX 1 R receptors in microglia. The results confirm that CB 2 R and OX 1 R may interact in both a heterologous expression system and in primary microglia from mouse brain.
Unfortunately, the 3D structure of heteromers formed by two class A GPCRs has not yet been solved. However, several determined structures display crystallographic interfaces potentially compatible with physiological interactions. Reliable models that have been deciphered using the approach of interfering peptides employed in this study, such as the adenosine A 1 and A 2A heterotetramer coupled to one G s protein and to one G i protein [30][31][32] Previous studies have demonstrated that OX 1 R can form complexes with other proteins with the same TM4-5 interface; a clear example is the heteromeric complex between the apelin receptor APJ and OX 1 R [33].
To have a framework to understand the functional regulation within the CB 2 R-OX 1 R hetereomer, we created two structural models of the TM4/5 interface using experimentally determined structures and molecular modeling (see Section 4). The models reveal that the orthosteric binding pockets of CB 2 R and OX 1 R are connected by a network of aromatic residues. It turns out that both the CB 2 R and the OX 1 R antagonists used in the present study place groups near TM5 that trigger a larger opening of the extracellular part of this helix than other antagonists [34,35]. Although the study of this network is beyond the scope of the present work, it is tempting to speculate that an antagonist of one receptor could concertedly push the aromatic residues in this network bringing changes to the orthosteric pocket of the second receptor leading to either cross-activation or cross-inactivation. Figure 5A shows how the SB334867 could affect the signal triggered by the CB 2 R agonist JWH133. Figure 5B shows how the CB 2 R antagonist SR144528 could affect the signal triggered by the endogenous OX 1 R agonist orexin-A. We propose that the opening of TM5 in either case, required to accommodate an antagonist, would modulate the interface. The fact that TM5 exhibits different orientations with different ligands ( Figure 5C) could be the basis of the opposite effects of the antagonists (cross-antagonism vs. positive modulation). Thus, the OX 1 R antagonist SB334867 allosterically stabilizes the active form of CB 2 R, potentiating the action induced by the agonist JWH133. By contrast, the CB 2 R antagonist SR144528 allosterically stabilizes an inactive state of the OX 1 R that hinders OXA binding, probably because the second extracellular loop that connects TMs 4 and 5 occludes its access.
Our heterodimeric models in Figure 4 imply the binding of a single G-protein (Supplementary Materials Figure S1). It is possible that both receptors bind a G-protein (G i or G q ) in an active complex, but it is not possible that both bind it simultaneously because they would clash sterically. This dimeric model already provides the framework to understand the allosteric regulations by different protomers. Further work would be required to understand the benefits of having a heterotetramer compared to a heterodimer. The results in transfected HEK-293T cells have demonstrated the formation of tetrameric complexes with one homodimer of CB 2 R and one homodimer of OX 1 R able to signal through both G i and a G q . A tetramer could eventually allow the binding of G q to one OX 1 protomer and G i to another. In fact, a tetramer assuming symmetric TM1 interfaces could simultaneously bind up to three G-proteins (Supplementary Materials Figure S1).
Some studies have demonstrated the exchange of G protein coupling to GPCRs when forming complexes in specific tissues. One of the first examples is that described by M Glass and collaborators, where cotreatment with a CB 1 R agonist and a dopaminergic D 2 R agonist increased cAMP levels in striatal neurons [36]. Thus, this indicates that two receptors that typically couple to the G i protein can couple to a G s protein when forming CB 1 R-D 2 R complexes in striatal neurons. Another example is D 1 R and D 2 R mediated signaling via the established G s/olf and G i/o , respectively. This is abolished when the heterodimer is stimulated, leading to new signal transduction by G q /11 in the striatum [37]. In this sense, we have demonstrated that the newly described complex functions through the OX 1 R and CB 2 R characteristic proteins, a G q and a G i , respectively, and, although the heteromer permits the OX 1 R to signal via G i as well.    The CB 2 R/SR144528-OX 1 R/orexin-A. Similarly, the CB 2 R antagonist SR144528 puts an aromatic ring at the region where most CB ligands place the alkyl chain (such as in JHW133), also towards TM5. The CB 2 R receptor is shown in gold and the OX 1 R receptor in blue. Small-molecule ligands and aromatic side chains connecting the pockets are shown as sticks; orexin-A with a thick green cartoon. Dotted boxes indicate TMs 4 and 5 of both receptors. In (C) detail of the CB 2 R/SR144528-OX 1 R/orexin-A complex viewed from the extracellular side superposing different orientations of the extracellular portion of TM5 and proximal region of the second extracellular loop (ECL2) (i) from the OX 2 R/orexin-B complex (PDB id 7L1U, cylinder and loop in light blue), which would clash sterically with the TM4 of the CB 2 R; and (ii) from the OX 1 R/SB334867 complex (PDB id 6TQ7, pink cylinder), which orients in a direction that does not directly face TM4 and better avoids the possible clashes with it TM4. Arrows indicate the movements of TM5 helices towards the TM5 modeled in the heteromer (see Section 4, dark blue) [33]. This complex shows a characteristic signaling where the OX 1 R selective antagonist potentiates the cannabinoid CB 2 R function in cAMP signaling while it blocks the indirect MAPK pathway. These interesting results are also observed in resting microglia. However, in activated microglia, the OX 1 R antagonist blockade became stronger, inducing a higher potentiation of CB 2 R cannabinoid signaling not only in cAMP but also in MAPK signaling. These results can be explained because in primary microglia from the transgenic AD model, the CB 2 -OX 1 receptor heteromer expression is significantly higher than in equivalent cells from control animals, as shown by PLA. This finding suggests that the heteromer could be a target to combat neurodegeneration. We have previously shown that primary microglia from this AD model have an activated phenotype that is, likely, neuroprotective [38][39][40]. This hypothesis would explain why cognitive deficits do not appear at birth but later in life. Future research should address whether the CB 2 -OX 1 receptor heteromer is a suitable target to skew microglia to the neuroprotective phenotype.

Cell Isolation and Culturing
HEK-293T cells (batch 70022180) were acquired from the American Type Culture Collection (ATCC). Cells were amplified and frozen in liquid nitrogen in several aliquots. Cells from each aliquot were used until passage 18. HEK-293T cells were grown in Dulbecco's Modified Eagle's Medium (DMEM) (Gibco) supplemented with 2 mM L-glutamine, 100 µg/mL sodium pyruvate, 100 U/mL penicillin/streptomycin, MEM non-essential amino acids solution (1/100) and 5% (v/v) heat-inactivated fetal bovine serum (FBS) (all supplements were from Invitrogen, Paisley, Scotland, UK) and maintained at 37 • C in a humid atmosphere of 5% CO 2 .
CD-1 strain mice handling, sacrifice, and further experiments were conducted according to the guidelines set in Directive 2010/63/EU of the European Parliament and the Council of the European Union that is enforced in Spain by National and Regional organizations; the 3R rule (replace, refine, reduce) for animal experimentation was also considered. Primary cultures of microglia were obtained from 2-3-day-old pups. Cells were isolated as described in [41] and plated at a confluence of 40,000 cells/0.32 cm 2 . Briefly, the samples were dissected, carefully stripped of their meninges and digested with 0.25% trypsin for 20 min at 37 • C. Trypsinization was stopped by repeated washes with Hanks balanced salt solution (HBSS composition: 1.26 mM CaCl 2 , 5 mM KCl, 0.44 mM KH 2 PO 4 , 0.5 mM MgCl 2 , 0.4 mM MgSO 4 , 137 mM NaCl, 0.34 mM Na 2 HPO 4 and 10 mM Hepes, pH: 7.4). Cells were brought to a single cell suspension by repeated pipetting followed by passage through a 100 µM-pore mesh. Cells were then resuspended in supplemented DMEM and seeded at a density of 3.5 × 10 5 cells/mL in 6-well plates or 96-well plates for functional assays and in twelve-well plates for immunocytochemistry or PLA assays. Cultures were maintained at 37 • C in a humidified 5% CO 2 atmosphere. The medium was replaced every 4-5 days. Immunodetection of specific markers (CD-11b) showed that microglia preparations contained at least 98% microglial cells [42]

Cell Transfection
HEK-293T cells were transiently transfected with the corresponding cDNA by the PEI (PolyEthylenImine, Sigma-Aldrich) method. Briefly, the corresponding cDNA diluted in 150 mM NaCl was mixed with PEI (5.5 mM in nitrogen residues), also prepared in 150 mM NaCl for 10 min. The cDNA-PEI complexes were transferred to HEK-293T cells and were incubated for 4 h in a serum-starved medium. The medium was then replaced by a fresh supplemented culture medium and cells were maintained at 37 • C in a humid atmosphere of 5% CO 2 . Forty-eight hours after transfection, cells were washed, detached, and resuspended in the assay buffer.

Transgenic Alzheimer's Disease (AD) Animal Model
APP Sw,Ind transgenic mice (line J9; C57BL/6 background) expressing human APP695 harboring the FAD-linked Swedish (K670N/M671L) and Indiana (V717F) mutations under the PDGF promoter were obtained by crossing APP Sw,Ind to non-transgenic (WT) mice [43]. APP Sw,Ind -derived embryos or pups, individually genotyped and divided into "APP Sw,Ind " and "control", were used for preparing primary cultures. Animal care and experimental procedures were in accordance with European and Spanish regulations (86/609/CEE; RD1201/2005). Mice were handled, as per law, by personnel with the ad hoc permit (issued by the Generalitat de Catalunya), which allows animal handling for research purposes.

Bioluminescence Resonance Energy Transfer (BRET) Assay and BRET with BiFC Assays
For the BRET assay, HEK-293T cells were transiently cotransfected with a constant amount of cDNA encoding for OX 1 R-Rluc (1 µg) and with increasing amounts of cDNA corresponding to CB 2 R-YFP (0.4 to 1.6 µg). As negative control, HEK-293T cells were transiently cotransfected with a constant amount of cDNA encoding for GABA B -Rluc (0.4 µg) and with increasing amounts of cDNA corresponding to CB 2 R-YFP (0.8 to 5 µg). For the BRET assays with Bimolecular fluorescence complementation (BiFC), HEK-293T cells were transiently cotransfected with a constant amount of cDNA encoding for proteins fused to Rluc hemiproteins (nRluc, cRluc), OX 1 R-nRluc8 (1.5 µg) and OX 1 R-cRluc8 (1.5 µg) and with increasing amounts of the cDNA corresponding to proteins fused to YFP hemiproteins (nYFP, cYFP), CB 2 R-nYFP Venus (0.3-2.5 µg), or CB 2 R-cYFP Venus (0.3-2.5 µg). To control the cell number, the sample protein concentration was determined using a Bradford assay kit (Bio-Rad, Munich, Germany) using bovine serum albumin (BSA) dilutions as standards. To quantify fluorescent proteins, cells (20 µg of total protein) were distributed in 96-well microplates (black plates with a transparent bottom) and fluorescence was read in a Fluostar Optima Fluorimeter (BMG Labtech, Ofenburg, Germany) equipped with a high-energy xenon flash lamp using a 10-nm bandwidth excitation filter at 485 nm. For BRET measurements, the equivalent of 20 µg of total protein cell suspension was distributed in 96-well white microplates with a white bottom (Corning 3600, Corning, NY). For BRET measurements, the equivalent to 20 µg cell suspension was distributed in 96-well microplates (white plates, Porvair, Leatherhead, UK) and 5 µM coelenterazine H was added (PJK GMBH, Kleinblittersdorf, Germany). One minute after coelenterazine H addition, the readings were collected using a Mithras LB 940 (Berthold, Bad Wildbad, Germany), which allowed the integration of the signals detected in the short-wavelength filter at 485 nm (440-500 nm) and the long-wavelength filter at 530 nm (510-590 nm). To quantify receptor-Rluc expression, luminescence readings were collected 10 min after the addition of 5 µM coelenterazine H. The net BRET is defined as [(long-wavelength emission)/(short-wavelength emission)]-C f , where C f corresponds to [(long-wavelength emission)/(short-wavelength emission)] for the Rluc construct expressed alone in the same experiment. Data in BRET curves that depict an equilateral hyperbola were fitted by a non-linear regression equation using GraphPad Prism software (San Diego, CA, USA). BRET values for specific interactions are given as milli BRET units (mBU: 1000 × net BRET). For BiFC assays, HEK-293T cells were transiently transfected with a constant amount of cDNA encoding for proteins fused to nVenus (OX 1 R-nYFP) or cVenus (CB 2 -cYFP) and incubated for 4 h in complete DMEM containing the interfering TAT peptides (with similar sequences to those in TM1 to TM7 for CB 2 R or TM4, TM5 or TM7 for OX 1 R). YFP resulting from complementation was detected by placing cells (20 µg protein) in 96-well microplates (black plates with a transparent bottom) and reading the fluorescence in a Fluostar Optima Fluorimeter (BMG Labtech, Offenburg, Germany) using a 30-nm bandwidth excitation filter (485 nm).

TAT-TM Peptides
Peptides with the sequence of the TM of CB 2 R and OX 1 R fused to the HIV TAT peptide (YGRKKRRQRRR) were used as oligomer-disrupting molecules (synthesized by Genemad Synthesis Inc. San Antonio, TX, USA). The cell penetrating TAT peptide allows for the intracellular delivery of fused peptides [20]. The TAT-fused TM peptide can then be inserted effectively into the plasma membrane because of the penetration capacity of the TAT peptide and the hydrophobic property of the TM moiety [44]. To obtain the right orientation of the inserted peptide, the HIV-TAT peptide was fused to the C-terminus or to the N-terminus as indicated: TAVAVLCTLLGLLSALENVAVLYLIL-YGRKKRRQRRR for CB 2 R TM1, YGRKKRRQRRR-YLFIGSLAGADFLASVVFACSFVNF for CB 2 R TM2, AVFLLKIGSVTMTFTASVGSLLLTAI-YGRKKRRQRRR for CB 2 R TM3, YGRKKRRQRRR-ALVTLGIMWVLSALVSYLPLMGW for CB 2 R TM4, YLLSWLLFIAFLFSGIIYTYGHVLW-YGRKKRRQRRR for CB 2 R TM5, YGRKKRRQRRR-TLGLVLAVLLICWFPVLALMAH for CB 2 R TM6, AFAFCSMLCLINSMVNPVIYAL-YGRKKRRQRRR for CB 2 R TM7, YGRKKRRQRRR-ILGIWAVSLAIMVPQAAVME for OX 1 R TM4, SSFFIVTYLAPLGLMAMAYFQIF-YGRKKRRQRRR for OX 1 R TM5, YASFTFSHWLVYANSAANPIIYNF-YGRKKRRQRRR for OX 1 R TM7.

cAMP Level Determination
The analysis of cAMP levels was performed in HEK-293T cells cotransfected with the cDNA for CB 2 R (1.5 µg) and the cDNA for the OX 1 R (1.5 µg), in primary microglia and primary neurons prepared from wild-type mice or the transgenic APP Sw/Ind AD mice model. In the case of HEK-293T cells, when indicated, were treated overnight with the required TAT peptides TM2, TM4, and TM5 (0.4 µM) or with 10 ng/mL pertussis toxin (PTX; ref: P7208-50UG, Sigma-Aldrich, over-night), 100 ng/mL cholera toxin (CTX; ref:C8052-5M, Sigma-Aldrich, 2 h) or 1µM G q inhibidor (YM 254890, 2 h). Two hours before the experiment, the medium was replaced by serum-starved DMEM medium. Cells growing in a medium containing 50 µM zardaverine were distributed in 384-well microplates (2000 HEK-293T or 4000 primary cells per well) treated at 37 • C with selective antagonists (1 µM SB334867 or 1 µM SR144528) for 15 min prior stimulation with OX 1 R and/or CB 2 R agonists (100 nM orexin-A and/or 100 nM JWH133); 15 min after agonists treatment, cells were treated (15 min) with 0.5 µM forskolin or vehicle. Homogeneous time-resolved fluorescence energy transfer (HTRF) measurements were performed using the Lance Ultra cAMP kit (PerkinElmer). Fluorescence at 665 nm was analyzed on a PHERAstar Flagship microplate reader equipped with an HTRF optical module (BMG Labtech). A standard curve for cAMP was obtained in each experiment.

Detection of Cytoplasmic Calcium Levels
HEK-293T cells were cotransfected with the cDNA for the protomers of the CB 2 R (1.5 µg), the cDNA for OX 1 R (1.5 µg) and with the cDNA for the GCaMP6 calcium sensor (1 µg) by the PEI method (Section 4.3). Forty-eight hours after transfection, HEK-293T cells plated in 6-well black, clear bottom plates, were incubated with Mg 2+ -free Locke's buffer (154 mM NaCl, 5.6 mM KCl, 3.6 mM NaHCO 3 , 2.3 mM CaCl 2 , 5.6 mM glucose, 5 mM HEPES, 10 µM glycine, pH 7.4). Online recordings were performed right after the addition of agonists. When indicated, cells were pre-treated with receptor antagonists for 10 min. Fluorescence emission intensity due to complex GCaMP6 was recorded at 515 nm upon excitation at 488 nm on the EnSpire ® Multimode Plate Reader for 150 s every 5 s at 100 flashes per well.

Extracellular Signal-Regulated Kinase (ERK) Phosphorylation Assays
HEK-293T cells were transfected with the cDNA encoding for CB 2 R and for OX 1 R. Two to four hours before initiating the experiment, the culture medium was replaced by serum-starved DMEM medium. Cells were treated at 37 • C with selective antagonists (1 µM SB334867 or 1 µM SR144528) for 10 min prior to stimulation with OX 1 R and/or CB 2 R agonists (100 nM orexin-A and/or 100 nM JWH133) for 7 min. Cells were then placed in an ice-water bath and washed twice with cold PBS and lysed with the addition of ice-cold lysis buffer (50 mM Tris-HCl pH 7.4, 50 mM NaF, 150 mM NaCl, 45 mM ß-glycerolphosphate, 1% Triton X-100, 20 µM phenyl-arsine oxide, 0.4mMNaVO4 and the protease inhibitor mixture (MERK, St. Louis, MO, USA)). Cellular debris were removed by centrifugation at 12,000 rpm for 10 min at 4 • C, and protein was adjusted to 1 mg/mL by the bicinchoninic acid method (ThermoFisher Scientific, Waltham, MA, USA) using a commercial bovine serum albumin dilution (BSA) (ThermoFisher Scientific) for standardization. Finally, cells were denatured by placing them at 100 • C for 5 min. ERK1/2 phosphorylation were determined by western blot. Equivalent amounts of protein (20 µg) were subjected to electrophoresis (10% SDS-polyacrylamide gel) and transferred onto PVDF membranes Membranes were washed three times with PBS-tween 0.05% and once with PBS and left to dry. Bands were analyzed using an Odyssey infrared scanner (LI-COR Biosciences).
Band densities were quantified using Fiji software, and the level of phosphorylated ERK1/2 was normalized using the total ERK1/2 protein band intensities. Results obtained are represented as the percent over basal (non-stimulated cells).
To determine the ERK1/2 phosphorylation in primary cultures, 50,000 cells/well were plated in transparent 96-well microplates and kept in the incubator for 14 days. Two hours before initiating the experiment, the medium was substituted by serum-starved DMEM medium. Cells were then treated at 37 • C with selective antagonists (1 µM SB334867 or 1 µM SR144528) for 7 min prior stimulation (25 • C) with OX 1 R and/or CB 2 R agonists (100 nM orexin-A and/or 100 nM JWH133). ERK1/2 phosphorylation was determined using an AlphaScreen ® SureFire ® kit (Perkin Elmer) following the instructions of the supplier and using an EnSpire ® Multimode Plate Reader (PerkinElmer, Waltham, MA, USA).

Proximity Ligation Assay (PLA)
Direct interaction between CB 2 R and OX 1 R was detected using the Duolink in situ PLA detection Kit (OLink; Bioscience, Uppsala, Sweden) following the instructions of the supplier. Primary neurons and microglia were grown on glass coverslips, fixed in 4% paraformaldehyde for 15 min, washed with PBS containing 20 mM glycine to quench the aldehyde groups and permeabilized with the same buffer containing 0.05% Triton X-100 (20 min). Samples were then successively washed with PBS. After 1 h of incubation at 37 • C with the blocking solution in a pre-heated humidity chamber, cells were incubated overnight in the antibody diluent medium with a mixture of equal amounts of rabbit anti-CB 2 R (ab230791, Abcam, Cambridge,UK) directly coupled to a plus DNA strand (obtained following the instructions of the Sigma-Aldrich supplier, ref: DUO92010-1KT) (1/100) and rabbit anti-OX 1 R (ab83960, Abcam Cambridge,UK) directly coupled to a minus DNA strand (obtained following the instructions of the Sigma-Aldrich supplier, ref: DUO92009-1KT) (1/100). Ligation and amplification were conducted as indicated by the supplier. Samples were mounted using the mounting medium with Hoechst (1/100; Sigma-Aldrich, St. Louis, MO, USA) to stain nuclei. Samples were observed in a Zeiss 880 confocal microscope (Carl Zeiss, Oberkochen, Germany) equipped with an apochromatic 63× oil immersion objective (N.A. 1.4) and 405 nm and 561 nm laser lines. For each field of view, a stack of two channels (one per staining) and four Z stacks with a step size of 1 µM were acquired. The number of neurons or microglia containing one or more red spots versus total cells (blue nucleus) was determined, and Bonferroni's multiple comparison post hoc test was used to compare the values (red dots/cell).

Statistical Analysis
GraphPad Prism 8 software (San Diego, CA, USA) was used for data fitting and statistical analysis. One-way ANOVA followed by post hoc Bonferroni's test was used when comparing multiple values. From PLA confocal images the number of red dots/cell was quantified using the Andy's algorithm Fiji's plug-in [45].

Molecular Modelling
We used experimentally determined structures to model the monomeric structures of both CB 2 R and OX 1 R in inactive and active states. The complex CB 2 R/JWH133 was modeled based on the active structure of CB 2 R bound to the structurally related hexaydrocannabinol (PDB id 6KPF; [46]). The complex CB 2 R/SR144528 was modeled based on the inactive structure of CB 2 R bound to antagonist AM10257 (PDB id 5ZTY; [47], which also shares the central pyrazole ring located in the same position [34].The complex OX 1 R/SB334867 has been determined experimentally (PDB id 6TQ7; [35]. The complex (N-terminal part of) OX 1 R/orexin-A was modeled based on the OX 2 R/orexin-B complex (PDB id 7L1U; [48]. We used DIMERBOW [49]) to explore for possible TM4/5 interfaces to model the CB 2 R/OX1 dimer. We selected the structure of the 5-HT 2C receptor (PDB id 6BQG; [50]), which reported a compact symmetric TM4/5 interface. In turn, we superposed one CB 2 R and one OX 1 R monomeric complex to the 5-HT 2C dimer to obtain the dimeric complexes CB 2 R/JWH133-OX 1 R/SB334867 and CB 2 R/SR144528-OX 1 R/orexin-A. Because these structures showed clashes between helices TMs 4-5 of both protomers, we created new monomeric models incorporating the TMs 4 and 5 and proximal parts of the second extracellular loop in the conformation observed in the 5-HT 2C receptor. All homology models of monomers were created using Modeller 10.3 [51]. All complexes (monomeric and dimeric) were energy minimized with AMBER22 using the ff19SB forcefield [52]. Ligands were parametrized using GAFF2 [53].