Cannabinoid tolerance relies on CB₁ receptor ubiquitination by NEDD4L

Significance

Cannabinoids, the active components of cannabis, lose efficacy after repeated use due to tolerance, limiting their therapeutic value and safety. Although short-term desensitization of cannabinoid receptors is well understood, the molecular basis of long-term cannabinoid tolerance has remained unclear. This study identifies a molecular mechanism by which cannabinoids trigger selective ubiquitination and proteasomal degradation of the CB₁ receptor in the brain. This process is driven by protein kinase C–dependent activation of the E3 ubiquitin ligase neural precursor cell-expressed developmentally downregulated 4-like and is required for behavioral cannabinoid tolerance in mice. These findings provide a molecular framework for cannabinoid tolerance, a process that can contribute to cannabis use disorder and cannabis hyperemesis syndrome, and suggest strategies to improve cannabinoid-based therapies.

Abstract

Cannabinoids, the active components of cannabis, exert numerous acute effects in the brain by engaging cannabinoid CB₁ receptors (CB₁Rs). However, tolerance emerges rapidly after repeated drug exposure, undermining the efficacy of cannabinoid-based therapies and contributing to cannabis-associated adverse effects. Although the processes of CB₁R short-term desensitization (i.e., receptor uncoupling and internalization) are well characterized, the mechanisms underlying CB₁R long-term tolerance (i.e., downregulation of receptor protein levels) remain elusive. Here, we identify a ubiquitin-dependent pathway that couples CB₁R activation to its proteasomal degradation. We show that cannabinoids engage a G_q/11-PLC-PKC signaling cascade that phosphorylates and activates the E3 ubiquitin ligase neural precursor cell-expressed developmentally downregulated 4-like (NEDD4L), promoting its recruitment to CB₁R and the ubiquitination of four specific lysine residues. This modification targets the receptor for proteasomal clearance, reducing neuronal CB₁R abundance in vitro and in the mouse brain. Using molecular, pharmacological, and circuit-specific rescue approaches, we demonstrate that preventing CB₁R ubiquitination stabilizes receptor levels and abolishes behavioral cannabinoid tolerance in mice without impairing acute drug responses. These findings reveal a molecular mechanism that controls CB₁R stability and identify NEDD4L-mediated ubiquitination as a central driver of cannabinoid tolerance.

Preparations of the hemp plant Cannabis sativa L. have been used medicinally for millennia, and nowadays there is a vigorous renaissance in the scientific and clinical study of their therapeutic effects (1–3). Thus, medicinal cannabis dispensation programs have been implemented in numerous countries worldwide as well as in most US states. In addition, pharmaceuticals based on Δ⁹-tetrahydrocannabinol (THC), the main psychoactive constituent of cannabis, and cannabidiol, a nonintoxicating cannabinoid, have been approved by various regulatory agencies, including the Food and Drug Administration, the European Medicines Agency and Health Canada, for antiemetic, anticachectic, antispastic, and anticonvulsant purposes (2, 4, 5). Cannabis is also one of the most commonly used drugs of abuse in the world, and several countries and US states have legalized or decriminalized this recreational use (5–7).

THC exerts its psychoactive effects by activating the cannabinoid CB₁ receptor (CB₁R), the primary molecular target of endocannabinoids and one of the most abundant G protein–coupled receptors (GPCRs) in the mammalian brain (8, 9). CB₁R expression is particularly high in discrete regions involved in the control of learning and memory (e.g., cortex, hippocampus), motor (e.g., basal ganglia, cerebellum) and emotional (e.g., amygdala) behaviors, sensory perception (e.g., olfactory bulb), and autonomic and endocrine functions (e.g., hypothalamus, pons, medulla), where it modulates a wide plethora of cellular functions, thereby determining many different physiopathological processes (10, 11). This wide distribution of CB₁Rs accompanies the diverse acute effects of cannabinoid drugs like THC. Of note, many of these effects of THC rapidly undergo tolerance upon prolonged exposure, progressively reducing the efficacy of cannabinoid-based medicines and increasing the risks associated with cannabis use, especially in uncontrolled, recreational settings. Substantial evidence has demonstrated the existence of cannabinoid tolerance in rodents, nonhuman primates and humans, as well as neuroadaptations occurring at CB₁R that underlie this process (5, 12, 13). Cannabinoid tolerance is largely attributed to pharmacodynamic changes, particularly a decrease in the number of functionally coupled and total CB₁R molecules, with only a minor pharmacokinetic component caused by enhanced cannabinoid biotransformation and excretion (12, 13). GPCR desensitization typically involves short and long-term molecular events (14). The mechanisms of CB₁R short-term desensitization at the plasma membrane (i.e., receptor uncoupling and internalization) are quite well understood. Thus, upon agonist-evoked CB₁R activation and subsequent heterotrimeric G protein dissociation at the plasma membrane, CB₁R is phosphorylated by GPCR kinases (GRKs) at residues S425 and S429 (hCB₁R numbering). This promotes the recruitment of β-arrestins-1/2, which disrupts CB₁R-G protein coupling and enables clathrin-mediated receptor internalization (15–18). By contrast, the mechanisms underlying CB₁R long-term desensitization (i.e., downregulation of receptor protein levels) remain basically unknown.

Here, using a wide array of in vitro and in vivo approaches, we show that CB₁R ubiquitination tags the receptor for proteasomal degradation, thereby decreasing neuronal receptor protein levels and eliciting behavioral cannabinoid tolerance in mice. Our data support a molecular model in which agonist-evoked CB₁R stimulation triggers protein kinase C (PKC) signaling, leading to phosphorylation of the E3 ubiquitin ligase “neural precursor cell-expressed developmentally downregulated 4-like” (NEDD4L; aka NEDD4-2) selectively at residue S448. This modification recruits NEDD4L to CB₁R, inducing receptor ubiquitination at four specific intracellular lysine residues (K225, K315, K326, and K434) and subsequent receptor proteasomal degradation. Collectively, these findings uncover a molecular mechanism of CB₁R downregulation by ubiquitin-dependent proteolysis, thus providing a molecular basis for understanding and potentially managing cannabinoid tolerance.

Results

CB₁R Undergoes Ubiquitin-Dependent Proteasomal Degradation In Vitro.

As a first approach to investigate the mechanisms of CB₁R degradation, we generated a doxycycline-inducible (Tet-On) HEK-293T cell line expressing 3xFLAG-hCB₁R and examined the contribution of the two major proteolytic pathways in eukaryotic cells, namely the ubiquitin–proteasome system and the autophagy-lysosomal pathway, to receptor proteostasis. Upon a 6-h doxycycline withdrawal, CB₁R downregulation was prevented by the proteasomal inhibitor MG-132 (20 µM), whereas the combination of the lysosomal inhibitors E64d and pepstatin A (10 µM each) had no effect (SI Appendix, Fig. S1A). CB₁R mRNA expression remained essentially unchanged during this 6-h period (SI Appendix, Fig. S1B), indicating that the observed decline of receptor protein levels results from a posttranscriptional mechanism. Because the proteasome degrades ubiquitinated substrates, we next asked whether CB₁R downregulation depends on ubiquitination. Pharmacological inhibition of UBA1, the primary ubiquitin-activating enzyme, using TAK-243 (0.1 µM) effectively impeded CB₁R degradation (SI Appendix, Fig. S1C). A similar effect was observed when valosin-containing protein (VCP)/p97, an ATPase that facilitates proteasomal degradation of ubiquitinated membrane proteins (19), was inhibited by NMS-873 (3 µM) (SI Appendix, Fig. S1D).

To directly assess CB₁R ubiquitination, we cotransfected wild-type (WT) HEK-293T cells with 3xFLAG-hCB₁R and 3xHA-hUbiquitin (or either construct alone as controls) and conducted immunoprecipitation (IP) assays using anti-FLAG or anti-HA antibodies. To ensure appropriate detergent-based solubilization for subsequent IP and immunodetection, we used a RIPA buffer lacking urea and containing low (0.1%) SDS. This aligns with well-established protocols in the GPCR ubiquitination field designed to balance detection stringency with solubilization constraints (20). This standard approach has been previously adopted for numerous GPCRs, including the β₂-adrenergic receptor (21), CXCR4 (22), MOR (23), mGluR7 (24), P2Y1R (25), and PAR1/2 (25–27). We observed a characteristic smear of high-molecular-weight ubiquitin-immunoreactive bands in the CB₁R precipitates (Fig. 1A), indicative of receptor ubiquitination. Treatment with MG-132 led to an accumulation of CB₁R molecules conjugated with lysine 48-polyubiquitin chains (Fig. 1B), consistent with receptor proteasomal targeting (28). Moreover, the cannabinoid agonist WIN55,212-2 (WIN; 1 µM) enhanced CB₁R ubiquitination (Fig. 1C), supporting a conceivable process of agonist-evoked receptor downregulation. Pharmacological inhibition of the ubiquitin–proteasome system or VCP/p97 with TAK-243 (Fig. 1D), MG-132 (Fig. 1E), or NMS-873 (Fig. 1F) prevented WIN-induced receptor degradation, whereas the lysosomal inhibitors E64d and pepstatin A had no effect (Fig. 1G).

Fig. 1.

CB₁R undergoes ubiquitin-dependent proteasomal degradation in vitro. (A) CB₁R ubiquitination in HEK-293T cells cotransfected with 3xFLAG-hCB₁R and HA-hUb. IP was conducted with anti-FLAG (Top) or anti-HA (Bottom) antibody. WCL: whole-cell lysate. A representative experiment is shown (n = 3 experiments). (B) Top, CB₁R ubiquitination in cells transfected with 3xFLAG-hCB₁R and treated for 6 h with vehicle or MG-132 (20 µM). IP was conducted with anti-FLAG antibody. Bottom, optical density values of K48-Ub chain-modified CB₁R (means ± SEM). P values were obtained by unpaired two-tailed Student’s t test (n = 6 experiments). (C) Left, CB₁R ubiquitination in cells cotransfected with 3xFLAG-hCB₁R and HA-hUb and treated for the indicated times with WIN55,212-2 (WIN; 1 µM). IP was conducted with anti-FLAG antibody. ERK phosphorylation was used as CB₁R activation readout. Right, optical density values of ubiquitinated CB₁R (means ± SEM). P values were obtained by one-way ANOVA with Dunnett’s test (n = 4 experiments). (D–G) Top, doxycycline-inducible 3xFLAG-hCB₁R-expressing cells were doxycycline-deprived and treated for 30 min with vehicle, TAK-243 (0.1 µM) (D), MG-132 (20 µM) (E), NMS-873 (3 µM) (F) or E64d plus pepstatin A (10 µM each) (G), and then for 2 h with vehicle or WIN (1 µM). Bottom, optical density values of CB₁R relative to GAPDH (means ± SEM). P values of each WIN-treated condition relative to its vehicle (shown above the dashed line) and between WIN-treated conditions (shown in the symbol-code box) were obtained by one-way ANOVA with Sidak’s test (n = 3 to 5 experiments). (H) Left, time-course of doxycycline withdrawal in cells expressing 3xFLAG-hCB₁R or 3xFLAG-hCB₁R-8KR. Right, optical density values of CB₁R relative to HSP90 (means ± SEM). P values were obtained by one-way ANOVA with Dunnett’s test (n = 3 experiments). (I) Doxycycline-inducible cells expressing 3xFLAG-hCB₁R or 3xFLAG-hCB₁R-8KR were doxycycline-deprived and treated for 2 h with vehicle or WIN (1 µM). Right, optical density values of CB₁R relative to GAPDH (means ± SEM). P values of each WIN-treated condition relative to its vehicle (shown above the dashed line) and between WIN-treated conditions (shown in the symbol-code box) were obtained by one-way ANOVA with Sidak’s test (n = 4 experiments).

Lysine residues constitute the primary sites for ubiquitin conjugation on proteins. hCB₁R contains 24 lysine residues, eight of which [K225 at intracellular loop 2 (ICL2); K315 and K326 at intracellular loop 3 (ICL3); and K402, K434, K451, K455, and K458 at the C-terminal domain (CTD)] face the cytoplasmic compartment and are therefore plausible targets for agonist-induced, biologically regulated receptor downregulation (SI Appendix, Fig. S1E). To test this idea, we generated an intracellular lysine-deficient CB₁R mutant in which all eight intracellular lysine residues were replaced with arginine (herein termed CB₁R-8KR). We observed that CB₁R-8KR ubiquitination was hindered (SI Appendix, Fig. S1F). Likewise, compared to WT CB₁R, CB₁R-8KR accumulated at the cell surface (SI Appendix, Fig. S1G) and triggered more sustained extracellular signal-regulated kinase (ERK) activation (SI Appendix, Fig. S1H). To directly evaluate CB₁R-8KR degradation, we generated a doxycycline-inducible (Tet-On) HEK-293T cell line expressing 3xFLAG-hCB₁R-8KR. Of note, this mutant receptor was resistant to both basal (Fig. 1H) and agonist-evoked (Fig. 1I) downregulation. The use of this site-specific ubiquitin-deficient mutant reinforces the specificity of CB₁R ubiquitination by ensuring that the detection of ubiquitinated receptor species is not confounded by potentially ubiquitinated coimmunoprecipitating proteins. Taken together, these data indicate that CB₁R ubiquitination and degradation via the ubiquitin–proteasome system are driven by agonist engagement and depend on the presence of intracellular lysine residues.

NEDD4L Ubiquitinates CB₁R In Vitro.

To identify the E3 ligase(s) responsible for CB₁R ubiquitination, we first queried the UbiBrowser 3.0 database (http://ubibrowser.bio-it.cn/ubibrowser_v30) (29). Remarkably, among more than 600 human E3 ubiquitin ligases, the nine members of the neural precursor cell expressed developmentally downregulated 4 (NEDD4) family emerged as the top-ranked putative hCB₁R-ubiquitinating enzymes, with NEDD4L displaying the highest confidence score (Fig. 2A). Consistent with this prioritization, analysis of a mouse proteome turnover atlas (30) revealed a strong negative correlation between NEDD4L abundance and CB₁R lifetime across different brain regions, while no such relationship was evident for the rest of the top-ranked E3 ligases (SI Appendix, Fig. S2).

Fig. 2.

NEDD4L ubiquitinates CB₁R in vitro. (A) UbiBrowser 3.0 prediction of E3 ligases that ubiquitinate hCB₁R (http://ubibrowser.bio-it.cn/ubibrowser_v30). (B) Coimmunoprecipitation experiments in HEK-293T cells cotransfected with 3xFLAG-hCB₁R and MYC-hNEDD4L. IP was conducted with anti-FLAG (Left) or anti-MYC (Right) antibody. WCL: whole-cell lysate. A representative experiment is shown (n = 3 experiments). (C) proximity ligation assays (PLA) in cells cotransfected with 3xFLAG-hCB₁R and/or MYC-hNEDD4L was performed with anti-CB₁R and anti-MYC antibodies. Note PLA-positive (red) dots evidencing CB₁R–NEDD4L complexes in doubly transfected cells. Cell nuclei were stained with DAPI (blue). A representative experiment is shown (n = 2 experiments). (D) CB₁R ubiquitination in NEDD4L knocked-down cells. Left, validation of NEDD4L silencing by RT-qPCR. Middle, IP with anti-FLAG antibody. Right, optical density values of ubiquitinated CB₁R (means ± SEM). P values were obtained by unpaired two-tailed Student’s t test (n = 3 experiments). (E) Left, after NEDD4L knockdown, doxycycline-inducible 3xFLAG-hCB₁R-expressing cells were doxycycline-deprived and incubated for 2 h with vehicle or WIN55,212-2 (WIN; 1 µM). Right, optical density values of CB₁R relative to HSP90 (means ± SEM). P values of each WIN-treated condition relative to its vehicle (shown above the dashed line) and between WIN-treated conditions (shown in the symbol-code box) were obtained by one-way ANOVA with Sidak’s test (n = 7 experiments). (F) Left, NEDD4L autoubiquitination in cells cotransfected with 3xFLAG-hCB₁R and MYC-hNEDD4L and treated for the indicated times with WIN (1 µM). IP was conducted with anti-MYC antibody. ERK phosphorylation was used as CB₁R activation readout. Right, optical density values of ubiquitinated NEDD4L (means ± SEM). P values were obtained by one-way ANOVA with Dunnett’s test (n = 3 experiments). (G) Left, coimmunoprecipitation experiments in cells cotransfected with 3xFLAG-hCB₁R and MYC-hNEDD4L and treated for the indicated times with WIN (1 µM). IP was conducted with anti-FLAG antibody. Right, optical density values of immunoprecipitated NEDD4L relative to CB₁R (means ± SEM). P values were obtained by one-way ANOVA with Dunnett’s test (n = 3 experiments).

To validate a potential physical association between CB₁R and NEDD4L, we performed coimmunoprecipitation experiments, which showed that the two proteins interact when ectopically expressed in HEK-293T cells (Fig. 2B; note that this experimental setup is not designed for evaluating CB₁R stability, as protein degradation could be masked by compensatory changes in protein expression or turnover arising from transient overexpression). The CB₁R–NEDD4L interaction was confirmed by PLA (Fig. 2C). Next, we assessed the functional relevance of this protein–protein interaction. siRNA-mediated knockdown of endogenous NEDD4L impaired both CB₁R ubiquitination (Fig. 2D) and agonist-evoked CB₁R degradation (Fig. 2E). Structurally, NEDD4L exhibits the characteristic modular architecture of NEDD4 family members, comprising i) an alternatively spliced N-terminal Ca²⁺/lipid-binding C2 domain, ii) four tryptophan-containing WW domains that typically recognize proline-rich motifs, and iii) a C-terminal “homologous to E6-AP carboxyl terminus” (HECT) catalytic domain responsible for autoubiquitination and subsequent ubiquitin transfer to substrate proteins (31). NEDD4 family ligases normally adopt an autoinhibited state, that converts to an active, open conformation upon specific upstream triggers, thereby preventing self-tagging and ensuring selective substrate targeting (32, 33). Hence, we asked whether, upon agonist engagement, CB₁R activates NEDD4L using ligase autoubiquitination as a functional readout. WIN increased NEDD4L autoubiquitination peaking at 15 min (Fig. 2F), coinciding with an enhanced CB₁R–NEDD4L interaction (Fig. 2G). Collectively, these data indicate that CB₁R signals NEDD4L activation, enabling the ligase to bind and ubiquitinate the receptor.

We subsequently tested whether other NEDD4 family members may contribute to CB₁R ubiquitination. Among them, NEDD4 is the closest NEDD4L homolog (31). Coimmunoprecipitation (SI Appendix, Fig. S3A) and PLA (SI Appendix, Fig. S3B) experiments supported a physical association between ectopically expressed CB₁R and NEDD4 in HEK-293T cells. In addition, siRNA-mediated knockdown of endogenous NEDD4 impaired basal CB₁R ubiquitination (SI Appendix, Fig. S3C). By contrast, upon WIN-evoked CB₁R stimulation, NEDD4 silencing did not affect receptor degradation (SI Appendix, Fig. S3D), and neither NEDD4 autoubiquitination (SI Appendix, Fig. S3E) nor its recruitment to CB₁R (SI Appendix, Fig. S3F) were altered by cannabinoid challenge. These findings suggest the existence of CB₁R-coupled signaling mechanisms that selectively activate NEDD4L but not NEDD4 (see below). Finally, we ruled out the involvement of the two other top-ranked NEDD4 family members predicted by UbiBrowser, namely ITCH and SMURF1, as their knockdown did not affect CB₁R ubiquitination (SI Appendix, Fig. S3G).

CB₁R-Evoked PKC Signaling Drives NEDD4L Activation In Vitro.

We subsequently investigated the signaling mechanisms linking CB₁R activation to NEDD4L function. NEDD4L phosphorylation is known to regulate both substrate recognition and ubiquitin ligase activity (32). Specifically, serine residues 342 and 448 of hNEDD4L (SI Appendix, Fig. S4A) can be phosphorylated by various protein kinases (32), several of which, including protein kinase A (PKA), Akt, Src, and PKC, are downstream effectors of CB₁R-evoked signaling (16). Thus, we assessed phosphorylation as a potential mechanism of CB₁R-dependent NEDD4L activation in HEK-293T cells. We found that WIN induced a time-dependent phosphorylation of NEDD4L-S448 peaking at 15 min (SI Appendix, Fig. S4B), which mirrors the kinetic profile of WIN-evoked NEDD4L activation. By contrast, WIN had no effect on NEDD4L-S342 phosphorylation (SI Appendix, Fig. S4B). To prove the functional relevance of NEDD4L-S448 phosphorylation, we generated phospho-mimetic (S448D) and phospho-resistant (S448A) NEDD4L mutants. Compared to WT NEDD4L, NEDD4L-S448D showed enhanced autoubiquitination (SI Appendix, Fig. S4C), increased recruitment to CB₁R (SI Appendix, Fig. S4D), and higher capacity to ubiquitinate the receptor (SI Appendix, Fig. S4E). Accordingly, NEDD4L-S448A tended to behave in the opposite manner (SI Appendix, Fig. S4 C–E). These data support that CB₁R-evoked signaling activates NEDD4L at least in part by phosphorylating its S448 residue.

We next tested which precise CB₁R/G protein–dependent pathways mediate NEDD4L activation using a panel of selective inhibitors. Strikingly, inhibiting canonical CB₁R/G_i/o protein-coupled signaling with pertussis toxin (PTX; 1 µg/mL), as confirmed by the abrogation of cannabinoid-evoked ERK phosphorylation, did not affect WIN-induced NEDD4L-S448 phosphorylation (SI Appendix, Fig. S5A). Likewise, when blocking well established CB₁R/G_i/o protein-dependent downstream pathways, we found that neither the ERK cascade inhibitor PD98059 (30 µM), the Src inhibitor PP2 (5 µM), the PKA inhibitor H89 (20 µM) (SI Appendix, Fig. S5B), the phosphatidylinositol-3-kinase (PI3K) inhibitor wortmannin (0.2 µM), the Akt inhibitor AKTi_1/2 (0.1 µM) nor the mechanistic target of rapamycin complex 1 (mTORC1) inhibitor rapamycin (30 nM) (SI Appendix, Fig. S5C) altered WIN-induced NEDD4L-S448 phosphorylation. By contrast, inhibition of phospholipase C (PLC) with U-73122 (3 µM) or PKC with Ro 31-8220 (3 µM) abolished WIN-evoked NEDD4L-S448 phosphorylation (SI Appendix, Fig. S4F). In further agreement with this notion, U-73122 and Ro 31-8220 (SI Appendix, Fig. S4G), but not PTX (SI Appendix, Fig. S5D), prevented WIN-induced CB₁R-degradation. The diacylglycerol analog phorbol-12-myristate-13-acetate (PMA; 1.5 µM) induced per se NEDD4L-S448 phosphorylation, whereas the Ca²⁺ ionophore A23187 (5 µM) did not (SI Appendix, Fig. S4H). Accordingly, the Ca²⁺ chelator BAPTA-AM (10 µM) failed to prevent WIN-evoked NEDD4L-S448 phosphorylation (SI Appendix, Fig. S5E). Likewise, PMA, but not A23187, promoted CB₁R downregulation (SI Appendix, Fig. S4I), and BAPTA-AM did not suppress WIN-induced CB₁R degradation (SI Appendix, Fig. S5F), indicating that the Ca²⁺ branch of PLC signaling is not involved in the observed PKC-mediated effects. Collectively, these data support that agonist-activated CB₁R triggers a G_q/11 protein/PLC/PKC signaling axis to phosphorylate NEDD4L at S448. This phosphorylation recruits NEDD4L to CB₁R, thus leading to receptor ubiquitination and proteasomal degradation.

To assess whether CB₁R undergoes similar ubiquitin-mediated degradation in a more native cellular environment, we extended our analysis to mouse primary neurons. Like in HEK-293T cells, WIN induced neuronal CB₁R ubiquitination (SI Appendix, Fig. S6A) as well as receptor degradation in a proteasome-dependent (SI Appendix, Fig. S6B), lysosome-independent (SI Appendix, Fig. S6C) manner. Cannabinoid challenge also enhanced the recruitment of NEDD4L to CB₁R (SI Appendix, Fig. S6D). In addition, agonist-evoked neuronal CB₁R degradation was G_i/o protein-independent (SI Appendix, Fig. S6E) and relied on PLC (SI Appendix, Fig. S6F) and PKC (SI Appendix, Fig. S6G) signaling.

CB₁R Is Ubiquitinated at Four Specific Intracellular Lysine Residues In Vitro.

To pinpoint the precise lysine residues targeted for CB₁R ubiquitination, we immunoprecipitated 3xFLAG-hCB₁R from HEK-293T cells coexpressing 3xHA-hUbiquitin. Sample elution was performed under stringent denaturing conditions (i.e., two sequential steps with 2% SDS at 70 °C) to ensure the removal of noncovalently associated proteins, leaving only the covalently modified receptor for analysis. The immunoprecipitates were proteolytically digested and the resulting peptides analyzed by LC–MS/MS-based proteomics. This approach revealed that only four intracellular lysine residues of hCB₁R [K225 (ICL2), K315 and K326 (ICL3) and K434 (CTD)] are ubiquitinated (Fig. 3A and SI Appendix, Fig. S7).

Fig. 3.

CB₁R is ubiquitinated at residues K225, K315, K326, and K434 in vitro. (A) Amino acid residues identified as ubiquitinated in hCB₁R. HEK-293T cells were cotransfected with 3xFLAG-hCB₁R and 3xHA-hUb. IP was conducted with anti-FLAG antibody, and samples were subjected to LC–MS/MS proteomics (B) AlphaFold2-Multimer model of hCB₁R (dark green; residues 103 to 472) in complex with the hNEDD4L-HECT domain (gray; residues 640 to 974), together with the predicted aligned error plot and confidence metrics. (C) Left, coimmunoprecipitation experiments in cells cotransfected with 3xFLAG-hCB₁R and MYC-hNEDD4L-WT, MYC-hNEDD4L-ΔHECT, or MYC-hNEDD4L-HECT. IP was conducted with anti-CB₁R antibody. WCL: whole-cell lysate. Right, optical density values of immunoprecipitated NEDD4L relative to CB₁R (means ± SEM). P values were obtained by one-way ANOVA with Tukey’s test (n = 3 experiments). (D) Left, CB₁R ubiquitination in cells cotransfected with MYC-hNEDD4L (or empty construct as control), HA-hUb, and 3xFLAG-hCB₁R, 3xFLAG-hCB₁R-2KR, 3xFLAG-hCB₁R-3KR, or 3xFLAG-hCB₁R-4KR. IP was conducted with anti-FLAG antibody. Right, optical density values of ubiquitinated CB₁R (means ± SEM). P values of each NEDD4L-expressing condition relative to its NEDD4L-empty condition (shown above the dashed line) and between NEDD4L-expressing conditions (shown in the symbol-code box) were obtained by one-way ANOVA with Sidak’s test (n = 7 experiments). (E) Left, CB₁R ubiquitination in cells transfected with 3xFLAG-hCB₁R, 3xFLAG-hCB₁R-2KR, 3xFLAG-hCB₁R-3KR, or 3xFLAG-hCB₁R-4KR, and treated for 15 min with vehicle or WIN55,212-2 (WIN; 1 µM). IP was conducted with anti-FLAG antibody. Right, optical density values of ubiquitinated CB₁R (means ± SEM). P values of each WIN-treated condition relative to its vehicle (shown above the dashed line) and between WIN-treated conditions (shown in the symbol-code box) were obtained by one-way ANOVA with Sidak’s test (n = 5 to 6 experiments). (F) Left, doxycycline-inducible cells expressing 3xFLAG-hCB₁R, 3xFLAG-hCB₁R-2KR, 3xFLAG-hCB₁R-3KR, or 3xFLAG-hCB₁R-4KR were doxycycline-deprived and treated for 2 h with vehicle or WIN (1 µM). Right, optical density values of CB₁R relative to vinculin (means ± SEM). P values of each WIN-treated condition relative to its vehicle (shown above the dashed line) and between WIN-treated conditions (shown in the symbol-code box) were obtained by one-way ANOVA with Sidak’s test (n = 9 experiments).

We subsequently leveraged AlphaFold multimer to model the CB₁R–NEDD4L complex (34–36). To simplify the search space, we trimmed the first 102 amino acids of hCB₁R because these residues are i) extracellular, ii) foreseen as inherently disordered by both AlphaFold and PrDOS (SI Appendix, Fig. S8 A and B), and iii) largely absent from currently resolved CB₁R structures (37). We then predicted complexes between this hCB₁R sequence and the distinct domains of hNEDD4L (i.e., C2, WW1, WW2, WW3, WW4, and HECT) in isolation. Remarkably, a model of the HECT domain (encompassing amino acids 640 to 974 of the ligase) bound to CB₁R, largely through the ICL3 of the receptor, gave reasonable confidence scores (pTM = 0.69, ipTM = 0.62, Multimer score = 0.63; Fig. 3B and SI Appendix, Fig. S8C). In this model, the HECT catalytic domain is conceivably positioned near the four cytoplasm-facing lysine residues of CB₁R identified as ubiquitinated in our LC–MS/MS analysis. Consistent with this prediction, coimmunoprecipitation experiments showed that the isolated HECT domain is sufficient to mediate the interaction with CB₁R (Fig. 3C). Together, these findings support a model in which the HECT domain of NEDD4L directly engages CB₁R to enable receptor ubiquitination at defined intracellular lysines.

We next assessed the relative contribution of each of the four ubiquitinated CB₁R lysine residues to overall receptor ubiquitination and degradation. For this purpose, based on the CB₁R domain ubiquitination predicted by the AlphaFold model (ICL3 > ICL2 > CTD), we generated a series of lysine-deficient hCB₁R mutants sequentially targeting i) K315 and K326 (herein termed CB₁R-2KR), ii) K225, K315, and K326 (herein termed CB₁R-3KR) or iii) K225, K315, K326, and K434 (herein termed CB₁R-4KR). We cotransfected HEK-293T cells with MYC-hNEDD4L (or without it as control), 3xHA-hUbiquitin and each 3xFLAG-hCB₁R mutant, and assessed CB₁R ubiquitination. A progressive decline in NEDD4L-induced ubiquitination was observed from CB₁R-WT (maximal ubiquitination) through CB₁R-2KR and CB₁R-3KR (intermediate ubiquitination) to CB₁R-4KR (no ubiquitination) (Fig. 3D). A similar stepwise pattern was found when measuring i) agonist-evoked CB₁R ubiquitination (Fig. 3E) and ii) agonist-evoked CB₁R protein degradation (Fig. 3F). These data provide further support to the site-specificity of CB₁R ubiquitination and indicate that each of the four ubiquitinated lysine residues contributes additively to CB₁R ubiquitin-dependent degradation.

CB₁R-Evoked PKC Signaling Drives NEDD4L Activation In Vivo.

Once the molecular mechanism was delineated in vitro, we evaluated the putative CB₁R–NEDD4L connection in the mouse brain, selecting the hippocampus and the dorsal striatum as two regions of paradigmatic CB₁R expression and functional relevance (8, 10, 38–40). First, we conducted coimmunoprecipitation studies on fresh tissue extracts, which revealed a specific association between the two proteins in both brain regions (Fig. 4A). To examine the effect of agonist-evoked receptor activation, mice were acutely injected with vehicle or THC (10 mg/kg, i.p.) and, 30 min later, hippocampal and dorsal-striatum extracts were prepared. THC administration enhanced CB₁R–NEDD4L association, as demonstrated by coimmunoprecipitation (Fig. 4B), and increased NEDD4L activation in those complexes, as evidenced by elevated NEDD4L-S448 phosphorylation (Fig. 4C). By contrast, NEDD4L-S342 phosphorylation remained unchanged (Fig. 4C), consistent with our in vitro observations. Similarly, cannabinoid-induced phosphorylation of NEDD4L-S448 was prevented by Ro 31-8220 (6 mg/kg, i.p.) (Fig. 4D) and mimicked by PMA (0.2 mg/kg, i.p.) (Fig. 4E). Taken together, these data indicate that agonist-evoked CB₁R stimulates NEDD4L activation in the mouse brain via PKC-dependent phosphorylation of S448, thus mirroring the mechanism found in vitro.

Fig. 4.

CB₁R-evoked PKC signaling drives NEDD4L activation in the mouse brain. (A) Coimmunoprecipitation experiments in the hippocampus (Left) and dorsal striatum (Right) of WT mice. IP was conducted with anti-CB₁R antibody. Tissue from CB₁R-KO mice was used as control. WTL: whole-tissue lysate. A representative experiment is shown (n = 3 pools of 2 mice per pool). (B) Coimmunoprecipitation experiments in the hippocampus (Top-Left) and dorsal striatum (Top-Right) from WT mice treated for 30 min with vehicle or THC (10 mg/kg, i.p). IP was conducted with anti-CB₁R antibody. A representative experiment is shown. Bottom, optical density values of immunoprecipitated NEDD4L relative to CB₁R (means ± SEM). P values were obtained by unpaired two-tailed Student’s t test (n = 3 pools of 2 mice per pool). (C) Left, WT mice were treated for 30 min with vehicle or THC (10 mg/kg, i.p) and NEDD4L phosphorylation at S448 and S342 was analyzed in the hippocampus and dorsal striatum. Right, phosphorylated NEDD4L relative to total NEDD4L (means ± SEM). P values were obtained by unpaired two-tailed Student’s t test (n = 8-10 mice per group). (D) Left, WT mice were treated for 30 min with vehicle or Ro 31-8220 (Ro 31; 6 mg/kg, i.p) and then for 30 min with vehicle or THC (10 mg/kg, i.p), and NEDD4L phosphorylation at S448 was analyzed in the hippocampus and dorsal striatum. Right, phosphorylated NEDD4L relative to total NEDD4L (means ± SEM). P values were obtained by one-way ANOVA with Sidak’s test (n = 4 to 5 mice per group). (E) Left, WT mice were treated for 30 min with vehicle or PMA (0.2 mg/kg, i.p.) and NEDD4L phosphorylation at S448 was analyzed in the hippocampus and dorsal striatum. Right, phosphorylated NEDD4L relative to total NEDD4L (means ± SEM). P values were obtained by unpaired two-tailed Student’s t test (n = 7 to 8 mice per group). In all panels, note that both long, full-length (~130 KDa) and short, ΔC2-domain (~110 KDa) isoforms of NEDD4L were detected.

CB₁R Proteasomal Degradation Underlies Behavioral Cannabinoid Tolerance In Vivo.

Pharmacological tolerance to cannabinoids is associated with reduced CB₁R protein expression across multiple brain regions (41–43). Hence, we asked whether prolonged cannabinoid administration promotes CB₁R proteasomal degradation in vivo. Because the precise functional role of CB₁R depends on the neurochemical identity of the cell expressing it, we first analyzed the two archetypal neuronal populations harboring the receptor in the mammalian brain, namely glutamatergic neurons and GABAergic neurons (39, 44). For this purpose, we used conditional Cre-loxP knockout mice in which the CB₁R-encoding gene (Cnr1) had been selectively deleted from dorsal telencephalic glutamatergic neurons [Cnr1^{fl/fl;Nex1-Cre} (herein termed Glu-CB₁R-KO) mice] or from forebrain GABAergic neurons [Cnr1^{fl/fl;Dlx5/6-Cre} (herein termed GABA-CB₁R-KO) mice] (44). These mutants, and their control littermates [CB₁R^fl/fl (herein termed CB₁R-floxed) mice], were treated for five consecutive days with vehicle or THC (10 mg/kg/d, i.p.), after which CB₁R protein expression was analyzed in the hippocampus and dorsal striatum. Sustained cannabinoid administration produced a comparable reduction of CB₁R protein levels in both brain regions of the two mutant lines (SI Appendix, Fig. S9 A–D) without altering CB₁R mRNA expression (SI Appendix, Fig. S9 E–H), thus pointing to a common posttranscriptional process of CB₁R downregulation in excitatory and inhibitory nerve terminals across the mouse brain. To directly test the possible involvement of proteasomal degradation, we injected WT mice for five consecutive days with vehicle or THC (10 mg/kg/d, i.p.) in the presence or absence of either MG-132 (5 mg/kg/d, i.p.) or another proteasomal inhibitor, the clinically relevant chemotherapeutic agent bortezomib (BTZ; 0.5 mg/kg/d, i.p.) (SI Appendix, Fig. S10A). Pharmacological blockade of the proteasome impeded the cannabinoid-evoked reduction of CB₁R protein in the hippocampus and dorsal striatum (SI Appendix, Fig. S10 B and C), while CB₁R mRNA levels remained unchanged across brain regions and pharmacological treatments (SI Appendix, Fig. S10 D and E). Together, these data support that repeated cannabinoid exposure triggers proteasome-dependent degradation of CB₁R in vivo.

We subsequently investigated whether CB₁R proteasomal degradation contributes to behavioral cannabinoid tolerance in mice assessing the classical “cannabinoid tetrad,” which comprises four hallmark responses to cannabinoid administration: hypothermia, analgesia, hypolocomotion, and catalepsy (45). Mice were treated as above with vehicle or THC plus vehicle or MG-132, and the cannabinoid tetrad was evaluated at both day 1 and day 5 right after each acute THC injection. On day 1, THC induced its characteristic hypothermic, analgesic, hypolocomotor, and cataleptic effects regardless of MG-132 coadministration (SI Appendix, Fig. S10 F, Upper row). On day 5, as expected, mice treated with THC alone had developed tolerance to all four behavioral effects (SI Appendix, Fig. S10 F, Lower row). By contrast, THC and MG-132-cotreated animals, presumably resistant to CB₁R protein degradation, retained full sensitivity to the acute analgesic, hypolocomotor, and cataleptic actions of the cannabinoid (SI Appendix, Fig. S10 F, Lower row). Tolerance to THC-induced hypothermia was only partially blocked by MG-132. This likely reflects the contribution of alternative THC targets (e.g., TRPV₁ and CB₂R) to the control of body temperature (46). Additionally, it may involve a proteasome-independent pathway for CB₁R degradation, potentially within body temperature-regulating neurons of the preoptic anterior hypothalamus (46). We could not assess behavioral traits reliably in BTZ and THC+BTZ-treated mice on day 5 because their overall motor activity was affected. Collectively, these data indicate that CB₁R proteasomal degradation is necessary for the development of behavioral cannabinoid tolerance in vivo.

CB₁R Ubiquitination at Four Specific Intracellular Lysine Residues Underlies Behavioral Cannabinoid Tolerance In Vivo.

Among the four behavioral tests that compose the cannabinoid tetrad, catalepsy -a core motor symptom of catatonia- is the one that involves the best-defined neuronal circuit. Specifically, THC-induced catalepsy in mice has been attributed to the selective activation of CB₁R molecules located on axon terminals of dopamine D₁ receptor (D₁R)-expressing medium spiny neurons (MSNs) of the dorsal striatum, which constitute the “direct,” striatonigral pathway controlling motor behavior (44, 47). Using RNAscope, we observed a robust colocalization of CB₁R and NEDD4L transcripts in the mouse dorsal striatum (where CB₁R displays its characteristic dorsolateral expression gradient) (48), conceivably within the somas of MSNs (the neurons that constitute the vast majority of striatal cells) (49) (Fig. 5A). Likewise, when interrogating a mouse whole-brain scRNA-seq atlas (50), we found an overt coexpression of the mRNAs encoding CB₁R and NEDD4L in striatal MSNs (SI Appendix, Fig. S11). Consistently, coimmunoprecipitation assays revealed a physical association between CB₁R and NEDD4L proteins in the substantia nigra (the region containing the axonal projections of D₁R-expressing MSNs) (48) (Fig. 5B). We therefore focused on this neuronal circuit to study the contribution of CB₁R ubiquitination to behavioral tolerance. For this purpose, we used conditional Cre-loxP knockout mice in which the Cnr1 gene had been selectively deleted from D₁R-expressing cells [Cnr1^{fl/fl;Drd1-Cre} (herein termed D₁R-CB₁R-KO) mice] (44). These mice and their control CB₁R-floxed littermates were bilaterally injected into the dorsal striatum with a Cre-dependent, double-floxed inverse open reading frame (DIO) AAV1/2-CAG-DIO-3xFLAG-hCB₁R-WT vector. This approach will restore CB₁R expression in D₁R-CB₁R-KO animals selectively within their infected Cre recombinase-expressing cells (that is, D₁R-containing, striatonigral MSNs) (Fig. 5C). 4 wk after surgery, to allow sufficient transgene expression, mice were divided into two treatment groups (vehicle or THC at 10 mg/kg/d, i.p.). Following four consecutive daily injections, the catalepsy test was assessed on day 5 starting 30 min after the final THC injection (Fig. 5C). Rescuing CB₁R expression in striatonigral MSNs restored both the sensitivity to the acute cataleptic effect of THC, as shown in control mice chronically treated with vehicle (Fig. 5D; gray and blue solid bars), and the development of cannabinoid tolerance, as shown in mice chronically treated with THC (Fig. 5D; gray and blue dashed bars). Immunohistochemical analysis confirmed that delivery of the CB₁R-encoding vector reinstated CB₁R protein expression in the substantia nigra of D₁R-CB₁R-KO mice (Fig. 5E, gray and blue solid bars; Fig. 5F, gray and blue top boxes) and that prolonged THC treatment decreased those CB₁R levels (Fig. 5E, gray and blue dashed bars; Fig. 5F, gray and blue bottom boxes).

Fig. 5.

CB₁R ubiquitination at four specific intracellular lysine residues underlies behavioral cannabinoid tolerance in mice. (A) Expression of CB₁R mRNA (green) and NEDD4L (red) in the brain of WT mice as assessed by RNAscope. Nuclei were stained with DAPI (blue). The dotted line depicts the high-magnification inset of the dorsal striatum shown on the right-hand side. A representative experiment is shown (n = 3 mice). (B) Coimmunoprecipitation of CB₁R and NEDD4L in the substantia nigra of WT mice. IP was conducted with anti-CB₁R antibody. Tissue from CB₁R-KO mice was used as control (the asterisks indicate nonspecific bands). WTL: whole-tissue lysate. A representative experiment is shown (n = 3 pools of 4 to 5 mice per pool). Note that both long, full-length (~130 KDa) and short, ΔC2-domain (~110 KDa) isoforms of NEDD4L were detected. (C) Top, timeline of the CB₁R expression-rescue experiments. D₁R-CB₁R-KO mice or their control CB₁R-floxed littermates were bilaterally injected into the dorsal striatum with AAV1/2-CAG-DIO-Empty, AAV1/2-CAG-DIO-3xFLAG-hCB₁R-WT, AAV1/2-CAG-DIO-3xFLAG-hCB₁R-8KR, or AAV1/2-CAG-DIO-3xFLAG-hCB₁R-4KR (vector schematic is shown). 4 wk later, animals were treated for four consecutive days with vehicle or THC (10 mg/kg/d, i.p.). On day 5, animals were treated with THC (10 mg/kg, i.p.) and catalepsy was evaluated. Bottom, schematic of the CB₁R expression-rescue strategy in each animal group. (D) Cataleptic response to the acute THC injection at day 5 in control mice and 3xFLAG-hCB₁R-WT, 3xFLAG-hCB₁R-8KR, or 3xFLAG-hCB₁R-4KR-rescued mice. P values of each chronic THC-treated condition relative to its chronic vehicle (shown above the columns) and between chronic THC-treated conditions (shown in the symbol-code box) were obtained by Mann–Whitney’s U test (n = 4 to 12 mice per group). (E) CB₁R immunoreactivity at day 5 in the substantia nigra of control mice and 3xFLAG-hCB₁R-WT, 3xFLAG-hCB₁R-8KR, or 3xFLAG-hCB₁R-4KR-rescued mice. P values of each chronic THC-treated condition relative to its chronic vehicle (shown above the columns) and between chronic THC-treated conditions (shown in the symbol-code box) were obtained by two-way ANOVA with Tukey’s test (n = 4 to 11 mice per group). (F) Representative images of CB₁R immunoreactivity (green) at day 5 in the substantia nigra of control mice and 3xFLAG-hCB₁R-WT, 3xFLAG-hCB₁R-8KR, or 3xFLAG-hCB₁R-4KR-rescued mice after chronic vehicle or chronic THC treatment. Nuclei were stained with DAPI (blue). The dotted line depicts the microscopy-counting area.

After setting up this conditional CB₁R expression-rescue model, we conducted an analogous experiment using vectors encoding CB₁R-8KR (AAV1/2-CAG-DIO-3xFLAG-hCB₁R-8KR) and CB₁R-4KR (AAV1/2-CAG-DIO-3xFLAG-hCB₁R-4KR), in which ubiquitination of either all the eight intracellular CB₁R lysine residues or only the four LC–MS/MS-identified ubiquitinated CB₁R lysine residues (i.e., K225, K315, K326, and K434) is respectively prevented. Expression of CB₁R-8KR or CB₁R-4KR (like CB₁R-WT, see above) in striatonigral MSNs of D₁R-CB₁R-KO mice restored the acute cataleptic response to THC (Fig. 5D; orange and pink solid bars). By contrast (unlike CB₁R-WT, see above), delivery of either CB₁R-8KR or CB₁R-4KR-encoding vectors to striatonigral MSNs impaired the development of tolerance to chronic THC exposure (Fig. 5D; orange and pink dashed bars). Consistent with this observation, we found that, upon prolonged cannabinoid treatment, CB₁R-8KR or CB₁R-4KR expression in D₁R-CB₁R-KO mice allowed higher receptor protein levels at the substantia nigra than in animals injected with the CB₁R-WT vector, conceivably owing to the resistance of both KR mutant receptors to proteasomal degradation (Fig. 5E, dashed bars; Fig. 5F, bottom boxes). In summary, these findings, using a neuron circuit-restricted, conditional CB₁R expression-rescue system, support that CB₁R ubiquitination at four specific intracellular lysine residues underlies cannabinoid-induced receptor degradation and behavioral tolerance in vivo.

Discussion

Tolerance is a fundamental adaptive process that attenuates cannabinoid sensitivity primarily through CB₁R downregulation rather than through changes in drug pharmacokinetics (12, 13). Here, we delineate a molecular cascade coupling agonist-evoked CB₁R activation to proteolytic degradation through the ubiquitin–proteasome system. This pathway is driven by PKC-dependent phosphorylation and activation of the E3 ubiquitin ligase NEDD4L, leading to site-specific ubiquitination of intracellular lysine residues within CB₁R. This chemical modification of the receptor promotes the development of behavioral cannabinoid tolerance in mice. Our findings apparently differ from previous reports pointing to endo-lysosomal targeting as the main route for CB₁R degradation. Many of those studies, however, used in vitro biotin-protection assays in combination with lysosomal inhibitors, thus capturing only a fraction of internalized receptors that undergo lysosomal degradation (51–53). By measuring bulk receptor protein abundance, our approach provides a more comprehensive view of CB₁R degradative fate. In any event, both proteolytic pathways may well coexist, operating in a coordinated and kinetically regulated manner. Given the high constitutive activity of CB₁Rs (54, 55), we speculate that endo-lysosomal targeting functions as a rapid, short-term regulatory mechanism, whereas sustained receptor activation preferentially engages proteasomal degradation to maintain longer-term homeostasis.

We demonstrate that CB₁R activation promotes receptor ubiquitination, a mechanism that was suggested from indirect evidence in two previous studies. Thus, proteasome-dependent CB₁R downregulation in the mouse medial prefrontal cortex, as induced by prolonged estrogen deprivation, was associated with upregulation of the E3 ubiquitin ligases Neurl1a/b and enhanced coimmunoprecipitation of the receptor with Neurl1a/b and ubiquitin (56). In addition, 13-docosenamide-induced CB₁R deubiquitination in mouse oligodendrocyte precursor cells increased receptor protein levels in concert with enhanced coimmunoprecipitation of the receptor with the deubiquitinase USP33 (57). Noteworthy, K226, K316, and K327, the three mCB₁R residues equivalent to hCB₁R K225, K315, and K326 (herein identified as ubiquitination sites in a targeted human cell-based experimental setting), were also found ubiquitinated in high-throughput proteomic analyses of mouse brain extracts (58, 59), thereby reinforcing the biological relevance of our LC–MS/MS data. To our knowledge, no previous study had reported the ubiquitination of the hCB₁R-K434/mCB₁R-K435 residue. Taken together, our findings establish CB₁R ubiquitination as a key mechanism to control receptor turnover in vitro and in vivo, and provide a molecular framework for future studies.

From a signal transduction perspective, our data support that agonist-evoked CB₁R engagement triggers receptor degradation in a G_i/o protein-independent, G_q/11 protein-dependent manner. This is consistent with previous reports showing that CB₁R can act through G_q/11 proteins in various cellular contexts (60–64). In this context, experiments conducted with COS7 cells have shown that CB₁R activates G₁₆, a G_q/11 protein family member that can couple various GPCRs to PLC stimulation (65). Another key insight from our signaling studies is that the PKC isoform involved in CB₁R degradation presumably belongs to the novel (diacylglycerol-sensitive, Ca²⁺-insensitive) rather than the conventional (diacylglycerol and Ca²⁺-sensitive) or atypical (diacylglycerol and Ca²⁺-insensitive) PKC subfamily, thus leaving PKCδ, PKCε, PKCη, and PKCθ as potential candidates (66). In this regard, inhibition or genetic deletion of PKCε in mice enhances behavioral responses to WIN55,212-2 (67). Moreover, S317 at rat CB₁R (equivalent to S316 at hCB₁R) undergoes PKC-dependent phosphorylation, resulting in receptor inactivation (68). This suggests a potential two-hit regulatory mechanism in which receptor phosphorylation precedes ubiquitination, as observed for another class-A GPCR like the β2-adrenergic receptor (21). In addition, owing to the immediate contiguity of K315 to S316 (hCB₁R numbering) at the ICL3 of the receptor, we cannot rule out a potential physical/functional interdependency between PKC-mediated receptor phosphorylation and ubiquitination processes.

At the functional level, the CB₁R–NEDD4L axis could not only determine receptor abundance, and therefore cannabinoid tolerance, but also contribute to the control of CB₁R-mediated synaptic plasticity, given that i) at least one form of CB₁R-mediated long-term depression, specifically long-term depression of inhibitory transmission, depends on local protein ubiquitination in rat hippocampal axons (69); ii) NEDD4L localizes to synapses established by Camk2a-expressing mouse striatal neurons (conceivably MSNs) (70) (syndive.org), and 16% of its interacting proteins reported in BioGRID (thebiogrid.org) are also synaptic; iii) the phosphorylation status of NEDD4L-S448 influences AMPA receptor activity and neuronal excitability (71); and iv) mutations in the NEDD4L gene cause diverse forms of epilepsy (72–74), a pathological situation in which CB₁R is also markedly dysregulated (75–77). Hence, whether neurotransmission-induced CB₁R activation engages NEDD4L to modulate functional plasticity via the ubiquitin–proteasome system represents an intriguing avenue for future research.

Substantial epidemiological, clinical, and brain-imaging evidence supports a pharmacodynamic model of cannabis tolerance in humans that relates downregulatory changes occurring at CB₁Rs to a blunted response upon repeated cannabis exposure (5, 12, 13). As cannabis tolerance depends on THC dose, frequency, and duration of use, it is more likely to occur in recreational users who consume high doses of cannabis continuously for a prolonged period (5, 12, 13). This makes these heavy users escalate drug intake to achieve the desired effects and avoid withdrawal symptoms, thereby increasing the risk of cannabis use disorder. In this situation, neuroadaptations within brain reward and cognitive-control systems may induce compulsive consumption, behavioral dependence, and substance addiction (5, 12, 13, 78, 79). Profound tolerance can also drive cannabis hyperemesis syndrome, a prevalence-increasing chronic condition characterized by cyclic episodes of abdominal pain, nausea, and severe vomiting, together with thermoregulatory shifts. The disorder is strongly associated with previous repeated cannabis use and only resolves upon cessation of drug intake (5, 12, 13, 80, 81). Chronic cannabis exposure is believed to desensitize the characteristic central antiemetic mechanisms of cannabinoids, increase stress, and interfere with thermoregulation, while cannabinoid overstimulation in peripheral tissues disrupts gastrointestinal motility through enteric nervous system dysregulation (5, 12, 13, 80, 81). Regarding the clinical setting, the expanding access to medicinal cannabis and prescription cannabinoids also highlights the challenge of managing drug-induced tolerance and functional impairment, which may make patients raise their cannabinoid dosages over time to maintain clinical benefits (5, 12, 13, 82–85). Thus, specific posology protocols based on slow dose escalation combined with low-dose regimens, aimed to allow symptom relief without causing unwanted psychoactive effects, are usually recommended (86). In this expanding landscape of recreational and medicinal cannabis use, the preclinical findings reported here advance our understanding of the mechanism of CB₁R action, open potential avenues for therapeutic interventions to mitigate cannabinoid tolerance, and provide directions for exploring how the interplay between CB₁R and the ubiquitin–proteasome system impacts brain physiopathology.

Materials and Methods

The experimental procedures used are described in detail in SI Appendix, Supporting Materials and Methods. Briefly, we employed adult male and female WT and genetically modified mice, including systemic and neuron population-specific CB₁R knockout models, maintained and processed according to approved animal welfare protocols. Brain tissue was collected for biochemical, molecular, and histological analyses, while primary neurons were isolated for cell-culture experiments. HEK-293T cells were cultured and engineered to express various CB₁R mutants. Constructs were generated by cloning and site-directed mutagenesis, and gene silencing was achieved using siRNA-mediated knockdown. RT-qPCR, western blotting, IP, PLA, RNAscope, and immunomicroscopy were conducted by standard procedures. CB₁R degradation and signaling assays were performed under pharmacological manipulation of various intracellular pathways. Ubiquitinated CB₁R species were characterized by LC–MS/MS-based proteomics, and receptor cell-surface expression was evaluated by biotinylation assays. In vivo studies included acute and repeated administration of cannabinoids and other pharmacological agents, followed by behavioral testing using the cannabinoid tetrad paradigm. Brain circuit-specific expression of CB₁R variants was achieved through stereotaxic delivery of AAV vectors into the dorsal striatum. Experimental design included randomization, blinded assessment, and statistical analyses appropriate for each dataset.

Supplementary Material

Appendix 01 (PDF)

Data, Materials, and Software Availability

Study data are included in the article and/or SI Appendix.