Targeting Peripheral CB₁ Receptors Reduces Ethanol Intake via a Gut-Brain Axis

SUMMARY

Endocannabinoids acting on the cannabinoid-1 receptor (CB₁R) or ghrelin acting on its receptor (GHS-R_1A) both promote alcohol-seeking behavior, but an interaction between the two signaling systems has not been explored. Here we report that the peripheral CB₁R inverse agonist JD5037 reduces ethanol drinking in wild-type mice but not in mice lacking CB₁R, ghrelin peptide or GHS-R_1A. JD5037 treatment of alcohol drinking mice inhibits the formation of biologically active octanoyl-ghrelin without affecting its inactive precursor desacyl-ghrelin. In ghrelin-producing stomach cells, JD5037 reduced the level of the substrate octanoyl-carnitine generated from palmitoyl-carnitine by increasing fatty acid β-oxidation. Blocking gastric vagal afferents abrogated the ability of either CB₁R or GHS-R_1A blockade to reduce ethanol drinking. We conclude that blocking CB₁R in ghrelin-producing cells reduces alcohol drinking by inhibiting the formation of active ghrelin and its signaling via gastric vagal afferents. Thus, peripheral CB₁R blockade may have therapeutic potential in the treatment of alcoholism.

In Brief (eTOC blurb)

Endocannabinoids and ghrelin are both implicated in promoting alcohol intake. Godlewski et al. show that cannabinoid-1 receptor (CB₁R) signaling in ghrelin-producing stomach cells modulates the formation of biologically active octanoyl-ghrelin and then acts via gastric vagal afferents to promote alcohol consumption. They further show that peripherally-restricted pharmacological inhibition of CB₁R reduces ethanol drinking in mice.

Graphical Abstract

INTRODUCTION

C57Bl6/J mice display hedonic traits, as reflected in their preference for eating diets high in fat and carbohydrates or drinking alcohol. Both of these preferences are driven, to a large extent, by endocannabinoids acting via its receptor CB₁R, as they are suppressed by pharmacological inhibition (Colombo et al., 2005; Di Marzo et al., 2001) or genetic deletion of CB₁R (Hungund et al., 2003; Ravinet Trillou et al., 2004; Wang et al., 2003). Endocannabinoids and CB₁R are obligatory components of the hypothalamic and limbic neural circuitries involved in the control of both food and drug seeking behavior (Solinas et al., 2008). Therefore, evidence that CB₁R in the CNS may be dispensable for the anorectic effect of CB₁R blockade was surprising (Bellocchio et al., 2013; DiPatrizio et al., 2011; Gomez et al., 2002), and it suggested the involvement of a periphery-to-brain axis in this effect, although the nature of the signaling pathway was not identified. Direct evidence for the dominant role of peripheral CB₁R in the control of food intake came from studies with the non-brain-penetrant CB₁R inverse agonist JD5037 (Chorvat et al., 2012), which was as effective as its brain-penetrant counterpart ibipinabant in reducing food intake in mice with diet-induced obesity (DIO), despite the documented absence of central CB₁R occupancy by JD5037 (Tam et al., 2012). Furthermore, the JD5037-induced hypophagia could be attributed to endogenous leptin, as the obesity-related hyperleptinemia thought to be responsible for maintaining leptin resistance (Li et al., 2013) was rapidly normalized by peripheral CB₁R blockade that decreased leptin production in adipocytes and increased leptin clearance in the kidney, thus restoring sensitivity to endogenous leptin (Tam et al., 2012). To test if periphery-to-brain signaling may also be involved in the control of alcohol drinking behavior by endocannabinoids, we analyzed the effects of JD5037 treatment on alcohol drinking by male C57Bl6/J mice, using the two-bottle free-choice and the drinking-in-the-dark paradigms.

RESULTS

Peripheral CB₁R Mediates High Alcohol Preference and Intake in C57Bl6 Mice

Male C57Bl6/J mice with access to water or a 15% ethanol solution displayed high preference for alcohol, resulting in an average daily intake of 11.5 ± 0.4 mg ethanol/g body weight (n = 84). The high alcohol preference and intake remained unaffected by daily gavage with vehicle, whereas both were sharply and similarly reduced during 5 days of treatment with daily oral doses of 3 mg/kg of the peripherally restricted CB₁R inverse agonist JD5037 (Tam et al., 2012) or 10 mg/kg of the brain-penetrant CB₁R inverse agonist rimonabant (Figure 1A), the difference in dosing reflecting the higher CB₁R binding affinity of JD5037 (K_d: 0.4 nM) compared to rimonabant (K_d: 3.2 nM). Alcohol intake was also significantly reduced by JD5037 treatment in a restricted access/drinking-in-the-dark paradigm that results in inebriating blood levels of ethanol (Figure 1B). JD5037 treatment did not affect total fluid intake, or preference for solutions containing saccharin, sucrose or quinine, suggesting that peripheral CB₁R blockade has no effect on the perception of thirst or the taste or caloric value of ethanol (Figure S1A), and also failed to affect the rate of alcohol metabolism (Figure S1B). The role of CB₁R in high alcohol preference and its inhibition by JD5037 or rimonabant was further supported by findings in CB₁R knockout (Cnr1^−/−) mice whose significantly lower ethanol preference compared to wild-type controls was unaffected by CB₁R blockade (Figure 1A). This less than 50% preference induced by CB₁R blockade or deletion may be interpreted as alcohol ‘avoidance’.

Figure 1. Effect of CB₁R Blockade on Alcohol Drinking in Wild-type Mice and in Mice Deficient in CB₁R, Ghrelin Peptide or Ghrelin Receptor. (see also Figures S1, S2 and S4 and Table S1).

(A) The effect of CB₁R blockade on alcohol preference, intake and total fluid intake in wild-type and Cnr1^−/− littermate mice. All 3 groups were treated daily with vehicle from day 2 to 5, and continued with vehicle (open circles), 10 mg/kg/day rimonabant (red) or 3 mg/kg/day JD5037 (blue) by oral gavage from day 6 to 10. Mice had free access to water or a 15% ethanol solution in water, using a two-bottle/free choice paradigm described in Methods. Points and vertical bars represent means ± SEM from the indicated number of experiments. For the bar graph on ethanol consumption, the vehicle groups for rimonabant and JD5037 were combined. Results were analyzed using one or two-way ANOVA, as appropriate. *P < 0.05, **P < 0.01, ***P < 0.001 relative to vehicle group. (B) Alcohol intake and blood ethanol concentrations in mice using a limited access/drinking in the dark paradigm described in Methods. Mean values indicated by horizontal lines; significance of difference between pre and post-JD5037 values, **P < 0.01. (C) Effect of peripheral CB₁R blockade on alcohol drinking in wild-type and Ghrl^−/− littermate mice. Treatments and symbols as described in panel A. (D) Effect of peripheral CB₁R blockade on alcohol drinking in wild-type and Ghsr^−/− littermate mice. Treatments and symbols as described in panel (A).

Endocannabinoid Modulation of Alcohol Drinking is Ghrelin-Dependent

To identify a possible hormonal signal that influences alcohol intake in a CB₁R-dependent manner, we measured the plasma levels of several appetitive and stress-related hormones in mice drinking either alcohol or water. Plasma levels of leptin, ACTH, corticosterone and neuropeptide Y were unaffected by alcohol drinking. However, total ghrelin plasma levels were significantly higher in alcohol-drinking than in control mice, in agreement with findings in a large, population-based study in human drinkers (Wittekind et al., 2018), and this increase was attenuated by JD5037 treatment (Table S1). Since ghrelin is known for promoting alcohol drinking (Jerlhag et al., 2009; Lee et al., 2018; Leggio et al., 2014; Zallar et al., 2018), we explored its role as a possible link between peripheral endocannabinoids and brain structures involved in the control of alcohol drinking. We analyzed alcohol preference and its modulation by peripheral CB₁R blockade in mice lacking ghrelin peptide or its receptor, which is the growth hormone secretagogue receptor-1A (GHS-R_1A). Both Ghrl^−/− and Ghsr^−/− mice had significantly lower alcohol preference and intake than their respective wild-type littermates, and daily treatment with JD5037 significantly reduced ethanol preference and intake in controls but not in the two knockout strains (Figure 1C–D). Food intake remained unaffected by JD5037 treatment in both wild-type and knockout strains (Figure S2).

These findings suggest that endocannabinoids acting via peripheral CB₁R promote alcohol drinking by inducing ghrelin production, its post-translational modification and/or its release from ghrelin-producing cells. Any of these mechanisms are predicated on the expression of CB₁R by ghrelin-producing cells. We used RNAscope ®, an RNA in situ hybridization technology, to confirm earlier findings of low Cnr1 expression in ghrelin-producing cells of the stomach mucosa (Engelstoft et al., 2013) (Figure S3A). Compared to other tissues, the mouse stomach contains very high levels of endocannabinoids, particularly anandamide, which are further increased in ethanol-drinking animals, whereas no such increase in EC levels was evident in the hypothalamus or plasma (Figure S3B). Endocannabinoids and their degrading enzymes are also present in MGN3–1 mouse gastric ghrelinoma cells (Figure S3C). These cells were selected for our study because they express CB₁R (Koyama et al., 2016) similar to primary gastric mucosal cells (Engelstoft et al., 2013). Additionally, they are capable of incorporating and metabolizing long-chain fatty acids to generate octanoic acid for ghrelin acylation, unlike some other ghrelin-producing cell lines that require octanoic acid supplementation to produce octanoyl-ghrelin (Bando et al., 2016).

CB₁R Blockade Inhibits the Octanoylation of Ghrelin

To become biologically active, ghrelin is octanoylated by ghrelin O-acyl transferase (GOAT) (Yang et al., 2008). Alcohol-drinking wild-type mice exhibited elevated plasma levels of both octanoyl-ghrelin and its inactive precursor desacyl-ghrelin. JD5037 treatment reversed the increase in octanoyl-ghrelin without affecting desacyl-ghrelin levels (Figure 2A). The selectivity of this effect makes it unlikely that it reflects a change in ghrelin release, in agreement with the lack of effect of CB₁R agonism on ghrelin secretion from ghrelin-producing primary stomach cells (Engelstoft et al., 2013). As to the alcohol-induced increase in total plasma ghrelin, alcohol intake is known to increase sympathetic nerve activity (Hering et al., 2011), and sympathetic stimulation increases ghrelin secretion via the β₁-adrenergic receptor (Zhao et al., 2010). Accordingly, treatment of alcohol-drinking mice with the β-adrenergic antagonist propranolol reduced alcohol preference and intake, in agreement with earlier observations (Gilpin and Koob, 2010), and decreased the plasma levels of both desacyl- and octanoyl-ghrelin (Figure S4).

Figure 2. Effect of Peripheral CB₁R Blockade on Plasma Acyl- and Desacyl-ghrelin in Alcohol Drinking Wild-type and Cnr1^−/− Mice and on Alcohol Drinking Behavior in Mboat^−/− Mice (see also^{Figure S3}).

(A) Plasma hormone levels measured by ELISA in control (no ethanol) and vehicle- or JD5037-treated alcohol-drinking wild-type mice and their similarly treated Cnr1^−/− littermates at the end of the 5-day treatment period. Significant difference from corresponding values in ‘no ethanol’ group (*) or between vehicle- and JD5037-treated alcohol group (#), with P values as indicated in the legend for Figure 1 and n values shown in brackets. (B) Mboat^−/− mice and their wild-type littermates were subjected to the two-bottle paradigm to monitor ethanol preference and intake in the absence or presence of JD5037 treatment (3 mg/kg/day) as described in the legend of Figure 1.

Peripheral CB₁R blockade did not affect octanoyl-ghrelin levels in Cnr1^−/− mice, confirming the distinct role of CB₁R in this effect (Figure 2A). Given the abundance of endocannabinoids in stomach tissue and their increase in alcohol drinking animals, decreases in octanoyl-ghrelin by JD5037 likely reflect the reversal of a tonic increase in ghrelin acylation by endocannabinoids. Selective targeting of ghrelin acylation by endocannabinoids via CB₁R is also supported by findings in mice lacking Mboat4, the gene encoding GOAT. Mboat4^−/− mice and their wild-type littermates had similar body weight and food intake. Plasma levels of acyl-ghrelin were 145.3 ±27.6 pg/mL (n = 11) in wild-type mice but undetectable (<10 pg/mL, n = 9) in Mboat4^−/− mice whereas desacyl-ghrelin levels were similar in the two strains (1.64 ± 0.43 ng/mL in wt vs 2.56 ± 0.67 ng/mL in ko, n.s.). Mboat4^−/− mice had significantly lower ethanol preference and intake than their wild-type littermates, and JD5037 treatment did not reduce alcohol preference and intake in these knockout animals as it did in wild-type counterparts (Figure 2B).

Selective regulation of octanoyl-ghrelin by CB₁R may result from altered gene expression or enzyme activity of GOAT, or changes in the availability of the GOAT substrate octanoyl-carnitine. Mboat4 expression was moderately reduced by 16 ± 5% (n = 16) in the stomach of JD5037-versus vehicle-treated, alcohol-drinking mice. Previously, however, a much greater, ~50% reduction in Mboat4 expression was reported to have no effect on plasma octanoyl-ghrelin levels (Kirchner et al., 2009). Therefore, it is unlikely that the modest decrease produced by JD5037 is responsible for the parallel decrease in octanoyl-ghrelin. CB₁R blockade is also unlikely to inhibit GOAT enzyme activity directly: JD5037 concentrations resulting in maximal CB₁R occupancy significantly reduced octanoyl-ghrelin levels in MGN3–1 cells incubated with 10 μM but not 50 or 250 μM octanoic acid (Figure S5A), whereas a synthetic GOAT inhibitor potently reduced octanoyl-ghrelin, which could not be overcome by increasing the substrate concentration to 250 μM (Figure S5B). Thus, inhibition of ghrelin acylation by JD5037 may be due to the depletion of octanoic acid, and this inhibition can be overcome with excess octanoic acid substrate. This mechanism is plausible as CB₁R blockade is known to increase energy expenditure by increasing fatty acid β-oxidation (FAO), as documented in white and brown fat tissue (Jbilo et al., 2005; Perwitz et al., 2010), liver (Tam et al., 2010), kidney (Udi et al., 2017) and sperm cells (Aquila et al., 2010). We sought to test whether a similar mechanism operates in ghrelin-producing stomach cells.

CB₁R Blockade Inhibits Ghrelin Acylation by Increasing the Oxidative Degradation of the Substrate Octanoyl-Carnitine

Mammals do not generate medium-chain fatty acids via lipogenesis, so octanoic acid could either be obtained through diet or derived metabolically from long-chain fatty acids from which it is generated by β-oxidation. The standard mouse chow used in our experiments does not contain octanoic acid, which was also undetectable in mouse plasma (not shown), suggesting that octanoic acid must be generated by uptake of long-chain fatty acids and subsequent β-oxidation in ghrelin producing cells. To test this, we incubated MGN3–1 cells with deuterated palmitic acid in the presence of 100 nM JD5037 or vehicle and followed its metabolic fate. The cellular level of [²H₃₁]palmitic acid was similar in the two groups, indicating that fatty acid uptake is unaffected by CB₁R blockade (Figure 3A). On the other hand, the conversion of palmitate to palmitoyl-carnitine, catalyzed by carnitine-palmitoyl transferase-1 (CPT1), the rate-limiting enzyme in fatty acid β-oxidation, was significantly increased in the presence of JD5037. The resulting increase in palmitoyl-carnitine:palmitic acid ratio indicates increased CPT-1 activity (Figure 3A). Furthermore, there was a progressive reduction of the ratios of [²H₂₆]myristoyl-carnitine, [²H]lauroyl-carnitine and [²22 H₁₄]octanoyl-carnitine in JD5037-treated vs. vehicle-treated cells, resulting in a ~50% decrease in the amount of [²H₁₄]octanoyl-carnitine generated from [²H₃₁]palmitoyl-carnitine in response to JD5037 treatment (Figure 3A). That this reduction was due to increased oxidative degradation rather than decreased production of octanoyl-carnitine was indicated by three lines of evidence.

Figure 3. Effects of CB₁R Blockade on Fatty Acid Uptake and Processing by MGN3–1 Stomach Ghrelinoma Cells (see also Figure S5 and Table S2).

(A) MGN3–1 cells seeded in 6-well plates at 2 × 10⁶ cells/plate were allowed to adhere for 2 days before changing the medium to BSA-free, low glucose (1.5 g/L) DMEM. Cells were then incubated with 100 nM JD5037 or its vehicle for 2 h followed by the addition of 50 M of [²H₃₁]palmitic acid for an additional 6 h before measuring the cellular uptake of [²H₃₁]palmitic acid and the generated labeled acyl-carnitine species by LC-MS/MS. (B) Effect of JD5037 on the conversion of octanoyl-carnitine to hexanoyl-carnitine in MGN3–1 cells incubated with 50 μM of either [²H₃₁]palmitic acid or [¹³C₄]octanoic acid and on the generation of acyl-ghrelin.

First, in MGN3–1 cells preincubated with either [²H]-labeled palmitic acid or [¹³C]-labeled octanoic acid, 100 nM JD5037 caused similar, significant reductions in the amount of labeled octanoyl-carnitine generated, whereas the amount of its immediate degradation product hexanoyl-carnitine was less reduced. The corresponding increase in hexanoyl:octanoyl ratio indicates increased metabolic degradation of octanoic acid, resulting in a decrease in ghrelin acylation (Figure 3B). Second, incubation of MGN3–1 cells with 10–100 nM JD5037 robustly increased the rate of oleic acid oxidation, as measured by the activity of a fluorescent extracellular O₂ consumption reagent (Figure 4A), and the effect of JD5037 was prevented in MGN3–1 cells with shRNA-mediated knockdown of Cnr1 (Figure 4B). In parallel experiments, exposure of MGN3–1 cells to JD5037 resulted in significant increases in the gene expression of two enzymes involved in the oxidation of long chain fatty acids, acyl-CoA dehydrogenase (Acadvl) and β-ketoacyl CoA hydrolase (Hadhb), which may have contributed to the increased rate of fatty acid oxidation (Figure S6). Third, treatment of MGN3–1 cells with selective inhibitors of three different enzymes in the fatty acid β-oxidation pathway, including oxfenicine for CPT1, triacsin-C for long chain fatty acylCoA synthase and 4-bromocrotonic acid for β-ketoacyl CoA thiolase, reduced acyl-ghrelin production and occluded the similar effect of JD5037 (Figure 5).

Figure 4. Effect of CB₁R Blockade on Fatty Acid β-oxidation and Ghrelin Production in MGN3–1 Cells. (see also Figures S3, S5, S6).

(A) Time course and rate of increase in oleate-driven respiratory activity in MGN3–1 cells. Cells were treated with 10 or 100 nM JD5037, 2.5 M of FCCP (positive control) or vehicle in the presence of the fluorescent extracellular O₂ consumption reagent. Oxidative respiration was recorded every 90 sec and the slope of the initial linear increase was calculated. Points and bars are means ± SEM from n = 11–15 experiments, as indicated. *P < 0.05 compared to vehicle. (B) Verification of Cnr1 knockdown by rtPCR in cells used in panel C. Cellular uptake of the constructs was verified by fluorescent microscopy (20x magnification) and the degree of knockdown was determined by rt-PCR, *P < 0.05, n = 4. (C) Oleate-driven oxidative activity in MGN3–1 cells with shRNA-mediated knockdown of Cnr1 expression and their mock-transfected controls, n = 3–8 experiments, as indicated, *P < 0.05.

Figure 5. Effects of Inhibitors of Fatty Acid Oxidation on Acyl- and Desacyl-ghrelin Production and its Inhibition by JD5037. (see also Figure S3).

Oxfenicine (1 mM), or 4-bromo-crotonic acid (50 M) were present in the medium for 2 h whereas triacsin C (5 μM) was present for 24 h before the experiments. The enzyme targets of each inhibitor are illustrated by the FAO scheme on the right (dashed red lines). Cell medium was collected for measurements of acyl- and desacyl-ghrelin. Columns and bars are means ± SEM, number of experiments indicated in brackets. *P < 0.05, **P < 0.01 compared to vehicle (ANOVA followed by Tukey’s multiple comparison test.

Ghrelin Signals via Vagal Afferents to Promote Alcohol Drinking Behavior

GHS-R_1A is expressed in brain regions associated with the control of food intake (Zigman et al., 2006) where it may be activated by blood-borne acyl-ghrelin. However, it is uncertain whether this occurs in mice, which lack a functional blood-to-brain transport mechanism for acyl-ghrelin (Banks et al., 2002). Alternatively, acyl-ghrelin released from ghrelin-producing cells of the stomach may act on GHS-R_1A expressed on vagal afferents (Burdyga et al., 2006; Date et al., 2002), which are in close apposition to ghrelin-producing neuroendocrine cells in the stomach mucosa (Gautron et al., 2011), with which they may form true synapses (Kaelberer et al., 2018). We have explored this latter possibility using mice with selective blockade of vagal afferent signaling by neonatal capsaicin treatment. These mice displayed permanent sensory afferent denervation verified by the complete loss of cholecystokinin (CCK)-induced anorexia (Figure 6A). In control (CCK+) mice with intact vagal innervation, treatment with either JD5037 or rimonabant significantly reduced alcohol preference and intake. In contrast, capsaicin-pretreated CCK-unresponsive (CCK-) mice displayed a statistically significant increase in both alcohol preference (pooled data during vehicle-treatment in CCK+: 69.8 ± 2.2%, n = 64, vs CCK-mice: 84.7 ± 1.8%, n = 61, P < 0.0001) and intake (12.4 ± 0.4 vs 16.2 ± 0.4 mg/g/day, P < 0.0001), which was unaffected by either JD5037 or rimonabant treatment (Figure 6B).

Figure 6. Effects of Blockade of Peripheral CB₁R or GHS-R_1A on Alcohol Intake and Preference in Control and Capsaicin-treated, Cholecystokinin (CCK)-Unresponsive Mice. (see also Figure S2).

(A) Effect of neonatal capsaicin or vehicle treatment on CCK-induced suppression food intake, measured as described in Methods. *P < 0.05, ***P < 0.001 indicate significant difference between pre- and post-CCK levels of intake of chocolate milk. (B) Effect of vehicle or CB₁R blockade by JD5037 (3 mg/kg/day) or rimonabant (10 mg/kg/day) treatment on alcohol preference and intake in CCK-responsive control and capsaicin-treated, CCK-unresponsive mice. Symbols and statistics as described in the Legend of Figure 1A. (C) Effect of the GHS-R_1A antagonist JMV2959 (JMV, 30 mg/kg/day) or its vehicle on alcohol preference and intake in CCK-responsive control and capsaicin-treated, CCK-unresponsive mice. (D) Effect of the peripherally restricted GHS-R_1A antagonist PF-5190457 (3 mg/kg/day) on alcohol preference and intake in control and capsaicin-treated CCK-unresponsive mice.

Similar to earlier findings (Gomez et al., 2015), the GHS-R_1A antagonist JMV2959 reduced alcohol drinking in control mice; however, it lost its effectiveness in CCK unresponsive mice (Figure 6C). Furthermore, the peripherally restricted GHS-R_1A antagonist PF-5190457 (Bhattacharya et al., 2014) also significantly reduced alcohol intake and preference in control but not in capsaicin-treated CCK-unresponsive mice (Figure 6D). Like JD5037, neither of these GHS-R_1A antagonists affected food intake (Figure S2).

Because neonatal capsaicin treatment can also affect sensory neurons outside of vagal afferents, we tested another group of mice with subdiaphragmatic surgical vagotomy followed by gastric banding to prevent the post-denervation dilation of the stomach. Although these mice lose both afferent and efferent vagal innervation to the stomach, sensory neurons outside vagal afferents remain intact. These mice also became unresponsive to CCK and, similar to capsaicin-treated mice (Figure 7A) displayed an increase in ethanol preference (sham: 70.0 ± 3.3%, n = 15 vs vagotomy: 91.1 ± 3.1%, n = 10, P = 0.0002) and intake (11.5 ± 0.7 vs 13.4 ± 0.7 mg/g/day, P = 0.08), which was no longer affected by CB₁R antagonists (Figure 7B). Together, these findings indicate that intact vagal afferent activity is required for modulation of alcohol drinking behavior by CB₁R signaling.

Figure 7. Effects of Subdiaphragmatic Vagotomy on the JD5037-Induced Decrease in Ethanol Intake and Preference in Control and CCK-Unresponsive Vagotomized Mice.

(A) Effect of sham operation and subdiaphragmatic vagotomy with pyloroplasty on CCK-induced suppression food intake, measured as described in Methods. ***P < 0.001 indicates significant difference between pre- and post-cholecystokinin-8 (CCK) levels of intake of chocolate milk. (B) Effect of pharmacological inhibition of CB₁R with 3 mg/kg JD5037 (JD) or its vehicle (V) on alcohol consumption in CCK-responsive sham-operated controls and in vagotomized, CCK-unresponsive mice. Symbols and statistics as described in the Legend of Figure 1D.

DISCUSSION

Endocannabinoids acting via CB₁R are integral components of the central neural circuits that control the rewarding effects of both food and drugs of abuse, including alcohol. CB₁R at such central sites used to be considered the target for both the hypophagic and anhedonic effects of CB₁R antagonists, but the brain-penetrance of globally acting CB₁R antagonists led to the whole-sale abandonment of this class of compounds as potential therapeutic agents. The present findings represent an important advance on three accounts. First, they demonstrate that addictive behavior such as alcohol preference and drinking can be disrupted via selective blockade of peripheral CB₁R in a manner that depends on afferent vagal signaling. Second, they identify ghrelin/GHS-R_1A signaling as the downstream target through which CB₁R affects drinking behavior. Third, the mechanism of the CB₁R/ghrelin interaction in ghrelin-producing stomach cells is unusual: CB₁R blockade inhibits the posttranslational activation of ghrelin by its GOAT-catalyzed octanoylation via depleting the substrate octanoyl-carnitine due to its increased oxidative degradation. To our knowledge, this is the first evidence for a GPCR-mediated selective regulation of ghrelin octanoylation.

Although the role of the gut/brain axis in the control of addictive behavior may strike one as surprising, recent evidence implicates the gut-brain axis not only in appetitive functions but also in the control of emotional and motivational states (Breit et al., 2018). In particular, vagal afferents originating in the right nodose ganglion were found to activate a parabrachio-nigral pathway mediating reward-related behaviors, whereas a different subset is likely involved in satiety and aversive behaviors (Han et al., 2018). The present observations are compatible with a scheme whereby an alcohol-induced, CB₁R-dependent release of acyl-ghrelin from ghrelin-producing cells in the stomach mucosa activates GHS-R_1A on vagal afferent terminals, triggering a signal to the brain to increase the drive to drink alcohol. The loss of the inhibitory effect of peripherally restricted antagonists of either CB₁R (JD5037) or GHS-R_1A (PF-5190457) on alcohol-drinking in mice with vagal deafferentation strongly supports the above scheme. Nevertheless, an alternative possibility is that vagal deafferentation interrupts a gut-brain-gut pathway that regulates GOAT activity. Studies are in progress to explore the existence of such a regulatory pathway.

Furthermore, a similar loss of the effect of the globally acting antagonists of CB₁R (rimonabant) or GHS-R1A (JMV-2959) on alcohol intake in vagal deafferented mice suggests that endocannabinoids acting through ghrelin regulate drinking behavior exclusively via the gut-brain axis; however, they do not negate findings that exogenous ghrelin administered i.c.v. can increase alcohol drinking via GHS-R_1A in the CNS (Jerlhag et al., 2009). It was proposed earlier that peripheral ghrelin is not involved in the control of alcohol drinking behavior based on the finding that neutralization of circulating acyl-ghrelin with a high affinity antibody do not affect alcohol drinking in mice (Jerlhag et al., 2014). However, this does not rule out the peripheral mechanism we propose here, as enteroendocrine cells in the GI tract have been shown to form true synapses with sensory afferent terminals (Kaelberer et al., 2018), so released acyl-ghrelin may directly activate GHS-R_1A on afferent nerve terminals before gaining access to the circulation.

These findings parallel some earlier observations of a similar dependence of ghrelin-induced hyperphagia on intact vagal afferent innervation in rodents (Date et al., 2002; Date et al., 2005; Senin et al., 2013) and humans (le Roux et al., 2005), although others found gut vagal afferents to be redundant for the orexigenic effect of peripherally administered ghrelin (Arnold et al., 2006), which may occur by activation of GHS-R_1A in the medullary dorsal vagal complex (Zigman et al., 2006). Ghrelin activation of GHS-R_1A on vagal afferents was shown to reduce the spontaneous firing rate of these neurons (Date et al., 2002), whereas the satiety hormone CCK increases gastric afferent firing (Date et al., 2002; Schwartz et al., 1997). Such findings led to the proposal that the orexigenic effect of ghrelin may involve suppression of the satiety signal of CCK (Burdyga et al., 2006). CCK also suppresses alcohol intake (Geary et al., 2004), which would be compatible with a similar mechanism for the ghrelin-induced increase in alcohol drinking. Such a mechanism is further supported by the present findings that a peripherally restricted GHS-R_1A antagonist that had been shown to increase vagal afferent nerve firing (Kong et al., 2016) reduces alcohol intake in mice with intact, but not absent, vagal afferents. The apparent inverse relationship between afferent vagal firing rate and the drive to drink alcohol is also compatible with the significant increase in alcohol preference and intake in mice with vagal deafferentation. Disruption of the vagal innervation of the stomach may explain increased alcohol drinking after Roux-en-Y gastric bypass surgery (Hajnal et al., 2012), which was associated with the loss of regulation of neuronal firing in the ventral tegmental area dopamine neurons by a ghrelin receptor antagonist (Sirohi et al., 2017).

The functional link we have uncovered between endocannabinoids and ghrelin may not be unexpected in view of the fact that both promote alcohol drinking via activation of CB₁R (Hungund et al., 2003; Wang et al., 2003) or GHS-R_1A (Lee et al., 2018; Leggio et al., 2014; Zallar et al., 2018), respectively, and there are also examples of intriguing similarities and/or interactions between endocannabinoids and ghrelin in the control of food intake (Alen et al., 2013; Kola et al., 2008) and peripheral energy metabolism (Kola et al., 2005; Lim et al., 2013). The mechanism of this interaction, however, is quite unique in ghrelin-producing stomach cells in which endocannabinoids acting via CB₁R regulate the availability of the fatty acid substrate for ghrelin acylation. It is well established that CB₁R agonists inhibit whereas antagonists promote fatty acid oxidation in various tissues, the latter being the primary mechanism of the peripherally mediated catabolic effect of CB₁R blockade. Here we presented several lines of evidence that blockade of CB₁R has a similar effect in ghrelin-producing cells where the increased flux through the fatty acid ß-oxidation pathway depletes the cellular level of octanoic acid available for ghrelin acylation. Our conclusion that CB₁R selectively targets ghrelin acylation is also supported by the loss of the CB₁R-mediated component of ethanol intake in Mboat4^−/− mice which produce desacyl- but not acyl-ghrelin. This also indicates that desacyl-ghrelin, which has been implicated in a number of peripheral effects mediated by non-conventional ghrelin receptors (Callaghan and Furness, 2014), is not involved in the control of alcohol drinking behavior in mice.

CB₁R at peripheral rather than central sites in the gut-brain axis have also been implicated in the hypophagic effects of the brain-penetrant CB₁R antagonist rimonabant (Bellocchio et al., 2013; Gomez del Pulgar et al., 2002), and, in particular, its preferential inhibition of the intake of palatable dietary fats (DiPatrizio et al., 2011). In all these studies, vagal deafferentation abolished the hypophagic effects of rimonabant, but the nature of the vagal afferent signaling affected by CB₁R blockade was not identified. In two other studies in rats on regular chow, overnight fasting increased food intake and plasma ghrelin levels and both were inhibited by rimonabant, suggesting a role for ghrelin in the orexigenic effects of endocannabinoids (Cani et al., 2004; Senin et al., 2013). However, a cause-and-effect relationship between the reduction in food intake and plasma ghrelin by rimonabant has not been unequivocally established in either study, and only total ghrelin was measured, which does not allow a distinction between ghrelin release and acylation as the target of CB₁R blockade. Despite these parallels between the mechanisms involved in the control of food intake and alcohol drinking, it is noteworthy that in the present experiments peripheral CB₁R blockade selectively reduced alcohol intake and preference without having any effect on food intake or the consumption of sucrose or saccharin-containing solutions. The lack of effect of JD5037 on food intake was not surprising, as its hypophagic effect reported in diet-induced obese mice is mediated by endogenous leptin, due to rapid reversal the obesity-induced leptin-resistance (Tam et al., 2012), and this mechanism is not expected to operate in leptin-sensitive lean mice. As for the role of CB₁R in promoting the intake of sweets, the suppression of such intake by rimonabant has been shown to be mediated at central sites (DiPatrizio and Simansky, 2008). The possible involvement of CB₁R on taste buds, which would be accessible to JD5037, is unlikely in view of a report that a CB₁R antagonist did not affect the gustatory nerve response to sucrose in lean mice (Niki et al., 2015). Therefore, the lack of effect of JD5037 on these ingestive behaviors supports the notion that CB₁R blockade affects the addictive/rewarding effects of alcohol rather than its caloric or tastant properties, and thus may have therapeutic potential in the pharmacologic management of alcoholism.

Limitations of Study

Numerous studies in rodent models of alcohol drinking indicate that rimonabant and other brain-penetrant CB₁R antagonists significantly reduce alcohol preference and intake. In two human studies there was a tendency for rimonabant to reduce alcohol relapse or reduce voluntary drinking by heavy drinkers, but these effects did not reach statistical significance (reviewed in (Maccioni et al., 2010)). The doses of rimonabant used in rodent studies (3–10 mg/kg/day) result in full occupancy of CB₁R whereas the dose used in all human studies to date was limited to 20 mg or 0.3 mg/kg/day to minimize the chances of psychiatric side effects, resulting in only 20–30% CB₁R occupancy (Hjorth et al., 2016). This limitation can be circumvented by peripherally restricted CB₁R antagonists that can be used at doses resulting in full receptor occupancy, which remains to be tested in human patients when peripheral CB₁R antagonists, which are in early stages of clinical development, become available for therapeutic use.

Another potential limitation stems from the finding that peripheral CB₁R blockade reduces alcohol preference and intake by limiting the availability of the octanoic acid substrate required for ghrelin acylation in the ghrelin-producing cells, which can be overcome by octanoic acid supplementation (see Figure S5A). Normal human plasma contains very low to negligible levels of octanoic acid (Christinat et al., 2016; Shrestha et al., 2015) as the majority of dietary medium-chain fatty acids is completely metabolized during their first pass through the liver. However, plasma levels of octanoic acid may be high enough to overcome the effect of CB₁R blockade in people consuming special diets reinforced by medium-chain fatty acids or containing high levels of coconut and its oils.

STAR ✯ METHODS

CONTACT FOR REAGENT AND RESOURCE SHARING

Further information and requests for resources and reagents should be directed to and will be fulfilled by the lead contact, George Kunos (george.kunos@nih.gov). Requests for MGN3–1 cells will be referred to the original provider, Professor Hiroshi Iwakura (hiwaku@wakayama-med.ac.jp) who has agreed to make this material available under a separate agreement to other scientists at non-profit organizations.

EXPERIMENTAL MODEL AND SUBJECT DETAILS

Animals

All animal procedures were approved by the Institutional Animal Care and Use Committee of NIAAA, NIH, and the experiments were carried out in accordance with the guidelines. C57BL/6J mice were purchased from the Jackson Laboratory (USA). Cnr1^−/− (Zimmer et al., 1999), Ghrl^−/− (Sun et al., 2003) and Ghsr^−/− mice (Sun et al., 2004) were generated as described and were propagated by heterozygote breeding, using corresponding wild-type littermates as controls. All 3 knockout strains had been backcrossed at least 8 times to a C57BL/6J background.

Mboat4^−/− mice were generated using CRISPR/Cas9-mediated genome engineering by Cyagen US Inc. (USA) on a contract with our laboratory (contract number: NTMCK-1700809-AJW-01), with exon 1 to exon 2 of the mouse Mboat4 gene (GenBank accession number: NM_001126314.2; Ensembl: ENSMUSG00000071113) selected as target sites. Cas9 mRNA and gRNA generated by in vitro transcription were then injected into fertilized C57BL/6J eggs for knockout mouse production. The founders were genotyped by PCR followed by DNA sequencing analysis, confirming a 6334 bp deletion in the knockouts. The positive founders were bred to the next generation (F1) and subsequently genotyped by PCR and DNA sequencing analysis. Three potential off-target sites have been identified by PCR and sequence analysis, and no mutation was detected.

Mice were housed 2–4 animals per cage in air-conditioned racks maintained at 22.0 ± 0.5°C and 50.0 ± 1.0% humidity, under a 12:12 light/dark cycle. They had ad libitum access to food (NIH-31 standard rodent chow) and water. Absence of known pathogenic organisms was verified by regular testing in sentinel mice. Only male mice were used in drinking paradigms to avoid the confounding effect of the oestrus cycle on the orexigenic effect of ghrelin (Clegg et al., 2007; Sakurazawa et al., 2013). Mice used in drinking paradigms were 8 weeks old and weighing 20–25 g at the onset of the study. Since breeding patterns were not synchronous, a small variation of age (± 1 week) was allowed to obtain a cohort of at least 8–10 mutant and wild-type mice per treatment group.

Cell lines

Stomach-derived MGN3–1 ghrelinoma cell line was derived from a male mouse (Iwakura et al., 2010). Cells were maintained in culture in high glucose (4.5 g/L) DMEM, containing 10% heat inactivated fetal bovine serum and 100 U/mL penicillin and 100 g/mL streptomycin sulfate at 37°C under a humidified atmosphere of 5% CO 2 and 5% O₂, as described in the original publications. Cells subcultured using Trypsin (0.25%) EDTA underwent no more than 6 passages during the course of the experiments. They exhibited high expression of Mboat4 and Ghrl genes and had the ability to produce acyl-ghrelin and desacyl-ghrelin. In most experiments, cells were subcultured onto 6 well plates at a density of 1.5×10⁶ cells/2 mL, or in 12 well plates at a density of 0.67×10⁶ cells/0.9 mL and allowed to adhere for 2 days. For gene expression and endocannabinoid measurements, cells were seeded onto 100-mm dishes at a density of 4.5×10⁶ cells/10 mL or 10⁷ cells/10 mL, respectively. On day 3, cells were washed 3 times with a sterile phosphate buffered saline and culture medium was replaced with treatment medium: BSA-free, low glucose (1.5 g/L) DMEM, containing 100 U/mL penicillin and 100 μg/mL streptomycin sulfate. Carnitine palmitoyltransferase 1 (CPT1) inhibitor oxfenicin 1 mM, mitochondrial 3-ketoacyl-CoA thiolase and acetoacetyl-CoA thiolase inhibitor 4-bromocrotonic acid 50 μM, CB₁ antagonist JD5037 10–100 nM, ghrelin O-acyltransferase (GOAT) inhibitor MRI-2443 1 nM-1 μM or vehicle was added to the medium for 2 hours, followed by albumin solution of [²H₃₁]palmitic acid 50 μM or [¹³C₄]octanoic acid 10 μM, 50 μM or 250 μM or their unlabeled counterparts for another 6 hours. The long-chain fatty-acyl CoA synthase inhibitor triacsin C (5 μM) or its vehicle was present in the cell medium for 24 h before the addition of the labeled fatty acid. For endocannabinoid measurement, cells were incubated with fatty acid amide hydrolase/monoacylglycerol lipase inhibitor JZL195 (1 μM) (Long et al., 2009) or its vehicle for 6 hours.

At the end of treatment, cell culture plates were placed on ice. Media were collected, centrifuged at 953 × g for 3 min at 4°C and acidifi ed with 1N HCl to a final concentration of 0.1 M for measurement of ghrelin. The acidification step was omitted in samples intended for endocannabinoid measurement. Cells were washed 3 times with ice-cold PBS and immediately frozen. Samples were kept at −80°C until assay.

METHOD DETAILS

Two bottle alcohol preference test

The test was performed as described before (Wang et al., 2003) with modifications. Animals were individually housed and acclimated to the paradigm for 5–7 days by having access to two identical water bottles and handled daily to minimize the stress associated with drug testing. Animals were first subjected to a gradual increase in ethanol concentration in drinking bottle (3%, 6%, 9%, 12%, 15%), while the other bottle contained water. The position of bottles was changed every day and alcohol and water bottles will be replaced every 4 days. Once alcohol concentration reached 15%, animals remained on the paradigm for 10 days. Starting on day 2, mice received vehicle by oral gavage for 4 days, followed by a daily treatment with the global or peripheral CB1 antagonist (rimonabant 10 mg/kg or JD5037 3 mg/kg, respectively) or vehicle (control group) for another 5 days, one hour before the dark period. The nonselective beta blocker propranolol (5 mg/kg), the GHS-R1A antagonist JMV2959 (30 mg/kg), the peripheral GHS-R1A antagonist PF-5190457 (3 mg/kg) or vehicle was administered intraperitoneally. All animals were sacrificed for blood and tissue collection 12–16 hours after treatment.

Two bottle taste preference test

Animals were individually housed, acclimated and handled the same way as described in the ‘Two bottle alcohol preference test’ subsection. Following acclimation, mice were given access to 2 bottles, one with water and the other containing 0.3% or 3% sucrose, 0.01% or 0.2% saccharin (w/v), or 2 mg/dL quinine in water, for 10 days. The position of bottles was changed every day and tastant and water bottles will be replaced every 4 days. Starting on day 2, animals were gavaged with vehicle for 4 days, followed by daily treatment with JD5037 3 mg/kg or were continued with vehicle (control group) for another 5 days.

Drinking in the dark

The procedure was performed as described (Rhodes et al., 2005), with modifications. Mice were individually housed and acclimated to the room for 5–7 days before testing and randomly assigned to treatment groups. Starting 3 hours into the dark cycle, water bottles were replaced with 20% ethanol in 25 × 100 mm glass tubes fitted with metal sippers, Access to the ethanol solution was limited to 4 hours every day. One hour before the dark period on day 4, animals were gavaged with JD5037 3 mg/kg or vehicle, and the alcohol session was repeated. Blood samples were collected from mandibular vein immediately after the drinking session. Alcohol concentration was measured in serum using a sample analyzer (Model GM7 Micro-Stat, Analox Instruments, USA).

Neonatal capsaicin treatment

The procedure was originally developed for rats (Jancso et al., 1977) and was subsequently adopted for mice (Shimizu et al., 1984). Neonatal (2-day old) mice were subjected to hypothermia on crushed ice and water to alleviate the neurotoxin-associated pain and distress. Once a lack of motion and/or blanching of the tips of the toes was noted, the pups were transferred to a cooling platform for the drug injection. A single dose 50 mg/kg of capsaicin or its vehicle was injected subcutaneously. Each animal received the same volume (20 μL) based on an average weight of all the pups. The drug was injected into the skin fold on either flank/side of the animal by insulin syringe equipped with the 30-gauge needle under the microscope to ensure the subcutaneous placement of the needle without damaging the abdominal organs. Pups remained on ice for up to 15 minutes to provide post-injection analgesia, then they were transferred to a warm water bath (32–35°C) for 20–30 minutes. Once pups regained consciousness and became active, they were returned to the nest in the cage on the warm water blanket.

Subdiaphragmatic vagal denervation

The procedure was performed under isofluorane anesthesia using sterile instruments as described (Erlanson-Albertsson and Lindqvist, 2008). Briefly, the abdomen was opened to expose the stomach and esophagus. Both vagal trunks were ligated and sectioned immediately below the diaphragm. Pyloroplasty was routinely performed in conjunction with vagal denervation in order to prevent food retention and fatal gastric dilatation. This involved a longitudinal incision across the pyloric sphincter that was closed transversely. After ligation, the intestine was returned to the abdomen, and the surgical site closed using 2 layers of vicryl suture. Sham-operation consisted of a mid-line abdominal incision and gentle manipulation of the stomach without disturbing its innervation. The completeness of vagal denervation was verified by the cholecystokinin test 10–14 days after the surgery.

Cholecystokinin test

In order to check the efficiency of capsaicin-induced deafferentation and subdiaphragmatic denervation, mice were subjected to the cholecystokinin test, which was performed as described elsewhere (Kopin et al., 1999). Briefly, mice were maintained on a 60:40 (vol/vol) solution of chocolate-flavored Ensure (Abbott Nutrition Abbott Laboratories, USA) diluted in water. The liquid diet was presented for 60 min in the morning (08:30–9:30 AM) and in the afternoon (4:30–5:30 PM) using 50-mL plastic centrifuge Falcon tubes (Corning Life Sciences, USA), adapted with leak-proof stainless-steel sippers. Water was freely available before and after the feeding periods. Once baseline chocolate intakes were stable for a period of 4 days, the cholecystokinin test was initiated after overnight fasting. Mice were injected intraperitoneally (i.p.) with 30 μg/kg sulfated cholecystokinin octapeptide (CCK-8) or its vehicle (saline) in a volume of 5 mL/kg body weight. Five min later chocolate-flavored Ensure was presented for 15 min. Injections of CCK-8 or saline into the same animals were repeated in a reversed order 4–5 days apart. Animals were habituated to the procedure with saline injections for 2 days before test.

Ethanol clearance assays

Mice were gavaged with JD5037 3 mg/kg or its vehicle. One hour later, animals were injected i.p. with ethanol (4 g/kg; 15% by volume) in saline. Blood was collected from mandibular vein 1 and 3 hours later into K₂EDTA microtainer tubes (Becton, Dickinson and Company, USA) and centrifuged at 3500 x rpm for 10 min. Plasma ethanol levels were measured using the colorimetric ethanol assay kit (Abcam, USA) in accordance with the manufacturers’ instructions.

Blood chemistry

Mouse plasma acylated ghrelin and unacylated ghrelin, leptin, corticosterone, ACTH and neuropeptide Y were determined using commercial sandwich ELISA assays (listed in Key Resources Table) in accordance with the manufacturers’ instructions and measured with the Spectra MAX 190 microplate reader (Molecular Devices, USA).

KEY RESOURCES TABLE

REAGENT or RESOURCE	SOURCE	IDENTIFIER
Antibodies
N/A
Bacterial and Virus Strains
mouse CNR1 shRNA silencing Adenovirus (Ad-m-Cnr1-shRNA)	Vector Biolabs	shADV-255795
Scramble shRNA with GFP Adenovirus (Ad-scramble-shRNA)	Vector Biolabs	1122
Biological Samples
N/A
Chemicals, Peptides, and Recombinant Proteins
DMEM high glucose, pyruvate	ThermoFisher Scientific	11995065
DMEM low glucose, pyruvate	ThermoFisher Scientific	11885084
DMEM, no glucose, no glutamine, no phenol red	ThermoFisher Scientific	A1443001
Fetal Bovine Serum (FBS), certified, heat inactivated	ThermoFisher Scientific	10082147
Penicillin-Streptomycin (5,000 U/mL), 100×	ThermoFisher Scientific	15070063
L-Glutamine(200 mM)	ThermoFisher Scientific	25030081
Trypsin (0.25%) EDTA	ThermoFisher Scientific	25200056
PBS (phosphate buffered saline)	ThermoFisher Scientific	10010
Oxfenicine (4-Hydroxy-D-phenylglycine)	MilliporeSigma	215333
Triacsin C from Streptomyces sp.	MilliporeSigma	T4540
4-bromocrotonic acid	Santa Cruz Biotechnology	sc-486695
MRI-2443	(Martinez-Grau, 2015)	N/A
JD5037	(Chorvat et al., 2012)	N/A
Rimonabant (SR141716)	NIDA Drug Supply Proqram	NOCD-082
JMV2959	MilliporeSigma	345888
PF-5190457	MilliporeSigma	PZ0270
(±) Propranolol hydrochloride	MilliporeSigma	P0884
[2H31] Palmitic acid	Cambridge Isotope Laboratories, Inc.	DLM-215–1
1,2,3,4-[13C4] Octanoic acid	Cambridge Isotope Laboratories, Inc.	CLM-2721–0.25
JZL 195	Cayman CHEMICAL	13668
Hydrochloric acid (HCI)	Avantor - Macron Fine Chemicals	2612–14
Sucrose	MilliporeSigma	S7903
Saccharin	MilliporeSigma	109185
Quinine hydrochloride dihydrate	MilliporeSigma	Q1125
Capsaicin	MilliporeSigma	M2028
Ensure	Abbott Nutrition Abbott Laboratories	57231
CCK octapeptide (CCK-8)	Tocris Bioscience	1166
p-hydroxymercuribenzoic acid (PHMB)	MilliporeSigma	55540
Aprotinin from bovine lung	MilliporeSigma	A1153
Dimethylsulfoxide (DMSO)	ATCC	4-X
QIAzol Lysis Reagent	QIAGEN	79306
DNase I DNase I, Amplification Grade	ThermoFisher Scientific	18068015
RNeasy Lipid Tissue Mini Kit	Qiagen	74804
iScript™ cDNA synthesiskit	BIO-RAD	9207996
Fast SYBR™ Green Master Mix	Applied Biosystems	4385617
Ethanol (ethyl alcohol, U.S.P. 200 proof, anhydrous)	The Warner Graham Company; product obtained through NIH Supply Center	6505001050000
UltraPureTM Agarose	ThermoFisher Scientific	16500500
Formalin solution, neutral buffered, 10%	MilliporeSigma	HT501128
RNAscope® LS Multiplex TSA Buffer	Advanced Cell Diagnostics	322809
GreenGlo Safe DNA Dye	Denville Scientific, Inc.	CA3600
TAE Buffer (Tris-acetate-EDTA) (50X)	ThermoFisher Scientific	B49
DirectPCR Lysis Reagent (Mouse Tail) 100ml	Viagen Biotech	102-T
Proteinase K Powder (≥30 U/mg)	ThermoFisher Scientific	17916
dNTP Mix(10 mM)	ThermoFisher Scientific	18427013
Taq DNA Polymerase, recombinant	ThermoFisher Scientific	10342020
Tween® 80	MilliporeSigma	P1754–25ML
0.9% Sodium chloride for injection, USP - 50ml	Hospira Inc., obtained through NIH Division of Veterinary Resources	N/A
Sodium hydroxide solution	MilliporeSigma	72068–100ML
Bovine Serum Albumin (BSA) solution, 30% in saline, fatty acid free, aseptically filled	MilliporeSigma	A9205–50ML
D-(+)-Glucose solution	MilliporeSigma	G8769–100ML
Qctanoic acid	MilliporeSigma	W279900-SAMPLE-K
Palmitic acid	MilliporeSigma	P0500–25G
Palmitoyl-L-carnitine	MilliporeSigma	61251–10MG
Myristoyl-L-carnitine	MilliporeSigma	61367–10MG
Lauroyl-L-carnitine	MilliporeSigma	39953–10MG
Octanoyl-L-Carnitine	MilliporeSigma	50892–10MG
Hexanoyl-L-carnitine	MilliporeSigma	07439–10MG
Anandamide-d4 (N-(2-Hydroxyethyl-1, 1, 2, 2-d4)-5Z, 8Z, 11Z, 14Z-eicosatetraenamide)	Tocris	3717
Chloroform, stabilized with amylene, Optima™, for HPLC and GC, Fisher Chemical	ThermoFisher Scientific	C297–4
Methanol, Optima™ LC/MS Grade, Fisher Chemical	ThermoFisher Scientific	A456–4
Water, Optima™ LC/MS Grade, Fisher Chemical	ThermoFisher Scientific	W6–4
Boc-Ala-OH (Boc-L-Alanine)	MilliporeSigma	15380–100G
Critical Commercial Assays
Ethanol assay kit	Abeam	ab65343
Fatty acid oxidation complete assay kit	Abeam	ab222944
Unacylated Ghrelin (mouse, rat) Express EIA Kit	Cayman CHEMICAL	10008953
Acylated Ghrelin (mouse, rat) Express EIA Kit	Cayman CHEMICAL	10006307
Mouse Leptin ELISA Kit	As One International, Inc.	KTE71186
Mouse/Rat Corticosterone ELISA	Alpco	55-CORMS-E01
Neuropeptide Y (NPY) (Human, Rat, Mouse) - EIA Kit	Phoenix Pharmaceuticals, Inc.	EK-049–03
ACTH (Rat, Mouse)-EIA Kit	Phoenix Pharmaceuticals, Inc.	EK-001–21
RNAscope® Multiplex Fluorescent Kit v2	Advanced Cell Diagnostics	323100
TSA Cy 3, Cy 5, TMR, Fluorescein Evaluation Kit	Perkin Elmer	NEL760001KT
Deposited Data
Experimental Models: Cell Lines
MGN3–1 ghrelinomacell line	(Iwakura et al., 2010)	N/A
Experimental Models: Organisms/Strains
C57BL/6J	The Jackson Laboratory	000664
Cnr1−/−	(Zimmer et al., 1999)	N/A
Ghrl−/−	(Sun et al., 2003)	N/A
Ghsr−/−	(Sun et al., 2004)	N/A
Mboat4−/−	Cyagen US Inc., This paper	N/A
Oligonucleotides
Mm_Acadvl_1_SG QuantiTect Primer Assay (Acadvl)	QIAGEN	QT00253127
Mm_Hadhb_1_SG QuantiTect Primer Assay (Hadhb)	QIAGEN	QT01059968
Mm Mboat4_va.1_SG QuantiTect Primer Assay (Mboat4)	QIAGEN	QT01562232
Mm_MGI:930008_1_SG QuantiTect Primer Assay (Ghrl)	QIAGEN	QT00137536
Mm_Rpl19_2_SG QuantiTect Primer Assay (Rpl19)	QIAGEN	QT01779218
mouse cannabinoid receptor 1 (Cnr1)	GeneCopoeia	MQP091551
RNAscope® Probe-Mm-Cnr1-C3	Advanced Cell Diagnostics	420721-C3
RNAscope® Probe-Mm-Mboat4	Advanced Cell Diagnostics	415311
RNAscope® Probe-Mm-Ghrl-C2	Advanced Cell Diagnostics	41 5301-C2
Mboat4−/− mice wt forward primer ATCCAGAACCTGGTATCCTGTAATA	This paper	N/A
Mboat4−/− mice wt reverse primer GTCAATCCCACTGACTCACAACACAG	This paper	N/A
Mboat4−/− mice mutant forward primer GAGTTCCTTGTCAAATCTCCAAAGCC	This paper	N/A
Mboat4−/− mice mutant reverse primer GTCAATCCCACTGACTCACAACACAG	This paper	N/A
Cnr1−/− wild type/mutant forward primer GTACCATCACCACAGACCTCCTC	(Zimmer et al., 1999)	N/A
Cnr1−/− wild type reverse primer GGATTCAGAATCATGAAGCACTCCA	(Zimmer et al., 1999)	N/A
Cnr1−/− mutant reverse primer AAGAACGAGATCAGCAGCCTCTGTT	(Zimmer et al., 1999)	N/A
Ghrl−/− wt forward primer GCTCTGGATGGACATGGCC	(Sun et al., 2003)	N/A
Ghrl−/− wt reverse primer CTGCGCATGTCTGCTGCTCC	(Sun et al., 2003)	N/A
Ghrl−/− mutant forward primer GAGGCTGAAGTTCAGATGTGCGGCG	(Sun et al., 2003)	N/A
Ghrl−/− mutant reverse primer CCTCGAATCAGCAACGGCTTGCCG	(Sun et al., 2003)	N/A
Ghsr−/− wt forward primer TCTTCGTGGTGGGCATCTCG	(Sun et al., 2004)	N/A
Ghsr−/− wt reverse primer CTTCCTCCCGATGAGACTGT	(Sun et al., 2004)	N/A
Ghsr−/− mutant forward primer AGCGCATCGCCTTCTATCGC	(Sun et al., 2004)	N/A
Ghsr−/− mutant reverse primer GCTCGACTTTGTCCAGGCC	(Sun et al., 2004)	N/A
Recombinant DNA
mouse CNR1 shRNA silencing Adenovirus (Ad-m-Cnr1-shRNA)	Vector Biolabs	shADV-255795
Scramble shRNA with GFP Adenovirus (Ad-scramble-shRNA)	Vector Biolabs	1122
Software and Algorithms
GraphPad Prism 7	GraphPad Software	N/A
MassHunter Workstation Optimizer software	Agilent Technologies	N/A
Other
N/A

For ghrelin measurement, blood was collected from mandibular vein into K₂EDTA microtainer tubes containing 1 mM p-hydroxymercuribenzoic acid (PHMB) and aprotinin 0,6 TIU per 1 mL of blood. The total volume of protease inhibitors was 10 μL/1mL of blood. Blood samples were then mixed by inversion, centrifuged at 3,500 rpm for 10 minutes at 4°C and acidified with 1N HCl to a final concentration of 0.1 M. For all other assays, blood was collected without other stabilizers. Plasma samples were assayed immediately or stored at − 80°C and used within 30 days after collection. For endocannabinoid and ethanol measurements, blood was allowed to clot for 30 min followed by centrifugation at 3,500 rpm for 10 minutes at 4°C. Serum samples were stored at −80°C until analysis.

Fatty acid oxidation assay

Oleate was used as substrate to evaluate fatty acid-driven oxygen consumption in MGN3–1 cells according to the manufacturer’s instructions. MGN3–1 cells were seeded in a Costar 96 well assay plates (black wall with clear flat bottom; Corning Incorporated, USA) at a density of 5 × 10⁵ cells/well in 200 μL culture medium and allowed to adhere to the plate for 24 hours. On the day of experiment cells were treated with JD5037 10 nM and 100 nM or its vehicle (DMSO). The mitochondrial oxidative phosphorylation uncoupler carbonylcyanide 4-(trifluoromethoxy) phenylhydrazone (FCCP, 2.5 μM) was used to achieve maximum respiratory activity. The oxygen consumption rate was measured using an Infinite F500 fluorescent microplate reader (Tecan, USA) (excitation 380 nm and emission 650 nm) and was expressed as the initial rate of increase in the fluorescent signal, as recommended by the manufacturer.

Adenoviral infection

Adenovirus expressing shRNA for silencing of mouse Cnr1 (Ad-m-Cnr1-shRNA) or containing a scrambled shRNA sequence as a control (Adv-scrambled-shRNA) which, under the control of U6 promoter with the GFP co-expressed under a CMV promoter, were purchased from Vector Biolabs (USA). Mouse MGN3–1 ghrelinoma cells were seeded at a density of 3 × 10⁵ / 200 μL per well in a 96-well plate to achieve full confluence for the measurement of fatty assay oxidation. After overnight cell attachment, the original cell culture medium was replaced with Ad-Cnr1-shRNA or Ad-scramble-shRNA virus-containing medium, which included 2 × 10⁷ plaque-forming units (pfu) per 200 L per well. After 24h incubation of MGN3–1 cells with Ad-Cnr1-shRNA or Ad-scramble-shRNA, the culture media were replaced with glucose-deprivation media for an additional 24h before performing fatty acid oxidation assay as described above. Cellular uptake of the constructs was verified by fluorescent microscopy and the degree of knockdown was determined by rt-PCR.

Quantitative real-time PCR

Total RNA was extracted with QIAzol Lysis Reagent and purified by RNeasy Lipid Tissue Mini Kit, followed by DNase I treatment and evaluation of RNA concentration by NanoDrop 2000c spectrophotometer (Thermo Scientific, USA). A fixed amount of 1 μg RNA was reverse-transcribed to cDNA with iScript cDNA synthesis kit. Quantitative real-time PCR was performed by a StepOne Plus real time PCR system (Applied Biosystems, USA) in a sample containing Fast SYBR™ Green Master Mix, 25 ng of single stranded cDNA transcript and QuantiTect primers against mouse acyl-Coenzyme A dehydrogenase, very long chain (Acadvl), against mouse hydroxyacyl-CoA dehydrogenase/3-ketoacyl-CoA thiolase/enoyl-CoA hydratase (trifunctional protein), beta subunit (Hadhb) or qPCR primer against mouse cannabinoid receptor 1 (Cnr1). Normalization was performed with the primer against mouse ribosomal protein L19 (Rpl19) as a housekeeping gene.

Genotyping

Genotyping was performed from tail or ear clippings using a REDExtract-N-Amp™ tissue PCR kit (MilliporeSigma, USA) according to the manufacturer’s protocol. The following forward (F) and reverse (R) primer sequences were used:

Ghs-r_1A:

1F (5’ TCTTCGTGGTGGGCATCTCG 3’) and 1R (5’-CTTCCTCCCGATGAGACTGT-3’) and also F2 (5’ AGCGCATCGCCTTCTATCGC) and R2 (5’-GCTCGACTTTGTCCAGGCC-3’) amplified a 557 bp and 380 bp segments of the wild type and Neo resistance cassette of the mutant mouse, respectively;

Ghrl:

1F (5’-GCTCTGGATGGACATGGCC-3’) and 1R (5’-CTGCGCATGTCTGCTGCTCC-3’) and also 2F (5’-GAGGCTGAAGTTCAGATGTGCGGCG-3’) and 2R (5’-CCTCGAATCAGCAACGGCTTGCCG-3’) amplified a 465 bp and 337 bp segments of the gene in wild type and mutant mouse, respectively;

Cnr1:

F (5’-GTACCATCACCACAGACCTCCTC-3’) and 1R (5’-GGATTCAGAATCATGAAGCACTCCA-3’) and also F (5’ GTACCATCACCACAGACCTCCTC 3’) and 2R (5’-AAGAACGAGATCAGCAGCCTCTGTT-3’) amplified a 350 bp and 150 bp segments of the gene in wild type and mutant mouse, respectively;

Mboat4:

1F (5’-ATCCAGAACCTGGTATCCTGTAATA-3’) and 1R (5’- GTCAATCCCACTGACTCACAACACAG-3’) and also 2F (5’- GAGTTCCTTGTCAAATCTCCAAAGCC-3’) and 2R (5’- GTCAATCCCACTGACTCACAACACAG-3’) amplified a 531 bp and 630 bp segments of the gene in wild type and mutant mouse, respectively.

Products were separated on 3% UltraPure™ Agarose gel in 1xTEA in the presence of GreenGlo Safe DNA Dye and bands were visualized on G:BOX scanner (Syngene, USA).

Fluorescent in situ hybridization.

Pieces of mouse stomach tissues were fixed in 10% neutral buffered formalin (NBF) for 24 hours, embedded in paraffin and sectioned (4  μm) onto glass slides. Tissue sections were processed for fluorescent in situ hybridization (FISH) using the RNAscope® Multiplex Fluorescent Kit v2 according to the manufacturer’s instructions (Wang et al., 2012) (Advanced Cell Diagnostics [ACD], USA). The following mouse RNAscope® FISH probes from ACD were used: Cnr1 in channel 3 (NM_007726.3, region 530 – 1458), Mboat4 in channel 1 (NM_001126314.2, region 435 – 1435), Ghrl in channel 2 (NM_021488.5, region 13 – 550). DAPI was part of the kit and was used to stain for nuclei. For fluorescence detection, sections were analyzed using a Zeiss LSM700 (USA) confocal microscope.

The fluorophores used to visualize the RNAscope® FISH probes were fluorescein, cyanin 3 and cyanin 5 (TSA Cyanin 3 & 5, TMR Fluorescein Evaluation Kit; Perkin Elmer, USA). Fluorophores were diluted in RNAscope® LS Multiplex TSA Buffer (ACD, USA). The fluorophore dilutions for each RNAscope® ISH probe were Cnr1-Fluorescein (1:750), Mboat4-Cyanin 3 (1:750), and Ghrl-Cyanin 5 (1:30000).

Endocannabinoid measurement

Endocannabinoids were extracted and quantified by liquid chromatography–tandem mass spectrometry (LC-MS/MS) as described (Mukhopadhyay et al., 2011). Stomach tissue or MGN- 3 cells were homogenized in 0.5 mL of ice-cold methanol/Tris buffer (50 mM, pH 8.0), 1:1, containing 7 ng of [2H4]arachidonoyl ethanolamide ([²H₄]AEA) as internal standard. Homogenates were extracted three times with CHCl₃:methanol (2:1, vol/vol), dried under nitrogen and reconstituted with methanol after precipitating proteins with ice-cold acetone. Liquid chromatography tandem mass spectrometry (LC-MS/MS) analyses were conducted on an Agilent 6410 triple quadrupole mass spectrometer (Agilent Technologies) coupled to an Agilent 1200 LC system. Analytes were separated using a Zorbax SB-C18 rapid-resolution HT column. Gradient elution mobile phases consisted of 0.1% formic acid in water (phase A) and 0.1% formic acid in methanol (phase B). Gradient elution (250 μL/min) was initiated and held at 10% B for 0.5 min, followed by a linear increase to 85% B at 1 min and maintained until 12.5 min, then increased linearly to 100% B at 13 min and maintained until 14.5 min. The mass spectrometer was set for electrospray ionization operated in positive ion mode. The source parameters were as follows: capillary voltage, 4,000 V; gas temperature, 350°C; drying gas, 10 L/min; nitrogen was used as the nebulizing gas. Collision-induced dissociation was performed using nitrogen. Level of each compound was analyzed by multiple reactions monitoring. The molecular ion and fragment for each compound were measured as follows: m/z 348.3→62.1 for AEA, m/z 352.3→66.1 for [²H₄]AEA and m/z 379.3→91.1 for 2-AG. Analytes were quantified using Mass-Hunter Workstation LC/QQQ Acquisition and MassHunter Workstation Quantitative Analysis software (Agilent Technologies). Levels of AEA and 2-AG in the samples were measured against standard curves. Values are expressed as pmol or nmol/g wet tissue or per number of cultured cells.

Acyl-carnitine measurement

Stable isotope-labelled acyl-carnitines were measured by LC-MS/MS. The sample extraction, chromatography and mass spectrometry conditions were set as described (Cinar et al., 2014). The extraction efficiency and chromatographic validation was performed using unlabeled acyl-carnitine standards obtained from MilliporeSigma (St Louis, MI). The extraction efficiency and chromatographic validation was performed using unlabeled acyl-carnitine standards obtained from MilliporeSigma (St Louis, MI). The labelled acyl-carnitines were generated in vitro in MGN3–1 cells after treatment with either [²H₃₁]palmitic acid or [¹³C₄]octanoic acid as detailed above. The cells were treated with vehicle, unlabeled fatty acids or labelled fatty acids, then cell extracts were injected to the LC-MS/MS under scanning mode to capture the corresponding mass [M+H]⁺ and chromatography of the generated acyl-carnitines. The labelled acyl-carnitines were analyzed by multiple reactions monitoring (MRM). In MRM mode, detection of each compound is based on fragmentation of the precursor ion [M+H]⁺ to yield a prominent product ion. Mass spectrometric conditions were optimized with serial injections of [²H₃₁]palmitic acid or [¹³C₄]octanoic acid-treated cell extracts by using MassHunter Workstation Optimizer software (Agilent Technologies, USA). Fragmentation of the molecular ion and mass spectrometric conditions of [²H₃₁]palmitoyl carnitine, [²H₂₆]myristoyl carnitine, [²H₂₁]lauroyl carnitine, [²H₁₄]octanoyl carnitine, [¹³C₄]octanoyl carnitine, [²H₁₀]hexanoyl carnitine and [¹³C₄]hexanoyl carnitine are listed in supplementary Table 2.

Drugs

JD5037 was synthesized as described (Chorvat et al., 2012). Sources of the combined fatty acid amide hydrolase/monoacylglycerol lipase inhibitor JZL195 (Long et al., 2009), JMV2959, cholecystokinin octapeptide (CCK-8), 4-bromocrotonic acid, rimonabant, [²H₃₁]palmitic acid, octanoic acid-1,2,3,4-¹³C₄ are provided in the Key Resources Table. All other chemicals including the peripheral GOAT inhibitor PF-51904577 were from MilliporeSigma (USA). For animal studies, (±)-propranolol hydrochloride was dissolved in saline and sterile filtered for i.p. injections. Other drugs were dissolved in DMSO:Tween 80:saline (5:2:93). Use of this vehicle is critical for keeping JD5037 in clear solution in order to maintain ~25% oral bioavailability. For in vitro studies, drugs were dissolved in DMSO and applied in a volume of 1 μL per 1 mL of cell medium. Protease inhibitors were prepared as 100x concentrated solutions: aprotynin was dissolved in saline, whereas PHMB was dissolved in saline containing 1.2% (v/v) of 10N NaOH and stored aliquoted for one month at −20°C. 10 μL of protease inhibitors was used per 1 mL of blood. Palmitic acid and octanoic acid were prepared as 250 mM stock solutions in 75% ethanol at 70°C for 10 min. To make fatty acid-BSA conjugat es, the fatty acid stock solution was added dropwise into 10% BSA to make a final stock of 10 mM. The solution was continuously mixed at 50°C for 10–60 min, then brought to room temperatur e and sterile filtered. 50 μM octanoate-BSA and palmitate-BSA complexes were used as a final concentration in most cell experiments.

When evaluating the effects of JF5037 and MRI-2443 at 10 μM or 250 μM octanoate-BSA in cell medium, stock solutions of octanoic acid were adjusted correspondingly. Capsaicin was dissolved in a solution of 10% ethanol, 10% Tween-80 in isotonic saline. The capsaicin/vehicle solutions were prepared aseptically. CCK-8 was dissolved aseptically in saline (Hospira Inc., USA).

GOAT inhibitor MRI-2443: (S)-N-(1-(4-(2-(4-amino-6-chloro-2-methylpyrimidin-5-yl)ethyl)piperidin-1-yl)-1-oxopropan-2-yl)cyclopropanecarboxamide (Martinez-Grau, 2015), was synthesized using commercially available starting materials and (2S)-2-(tert-butoxycarbonylamino)propanoic acid (Boc-L-Alanine) using the procedures outlined in PCT/2014/064202. The structure and purity were verified by NMR spectroscopy and LC-MS. ¹H-NMR (400 MHz, CDCl3): δ 6.84 (d, J = 7.1 Hz, 1H), 5.05 (s, 2H), 4.90 (dd, J = 11.2, 6.6 Hz, 1H), 4.60–4.55 (m, 1H), 3.91–3.86 (m, 1H), 3.02 (q, J = 10.0 Hz, 1H), 2.60–2.53 (m, 2H), 2.43 (s, 3H), 1.90–1.80 (m, 3H), 1.64–1.56 (m, 1H), 1.47 (t, J = 7.6 Hz, 2H), 1.39 (dd, J = 7.6, 3.9 Hz, 1H), 1.28 (q, J = 8.0 Hz, 3H), 1.25–1.12 (m, 2H), 0.92 (d, J = 3.7 Hz, 2H), 0.71 (d, J = 7.0 Hz, 2H). LC-MS detection was carried out on Agilent 1200 using Luna C18 3 um (3 × 75 mm). The mobile phase was 4% to 100% acetonitrile (0.05% TFA) standard gradient. LC-MS m/z for C19H28ClN5O2 (MH+) 394.1.

QUANTIFICATION AND STATISTICAL ANALYSIS

Values are presented as mean ± SEM, with the number of replicates and the level of significance reported in figures, figure legends and supplementary tables. Statistical data analysis was performed using GraphPad Prism 7 for Windows (GraphPad Software, USA). A two-tailed Student’s t-test for paired or unpaired data was used for comparison of values between two groups. Data were assumed to have normal distribution as indicated by the D’Agostino & Pearson normality test. For multiple groups, ordinary one-way ANOVA followed by the Dunnett’s post-hoc test was applied. Time-dependent variables were analyzed by two-way ANOVA with Bonferroni’s correction when one group was compared to the control or Tukey’s multiple comparisons test. Differences were considered significant when P < 0.05.

Sample sizes for in vivo experiments were based on prior experience that 8 to 10 mice per treatment group provided reproducible values and differences. A much larger group of mice was used in the initial phase of the project (n = 17–24 mice per group, see Figure 1A) to obtain biological samples for screening purposes before ghrelin was selected as the focus of the study. Sample size was also dependent on the availability of animals through in-house breeding, which aimed to generate age- and weight-matched male littermates of 8–10 mice per treatment group. Randomization of animals into different treatment groups was done after the initial control drinking period to assure that mean values of alcohol intake and preference were similar in the vehicle versus drug treatment groups. The number of animals that underwent subdiaphragmatic vagotomy with pyloroplasty was limited by the Institutional Animal Care and Use Committee. The investigator monitoring the drinking pattern of these mice was blinded to their sham or vagotomized status. In vitro experiments were performed in duplicates. Cells from two dishes (acyl carnitine measurement) or three dishes (endocannabinoid measurement) were combined as a single biological replicate for more accurate detection. The initial slope of the fluorescence intensity in the Fatty Acid Oxidation assay was assessed from the same linear portion of the slope between 3 and 9 minutes, using GraphPad Prism. No data were excluded from in vitro studies, whereas a total of 10 mice were excluded from the drinking studies because of excessive weight loss caused by malocclusion (5 mice), seizures (2 mice) or complications of surgery (3 mice).

Highlights.

• CB₁R and ghrelin work together to promote alcohol intake via a gut-brain axis

• Alcohol intake is inhibited by CB₁R blockade, but only when ghrelin signaling is intact

• Inhibition of peripheral CB₁R reduces plasma levels of biologically active ghrelin

• CB₁R blockade promotes the degradation of the substrate needed to activate ghrelin

Context and Significance.

Endocannabinoids and the stomach-derived protein ghrelin both act as signaling molecules, and both are implicated in promoting appetite, including that for alcohol. Godlewski et al. show that these two molecules are functionally linked as they find that endocannabinoids promote the formation of biologically active ghrelin in the stomach, where the latter then promotes a gut-brain signaling pathway that leads to increased alcohol intake. They further show that peripherally-restricted pharmacological inhibition of this pathway reduces voluntary alcohol drinking in mice. As direct pharmacological targeting of the brain is undesired given the potential for adverse neuropsychological effects, the peripherally-centered approach described here may be a safer approach for the treatment of alcoholism.