CaMKII Modulates Diacylglycerol Lipase-α Activity in the Rat Nucleus Accumbens after Incubation of Cocaine Craving

Abstract Relapse is a major challenge to the treatment of substance use disorders. A progressive increase in cue-induced drug craving, termed incubation of craving, is observed after withdrawal from multiple drugs of abuse in humans and rodents. Incubation of cocaine craving involves the strengthening of excitatory synapses onto nucleus accumbens (NAc) medium spiny neurons via postsynaptic accumulation of high-conductance Ca2+-permeable AMPA receptors. This enhances reactivity to drug-associated cues and is required for the expression of incubation. Additionally, incubation of cocaine craving is associated with loss of the synaptic depression normally triggered by stimulation of metabotropic glutamate receptor 5 (mGlu5), leading to endocannabinoid production, and expressed presynaptically via cannabinoid receptor 1 activation. Previous studies have found alterations in mGlu5 and Homer proteins associated with the loss of this synaptic depression. Here we conducted coimmunoprecipitation studies to investigate associations of diacylglycerol lipase-α (DGL), which catalyzes formation of the endocannabinoid 2-arachidonylglycerol (2-AG), with mGlu5 and Homer proteins. Although these interactions were unchanged in the NAc core at incubation-relevant withdrawal times, the association of DGL with total and phosphorylated Ca2+/calmodulin-dependent protein kinase IIα (CaMKIIα) and CaMKIIβ was increased. This would be predicted, based on other studies, to inhibit DGL activity and therefore 2-AG production. This was confirmed by measuring DGL enzymatic activity. However, the magnitude of DGL inhibition did not correlate with the magnitude of incubation of craving for individual rats. These results suggest that CaMKII contributes to the loss of mGlu5-dependent synaptic depression after incubation, but the functional significance of this loss remains unclear.


Introduction
Relapse is often triggered by exposure to drug-associated cues even after prolonged abstinence (O'Brien, 2005). During abstinence, cue-induced craving intensifies over time, a phenomenon termed incubation of craving. Incubation of cocaine craving has been well documented in rats (Pickens et al., 2011) and also studied in humans (Parvaz et al., 2016), where cue-induced craving peaks between 1 and 6 months before declining by 1 year of abstinence. This trajectory is similar in rats, where craving is maximal after ;1 month, remains high for months, and then declines, although it remains significantly elevated, compared with withdrawal day 1 (WD1), after 180 d (Lu et al., 2004).
Incubation involves adaptations in regions related to motivation and reward including the nucleus accumbens (NAc; Li et al., 2015;Wolf, 2016;Wright and Dong, 2020). The NAc serves as an interface between cortical and limbic inputs and motor-related outputs, thereby contributing to the execution of motivated behaviors (Meredith et al., 2008;Sesack and Grace, 2010). Medium spiny neurons (MSNs), the principal cell type and projection neurons of the NAc, receive glutamatergic projections from cortical and limbic regions. Incubation of cocaine craving is associated with the incorporation of high-conductance calcium-permeable AMPA receptors (CP-AMPARs) into excitatory synapses of NAc core and shell MSNs (Wolf, 2016). In the core, this begins after ;1 month of withdrawal (Wolf and Tseng, 2012). Once present, CP-AMPAR activation is required for the expression of incubation (Conrad et al., 2008;Lee et al., 2013;Loweth et al., 2014;Ma et al., 2014). It is theoretically possible that the expression of incubation also involves a reduction in mechanisms that normally constrain excitatory transmission in the NAc, such as long-term depression (LTD).
Here, we used coimmunoprecipitation (co-IP) to assess associations between DGL and other components of the signalosome and also assayed enzymatic activity of DGL in the NAc core. Compared with saline controls, associations between total and phosphorylated Ca 21 /calmodulin-dependent protein kinase II (CaMKII) and DGL were increased at incubation-relevant withdrawal times. This was accompanied by decreased DGL activity, as predicted by studies demonstrating that CaMKII phosphorylation reduces DGL activity (Shonesy et al., 2013;Park et al., 2017). However, our results fall short of demonstrating that loss of this synaptic depression contributes to incubated craving. In fact, other findings argue that this synaptic depression promotes reward seeking (Novak et al., 2010;Bilbao et al., 2020;Mitra et al., 2021).

Subjects and surgery
All animal procedures were performed in accordance with the regulations of the animal care committees of Rosalind Franklin University and the Oregon Health & Science University, and with the National Institutes of Health (NIH) Guide for the Care and Use of Laboratory Animals. Male Sprague Dawley rats (Envigo) weighing 275-300 g were housed three to a cage under a reverse 12 h light/dark cycle with food and water available ad libitum. One week after arrival, rats were implanted with a jugular catheter under ketamine/xylazine anesthesia (80/10 mg/kg, i.p.). Before surgery, rats received an injection of the analgesic Banamine (flunixin meglumine, 2.5 mg/kg, s.c.) to minimize postoperative discomfort. Thereafter, rats were single housed for 7 d before beginning self-administration training. Catheters were flushed daily with cefazolin (100 mg/ml in sterile 0.9% saline; 0.15 ml) during this period to prevent infection and maintain catheter patency.

Drug self-administration and cue-induced seeking tests
Cocaine was obtained from the National Institute on Drug Abuse and dissolved in 0.9% saline. Rats self-administered cocaine (0.5 mg/kg/infusion in a 100 ml/kg volume over 3 s) or saline, during 10 consecutive sessions (6 h/d starting at zeitgeber time 15), in a chamber equipped with two nose-poke holes. Active hole nose pokes resulted in intravenous delivery of the drug paired with a 20 s light cue (white light illuminating the active hole) on a fixed ratio 1 schedule. Each infusion was followed by a 20 s time-out period. Nose poking in the inactive hole had no consequences. After 10 d of training, rats underwent forced abstinence in home cages and were handled weekly until being used for experiments on or after WD40. This regimen reliably results in incubation of cocaine craving that is robustly expressed during this withdrawal period (Conrad et al., 2008;Loweth et al., 2014). Therefore, we did not assess incubation of craving in rats used in some of our biochemical studies. However, two cohorts of rats, used for DGL activity assays, underwent cue-induced seeking tests after the same cocaine self-administration regimen to verify incubation of craving. During these seeking tests, responses in the previously active hole delivered the light cue but no drug. Responding under these conditions is our operational measure of cue-induced cocaine craving. One cohort underwent seeking tests on WD1 and WD40, and the other on WD1 and WD42.

Coimmunoprecipitation
Rats were decapitated and bilateral NAc core was rapidly dissected from two 1 mm slices prepared with a brain matrix (ASI Instruments) using a 1.5 mm biopsy punch (Thermo Fisher Scientific). NAc punches were homogenized in 500 ml of lysis buffer consisting of 25 mM HEPES, pH 7.4, 500 mM NaCl, 2 mM EDTA, 20 mM NaF, 1 mM PMSF, 0.1% NP-40 (v/v), 1 mM NaOV, 1Â protease inhibitor cocktail Set 1 (Millipore), 1% Triton, and 0.5% deoxycholate. Three micrograms of DGL-a antibody (a gift from Dr. Ken Mackie, Indiana University, Bloomington), validated in previous studies (Katona et al., 2006;Keimpema et al., 2010), was incubated overnight at 4°C with protein A/G agarose slurry. The pellet containing antibody-coated beads was then incubated overnight at 4°C with 100 mg of NAc core tissue homogenate. The agarose-bound antibody was pelleted by centrifugation to isolate the bound fraction (two rounds of immunoprecipitation were performed to maximize recovery). The combined bound fraction was suspended in 2Â Laemmli sample buffer with 1Â XT Reducing Agent (Bio-Rad) in a volume equal to the input tissue volume. Samples were heated to 100°C for 3 min and stored at À20°C. Samples were run on 4-12% Bis-Tris gels (BIO-RAD) and transferred to PVDF membranes for immunoblotting. Membranes were then washed in distilled H 2 O (dH 2 O) and blocked with 1% normal goat serum with 5% nonfat dry milk in 0.05% Tween-20 in TBS, pH 7.4, for 1 h at room temperature. Membranes were incubated overnight at 4°C with antibodies to mGlu5 (1:10,000; catalog #AB5675, Millipore Sigma), CaMKII (1:1000; catalog #3362, Cell Signaling Technology), p-Thr286 CaMKII (1:1000; catalog #p1005-286, Phospho Solutions), p-Ser (1:120; catalog #sc-81 514, Santa Cruz Biotechnology), Homer1b/c (1:200; catalog #sc-55 463, Santa Cruz Biotechnology), or Homer2 (1:500; catalog #H00009455-B01P, Abnova). The mGlu5 dimer band (;260 kDa) was analyzed because it represents the functional pool of these receptors (Jingami et al., 2003). GAPDH was used as a loading control when immunoblotting total tissue homogenates. Secondary antibodies to rabbit or mouse IgG light chain were used (catalog #211-032-171 or #115-035-174, Jackson ImmunoResearch) in co-IP studies to prevent interference with the Homer band (;45 kDa) by the heavy chain (;50 kDa). Visualization was achieved by chemiluminescence (GE Healthcare). Immunoblots were analyzed with TotalLab (Life Sciences Analysis Essentials). Data were excluded only if imperfections in the gel or blot interfered with analysis. Across all blots, data points for two samples were removed from analysis. These consisted of one cocaine and one saline data point from the synaptoneurosome preparation.

Synaptoneurosome preparation
Rats were decapitated and bilateral NAc tissue (primarily core) was rapidly dissected from a 2 mm coronal slice prepared with a brain matrix (ASI Instruments). Immediately following dissection, synaptoneurosomes were prepared according to published protocols (Most et al., 2015;Workman et al., 2015;Werner et al., 2018). NAc punches were homogenized in 500 ml of homogenization buffer [HB; 20 mM HEPES, 0.5 mM EGTA, 1Â Proteasome Inhibitor Cocktail Set 1 (Millipore)]. Homogenates were passed through a 100-mm-pore filter and then through a 5-mm-pore filter (Millipore; both filters were prewashed with HB). After homogenates were passed through each filter, filters were washed with 50 ml of HB, and the washes were added to homogenates to maximize yield. Homogenates were then centrifuged at 14,000 Â g for 20 min at 4°C. The pellet, which contains the synaptoneurosomes, was frozen on dry ice, stored at À80°C, and ultimately lysed in lysis buffer [0.605 Â g Tris-HCl, 0.25 Â g sodium deoxycholate, 0.876 Â g NaCl, 1 mg/ml PMSF, 5 ml of 20% SDS, and 1Â Protease Inhibitor Cocktail Set 1 (Millipore) in 100 ml of dH 2 O] for immunoblotting. NAc synaptoneurosomes prepared from individual rats (10 mg protein/lane) were mixed 1:1 with 2Â sample treatment buffer (catalog #161-0737, BIO-RAD) and analyzed by SDS-PAGE and immunoblotting. b -Tubulin was used as a loading control.

DGL activity
Rats were decapitated, and bilateral NAc was rapidly dissected as described above for synaptoneurosome preparation. Lipase activity was subsequently assayed using fluorescence resonance energy transfer (FRET) as adapted from a previous report (Johnston et al., 2012). Briefly, 600 ml of homogenization buffer [50 mM HEPES, pH 7.0, 250 mM sucrose, 1 mM Roche PhosStop (Millipore Sigma), 0.5 mM tris (2-carboxyethyl)phosphine (TCEP), 0.01 mM leupeptin, 0.001 mM pepsin] was added to punched NAc tissue and macerated by a 27 gauge needle. The solution was homogenized with an electric pestle and then centrifuged for 30 min. Supernatant was removed from the pellet and discarded. The pellet was resuspended in 300 ml of membrane resuspension buffer (same as homogenization buffer but without sucrose), then vortexed and briefly sonicated on ice. Using a Bradford assay, protein concentration of the punched NAc tissue was determined. A 96-well plate was then prepared on ice with protein samples in triplicate at 40 mg total protein/well. As a negative control, tissue from cerebellum was run in the absence and presence of the DAGL inhibitor DO34 (2 mM). The FRET reporter compound MRJ20 (compound #17 in Johnston et al., 2012), a FRETbased substrate for DGL, was made in-house (Vanderbilt Chemical Synthesis Core, Nashville, TN). MRJ20 was added to the assay buffer (50 mM HEPES, pH 7.0, 0.5 mM TCEP, 1 mM PhosStop) to reach a final concentration of 2 mM. The assay buffer was then vortexed and sonicated on ice for 2 min. The plate was then taken to a prewarmed fluorescent plate reader (Synergy H4, BioTek), and the recently sonicated assay buffer was pipetted into the wells at 150 ml. The reader maintained a constant 37°C temperature and took fluorescent measurements every minute for 30 min. Samples were then analyzed for relative fluorescent units (RFU) over the 30 min, and a line of best fit for RFU per minute determined relative slopes (DRFU/minute) for each sample (GraphPad Prism).

Statistical analyses
Data are expressed as the mean 6 SEM. Groups were compared with unpaired (between group) or paired (within group) t tests. Pearson correlation coefficients were used to assess the linear correlation between the magnitude of incubation of craving and DGL activity. Differences between experimental groups were considered statistically significant when p , 0.05.

Results
Incubation of cocaine craving is not associated with changes in DGL expression or its association with mGlu5 receptors and Homer scaffolding proteins Three cohorts of rats underwent extended-access saline or cocaine self-administration (10 sessions of 6 h/d; the three cohorts are depicted in Fig. 1A top, Fig. 1A bottom, and Fig. 2A). In all cases, biochemical analyses of NAc were conducted on or after WD40, when incubation of cocaine craving has plateaued (Lu et al., 2004) and stable CP-AMPAR elevation (Wolf and Tseng, 2012) and loss of mGlu5-dependent synaptic depression (McCutcheon et al., 2011;Scheyer et al., 2018) have been observed. The core subregion was analyzed, because of its critical role in the incubation of cocaine craving (Conrad et al., 2008;Guillem et al., 2014;Loweth et al., 2014). In NAc homogenates prepared from the first cohort of rats (Fig. 1A, top), we found that DGL levels did not differ between cocaine and saline groups (Fig. 1B, top). NAc tissue from the second cohort (Fig. 1A, bottom) was used to prepare synaptoneurosomes, a subcellular fraction enriched for the postsynaptic density (Hollingsworth et al., 1985;Quinlan et al., 1999;Most et al., 2015;Workman et al., 2015), on WD40. DGL levels in NAc synaptoneurosomes likewise did not differ between cocaine and saline groups (Fig. 1B, bottom).
NAc tissue from the third cohort was used to assess the DGL associations with Homer2 and mGlu5 by immunoprecipitating DGL and measuring bound Homer and mGlu5 levels (Fig. 2). mGlu5 and its association with Homer scaffolding proteins are necessary for 2-AG signaling at excitatory synapses (Roloff et al., 2010) and DGL associations with Homer scaffolding proteins are required for membrane-bound DGL activity (Jung et al., 2007). Although DGL does not directly bind to mGlu5, we detected bound mGlu5 in DGL immune complexes (Fig. 2B, Extended Data Fig. 2-1A) and also detected bound DGL in mGlu5 immune complexes (Extended Data Fig. 2-1A), presumably via intermediate Homer interactions. Homer1b/c was not detected in our DGL immune complexes in NAc core (Extended Data Fig. 2-1B,C). We therefore focused on Homer2, which has been implicated in the actions of cocaine in the NAc (Szumlinski et al., 2008). DGL associations with Homer2 and mGlu5 were unaffected by cocaine self-administration and prolonged withdrawal (Fig. 2B). These results suggest that impairments in physical association between these components of the 2-AG signalosome are not responsible for impaired mGlu5-dependent synaptic depression after incubation of cocaine craving.

Incubation of cocaine craving is associated with increased interactions between CaMKII and DGL
The loss of mGlu5-dependent synaptic depression after incubation of cocaine craving may be attributable to interactions that modulate DGL activity. For instance, CaMKIIa and CaMKIIb bind DGL, and studies in dorsal striatum have found that CaMKII phosphorylation of DGL inhibits its activity and, conversely, that 2-AG-dependent depolarization-induced suppression of excitation is augmented by pharmacological inhibition of CaMKII (Shonesy et al., 2013). Furthermore, increased CaMKII activity is implicated in stress-induced impairment of endocannabinoid-mediated synaptic depression in the lateral habenula (Park et al., 2017). To determine whether interactions between DGL and CaMKII are affected during late withdrawal from cocaine self-administration, different . Associations between components of the 2-AG signalosome in the NAc core are unchanged after prolonged withdrawal from extended-access cocaine (Coc) self-administration. A, Rats underwent extended-access self-administration of cocaine or saline (6 h/d for 10 d), destined for the preparation of NAc core homogenates on WD55. Shown are the number of infusions (left graph) or the number of active hole (AH) and inactive hole (IH) nose pokes (right graph) during the 10 d of self-administration training. N values are given in the figure. B, Coimmunoprecipitation experiments assessing the physical associations between DGL and mGlu5 or Homer2 in NAc tissue obtained on WD55 (also see Extended Data Figure 2-1). DGL was immunoprecipitated from NAc core homogenates, and the bound fraction was immunoblotted for mGlu5 or Homer2. Data are expressed as the percentage of saline (Sal) controls (n = 9-10 rats/group; mean 6 SEM). Representative lanes from immunoblots (cropped) are shown for a saline rat (left) and a cocaine rat (right). Arrows indicate bands analyzed, and lines indicate the location of the molecular weight marker. IB, Immunoblot; IP, Immunoprecipitation. Saline and cocaine groups were compared with unpaired t tests (n.s.). . Data are expressed as the percentage of controls (n = 7-9 rats/group; mean 6 SEM). Representative lanes from immunoblots (cropped) are shown for a saline rat (left) and a cocaine rat (right). Arrows indicate bands analyzed, and lines indicate the location of the molecular weight marker. Saline and cocaine groups were compared with unpaired t tests (n.s.). aliquots of the same tissue samples used for the experiment depicted in Figure 2 were used to immunoprecipitate DGL and measure bound CaMKII by immunoblotting. DGL associations with total CaMKIIa (t (14) = 2.59, p = 0.021), total CaMKIIb (t (14) = 2.30, p = 0.037), phosphorylated (Thr286) CaMKIIa (t (14) = 2.14, p = 0.050), and phosphorylated CaMKIIb (t (14) = 2.59, p = 0.014) were all increased in the cocaine group relative to saline controls (Fig. 3). CaMKII phosphorylates DGL on two serine residues (Shonesy et al., 2013). Therefore, to determine whether CaMKII associations with DGL also led to detectable increases in DGL phosphorylation, other aliquots of these DGL immunoprecipitated samples were subjected to SDS-PAGE followed by immunoblotting with an antibody that recognizes phosphorylated serine residues. A band was detected at the molecular weight of DGL (;100 kDa), but its intensity did not differ between saline and cocaine groups (Extended Data Fig. 3-1). However, these data are not conclusive. Whereas the lysis buffer for initial homogenization of tissue contained a reagent for inhibiting phosphatases (20 mM NaF), our washing buffer for processing the bound fraction during the immunoprecipitation protocol did not, because at the time the tissue was processed, we did not intend to use it to detect protein phosphorylation. Therefore, DGL might have been dephosphorylated during the lengthy immunoprecipitation protocol (see Materials and Methods).

DGL activity is reduced in the NAc core after incubation of cocaine craving
Based on earlier studies (Shonesy et al., 2013;Park et al., 2017), the observation of increased CaMKII association with DGL suggests that DGL activity will be reduced after prolonged withdrawal from extended-access cocaine self-administration. To test this, a new cohort of rats was generated (nine saline and nine cocaine; Fig. 4A). Cue-induced seeking tests were conducted on WD1 and WD40 to confirm incubation of craving (t (16) = 3.748, p = 0.002; Fig. 4B). On WD45, NAc tissue (primarily core) was harvested and snap frozen. DGL activity was subsequently assayed using FRET (Johnston et al., 2012). Compared with saline controls, the cocaine group was found to have reduced DGL activity (t (14) = 2.54, p = 0.024; Fig. 4C). Furthermore, in this initial cohort there was a trend for an inverse relationship between DGL activity in late withdrawal (WD40) and the magnitude of incubation of craving for rats in the cocaine group, expressed as the ratio of active hole nose pokes during 30 min seeking tests conducted on WD40 versus WD1 for each rat (p = 0.08; Fig. 4D).
We attempted to replicate this using an additional cohort of late withdrawal cocaine rats (Fig. 5A) for which incubation of craving was demonstrated (Fig. 5B). In this second cohort, we observed no significant relationship between DGL activity and the magnitude of incubation (Fig. 5D). We also assessed this relationship after combining the two cohorts. The second cohort was run over a year after the first, and required re-establishment of the assay. Perhaps for this reason, absolute RFU/minute values for the two cohorts differed substantially (compare Fig. 4C with 5C and Extended Data Fig. 5-1). Because the second cohort did not contain saline controls (they were not run because our focus was on understanding the correlation between DGL activity and incubation in the cocaine group), we could not combine the two cohorts by normalizing to saline control values. As an alternative approach, mean values (DRFU/minute) were calculated for each cohort, and individual rat values were expressed as a percentage of the cohort mean. We found no significant difference between DGL activity and incubation when the cohorts were combined in this manner (Fig. 5E). Overall, these results show that DGL activity is reduced in the NAc core after prolonged withdrawal from extended-access cocaine self-administration, and that the level of DGL activity in late withdrawal cocaine rats does not correlate with the magnitude of incubation.

Discussion
Enduring vulnerability to relapse is encoded by synaptic plasticity in the NAc (Lüscher, 2016;Scofield et al., 2016;Wolf, 2016). An important form of synaptic depression in the NAc is elicited by postsynaptic mGlu5 stimulation, resulting in the generation of eCBs, which travel in a retrograde fashion to stimulate presynaptic CB1 receptors and reduce glutamate release (see Introduction). After the incubation of cocaine craving, this mGlu5/CB1R-dependent synaptic depression is lost in the NAc core (McCutcheon et al., 2011;Scheyer et al., 2014Scheyer et al., , 2018. This is accompanied by reduced surface mGlu5 and reduced mGlu5-Homer interactions (Loweth et al., 2014). Coupled with intact synaptic depression in the NAc of cocaine-incubated animals in response to direct CB1R activation (McCutcheon et al., 2011), Figure 3. The association of CaMKII with DGL is increased in the NAc core after prolonged withdrawal from extended-access cocaine (Coc) self-administration. Coimmunoprecipitation experiments assessed the physical associations between DGL and phosphorylated or total CaMKIIa and CaMKIIb in NAc core homogenates prepared on WD55 (same rats for which behavioral data are shown in Fig. 2). Also see Extended Data Fig. 3-1. Data are expressed as the percentage of the saline (Sal) control group (n = 6-10 rats/group; mean 6 SEM; unpaired t tests, *p , 0.05, # p = 0.0501 vs saline). Representative lanes from immunoblots (cropped) are shown for a saline rat (left) and a cocaine rat (right). Arrows indicate bands analyzed, and lines indicate the location of the molecular weight marker. IB, Immunoblot. these results suggest a postsynaptic locus for the loss of mGlu5/CB1R-dependent synaptic depression.
Here, we assessed the postsynaptic mGlu5 signaling complex that mediates the generation of 2-AG (the 2-AG signalosome; Piomelli, 2014) in the NAc core subregion. No changes were found in DGL protein levels or its physical association with mGlu5 and Homer proteins (Figs. 1,  2), indicating that the loss of mGlu5/CB1R-dependent synaptic depression does not reflect reductions in physical associations between these components of the 2AGsignalasome. However, we observed increased associations between DGL and CaMKII (Fig. 3), an interaction previously demonstrated to result in DGL phosphorylation and reduced DGL activity (Shonesy et al., 2013). For these biochemical studies (Figs. 1-3), we did not demonstrate incubation of craving in the rats used, but we made measurements at withdrawal times (WD ! 40) when incubation is reliably demonstrated after the same regimen (Conrad et al., 2008;Loweth et al., 2014;Figs. 4, 5). Next, we measured the enzymatic activity of DGL, the enzyme that catalyzes 2-AG production (Piomelli, 2014). As might be expected from our observation of increased DGL-CaMKII association and previous results mentioned above (Shonesy et al., 2013), we observed a decrease in DGL activity in NAc core tissue obtained from rats demonstrated to have undergone incubation of cocaine craving compared with saline controls (Fig. 4). Although we lack evidence for a causal relationship between the increased DGL-CaMKII association and the observed reduction in DGL activity, it is tempting to speculate that the increased DGL-CaMKII association may contribute to the loss of mGlu5/CB1R-dependent synaptic depression after prolonged cocaine withdrawal. Finally, we found that DGL activity in late withdrawal (WD40 or WD42) was not significantly correlated with the magnitude of incubation (expressed as the ratio of active hole nose pokes during seeking tests conducted on WD40 or WD42 vs WD1) for individual cocaine rats (Figs. 4, 5).
We note that there are several limitations to our study. First, we did not include cocaine WD1 rats or yoked cocaine controls in our biochemical studies. These groups would be necessary to establish that observed changes are specific to incubation and not merely the result of cocaine self-administration or cocaine exposure, respectively. A second limitation of our study is that the use of homogenates or synaptoneurosomes precludes the Figure 4. Reduction of DGL activity in the NAc after prolonged withdrawal from extended-access cocaine (Coc) self-administration. A, Training. Rats self-administered saline or cocaine for 6 h/d for 10 d (n = 9 rats/group). Shown are the number of infusions (left graph) or number of active hole (AH) and inactive hole (IH) nose pokes (right graph). B, Seeking tests on WD1 and WD40. During the 30 min test, AH nose pokes led to contingent presentation of the 20 s light cue previously paired with cocaine infusion, but no cocaine. Shown are AH and IH nose pokes during the seeking test (mean 6 SEM; paired t test, **p , 0.01 vs WD1). C, DGL activity. Five days after the WD40 seeking test, bilateral NAc punches comprised primarily of NAc core were snap frozen. Fluorescence resonance energy transfer assay of DGL activity was then conducted and results expressed as DRFU/minute. The cocaine group shows reduced DGL activity (n = 7-9 rats/group; mean 6 SEM; unpaired t test, *p , 0.05 vs saline). D, Linear regression evaluating the relationship between DGL activity in NAc core of cocaine animals versus the magnitude of incubation of craving expressed as WD40/WD1 AH nose pokes (n = 7 rats; R = À0.692, p = 0.085). detection of potentially cell type-specific (D 1 vs D 2 MSNs) or input-specific adaptations in DGL activity. This could be one reason why we failed to identify a clear role for DGL adaptations in the incubation of craving. For example, one prominent incubation-related adaptation, synaptic insertion of CP-AMPARs, can occur on both D 1 and D 2 MSNs (Terrier et al., 2016;Wolf, 2016) but demonstrates input specificity (Lee et al., 2013;Ma et al., 2014;Pascoli et al., 2014;Terrier et al., 2016). A third limitation is that saline controls were not included in the experiment shown in Figure 5, as discussed extensively in Results. Finally, male rats were used for this study, in accordance with the NIH grant supporting this project. Given recent work showing sex differences in the incubation of cocaine craving (Kerstetter et al., 2008;Nicolas et al., 2019), future studies should explore similar questions in female rats.

Role for CaMKII in loss of mGlu5/CB1R-dependent synaptic depression after incubation of cocaine craving
CaMKII is a Ca 21 -activated enzyme that is best known for its critical role in early and late phases of long-term potentiation (LTP) but also participates in many other mechanisms regulating synaptic function (Lisman et al., 2002(Lisman et al., , 2012Hell, 2014;Bayer and Schulman, 2019). There are four isoforms derived from four genes (a, b , g , and d ), with the CaMKIIa and CaMKIIb isoforms mainly expressed in the brain. The regulatory site of each isoform contains a phosphorylation site (Thr286 for CaMKIIa and Thr287 for CaMKIIb ) that can generate autonomous kinase activity (see reviews cited above).
We explored a role for CaMKII in the loss of mGlu5/ CB1R-dependent synaptic depression after incubation of craving based on studies in dorsal striatum showing that CaMKIIa and CaMKIIb bind DGL, CaMKII phosphorylation of DGL inhibits its activity, and, conversely, that 2-AG-dependent depolarization-induced suppression of excitation is augmented by pharmacological inhibition of CaMKII (Shonesy et al., 2013). Furthermore, increased CaMKII activity is implicated in stress-induced impairment of endocannabinoid-mediated synaptic depression in the lateral habenula (Park et al., 2017). Consistent with these prior findings, we found increased associations between DGL and both total and phosphorylated CaMKIIa Figure 5. DGL activity in the NAc after prolonged withdrawal from extended-access cocaine (Coc) self-administration does not correlate with the magnitude of incubation of craving. A, Training. Rats self-administered cocaine for 6 h/d for 10 d (n = 9 rats). Shown are the number of infusions (left graph), and the number of active hole (AH) and inactive hole (IH) nose pokes (right graph). B, Seeking tests on WD1 and WD42. Shown are AH and IH nose pokes during the 30 min seeking test (mean 6 SEM; paired t test, **p , 0.01 vs WD1). C, Five days after the WD42 seeking test, bilateral NAc punches comprised primarily of NAc core were snap frozen and DGL activity was determined using fluorescent resonance energy transfer (Extended Data Fig. 5-1). D, Linear regression evaluating DGL activity versus the magnitude of incubation of craving expressed as WD42/WD1 AH nose pokes. No significant correlation between DGL activity and incubation was found (n = 9 rats; R = 0.393; n.s.). E, Combined linear regression evaluating DGL activity versus the magnitude of incubation of craving from both cocaine cohorts (shown in Fig. 4A and Fig. 5A). From comparison of Fig. 4C and Fig. 5C, it is apparent that absolute RFU per minute values for the two cohorts differed substantially (the cohorts were run over a year apart). To enable the data from the cohorts to be combined, mean values (DRFU/minute) were calculated for each cohort and individual rat values (DRFU/minute) were expressed as percentage of the cohort mean ( and CaMKIIb in NAc core tissue from cocaine rats (examined at an incubation-relevant late withdrawal time) versus saline controls. Furthermore, we observed a significant reduction in DGL activity in rats that had undergone incubation of craving compared with saline controls. These results suggest that CaMKII contributes to loss of mGlu5/ CB1R-dependent synaptic depression after incubation. In theory, slice physiology experiments could directly test this relationship by determining whether acute CaMKII inhibition restores mGlu5/CB1R-dependent synaptic depression in NAc core MSNs from rats that have undergone incubation of cocaine craving. A practical concern is that commercially available inhibitors are not effective against the autophosphorylated enzyme, although new inhibitors are under development (Nassal et al., 2020).
Our findings are not the first to implicate CaMKII in neuroadaptations in the NAc after cocaine self-administration. For example, CaMKIIa, but not CaMKIIb , mRNA levels increased in NAc shell after cocaine self-administration and in cocaine-yoked controls relative to saline self-administering animals, likely via histone H3 acetylation; the increase correlated with motivation for cocaine measured using a progressive ratio schedule, while knockdown of CaMKIIa in the NAc shell decreased motivation for cocaine (Wang et al., 2010). In another study, cocaine self-administration followed by extinction training did not alter CaMKII phosphorylation in NAc shell or core; however, cocaine-primed reinstatement was associated with an increase in NAc shell CaMKII phosphorylation, and intra-NAc shell CaMKII inhibition via KN-93 attenuated reinstatement (Anderson et al., 2008). During abstinence, increased CaMKIIa phosphorylation in the NAc was found 24 h after discontinuing short-access cocaine self-administration relative to yoked saline controls, but this was not observed in a yoked cocaine group, immediately after discontinuing self-administration, or after selfadministration and 7 d of abstinence (Caffino et al., 2014). In an incubation of cocaine craving study in which rats self-administered saline or cocaine under extended-access conditions, increased ratios of phosphorylated to total CaMKIIa and CaMKIIb were found in NAc postsynaptic density fractions from cocaine rats on WD45 versus cocaine rats on WD1 and saline rats (Ferrario et al., 2011). CaMKII is also implicated in responding to other drugs of abuse. For example, CaMKII activity in NAc shell is required for the enhanced amphetamine intake and sensitized locomotor responding observed in rats previously sensitized to noncontingent amphetamine (Loweth et al., 2008(Loweth et al., , 2010(Loweth et al., , 2013. How might CaMKII become activated during abstinence? While classical studies identified CaMKII activation following acute stimulation in slices (i.e., in conjunction with LTP), CaMKII may also contribute to homeostatic plasticity (Chen et al., 2014;Hell, 2014). Most notably, in cultured hippocampal neurons, prolonged AMPAR blockade (24 h) increases CaMKIIb activity, which in turn scales up synaptic levels of homomeric GluA1 receptors; when blockade is discontinued, a presynaptic component of homeostatic adaptation (increased release probability) is revealed in response to Ca 21 entry via CP-AMPARs and the resulting retrograde signaling (Thiagarajan et al., 2002(Thiagarajan et al., , 2005Groth et al., 2011). We have suggested that AMPAR upregulation during cocaine abstinence reflects a homeostatic scaling up of synaptic strength, triggered by relative inactivity during abstinence versus the period of cocaine exposure; while this may normalize synaptic transmission during abstinence, it has the unfortunate consequence of enhancing excitatory responses to drug-related cues when they are presented after abstinence (Boudreau and Wolf, 2005;Conrad et al., 2008;Sun and Wolf, 2009). Likewise, the observation that sustained food restriction upregulates reward responding through the upregulation of NAc CP-AMPARs has been hypothesized to represent a homeostatic response (Carr, 2020). The finding of increased CaMKII phosphorylation in NAc postsynaptic density fractions prepared after incubation of cocaine craving (Ferrario et al., 2011) combined with increased phosphorylation of the DGL-bound CaMKII pool in the present study suggests that CaMKII may play a role after cocaine withdrawal that to some extent parallels that observed after AMPAR blockade in hippocampal neurons (above). The reduction in DGL activity that is observed in parallel with CaMKII activation could be "collateral damage" or it could contribute to the homeostatic response by removing a braking effect on excitatory synaptic transmission. It is also possible that CaMKII activation after incubation of craving is functionally relevant to DGL regulation, but that different homeostatic cascades are involved in CP-AMPAR upregulation (Wang et al., 2018;Loweth et al., 2019).
Relationship between mGlu5/CB1R-dependent synaptic depression and cocaine seeking As DGL activity was reduced in rats that underwent incubation of cocaine craving compared with saline controls (above), we tested for a correlation between DGL activity and the magnitude of incubation of craving (expressed relative to baseline craving on withdrawal day 1) for individual cocaine rats in two separate cohorts. Because the expression of incubation requires excitatory synaptic transmission onto NAc core MSN (Wolf, 2016), and DGL activity leading to 2-AG formation leads to reduced glutamate transmission (i.e., synaptic depression), we hypothesized that cocaine rats with the lowest DGL activity might show the strongest incubation of craving. However, we failed to observe a significant correlation between DGL activity and cocaine seeking. This does not rule out a contribution of reduced DGL activity to incubation. It is possible that reduction of DGL activity below a threshold level plays a permissive role in the enhancement of cue-induced cocaine seeking.
However, a number of lines of evidence argue against this possibility. First, mGlu5/CB1R-dependent synaptic depression has been linked to the promotion of reward seeking in mice with constitutive knockdown of mGlu5 in D 1 receptor-expressing MSNs. These mice do not express mGlu5-dependent synaptic depression or demonstrate cue-induced reinstatement of cocaine or saccharin seeking; however, after 2-AG elevation with monoacylglycerol lipase (MAGL) inhibition, reinstatement of saccharin seeking is restored (Novak et al., 2010;Bilbao et al., 2020). Furthermore, it was found that DGL and MAGL were dysregulated after prolonged abstinence (30 d) from cocaine self-administration, and that intra-NAc administration of a DGL inhibitor at this withdrawal time reduced cue-induced cocaine craving, while inhibition of MAGL (to increase 2-AG levels) had the opposite effect (Mitra et al., 2021). We note that DGL levels were increased at WD30 in this prior study (Mitra et al., 2021), in contrast to the lack of change reported here. This is most likely because of the fact that Mitra et al. (2021) measured DGL in NAc shell, whereas we analyzed NAc core; there were also significant differences in the cocaine self-administration regimen. While the work cited above focused on synaptic depression expressed presynaptically via CB1R stimulation, other studies link mGlu5-dependent reinstatement of cocaine seeking to postsynaptically expressed LTD mediated via AMPAR internalization (Schmidt et al., 2013;Benneyworth et al., 2019).
Additional evidence against a causal relationship between loss of mGlu5/CB1R-dependent synaptic depression and elevation of cue-induced drug craving comes from studies on the incubation of methamphetamine craving. Dihydroxyphenylglycine-induced, mGlu5-dependent synaptic depression in the NAc core was lost during the first week of abstinence from extended-access methamphetamine self-administration, which corresponds to the rising phase of incubation of methamphetamine craving, but at later withdrawal times this synaptic depression was restored whereas craving remained at high "incubated" levels (Murray et al., 2021). These results argue against a role for mGlu5/CB1R-dependent synaptic depression in maintaining incubated craving for methamphetamine. It is also interesting to note that associations between CaMKII and DGL were unaltered during the first week of methamphetamine withdrawal when mGlu5/CB1R-dependent synaptic depression was lost (Murray et al., 2021), in contrast to the present results with cocaine. This adds to mechanistic distinctions between incubation of cocaine and methamphetamine craving, although commonalities also exist (Murray et al., 2019).
Overall, these results argue that synaptic depression can promote (Novak et al., 2010;Bilbao et al., 2020;Mitra et al., 2021) or be dissociated from (Murray et al., 2021) drug or natural reward seeking. Thus, the functional significance of the observed reduction in DGL activity observed after prolonged cocaine withdrawal remains unclear. Likewise, no consensus has emerged from work that has tested the effect of manipulation of eCB signaling on seeking for cocaine and other drugs of abuse. While reducing endocannabinoid signaling through negative allosteric modulation has shown therapeutic promise, so too has enhancement of endocannabinoid signaling through the inhibition of endocannabinoid degradation (Galaj and Xi, 2019). The complex nature of effects observed after pharmacological targeting of CB1Rs is likely to reflect their widespread expression (e.g., they are expressed not only on glutamate terminals but also on GABA terminals throughout the brain (Augustin and Lovinger, 2018). Furthermore, within the NAc, in addition to mediating synaptic depression onto MSNs, endocannabinoids also mediate synaptic depression onto parvalbumin-containing interneurons that regulate MSN activity (Manz et al., 2020) and contribute to the modulation of dopamine release (Mateo et al., 2017).

Conclusion
The loss of mGlu5/CB1R-dependent synaptic depression in the NAc core after incubation of cocaine craving is likely attributable to multiple postsynaptic mechanisms, including reduced surface expression of mGlu5 and reduced mGlu5-Homer interactions (Loweth et al., 2014) and, as shown here, a reduction in DGL activity that may result from increased CaMKII-DGL association. Although our findings and related work in the literature suggest that the loss of mGlu5/CB1R-dependent synaptic transmission is unlikely to be a driver of the incubation of cocaine craving, it is expected that loss of a major form of synaptic plasticity would have significant consequences for NAc function. Future studies should continue to explore this problem.