|Home | About | Journals | Submit | Contact Us | Français|
The transcription factor ΔFosB and the brain-enriched protein kinase CaMKIIα (calcium/calmodulin-dependent protein kinase II) are induced in the nucleus accumbens (NAc) by chronic exposure to cocaine or other psychostimulant drugs of abuse, where the two proteins mediate sensitized drug responses. Although ΔFosB and CaMKIIα both regulate AMPA glutamate receptor expression and function in NAc, dendritic spine formation on NAc medium spiny neurons (MSNs), and locomotor sensitization to cocaine, no direct link between these molecules has to date been explored. Here, we demonstrate that ΔFosB is phosphorylated by CaMKIIα at the protein-stabilizing Ser27, and that CaMKII is required for the cocaine-mediated accumulation of ΔFosB in rat NAc. Conversely, we show that ΔFosB is both necessary and sufficient for cocaine induction of CaMKIIα gene expression in vivo, an effect selective for D1-type MSNs in the NAc shell subregion. Furthermore, induction of dendritic spines on NAc MSNs and increased behavioral responsiveness to cocaine after NAc overexpression of ΔFosB are both CaMKII-dependent. Importantly, we demonstrate for the first time induction of ΔFosB and CaMKII in the NAc of human cocaine addicts, suggesting possible targets for future therapeutic intervention. These data establish that ΔFosB and CaMKII engage in a cell type- and brain region-specific positive feed-forward loop as a key mechanism for regulating the brain’s reward circuitry in response to chronic cocaine.
Increasing evidence supports the view that changes in gene expression contribute to mechanisms of drug addiction (Robison and Nestler, 2011). One important mediator of these changes is ΔFosB, a Fos family transcription factor (Nestler, 2008). Chronic administration of virtually any drug of abuse induces the long-lasting accumulation of ΔFosB in nucleus accumbens (NAc), a limbic region essential for reward behaviors. Such induction appears specific to the class of NAc medium spiny neuron (MSN) that expresses D1 dopamine receptors. Inducible overexpression of ΔFosB in these D1-type NAc MSNs increases locomotor and rewarding responses to cocaine and morphine (Kelz et al., 1999; Zachariou et al., 2006), including increased cocaine self-administration (Colby et al., 2003). Furthermore, genetic or viral blockade of ΔFosB transcriptional activity reduces the rewarding effects of these drugs (Zachariou et al., 2006), indicating that this sustained induction of ΔFosB is a critical mediator of the lasting changes induced in NAc by chronic drug administration.
The unusual stability of ΔFosB (relative to all other Fos family proteins) is both an intrinsic property of the molecule, due to the truncation of degron domains present in full-length FosB (Carle et al., 2007), and a regulated process. ΔFosB is phosphorylated in vitro and in vivo at Ser27, and this reaction further stabilizes ΔFosB, ~10-fold, in cell culture and NAc in vivo (Ulery-Reynolds et al., 2009). Although Ser27-ΔFosB has been shown to be a substrate for casein kinase-2 in vitro (Ulery et al., 2006), its mechanism of in vivo phosphorylation remains unknown.
Calcium/calmodulin-dependent protein kinase II (CaMKII) is a highly-expressed serine/threonine kinase whose α and β isoforms form dodecameric homo- and hetero-holoenzymes in vivo, and are essential for multiple forms of neuroplasticity (Lisman et al., 2002; Colbran and Brown, 2004). CaMKIIα is induced selectively in NAc shell by chronic amphetamine (Loweth et al., 2010), and pharmacological blockade of CaMKII activity in NAc shell reduces behavioral sensitization to amphetamine (Loweth et al., 2008) and cocaine (Pierce et al., 1998), while viral overexpression of CaMKIIα in this NAc subregion enhances locomotor sensitization to and self-administration of amphetamine (Loweth et al., 2010). CaMKIIα may affect reward behaviors via modulation of AMPA glutamate receptor subunits (Pierce et al., 1998), as CaMKIIα activity has long been associated with AMPA receptor function and synaptic-targeting in several forms of neuroplasticity (Malinow and Malenka, 2002).
This literature demonstrates several parallels between ΔFosB and CaMKII: both are necessary and sufficient for multiple behavioral effects of drugs of abuse, both upregulate dendritic spines in various neuronal cell types in vivo (Jourdain et al., 2003; Maze et al., 2010), and both exert at least some of their behavioral effects through modulation of AMPA receptors (Kelz et al., 1999; Malinow and Malenka, 2002; Vialou et al., 2010). Despite these parallels, no functional link between ΔFosB and CaMKII is known. Here, we establish reciprocal regulation between ΔFosB and CaMKII, and demonstrate that the two proteins form a D1-type MSN-specific feed-forward loop in NAc shell that is induced by cocaine and regulates a range of cocaine responses in vivo.
Adult (8 weeks) male rats were administered 20 mg/kg cocaine or saline vehicle IP once per day for seven days. 24 hr after the last injection, NAc shell and core were microdissected (Fig 1A) and flash frozen. iTRAQ analyses were performed as previously described (Ross et al., 2004; Davalos et al., 2010).
Adult (8 weeks) male rats were administered 10 mg/kg cocaine or saline vehicle IP once per day for seven days in locomotor recording chambers. Locomotor responses to a single injection of cocaine (5 mg/kg IP) was recorded in those animals treated previously with cocaine (called “chronic”) and a portion of those treated with saline (called “acute”), and locomotor responses to saline alone was recorded in the remaining chronic saline-treated animals (called “saline”). Locomotor activity assays were performed as described (Hiroi et al., 1997). Briefly, adult male rats were placed in 18”×24” PAS open-field recording boxes (San Diego Instruments) for 30 min to habituate, were given a single IP injection of saline and monitored for an additional 30 min, and were given a single IP injection of 5 mg/kg cocaine and monitored for 30 min.
24 hr following this final injection, rats were decapitated without anesthesia to avoid effects of anesthetics on neuronal protein levels and phospho-states. Brains were serially sliced in a 1.2 mm matrix (Braintree Scientific) and target tissue was removed in phosphate buffered saline containing protease (Roche) and phosphatase (Sigma Aldrich) inhibitors using a 14 gauge punch for NAc core and a 12 gauge punch of the remaining tissue for NAc shell (See Fig 1A) and immediately frozen on dry ice. Samples were homogenized by light sonication in modified RIPA buffer: 10 mM Tris base, 150 mM sodium chloride, 1 mM EDTA, 0.1% sodium dodecyl sulfate, 1% Triton X-100, 1% sodium deoxycholate, pH 7.4, protease and phosphatase inhibitors as above. After addition of Laemmli buffer, proteins were separated on 4–15% polyacrylamaide gradient gels (Criterion System, BioRad), and Western blotting was performed using the Odyssey system (Li-Cor) according to manufacturer protocols.
Adult (8 weeks) male rats were administered 10 mg/kg cocaine or saline vehicle IP once per day for seven days. 14 days after the final injection, animals treated with saline were given another saline injection (called “saline), and animals treated with cocaine were given another saline injection (called 14 day withdrawal or “14d WD”) or a single injection of cocaine (called “14d WD Chal” for challenge). One hr after the final injection, animals were decapitated and Western blotting performed as in Experiment 2.
Rats were trained to self-administer 0.5 mg/kg/infusion of cocaine in one-hr sessions under a fixed-ratio 1 schedule for nine days. After nine baseline sessions, the rats were divided into two groups balanced by cocaine intake on the last two sessions. One group of rats was allowed to self-administer cocaine (0.5 mg/kg/infusion) in one-hr sessions (short access, ShA) while the other group of rats self-administered cocaine in six-hr sessions (long access, LgA) for ten additional days (escalation sessions).
Brain sections were processed for immunohistochemistry as described (Perrotti et al., 2004). Brains were perfused 18–24 hr after the last exposure to drug, resulting in the degradation of any residual full-length FosB protein such that all remaining immunoreactivity reflects ΔFosB. This degradation was confirmed by Western blotting, which showed no significant staining with an antibody directed against the C terminus of full-length FosB that does not recognize ΔFosB (data not shown). After slicing into 35 µm sections, the number of ΔFosB immunopositive cells was quantified by a blinded observer in two sections through the NAc of each rat, and mean values per 40× field were then calculated by region for each animal. Each animal was considered an individual observation for statistical analysis. Regions of interest were identified using Paxinos and Watson (Paxinos and Watson, 2007).
Quantification of CaMKIIα immunoreactivity was performed using a Licor system as described (Covington et al., 2009). Integrated intensities of CaMKII and GAPDH were determined with Odyssey software. Results are presented as integrated intensity values per mm2 and are presented as means ± s.e.m. (n = 4–10 per group). Values for GAPDH were used as reference to normalize CaMKII intensity for slice thickness and conditions.
Postmortem human brain tissues were obtained from the Quebec Suicide Brain Bank (Douglas Mental Health University Institute, Montreal, Quebec, Canada). The preservation of tissue proceeded essentially as described (Quirion et al., 1987). Briefly, once extracted, the brain is placed on wet ice in a Styrofoam box and rushed to the Quebec Suicide Brain Bank facilities. Hemispheres are immediately separated by a sagittal cut in the middle of the brain, brain stem, and cerebellum. Blood vessels, pineal gland, choroid plexus, half cerebellum, and half brain stem are typically dissected from the left hemisphere which is then cut coronally into 1 cm-thick slices before freezing. The latter half cerebellum is cut sagittally into 1cm-thick slices before freezing. Tissues are flash frozen in 2-methylbutane at −40°C for ~60 sec. All frozen tissues are kept separately in plastic bags at −80°C for long-term storage. Specific brain regions are dissected from frozen coronal slices on a stainless steel plate with dry ice all around to control the temperature of the environment. Western blotting was performed as described in Experiment 2.
The cohort was composed of 37 male and 3 female subjects, ranging in age between 15–66 years. All subjects died suddenly without a prolonged agonal state or protracted medical illness. In each case, the cause of death was ascertained by the Quebec’s Coroner office, and a toxicological screen was conducted with tissue samples to obtain information on medication and illicit substance use at the time of death. The subject group consisted of 20 individuals who met the SCID-I criteria for cocaine dependence. The control group was comprised of 20 subjects with no history of cocaine dependence and no major psychiatric diagnoses. All subjects died suddenly from causes that had no direct influence on brain tissue. Groups were matched for mean subject age, refrigeration delay, and pH. For all subjects, psychological autopsies were performed as described previously (Dumais et al., 2005), allowing us to have access to detailed case information on psychiatric and medical history, as well as other relevant clinical and sociodemographic data. In brief, a trained interviewer conducted the Structured Clinical Interview for DSM-IV Psychiatric Disorders (SCID-I) with one or more informants of the deceased. A panel of clinicians reviewed SCID-I assessments, case reports, coroner’s notes, and medical records to obtain consensus psychiatric diagnoses.
Adult (8 weeks) male rats were administered 10 mg/kg cocaine or saline vehicle IP once per day for seven days. 24 hr after the last injection, NAc shell and core were microdissected. Chromatin immunoprecipitation (ChIP) was performed pooling bilateral NAc punches of shell or core from seven rats per group in 14 total groups (98 animals total, 7 cocaine pools, 7 saline pools). Tissues were cross-linked, washed, and stored at −80°C until chromatin shearing by sonication. Sheared chromatin was incubated overnight with antibodies previously bound to magnetic beads (Dynabeads M-280, Invitrogen). Non-immune IgG was used as a control. After reverse cross-linking and DNA purification, qPCR was used to measure levels of CaMKIIα promoter DNA. Primers were designed to amplify a region containing an AP-1 consensus sequence located ~450 bp prior to the transcription start site (Forward: ACTGACTCAGGAAGAGGGATA; Reverse: TGTGCTCCTCAGAATCCACAA).
Male bitransgenic mice derived from NSE-tTA (line A) × TetOp-ΔfosB (line 11) and NSE-tTA (line B) × TetOp-FLAG-ΔfosB (line 11) mice (Chen et al., 1998; Kelz et al., 1999; Werme et al., 2002; Zachariou et al., 2006) were conceived and raised on 100 µg/ml doxycycline to suppress ΔFosB expression during development. Littermates were divided at weaning: half remained on doxycycline and half were switched to water, and the animals were used 8 to 11 weeks later when transcriptional effects of ΔFosB are maximal (Kelz et al., 1999; McClung and Nestler, 2003). For transcriptional analyses, mice were rapidly decapitated, and brains were removed and placed on ice. Dissections of NAc were taken with a 14-gauge needle punch and quickly frozen on dry ice until RNA was extracted. RNA isolation, qPCR, and data analysis were performed as previously described (LaPlant et al., 2009). Briefly, RNA was isolated with TriZol reagent (Invitrogen), further purified with the RNAeasy micro kit from Qiagen, and checked for quality with Agilent’s Bioanalyzer. Reverse transcription was performed using iScript (BioRad). qPCR was carried out with an Applied Biosystems 7900HT RT PCR system with the following cycle parameters: 10 min at 95°C; 40 cycles of 95°C for 1 min, 60°C for 30 sec, 72°C for 30 sec; graded heating to 95°C to generate dissociation curves for confirmation of single PCR products. Immunohistochemical analyses of ΔFosB and CaMKIIα protein expression were performed as described in Experiment 4.
Adult (8 weeks) male rats were administered 10 mg/kg cocaine or saline vehicle (“vehicle” group) IP once per day for seven days. 30 min prior to each cocaine injection, rats were IP administered either the D1 receptor antagonist SCH 23390 (0.5 mg/kg, “D1 Ant” group), or the D2 receptor antagonist eticlopride (0.5 mg/kg, “D2 Ant” group), or a saline control injection (“cocaine” group). 24 hr after the final injection, animals were decapitated and proteins quantified by Western blotting as per Experiment 2.
Stereotaxic surgery was performed on adult male rats (8 weeks) to inject AAV-GFP (green fluorescent protein) or AAV-GFP-ΔFosB (Maze et al., 2010). 33 gauge needles (Hamilton) were employed for all surgeries, during which 0.5 µl of purified high-titer virus was bilaterally infused over a 5 min period of time, followed by an additional 5 min post-infusion rest period. All distances are measured relative to Bregma: 10° angle, AP = +1.7 mm, Lat = 2.5 mm, DV = −6.7 mm. 14 days after surgery, animals were given a single IP injection of 10 mg/kg cocaine in locomotor monitoring chambers to assess the behavioral effects of ΔFosB overexpression. 24 hr after this final injection, rats were decapitated as per Experiment 2, and tissue microdissection was performed under fluorescence microscopic guidance to obtain GFP-positive NAc tissue. Western blotting was then performed as per Experiment 2.
Stereotaxic injection of AAV-GFP or AAV-GFP-ΔJunD was performed as per Experiment 8. 14 days after surgery, animals were administered 10 mg/kg cocaine or saline vehicle IP once per day for seven days in locomotor recording chambers. Locomotor responses to a single injection of cocaine (5 mg/kg IP) or saline was recorded. 24 hr after this final injection, rats were decapitated, tissue harvested, and Western blots performed as in Experiment 9.
Recombinant CaMKIIα and ΔFosB were purified from insect cells (Brickey et al., 1990; Jorissen et al., 2007), and protein kinase assays were performed (Colbran, 1993), as previously described. Briefly, CaMKII was preincubated on ice with 2.5 µM (or indicated concentration) of ΔFosB, 1 mM Ca2+, 40 mM Mg2+, 15 µM calmodulin, and 200 mM HEPES pH 7.5. Phosphorylation was initiated by addition of 200 µM ATP with or without [γ-32P]ATP and allowed to proceed for 10 min at room temperature (Fig 5A&B) or 2 min on ice (Fig 5C&D). Products were resolved by Western blotting (Fig 5A&B) or by autoradiogram and scintillation counting (Fig B–D).
In vitro kinase assays were performed as per Experiment 11, proteins were separated by SDS-PAGE, and bands corresponding to ΔFosB were cut out and subjected to tandem mass spectrometry. The m/z assignments of the corresponding ion fragments in all of the panels are labeled on top of the ion peaks. Not all fragment ions are labeled due to space limitations. Generally, the text for the fragment ion labels are colored in black except when they directly confirm or add evidence to the presence of the phosphorylation sites of interest, in which case they are marked in red. Evidence for backbone fragmentation products are presented in the sequence readout of the phosphopeptide with the detected site of phosphorylation residue indicated in red with a single amino acid letter designation. The numeric description of the observed fragment ions are also marked on the peptide sequence as b and y ions. The zoom factors for the sections of the m/z axis to show the lower intensity fragment ions are marked at the top of each fragment mass spectra. The fragment ions shown in panel H confirms the presence of Ser27 phosphorylated isoform, however, within a mixture of other phosphorylated isoforms at sites Ser28, Ser31, Ser34, and Thr37. The presence of pa5, pa5-P, pb5, and pb5-P ions uniquely confirm the phosphorylation of the Ser27 residue.
Standard peptides were designed mimicking the phospho and non-phospho forms of Ser27 ΔFosB. After synthesis and purification, each “heavy” idiotypic peptide was dissolved in a 50/50 acetonitrile/water buffer and sent for amino acid analysis to determine absolute concentration on the synthetic peptide stock solution. Each “heavy” peptide was then directly infused into the 4000 QTRAP mass spectrometer (MS) to determine the best collision energy for MS/MS fragmentation and two to four MRM transitions. Next, the neat “heavy” peptides were subjected to LCMS on the 4000 QTRAP to ensure peptide separation. The instrument was run in the triple quadrupole mode, with Q1 set on the specific precursor m/z value (Q1 is not scanning), and Q3 set to the specific m/z value corresponding to a specific fragment of that peptide. In the MRM mode, a series of single reactions (precursor / fragment ion transitions where the collision energy is tuned to optimize the intensity of the fragment ions of interest) were measured sequentially, and the cycle (typically 1–2 sec) was looped throughout the entire time of the HPLC separation. MRM transitions were determined from the MS/MS spectra of the existing peptides. Two transitions per peptide, corresponding to high intensity fragment ions, were then selected and the collision energy optimized to maximize signal strength of MRM transitions using automation software. Peaks resulting from standard peptides and ΔFosB samples exposed to CaMKII or control were then compared to determine the absolute abundance of each peptide form in the reaction. Data analysis on LC-MRM data is performed using AB Multiquant 1.1 software.
Transgenic mice overexpressing T286D CaMKII (Mayford et al., 1996; Kourrich et al., 2012) and wild type littermates were raised in the absence of doxycycline to allow transgene expression. Adult mice were administered 20 mg/kg cocaine or saline IP once daily for 14 days. 24 hr after the final injection, animals were decapitated and immunohistochemistry and quantification of ΔFosB expression was performed as in Experiment 4.
Adult male mice (8 weeks) were stereotaxically injected in NAc with HSV-GFP, HSV-GFP-ΔFosB (Olausson et al., 2006), HSV-GFPAC3I, or HSV-GFPAC3I-ΔFosB. In these constructs, AC3I, a peptide-based inhibitor of CaMKII activity, is fused to the C-terminus of GFP. GFPAC3I was cloned by PCR using the pMM400-vector containing GFPAC3I as a template with the following primers: GFP-AC3I-F: 5’ CC GCTAGC GCCGCCACC ATGGTGAGCAAGGGCGAGGAGCTGT 3’ (clampNheIKozakmet); GFP-AC3I-R: 5’ CC TCCGGA TTACAGGCAGTCCACGGCCT 3’ (clampBspEIstop). The resulting PCR product was inserted into the p1005+ and p1005+-Δ FosB vectors using NheI and BspEI sites. The construct was validated by sequencing. Stereotaxic coordinates were: 10° angle, AP = +1.6 mm, Lat = +1.5 mm, DV = −4.4 mm (Barrot et al., 2002). Perfusion and brain sectioning was performed as per Experiment 4.
Spine analysis was performed as described (Christoffel et al., 2011). Briefly, dendritic segments 50–150 µm away from the soma were randomly chosen from HSV-infected cells that express GFP. Images were acquired on a confocal LSM 710 (Carl Zeiss) for morphological analysis using NeuronStudio with the rayburst algorithm. NeuronStudio classifies spines as thin, mushroom, or stubby based on the following values: (1) aspect ratio, (2) head to neck ratio, and (3) head diameter. Spines with a neck can be classified as either thin or mushroom, and those without a significant neck are classified as stubby. Spines with a neck are labeled as thin or mushroom based on head diameter.
Adult male mice were injected with viruses as per Experiment 15, and locomotor responses to a single 5 mg/kg injection of cocaine was measured as per Experiment 9. Locomotor data are expressed as total beam breaks over 30 min after cocaine injection.
Male Sprague Dawley rats (250–275 g; Charles River Laboratories) were pair-housed. Eight-week-old C57BL/6J male mice (The Jackson Laboratory) were group housed with a maximum of five animals per cage. All animals were habituated to the animal facility for ≥1 week before experimental manipulations and housed in climate-controlled rooms (23–25°C) on a 12 hr light/dark cycle (lights on at 7:00 A.M.) with access to food and water ad libitum. Experiments were conducted in accordance with guidelines of the Society for Neuroscience and the institutional animal care and use committee (IACUC) at Mount Sinai.
Drugs were administered IP and dissolved in sterile saline, including cocaine (5–20 mg/kg per 10 µl for mice, per 1 ml for rats, NIDA) and SCH 23390 or eticlopride hydrochloride (0.5 mg/kg per 1 ml, Tocris). For stereotaxic surgery, mice were anesthetized with a “cocktail” of ketamine (100 mg/kg) and xylazine (10 mg/kg) (Henry Schein) in sterile saline.
CaMKIIα (total): Upstate 05–532, 1:5,000
CaMKII phospho-Thr286: Promega V111A, 1:1,000
ΔFosB (total): Cell Signaling 5G4, 1:250
ΔFosB phospho-Ser27: Phosphosolutions, 1:500
GluA1 (total): Abcam, Ab31232, 1:1,000
GluA1 phospho-Ser831: Millipore N453, 1:1,000
GluA1 phospho-Ser845: Chemicon Ab5849, 1:2,000
GluA2: Millipore 07–598, 1:2,000
NR2A: Sigma HPA004692, 1:2,500
NR2B: Millipore Ab1557P, 1:1,000
All statistical analyses were performed using the Prism 6 software package (GraphPad). Student’s t-tests were used for all pair-wise comparisons (indicated in Results where t value is given), and one-way ANOVAs were used for all multiple comparisons (indicated in results section where F value is given).
Many studies have indicated that MSNs in the NAc shell and core have different biochemical and physiological responses to chronic exposure to drugs of abuse (Kourrich and Thomas, 2009; Loweth et al., 2010) and that the two subregions differentially regulate drug-seeking behaviors (Ito et al., 2004). To determine the differential effects of cocaine on the protein constituents of NAc shell vs. core, we used Multiplexed Isobaric Tagging (iTRAQ) and tandem mass spectroscopy (MS/MS). Adult male rats were injected IP with cocaine (20 mg/kg) or saline daily for 7 days; 24 hr after the last injection, NAc shell and core were microdissected (Fig 1A) and flash frozen. Proteins in these samples were then quantified using iTRAQ. All four CaMKII isoforms displayed large increases in expression after cocaine treatment that were specific to NAc shell compared to core. Several protein phosphatases, including PP1 catalytic and regulatory subunits and PP2A, which have been previously associated with various CaMKII substrates in other systems (Colbran, 2004), followed a similar pattern. These findings provided novel, unbiased evidence that the CaMKII signaling pathway is prominently regulated by cocaine in NAc in a shell-specific manner.
To validate this finding more quantitatively, we treated rats as above with cocaine (at varying doses) or saline and measured locomotor responses to a cocaine (5 mg/kg) or saline challenge dose. Repeated exposure to 10 mg/kg cocaine resulted in the typical pattern of locomotor sensitization (Fig 1B). Further studies with this dosing regimen revealed, by use of Western blotting, that repeated cocaine induces CaMKIIα selectively in NAc shell 24 hr after the final injection of cocaine (Fig 1C and D; p=0.0019; F=7.943; df=29). In addition, phosphorylation of the canonical CaMKII substrate Ser831 of the GluA1 subunit of the AMPA receptor was significantly increased in NAc shell and not core (p=0.0261; F=4.208; df=28), while CaMKIIα Thr286 autophosphorylation had a strong but not significant trend toward induction in shell only (Fig 1D). Several other glutamate receptors were unaffected. In contrast to these measures of CaMKII, the same tissue samples displayed induction of ΔFosB in both shell (p=0.0260; F=4.189; df=29) and core (p=0.0350; F=3.807; df=29) of the NAc (Fig 1C and D), consistent with previous findings (Perrotti et al., 2008).
Since several prior studies of cocaine regulation of AMPA receptors analyzed animals after ~14 days of withdrawal from chronic cocaine (see Discussion), we repeated these biochemical analyses at this time point. We found that, 14 days after the final injection of cocaine, ΔFosB remains elevated in NAc (p=0.0288; F=4.258; df=22), while neither CaMKII nor phosphorylation of GluA1 Ser831 remains increased (Fig 1E). However, 1 hr after a single 10 mg/kg challenge dose of cocaine, levels of total CaMKII (p=0.0330; F=3.947; df=26) and of GluA1 Ser831 (p=0.0213; F=4.509; df=27) phosphorylation are both elevated to a degree similar to that found after initial chronic cocaine exposure (Fig 1E). These data indicate that NAc shell neurons are primed for CaMKII induction during extended periods of abstinence, perhaps via direct priming of the CaMKII gene promoter (see Discussion). Moreover, the fact that ΔFosB induction is more persistent than CaMKII induction suggests the existence of additional mechanisms, whether chromatin-based or otherwise, that exert a “brake” on CaMKII regulation, as covered in the Discussion.
To further strengthen these observations, we explored models of cocaine self-administration, which involve volitional drug intake. Adult male rats were given either short or long access to cocaine; as expected (Ahmed and Koob, 1998), only long access conditions led to escalating self-administration of the drug (Fig 2A). ΔFosB was induced to a greater extent by long vs. short access to cocaine in both NAc shell (p=0.0011; F=11.12; df=17) and core (p=0.0004; F=13.86; df=17). In contrast, CaMKIIα was induced in NAc shell only by long access to cocaine (Fig 2B and C; p=0.0236; F=4.957; df=16). It is interesting to compare the average daily cocaine intake across short-access animals (~12 mg/kg IV), long-access animals (~70 mg/kg IV), and experimenter-administered animals (10 mg/kg), and ask why the latter elicits robust induction of ΔFosB and CaMKII whereas short-access does not. This discrepancy is likely due to differences in peak cocaine levels (experimenter-administered cocaine is given as a single bolus IP, whereas self-administered cocaine is delivered via multiple IV doses), or by differences in length of drug exposure (7 days for experimenter administration, 19 days for self-administration).
Despite the large literature on ΔFosB and CaMKII in cocaine action, there are no studies of these proteins in human cocaine users. Here, we present the first evidence that levels of both ΔFosB (p=0.0316; t=1.921; df=34) and CaMKII (p=0.0444; t=1.755; df=32) are increased in NAc of cocaine-dependent humans (Fig 2D, Table 1). These data indicate that our examination of ΔFosB and CaMKII induction by cocaine in rodent NAc is clinically relevant to human cocaine addiction.
The finding that both CaMKII and ΔFosB are upregulated by cocaine in the rodent NAc led us to determine whether ΔFosB might regulate transcription of the CaMKII gene. We had previously reported CaMKIIα as a possible target for ΔFosB in an unbiased microarray analysis of NAc (McClung and Nestler, 2003), but this finding was not further validated in that study. We first used quantitative ChIP (qChIP—ChIP followed by quantitative PCR) to determine whether ΔFosB binds to the CaMKIIα gene promoter in NAc of adult male rats, and found strikingly that this binding is significantly increased, by chronic cocaine administration, in the shell (p=0.0133; t=2.901; df=12), but not the core, subregion (Fig 3A). To further understand the mechanisms related to this subregion-specific difference in ΔFosB binding to the CaMKIIα promoter, we used qChIP to characterize the state of histone modfications at this genomic region. Prior studies demonstrated cocaine induction of H3 acetylation at the CaMKIIα promoter in total mouse NAc (Wang et al., 2010). In contrast, we found that cocaine decreases H3 acetylation at the CaMKIIα promoter selectively in NAc core (Fig 3B; p=0.0213; t=2.726; df=10), with no change apparent in shell, consistent with subregion-specific chromatin alterations beyond ΔFosB binding. qChIP for the repressive mark, dimethylated H3 lysine 9 (H3K9me2), revealed trends for decreases in both the shell and core subregions (Fig 3C).
To determine whether ΔFosB regulates CaMKIIα transcription in vivo, we utilized two bitransgenic mouse lines that inducibly overexpress ΔFosB specifically in D1 vs. D2-type MSNs in a manner controlled by doxycycline administration in drinking water (Chen et al., 1998; Kelz et al., 1999; Werme et al., 2002). Adult male mice overexpressing ΔFosB solely in D1-type MSNs had significantly increased levels of CaMKIIα mRNA in NAc (p=0.0337; t=1.996; df=13), an effect not seen in mice overexpressing ΔFosB predominantly in D2-type MSNs (Fig 3D). The increase in CaMKIIα mRNA, induced by ΔFosB expression in D1-type MSNs, was accompanied by a concomitant increase in CaMKIIα protein in both NAc shell (p=0.0030; t=3.578; df=14) and core (p=0.0392; t=2.275; df=14; Figs 3E and F). These data demonstrate that ΔFosB is capable of driving CaMKIIα gene expression in D1-type MSNs in both subregions, although Figure 3B suggests that cocaine-mediated chromatin changes at the CaMKIIα promoter (e.g., reduced acetylation) prevent ΔFosB from upregulating CaMKII in the core subregion after cocaine.
Because our transgenic mouse data indicated that ΔFosB induction of CaMKII gene expression is specific to D1-type MSNs in NAc, we next sought to determine whether cocaine-dependent upregulation of CaMKII requires activation of the D1 dopamine receptor. Adult male rats were administered chronic cocaine or saline as before, but 30 min prior to each injection, rats in the cocaine group were given IP injection of saline, the D1 antagonist SCH 23390 (0.5 mg/kg), or the D2 receptor antagonist eticlopride (0.5 mg/kg). Animals were analyzed 24 hr after the last injection of cocaine. Western blotting revealed that the D1, but not the D2, antagonist completely blocked the cocaine-mediated increase in ΔFosB (p<0.0001; F=18.96; df=18), as previously reported (Nye et al., 1995), as well as in CaMKII (p=0.0005; F=10.99; df=18; Fig 3G and H). These data support the hypothesis that cocaine engages a ΔFosB-mediated increase in CaMKII gene expression specifically in D1-type MSNs of NAc shell. It would be important in future studies to demonstrate directly this cell-type specific effect of cocaine on CaMKII expression within this brain region.
To complement the use of bitransgenic mice, we next studied the role of ΔFosB in mediating cocaine induction of CaMKIIα by use of viral-mediated gene transfer in rats. We bilaterally injected adeno-associated viral (AAV) particles into NAc shell of adult male rats (where shell can be selectively targeted) to overexpress ΔFosB plus GFP or GFP alone. The animals were then given a single IP injection of 10 mg/kg cocaine. The animals overexpressing ΔFosB/GFP exhibited an increased locomotor response compared to animals overexpressing GFP alone (Fig 4A). 24 hr after the single cocaine injection, GFP-positive NAc tissue was excised from these animals by dissection under a fluorescent light source. Western blotting of this tissue (Fig 4B and C) revealed strong ΔFosB overexpression as well as a significant increase in total CaMKIIα protein compared to GFP animals (p=0.0070; t=2.894; df=30), similar to the induction seen with chronic cocaine administration. In addition, CaMKIIα autophosphorylation at Thr286 (indicative of enzyme activation) was increased by ΔFosB overexpression (p=0.0330; t=2.243; df=28), as was phosphorylation of the CaMKII substrate, Ser831 of GluA1 (p=0.0540; t=2.012; df=28), again mimicking the actions of chronic cocaine (Fig 1C and D). Taken together, these data provide further evidence that ΔFosB expression in NAc shell is sufficient for locomotor sensitization to cocaine and for CaMKII induction and activation in this subregion.
We used a similar approach to determine whether ΔFosB is also necessary for cocaine-mediated induction of CaMKIIα in the NAc shell. AAV was used to overexpress a truncated JunD protein, termed ΔJunD, which is a negative regulator of ΔFosB transcriptional activation (Winstanley et al., 2007) plus GFP or GFP alone. Two weeks later, when transgene expression is maximal, animals were given cocaine (10 mg/kg) or saline daily for 7 days, and tested for locomotor responses to a cocaine challenge (5 mg/kg) 24 hr after the last chronic injection (Fig 4D). ΔJunD overexpression prevented locomotor sensitization to cocaine, and also prevented CaMKIIα induction and activation in NAc shell (Fig 4E and F; p=0.0437; F=2.997; total df=38), indicating that ΔFosB transcriptional activity is necessary for cocaine-mediated induction of CaMKIIα in this subregion. Interestingly, we found that ΔJunD reduced levels of ΔFosB under both saline and cocaine-treated conditions (p=0.0004; F=8.110; df=35), raising the novel possibility that ΔFosB depends on AP-1 activity for its own expression levels.
Using in vitro protein kinase assays, we determined that purified ΔFosB is a robust substrate for CaMKIIα. Incubation of His6-ΔFosB with CaMKIIα and ATP caused an upward shift in electrophoretic mobility of ΔFosB (Fig 5A); the several resulting bands suggested multiple sites of phosphorylation. Similar in vitro kinase assays using [γ-32P]ATP showed incorporation of radiolabeled phosphate into the shifted ΔFosB bands (Fig 5B), demonstrating direct phosphorylation of the protein. We generated a phospho-specific antibody to the previously characterized Ser27 of ΔFosB (Ulery et al., 2006). While this antibody does not produce a signal against brain extracts that contain Ser27-phosphorylated ΔFosB (data not shown), we were able to detect Ser27 phosphorylation in the in vitro kinase assay using CaMKII (Fig 5B). Kinetic analyses of the CaMKII phosphorylation of ΔFosB indicate that it is a potent substrate for the kinase (Fig 5C), with an apparent KM of 5.7±2.0µM and KCAT of 2.3±0.3min−1. These results are comparable to many well-characterized in vivo substrates of CaMKII (Colbran and Brown, 2004). In addition, we determined that CaMKII phosphorylates ΔFosB with a stoichiometry of 2.27±0.07 mol/mol (Fig 5D), indicating that there are at least three sites of CaMKII phosphorylation within the His6-ΔFosB protein, in agreement with Fig 5A.
To investigate individual sites of phosphorylation, we employed MS analyses of samples from our in vitro kinase assays. Fig 5E demonstrates ΔFosB phosphorylation at the previously characterized Ser27 and at several additional sites (data not shown). Given the prior functional characterization of Ser27, we focused on this site by generating labeled synthetic peptides mimicking the phospho- and nonphospho-states of Ser27, then used known quantities of these peptides as standards in MRM analyses of ΔFosB before and after in vitro phosphorylation by CaMKII. Subsequent quantitation (Fig 5F) confirms that Ser27 is a potent substrate for CaMKII. These results indicate that, among multiple phosphorylated residues within ΔFosB, Ser27 is a particularly effective substrate for CaMKII.
Since CaMKII can phosphorylate ΔFosB in vitro at a site that dramatically enhances its stability in vitro and in vivo (Ulery et al., 2006; Ulery-Reynolds et al., 2009), we determined whether CaMKII activity controls ΔFosB levels in NAc in vivo. To address this question, we first used a mouse line overexpressing a calcium-independent mutant of CaMKIIα (T286D) in multiple brain regions including NAc (Mayford et al., 1996; Kourrich et al., 2012). We injected age-matched adult male mutant and wildtype littermates with 20 mg/kg cocaine or saline once daily for 14 days, then analyzed the animals one day after the final injection. We found that basal levels of ΔFosB were increased in the mutant animals in NAc shell (p=0.0001; F=9.207; df=37), but not core (Fig 5G and H). Surprisingly, cocaine-dependent induction of ΔFosB was blocked in the mutant animals in both shell and core, suggesting that, although CaMKII may directly regulate ΔFosB stability in NAc shell, it may also lie upstream of ΔFosB in cocaine-activated pathways in both NAc subregions.
Cocaine induction of dendritic spines on NAc MSNs is one of the best established drug-induced adaptations in this brain region, and such spine induction has been correlated with sensitized behavioral responses to the drug (Robinson and Kolb, 2004; Russo et al., 2010) and reported to be selective for D1-type MSNs (Lee et al., 2006). We have demonstrated recently that cocaine induction of dendritic spines in NAc is dependent on ΔFosB and its downstream transcriptional program (Maze et al., 2010). Though there is an extensive literature concerning the involvement of CaMKII in dendritic spine morphology and induction in other brain regions and experimental systems (Jourdain et al., 2003; Penzes et al., 2008; Okamoto et al., 2009), its role in NAc MSN spine formation has not been studied. Therefore, we determined whether CaMKII activity is required for ΔFosB-mediated induction of MSN dendritic spines by utilizing HSV-mediated overexpression of the CaMKII inhibitor peptide AC3I fused to GFP, a construct previously shown to inhibit CaMKII activity in vivo (Zhang et al., 2005; Klug et al., 2012). Viral overexpression of ΔFosB in NAc shell of adult mice induced a significant increase in MSN dendritic spine density (p<0.0001; F=8.558; df=59; Fig 6A and B) as previously reported (Maze et al., 2010), and this increase was driven primarily by thin (p=0.0027; F=5.319; df=59) and stubby (p=0.0378; F=2.988; df=59) spine types (both thought to be immature spines) (Fig 6C–E). No effect was seen on more mature, mushroom-shaped spines. However, when GFP-AC3I was coexpressed, ΔFosB induction of spines was completely abrogated (Fig 6A–E), indicating that CaMKII activity is required for ΔFosB induction of dendritic spines in NAc shell.
We next used the same viral tools to determine whether CaMKII activity is required for ΔFosB’s effects on behavioral sensitivity to cocaine. 72 hr after viral injection into NAc shell, animals were given a single injection of 5 mg/kg cocaine and their locomotor activity was recorded. As previously shown with more extended AAV overexpression of ΔFosB (Fig 4A), HSV-mediated overexpression of ΔFosB increased locomotor sensitivity to cocaine (p=0.0002; F=8.823; df=37; Fig 6F). As with induction of dendritic spines, inhibition of CaMKII activity by coexpression of GFP-AC3I completely blocked the ΔFosB-mediated increase in cocaine sensitivity, indicating that CaMKII activity is required for ΔFosB-induced alterations in cocaine’s behavioral effects.
The present study delineates a novel feed-forward mechanism where cocaine induces ΔFosB in NAc, which upregulates transcription of the CaMKIIα gene selectively in NAc shell. CaMKIIα subsequently phosphorylates and stabilizes ΔFosB leading to greater ΔFosB accumulation and to further CaMKIIα induction (Fig 6G). The co-escalating levels of the two proteins during chronic exposure to cocaine then contribute in essential ways to sensitized behavioral responses to the drug. This is a particularly appealing hypothesis as both ΔFosB and CaMKII have each been demonstrated previously to be required for increased behavioral responses to cocaine (Pierce et al., 1998; Peakman et al., 2003), and we replicate this finding for ΔFosB in NAc shell specifically using a viral approach (Figs 4 and and66).
Although transgenic ΔFosB overexpression in D1-type MSNs can drive CaMKII induction in both NAc shell and core of cocaine-naïve animals, in the context of cocaine, accumulation of endogenous ΔFosB, which occurs in both subregions, drives induction of CaMKII specifically in NAc shell. This difference could relate to the higher levels of ΔFosB induced in our bitransgenic model, however, it might also reflect the ability of cocaine to differentially alter the CaMKIIα promoter in shell vs. core MSNs to either promote ΔFosB binding in the former or exclude it in the latter subregion. In fact, our ChIP data, which reveal a cocaine-mediated deacetylation of histones at the CaMKIIα gene promoter in NAc core only, support the possible involvement of a chromatin mechanism. In keeping with this hypothesis, ΔFosB overexpression in D1-type MSNs was able to drive CaMKIIα induction in NAc core in the absence of cocaine (Fig 3F), suggesting that there are active modifications of the CaMKIIα promoter that prevent this induction during chronic cocaine exposure. Regulation of the chromatin landscape at the CaMKII promoter might also explain why CaMKII is induced by a challenge dose of cocaine in NAc shell of chronic cocaine-withdrawing rats (Fig 1E) but not of drug-naïve animals (Fig 1D). This could represent an epigenetic “gene priming” effect of ΔFosB (Robison and Nestler, 2011), and might thus be one molecular mechanism of the incubation of cocaine craving (Pickens et al., 2011). However, for this chromatin change to be causally linked to incubation of craving, it would have to increase over time. It will be interesting to determine if this is the case, and to study whether other genes show ΔFosB-dependent, subregion-specific regulation by cocaine. It is also important to note that the feed-forward loop we describe does not lead to an endless accumulation of CaMKII or ΔFosB (Fig 1E); uncovering the molecular “brake” responsible for this is an important goal of future studies.
The known functions of ΔFosB and CaMKII in several experimental systems and brain regions converge at many levels (Fig 6F). Both molecules are intimately linked to dendritic spine growth: CaMKII interacts with the actin cytoskeleton (Okamoto et al., 2009), regulates spine head size (Matsuzaki et al., 2004), and is both necessary and sufficient for plasticity-induced increases in filopodia and synapse number in hippocampal organotypic slice cultures (Jourdain et al., 2003), while ΔFosB is both necessary and sufficient for cocaine-induced dendritic spine formation in NAc MSNs (Maze et al., 2010). Additionally, both molecules have been associated with regulation of AMPA glutamate receptors. CaMKII does not regulate total levels of AMPA receptor subunits, but drives the insertion of AMPA receptors into synapses and increases AMPA channel conductance by phosphorylating GluA1 at Ser831 in hippocampal pyramidal neurons in culture and in vivo (reviewed in (Malinow and Malenka, 2002; Colbran and Brown, 2004)). Such increased trafficking of GluA1 to the synapse has been implicated in chronic cocaine action as well (Boudreau and Wolf, 2005). Moreover, behavioral responses to AMPA receptor activation in NAc are enhanced by CaMKIIα overexpression in a D1 dopamine receptor-dependent manner (Singer et al., 2010). Long-term D1-specific overexpression of ΔFosB has been shown to induce GluA2 transcription in NAc (Kelz et al., 1999), which dampens AMPA responses mediated via GluA1, while we show here that shorter-term ΔFosB overexpression—as well as shorter-term cocaine exposure—have no effect on this subunit (Fig 1). Nevertheless, we have found recently that short-term ΔFosB overexpression nevertheless reduces AMPA responses in D1-type MSNs in NAc (Grueter et al., 2013). These data suggest temporally distinct mechanisms that might constitute a time-dependent series of neuroadaptations to cocaine that underlie different aspects of addiction progression not yet well understood. At the behavioral level, both CaMKII and ΔFosB are required for locomotor sensitization to cocaine (see above), and both are required for sustained cocaine self-administration in rodents (Colby et al., 2003; Wang et al., 2010), suggesting that the two proteins are important for both short- and long-term behavioral adaptations to drug exposure, albeit via partly distinct underlying mechanisms. Presumably, ΔFosB and CaMKII regulate such complex behavioral adaptations through changes in NAc synaptic function, although much further work is needed to directly link synaptic phenomena to behavioral change.
The CaMKII holoenzyme simultaneously interacts with a variety of synapse-associated proteins (Robison et al., 2005) which are thought to regulate its targeting to the postsynaptic density (PSD), a phenomenon suggested to be important for synaptic plasticity. In particular, the interaction of CaMKII with the GluN2B subunit of the NMDA-type glutamate receptor was recently shown to regulate both synaptic plasticity and learning (Halt et al., 2012). While the AC3I peptide mimics the autoinhibitory domain of CaMKII, and thus inhibits enzyme catalytic activity, it also blocks multiple protein-protein interactions (Strack et al., 2000; Robison et al., 2005). Thus, the behavioral and morphological effects of HSV-GFP-AC3I reported here could occur through reduced phosphorylation of CaMKII target proteins, changes in CaMKII targeting, or a change in CaMKII’s proposed structural role at synapses (Lisman et al., 2002).
The restriction of the proposed ΔFosB-CaMKII loop to the NAc shell is of special note, as recent work has demonstrated several physiological differences between the NAc shell and core in response to cocaine administration, a notion confirmed by our unbiased iTRAQ (Table S1) data. MSNs in NAc shell show a depression in firing capacity after chronic cocaine that is sustained for weeks, while core MSNs from the same animals display a transient (1–3 day) increase in firing capacity that returns to basal levels within 2 weeks (Kourrich and Thomas, 2009). In addition, numerous synaptic proteins are differentially regulated in NAc shell vs. core of animals exposed to chronic cocaine, including GluA2 (Knackstedt et al., 2010). As chronic amphetamine induces CaMKIIα specifically in NAc shell (Loweth et al., 2010), it is not surprising that we find a similar effect with cocaine. However, as ΔFosB is induced in both the NAc shell and core by chronic cocaine (Perrotti et al., 2008), and since we show that CaMKIIα induction in shell is ΔFosB-dependent, our findings provide new evidence for distinct transcriptional mechanisms at the CaMKIIα promoter between these two subregions, which are responsible for the selective induction of CaMKIIα in shell.
A great deal of recent work has focused on delineating differences between D1- and D2-type NAc MSNs. Though both D1 and D2 receptors are involved in the rewarding effects of cocaine (Self, 2010), recent work demonstrates that optogenetic activation of D1-type MSNs increases behavioral responses to cocaine, while D2-type MSN activation has the opposite effect (Lobo et al., 2010). In line with these findings, D1-receptor knockout mice are deficient in acquisition of cocaine self-administration (Caine et al., 2007), while D2 knockouts are not (Caine et al., 2002). D1 agonist administration directly into NAc triggers cocaine-seeking behavior in reinstatement paradigms (Self, 2010). Interestingly, this effect requires D1-receptor-dependent increases in CaMKII activity in the NAc shell, but not core (Anderson et al., 2008), a result that dovetails nicely with the D1- and shell-specific ΔFosB-CaMKII loop proposed here.
We previously reported that Ser27 in ΔFosB can be phosphorylated by casein kinase-2 (Ulery et al., 2006), however, we establish here that CaMKII phosphorylates ΔFosB at this and other sites with far greater kinetics and stoichiometry and can replicate the higher apparent Mr observed for ΔFosB (Fig 5A) with cocaine exposure in vivo (Nestler, 2008). We already know that Ser27 phosphorylation increases ΔFosB stability and transcriptional activity (Ulery et al., 2006; Ulery and Nestler, 2007; Ulery-Reynolds et al., 2009). Future work will now focus on the identification and the functional consequences of novel sites of ΔFosB phosphorylation indicated by the present study.
The feed-forward loop described here provides a plausible new mechanism by which repeated administration of cocaine drives progressive abnormalities in the NAc. As such, this biochemical pathway may provide an important target for future therapeutic intervention in addictive disorders. Because CaMKII is ubiquitous and required for many basal neuronal and behavioral functions, direct use of CaMKII inhibitors has been avoided as an addiction treatment. Our data suggest that more subtle targeting of the mechanism of CaMKII induction, which is specific to an individual cell type and subregion of the brain’s reward circuitry, could provide a therapeutic target that would avoid the complications of systemic CaMKII inhibition.
This work was supported by grants from the National Institute on Drug Abuse (EJN), NIDA-Yale Proteomics Center DA018343 (AJR and EJN), and Hartwell Foundation (AJR). The authors would like to thank Gabby Rundenko for the generous gift of purified ΔFosB and Roger Colbran for the generous gift of purified CaMKIIα.