CREB-mediated alterations in the amygdala transcriptome: coordinated regulation of immune response genes following cocaine

Abstract

The neuronal circuitry underlying stress- and drug-induced reinstatement of cocaine-seeking has been relatively well characterized; however, less is known regarding the long-term molecular changes following cocaine administration that may promote future reinstatement. The transcription factor cAMP response element-binding protein (CREB) is necessary for stress- but not cocaine-induced reinstatement of conditioned reward, suggesting that different molecular mechanisms may underlie these two types of reinstatement. To explore the relationship between this transcription factor and reinstatement, we utilized the place-conditioning paradigm to examine alterations in gene expression in the amygdala, a neural substrate critically involved in stress-induced reinstatement, following the development of cocaine reward and subsequent extinction. Our findings demonstrate that the amygdala transcriptome was altered by CREB deficiency more than by previous cocaine experience, with an over-representation of genes involved in the immune response. However, a subset of genes involved in stress and immune response demonstrated a drug×genotype interaction, indicating that cocaine produces different long-term alterations in gene expression depending on the presence or absence of CREB. This profile of gene expression in the context of addiction enhances our understanding of the long-term molecular changes that occur throughout the addiction cycle and identifies novel genes and pathways that might lead to the creation of better therapeutic agents.

Conditioned place preference, CREB, gene expression, immune response, stress

Introduction

Drug addiction is a psychiatric disorder characterized by a transition from recreational to compulsive drug use that continues in spite of severe negative consequences (O'Brien, 2003). Despite attempts by individuals to quit, the desire or need to resume drug-taking can last for months or years (Sinha +6; Li, 2007). The persistence of addiction over time and the propensity to relapse suggest that exposure to drugs results in long-term adaptations in the brain that probably involve alterations in transcription and genetic regulation.

Behavioural reinstatement paradigms are used to identify factors and brain regions that underlie relapse to drug-seeking induced by exposure to drug, drug-associated cues, or stressors. Our laboratory uses the place-conditioning paradigm to study stress-induced reinstatement and has previously demonstrated that acute exposure to forced swim stress (FSS) induces reinstatement of cocaine conditioned place preference (CPP) in wild-type mice (Kreibich +6; Blendy, 2004). In addition, cocaine conditioning increases phosphorylation of the transcription factor CREB (pCREB) in the amygdala upon re-exposure to the CPP boxes on reinstatement day, and acute exposure to FSS at this time further augments these changes. Last, mice deficient in CREB protein show deficits in FSS-induced reinstatement of CPP. In contrast, they show robust cocaine-induced reinstatement. This deficit in stress- but not drug-induced reinstatement indicates a specific requirement for CREB in stress-induced behavioural responses to drugs of abuse. Thus, CREB-induced changes in the amygdala may occur following drug administration and act to promote stress-induced reinstatement in the place-conditioning paradigm.

While the signalling pathways controlled by CREB have been strongly implicated in drug reward and relapse, the identification and regulation of CREB target genes has been less well studied. A variety of candidate genes have been identified that may promote and maintain addictive behaviours (Lu et al.2003), including ΔFosB (Ang et al.2001; Chen et al.1995; McClung +6; Nestler, 2003), brain-derived neurotrophic factor (BDNF) (Grimm et al.2003; Lu et al.2004), dynorphin (Redila +6; Chavkin, 2008; Shippenberg et al.2007) and corticosterone-releasing factor (CRF) (Maj et al.2003; Sarnyai et al.2001; Zorrilla et al.2001). Of interest, several of these (BDNF, dynorphin, CRF) are potential CREB target genes based on the presence of a CRE element in their promoter DNA sequences (Briand +6; Blendy, 2010).

To date most studies examining large-scale changes in gene expression that occur following drug exposure have been focused on the mesolimbic dopamine system, which includes the ventral tegmental area (VTA) and the nucleus accumbens (NAc) (Freeman et al.2010; Krasnova et al.2008; McClung +6; Nestler, 2003; Yuferov et al. 2003, 2005). The VTA and NAc are critically important in initial reward (Pierce +6; Kumaresan, 2006; Roberts et al.1977; Wise, 2004) as well as reinstatement of drug-seeking (Kalivas +6; McFarland, 2003; McFarland et al.2004; McFarland +6; Kalivas, 2001). However, the amygdala has been shown to be particularly important in mediating stress-induced reinstatement (Erb et al.2001; Leri et al.2002; McFarland et al.2004), and the long-term changes that occur in this brain region during abstinence and reinstatement have been largely unexplored.

In the current study, we aimed to generate a broader inventory of genes that may underlie the prolonged persistence of addictive behaviours and identified novel targets involved in this cycle of addiction. We utilized expression profiling to examine changes in gene expression following cocaine administration, which might contribute to the reinstatement of place preference. We focused our gene analysis on the amygdala of wild-type and CREBαΔ mice following cocaine conditioning and extinction but prior to reinstatement. This time-point was chosen to identify gene changes that occur during extinction and could potentially underlie subsequent reinstatement behaviour elicited by a stressor, while avoiding the acute changes in gene expression that occur following stress exposure. A number of novel genes that are classically involved in stress and immune response were identified. These findings may lead to a better understanding of the long-term genetic alterations in the amygdala that act to promote reinstatement behaviour.

Materials and methods

Animals

Mice (aged 3–6 months; 20–40 g; mixed sexes) were group-housed in a 21°C humidity-controlled animal facility approved by the AAALAC (Association for Assessment and Accreditation of Laboratory Animal Care International) with food and water available ad libidum. All experiments were performed in accordance with NIH guidelines for ‘Guiding principles in the care and used of animals’.

Both wild-type and CREBαΔ mice were maintained as a 129SvEvTac:C57BL/6J F1 hybrid strain, obtained from crossing mice heterozygous for the CREB mutation from each parental strain. Both parental strains (129SvEvTac and C57BL/6J) had been backcrossed with vendor-supplied wild-type mice for >20 generations.

Place conditioning procedure

Place conditioning boxes were divided into two sides (20×20×20 cm); one side consisted of a striped wall with plastic flooring, and the other side, solid grey-coloured walls with a metal grid floor. The solid side was illuminated throughout the 10-d paradigm, while the striped side was dark.

Preconditioning day

Mice were placed on one side of the box and allowed to roam freely between both sides for 900 s. Time spent on each side was recorded and data were used to separate the mice into groups that had average bias on each side.

Pairing days

For days 2–9, mice underwent conditioning, with the cocaine group receiving cocaine (20 mg/kg i.p.; NIDA Drug Supply, USA) on one side of the box and saline on the other, and the saline group receiving saline on both sides of the apparatus.

Test day

Mice were given a saline injection, placed into the box, and allowed to roam freely between both sides for 900 s. Time spent on each side was recorded, and data were expressed as time spent on the paired side minus time spent on the unpaired side.

Extinction

For days 11–22, mice were given saline injections on both sides of the conditioning boxes.

Extinction test day

On day 23, mice were given a saline injection, placed into the box and allowed to roam freely between both sides for 900 s. Time spent on each side was recorded and data were analysed using a two-way ANOVA.

Brain dissections

Mice were killed 24 h following extinction test by cervical dislocation and brains were rapidly removed and dissected on ice. Brains were first sliced into 1-mm slices using a mouse brain matrix (Braintree Scientific, USA) and specific regions were identified using coordinates from the mouse stereotaxic atlas (amygdala: bregma −1.2 mm; cortex: bregma 0.15 mm) (Franklin +6; Paxinos, 2007). All amygdala subregions, such as the central and basolateral nuclei, were included in the dissection. Tissue was then macrodissected and immediately frozen in liquid nitrogen.

Expression profiling

RNA was extracted from brain tissue by homogenization in 800 µl Trizol and 160 µl chloroform. Samples were sedimented at 13 000 rpm for 15 min and the aqueous layer removed. RNA was purified using an RNeasy Mini kit (Qiagen, USA). RNA concentration was determined using a Nanodrop spectrometer (Nanodrop Technologies, USA) and quality was evaluated using a RNA NanoChip on the Bioanalyzer (Agilent Technologies, USA). Next, 1000 ng total RNA was amplified and labelled with Cy3 using the Low RNA Input Linear Amp kit PLUS, One-Color (Agilent Technologies). After purification, 1.65 µg cRNA was fragmented and hybridized to the Whole Mouse Genome Oligo Microarray G4122A (Agilent Technologies) for 17 h at 65°C. One animal per group was hybridized to one array tile (n=4 per group). Following hybridization, the slides were washed and scanned with an Agilent G2565BA microarray scanner. Images were analysed with Feature Extraction 9.5 (Agilent Technologies). Mean foreground intensities were obtained for each spot and imported into the mathematical software package ‘R’, which normalized the data using Limma quantile normalization (Smyth 2004, Bolstad et al.2003). Complete statistical analysis was then performed in ‘R’ using both the LIMMA (linear models for microarray data) and SAM (significance analysis of microarrays) packages. Hierarchical clustering was performed on the samples using ‘pvclust’ (Suzuki +6; Shimodaira, 2006). This package calculates p values for hierarchical clustering via multiscale bootstrap resampling. Differentially regulated genes were determined for four comparisons: wild-type (WT)/cocaine (Coc)–WT/saline (Sal), CREBαΔ mutant (MT)/Coc–MT/Sal, MT/Coc–WT/Coc, and MT/Sal–WT/Sal. Further analysis of the dataset to determine functional classes of these genes was completed using the UCSC mouse genome browser and EntrezGene.

Functional analysis of genes

GO (Gene ontology) biological functions were determined from gene lists in each array by NIH DAVID (Dennis et al.2003; Hosack et al.2003). Significance was determined using a modified Fisher's exact test (EASE) score (Hosack et al.2003).

Quantitative real-time polymerase chain reaction (qPCR)

To verify the differentially expressed genes found in the microarray, a separate cohort of mice were taken through an identical behavioural paradigm and their brain tissue obtained as described for the array; however, no amplification was required for the biological validation of gene expression changes using qPCR. RNA (500 ng) was used for cDNA synthesis using 1 µg oligo(dT) primer (Operon, USA) and Superscript II reverse transcriptase with its accompanying reagents (Invitrogen, USA). All qPCR reactions were run using the Stratagene MX3000 and the MXPro qPCR software. Reactions were assembled using Applied Biosystems 2× SYBR-Green master mix along with 300 nm primers (final concentration) in accordance with the manufacturer's instructions, except that the total reaction volume was scaled down to 25 µl. Cycling parameters were 95°C for 10 min and then 40 cycles of 95°C (30 s) and 60°C (1 min), followed by a melting curve analysis. All reactions were performed in triplicate and the median cycle threshold was used for analysis. Cycle threshold values were normalized to TATA binding protein (TBP). Two-way ANOVAs were used to confirm significance and direction of fold changes predicted by the array. Primer sequences are available upon request.

Results

Both wild-type and CREBαΔ mutant mice show development and extinction of cocaine conditioned place preference as previously demonstrated

Following conditioning, both wild-type and CREBαΔ groups developed preference for the cocaine-paired context (two-way ANOVA: F_1,28=33.96, p<0.0001; Bonferroni post-hoc: * p<0.01 from corresponding saline-treated animals), and this preference was no longer evident following 12 d of extinction (Fig. 1). Mice were killed 24 h following extinction test day but prior to reinstatement in order to identify candidate genes whose expression might be changed throughout conditioning and extinction, while avoiding the confounding effects of an acute stressor on gene expression.

Fig. 1

Experimental timeline and behavioural data associated with microarray study. Both CREBαΔ mutant (MT) and wild-type (WT) mice were tested in the CPP paradigm to examine both reward and extinction of the reward behaviour. Twenty-four hours following the extinction test day, mice were sacrificed for microarray analysis. Both WT and MT mice demonstrated place preference to 20 mg/kg cocaine (significant effect of drug on test day: F_1,28=33.96, p<0.0001, two-way ANOVA, * p<0.01 from corresponding saline-treated animals, Bonferroni post-hoc). Following extinction, WT and MT mice no longer showed preference for either side of the CPP chamber (not significant, two-way ANOVA). These data are similar to results observed in the second cohort of mice used for the biological replication of the microarrays. Data are expressed as mean±s.e.m. (n=8 per group).

CREB genotype plays a key role in the expression of genes in the amygdala

Microarray analysis was performed on the amygdala of both wild-type (WT) and CREBαΔ mutant (MT) mice exposed to the place-conditioning paradigm. Comparisons were designed to determine the effect on gene expression of genotype (MT/Sal–WT/Sal), drug treatment in wild-type (WT/Coc–WT/Sal) and mutant mice (MT/Coc–MT/Sal), and the interaction between drug and genotype (MT/Sal–MT/Sal vs. MT/Coc–WT/Coc) (Fig. 2a). Hierarchical clustering revealed a significant impact of the CREB genotype on gene expression, with 807 genes differentially expressed in the MT/Sal–WT/Sal comparison (Fig. 2b). More than half of the genes found in the MT/Coc–MT/Sal comparison (24/47 genes), as well as a majority of genes in the MT/Coc–WT/Coc comparison (21/29 genes), were identical to differentially expressed genes found in the MT/Sal–WT/Sal comparison, further supporting the observation that CREB plays a significant role in gene expression in this brain region (Fig. 2b).

Fig. 2

Overview of results found in microarray examining the effects of the CREBαΔ genotype following extinction of cocaine place preference. (a) Multiple comparisons were made between all four experimental groups with each bracket representing one comparison. (b) A Venn diagram representing the number of differentially expressed genes and their overlap with other pairwise comparisons (fold change >1.4, 20% false discovery rate as determined by significance analysis of microarrays). WT, Wild-type; Sal, saline, Coc, cocaine; MT, CREBαΔ mutant.

Clustering analysis also indicated that cocaine treatment did not have a significant effect on basal gene expression in the amygdala, as mice given cocaine did not cluster separately from those receiving saline, regardless of genotype. SAM further supported this finding as no genes were differentially expressed in the WT/Coc–WT/Sal comparison (Fig. 2b, see also Supplementary Table 1). However, 47 genes were differentially expressed by SAM analysis in the MT/Coc–MT/Sal comparison, indicating that prior cocaine treatment in the mutants, but not the wild-type mice, influences gene transcription following extinction of cocaine preference.

Table 1 lists the top 10 differentially expressed genes found in each comparison (for full list of genes, see Supplementary Table 1). In the MT/Sal–WT/Sal comparison, there is both an up-regulation and a down-regulation in the expression of various genes in the mutant animals. This indicates that while CREB is a positive regulator of transcription and thus reduced levels of this protein are likely to down-regulate gene expression, compensatory mechanisms may come into play to increase expression of other genes. Furthermore, CREBαΔ mutants treated with cocaine exhibited down-regulated gene expression compared to wild-type mice treated with cocaine. However, many of these genes were also down-regulated to a similar degree in the MT/Sal–WT/Sal comparison (Supplementary Table 1), suggesting that this down-regulation is simply due to the absence of CREB and is not affected by cocaine. Last, mutants treated with cocaine exhibited down-regulated gene expression compared to mutants treated with saline, further supporting the observation that cocaine treatment alters gene expression only in mutant animals as there were no differences in the WT/Coc–WT/Sal comparison.

Table 1

List of top 10 up-regulated and down-regulated genes in each comparison (fold change>1.5, FDR 20%)

Coc, Cocaine; MT, CREBαΔ mutant; Sal, saline; WT, wild-type.

Table 2 lists the top 10 differentially expressed genes similar between two different comparisons: MT/Coc–WT/Coc vs. MT/Sal–WT/Sal and MT/Sal–WT/Sal vs. MT/Coc–MT/Sal. The first comparison demonstrates that lack of CREB causes comparable down-regulation in expression of these genes regardless of drug treatment. The second comparison again highlights a more interesting change: saline-treated mutants demonstrate increased expression of certain genes compared to saline-treated wild-types and cocaine-treated mutants. This suggests that at baseline, mutants show greater expression of these genes and that this enhanced expression is diminished following cocaine conditioning and extinction.

Table 2

List of top 10 up-regulated and down-regulated genes similar between comparisons at extinction time point (fold change>1.5, FDR 20%)

Coc, Cocaine; MT, CREBαΔ mutant; Sal, saline; WT, wild-type.

Analysis of gene functions in the genotype comparison: MT/Sal–WT/Sal

GO was used to analyse patterns of functionality among the differentially expressed genes in the MT/Sal–WT/Sal comparison (Table 3). The highest-scoring biologically relevant category found to be over-represented in the genotype comparison was ‘immune response’ (p<1.19×10⁻⁶, EASE score), followed by ‘transcription’ as well as ‘G-protein-coupled receptor binding’. Of interest, analysis of putative CRE sites in the promoters of differentially expressed genes in this comparison demonstrated that over half did not have a CRE site in their promoter (data not shown), indicating that many of these genes are likely to be regulated by CREB indirectly. As there were very few genes differentially expressed in the other group comparisons, analysis of the GO biological function was limited and produced no significant results.

Table 3

Functional analysis of gene changes: MT/Sal compared to WT/Sal

MT, CREBαΔ mutant; Sal, saline; WT, wild-type.

Verification by qPCR of genes found to be differentially expressed in the amygdala

Sixteen candidate genes from a range of functional categories were sampled from the three lists of comparisons that demonstrated changes in gene expression: MT/Sal–WT/Sal, MT/Coc–WT/Coc, and MT/Coc–MT/Sal (Table 4; Supplementary Table 1). All four group comparisons for each of these 16 genes were consolidated to better visualize how CREB and cocaine alter patterns of gene expression (Table 5). The fold-change values across all four comparisons suggested that while many of the genes demonstrated a genotype effect (typically decreased expression in CREBαΔ mice regardless of drug experience), some of the genes demonstrated a possible drug×genotype interaction. Some of these comparisons are not represented in other tables either because their fold change was <1.5 or their false discovery rate (FDR) was >20%. Changes in these 16 candidate genes were then verified by qPCR in a separate cohort of mice that underwent the same drug conditioning and extinction paradigm.

Table 4

Sixteen genes representing a range of functional categories sampled from the four group comparisons on the microarray

Table 5

Validation of the 16 genes of interest

Coc, Cocaine; MT, CREBαΔ mutant; Sal, saline; WT, wild-type.

Trends were predicted based on the four pairwise comparisons and fold changes of the microarray. Genes were validated in a new cohort of mice using two-way ANOVAs. Genes that were validated in the new cohort are highlighted.

Overall, the validation rate of the microarray in a second cohort of mice, which served as the biological replicate, was 75% (see Table 5 where validated genes are highlighted). Genes that indicated a genotype effect and were validated as such included Camk1 (calcium/calmodulin-dependent protein kinase I), Crem (cAMP responsive element modulator), Dhrs7 (dehydrogenase/reductase member 7), IL18 (interleukin-18), Pecr (peroxisomal trans-2-enoyl-CoA reductase), Sln (sarcolipin), and Zfp367 (zinc finger protein 367) (Fig. 3). CREBαΔ mice displayed decreased gene expression regardless of drug experience for all genes except Crem (data not shown), which was, and has previously been shown to be, up-regulated in CREBαΔ mice (Blendy et al.1996). Both Dhrs7 and Pecr are genes that appeared to be unique to the MT/Coc–WT/Coc comparison. However, when qPCR was used to verify the microarray results these genes were shown to be down-regulated in the absence of CREB regardless of drug treatment. This makes sense in light of the comparisons in Table 5, which demonstrate a −1.3-fold change for both genes in the MT/Sal–WT/Sal comparison. This difference did not appear on the MT/Sal–WT/Sal comparison list of 807 genes, which used a fold-change cut-off value of 1.4 (Supplementary Table 1), emphasizing the importance of analysing the trend in all four group comparisons when attempting to understand how these genes are regulated by both CREB as well as cocaine conditioning (Table 5). Genes that suggested an interaction effect and were validated as such included Avpr1a (arginine vasopressin receptor 1A), IL10 (interleukin-10), Sgk2 (serum/glucocorticoid regulated kinase 2), Stat1 (signal transducer and activator of transcription 1), and Tnfrsf1b (tumour necrosis factor receptor superfamily, member 1b) (Fig. 4). These genes are all involved in stress and immune response and they all demonstrate the same pattern of expression; saline-treated CREBαΔ mice exhibited increased expression of these genes relative to wild-type mice, and treatment with cocaine abolished this difference.

Fig. 3

Biological replication by qPCR of genes in the amygdala that demonstrated a genotype effect in the microarray. All data are expressed as mean±s.e.m with n=7–8 per group. Two-way ANOVAs were used to determine statistical significance. Camk1 (calcium/calmodulin-dependent protein kinase I): overall effect of genotype (F_1,24=47.44, *** p<0.0001) and drug (F_1,24=5.12, * p<0.05). Dhrs7 (dehydrogenase/reductase member 7): effect of genotype (F_1,25=16.63, *** p<0.001). IL18 (interleukin-18): effect of genotype (F_1,27=67.33, *** p<0.0001). Pecr (peroxisomal trans-2-enoyl-CoA reductase): effect of genotype (F_1,26=13.73, ** p<0.001). Sln (sarcolipin): effect of genotype (F_1,24=86.51, *** p<0.0001) and drug (F_1,25=4.927, * p<0.05). Zfp367 (zinc finger protein 367): effect of genotype (F_1,25=302.6, *** p<0.0001). TBP, TATA binding protein.

Fig. 4

Biological replication by qPCR of genes in the amygdala that demonstrated a drug×genotype interaction on the microarray. All data are expressed as mean±s.e.m with n=7–8 per group. Two-way ANOVAs were used to determine statistical significance. All Bonferroni/Dunn post-hoc tests revealed a significant difference between saline-treated wild-type and CREBαΔ mutant mice (* p<0.05). Avpr1a (arginine vasopressin receptor 1A): drug×genotype interaction (F_1,26=10.71, p<0.01). IL10 (interleukin-10): drug×genotype interaction (F_1,26=8.64, p<0.01). Sgk2 (serum/glucocorticoid regulated kinase 2): drug×genotype interaction (F_1,27=6.184, p<0.05). Stat1 (signal transducer and activator of transcription 1): drug×genotype interaction (F_1,27=8.957, p<0.01). Tnfrsf1b (tumour necrosis factor receptor superfamily, member 1b) drug×genotype interaction (F_1,27=10.04, p<0.01). TBP, TATA binding protein.

To determine whether this effect was specific to the amygdala, we examined a subset of these genes in an area of the cortex consisting mainly of motor and somatosensory areas, which are typically not associated with drug or stress response. The drug×genotype interaction observed for IL10 and Stat1 in the amygdala was no longer present in the cortex, while Sgk2 appeared to have a subtle but significant genotype effect. We tested IL18 as well because it is also known to be involved in stress and immune response. This gene was down-regulated in the amygdala of CREBαΔ mutant mice regardless of drug experience in contrast to the other stress and immune response genes that demonstrated a clear drug×genotype interaction (Fig. 5). These data suggest that depletion of CREB elicits widespread down-regulation of certain genes in the brain; however, many of the drug×genotype interactions observed following our behavioural paradigm are specific to the amygdala: a brain region critically involved in the stress response. Abcb1b (ATP-binding cassette, sub-family B, member 1B), Gpr21 (G protein-coupled receptor 21), Pomc1 (pro-opiomelanocortin-alpha), and Slc6a3 (dopamine transporter) were also tested (individual data not shown), but none of these gene changes were validated in a new cohort of mice. Results of all 16 genes that were examined in both cohorts of mice are summarized in Table 5.

Fig. 5

qPCR of genes in the cortex to determine how patterns of gene expression in this brain region compare to those observed in the amygdala. All data are expressed as mean±s.e.m with n=7–8 per group. Two-way ANOVAs were used to determine statistical significance. IL18 (interleukin-18): effect of genotype (F_1,25=48.78, *** p<0.0001). Sgk2 (serum/glucocorticoid regulated kinase 2): effect of genotype (F_1,27=5.097, * p<0.05). TBP, TATA binding protein.

Discussion

The molecular mechanisms underlying the transition from initial drug use to the compulsive drug-taking that characterizes an addictive state have only been partially identified. Exposure to drugs of abuse results in long-term adaptations in the brain, which may result from activation of transcription factors such as CREB and concomitant alterations in gene expression. Studies in humans and animal models indicate that stress can increase vulnerability to addiction as well as enhance susceptibility to relapse. To identify gene expression changes that may underlie stress-induced drug relapse, we focused our expression profile analysis on alterations in the amygdala that occur following extinction of cocaine CPP but prior to reinstatement. We focused our analysis on CREB and its requirement in cocaine-induced changes in gene expression since mice lacking this transcription factor do not exhibit stress-induced reinstatement of cocaine place preference. Our data demonstrate that CREB regulates a wide variety of genes in the amygdala. Approximately half of these genes do not contain CRE elements in their promoters, suggesting that CREB is influencing changes in gene transcription both directly and indirectly.

Functional analysis revealed that genes involved in the immune response are differentially regulated by CREB in the amygdala. This suggests that genes classically involved in immune function could be changing stress-related behaviours in the CREBαΔ mice. Recent evidence has linked immune dysfunction to psychological stress in both humans and rodents (Godbout +6; Glaser, 2006). Furthermore, the cytokine theory of depression suggests that enhanced production of pro-inflammatory cytokines is associated with the pathogenesis of depression (Roque et al.2009). For instance, the pro-inflammatory cytokine IL18 is over-expressed in human stress disorders, including depression and panic disorder (Takeuchi et al.1999). In rodent models, levels of IL18 mRNA are significantly increased in subordinant rats following a social dominance paradigm, demonstrating a link between stress and cytokine gene expression in the brain (Kroes et al.2006).

In the present study, IL18 was significantly decreased in CREBαΔ mice regardless of drug treatment. IL18 is positively regulated by corticotrophin-releasing factor (CRF) in vitro (Park et al.2005; Yang et al.2005) and has a putative CRE element in its promoter, located at position +B197 (Zhang et al.2005). The decreased IL18 gene expression observed in the mutant mice might be playing a role in their altered behaviours by imparting them with a resilience to stress. Indeed, previous findings have shown that CREBαΔ mice have a blunted stress response (Conti et al.2002) and do not exhibit stress-induced reinstatement of cocaine CPP (Kreibich +6; Blendy, 2004). However, future studies must be completed to examine the role of immune-related genes in the altered stress responses of CREBαΔ mice.

Other stress- and immune-related genes demonstrated a drug×genotype interaction. Among these is Avpr1a, a G-protein-coupled receptor that binds the hormone arginine vasopressin when it is released from the hypothalamus. Avpr1a has been implicated in regulating aggression, social bonding, and maternal behaviours (Goodson +6; Bass, 2001). Furthermore, increases in Avpr1a are linked to increased anxiety, while decreases in this receptor are linked to decreased anxiety (Bielsky et al. 2004, 2005). Interestingly, our data show increased baseline expression of this gene in the amygdala of CREBαΔ mice, which display increased anxiety despite their antidepressant phenotype (Graves et al.2002; Gur et al.2007).

Another gene of interest identified in this analysis is IL10, an anti-inflammatory interleukin that may play a key role in modulating depressive-like behaviours (Roque et al.2009). IL10, along with other cytokines, is known to elicit phosphorylation of the transcriptional activator, Stat1, which also demonstrated an interaction on the microarray (Zocchia et al.1997). Tnfrsf1b is the receptor for tumour necrosis factor (TNF), increased production of which has also been observed in depressed patients (Raison et al.2006). Additionally, a recent microarray study examining expression changes in the NAc following abstinence from cocaine self-administration identified a TNF-centred network of genes, suggesting that long-term changes following cocaine administration may involve alterations in TNF signalling (Freeman et al.2010). A subunit of the transcription factor nuclear factor-κB (NF-κB), a central mediator of the immune response, was also identified as being up-regulated by chronic cocaine administration (Ang et al.2001). Together, these studies provide further evidence linking immune-related pathways to cocaine-induced adaptations and also highlight the importance of microarray studies in identifying novel targets and pathways not classically involved in drug addiction and therefore often overlooked.

Overall, CREBαΔ mice displayed higher levels of these various immune-related genes at baseline compared to wild-type mice, but cocaine treatment abolished this difference. The drug×genotype interaction observed using qPCR is consistent with the more general observation that there were 807 genes differentially expressed in the MT/Sal–WT/Sal comparison of the microarray relative to only 29 in the MT/Coc–WT/Coc comparison. This suggests that differences in gene expression between CREBαΔ and wild-type mice diminish if the animals have undergone development and extinction of cocaine CPP. Since only CREBαΔ mice demonstrate decreased expression of stress- and immune-related genes following cocaine, this response might be contributing to their greater stress resilience. However, it is unclear whether this altered gene expression has a direct effect on subsequent behavioural responses or whether these changes are simply a downstream readout of alterations in certain signalling pathways, possibly immune-related, that are affecting the behaviour.

It was surprising to observe up-regulated gene expression at baseline in animals lacking an activating transcription factor. This up-regulation may be an indirect effect of CREB deletion, such as decreased transcriptional repression or increased availability of the co-activator CREB binding protein (CBP), which may promote binding to other transcription factors that directly up-regulate these genes (Kamei et al.1996; Manna +6; Stocco, 2007). Since this up-regulation was no longer evident following cocaine CPP and extinction, the drug exposure and experience with conditioning may have blunted this increased gene expression via CREB-independent mechanisms. This idea is consistent with the microarray results, which demonstrated that 47 genes were down-regulated in CREBαΔ mice following cocaine.

One such CREB-independent mechanism could be driven by glucocorticoid receptors (GRs). GRs are known to be involved in mediating the behavioural responses to cocaine as well as the cross-sensitization between drugs and stress (Ambroggi et al.2009; de Jong +6; de Kloet, 2009; de Jong et al.2009; Deroche-Gamonet et al.2003; Izawa et al.2006). GRs are also involved in combating inflammation by repressing key inflammatory transcription factors such as activator protein 1 (AP-1) and NF-κB and reducing the expression of pro-inflammatory genes (Almawi +6; Melemedjian, 2002; Hosoi et al.2003; Newton +6; Holden, 2007; Paliogianni et al.1993). Thus, GRs may be a common link between cocaine, stress and immune function.

The expression profiling study presented above is one of the first to investigate CREB's involvement in the long-term alterations in gene expression that take place in the amygdala following cocaine conditioning and extinction. Elucidating changes in gene expression that occur during cocaine administration as well as during subsequent abstinence may lead to a better understanding of why the amygdala is necessary for reinstatement of drug-seeking following exposure to a stressor. Identification of novel genes and pathways in this brain region could prove useful in creating therapeutic agents designed to promote abstinence in human addicts and diminish the likelihood of relapse following a stressful life event.

Note

Supplementary material accompanies this paper on the Journal's website.

Acknowledgements

This work was supported by National Institutes of Health Grants T32-DA028874 (L.E.E.), F31-DA019757 (J.N.C.) and DA-011649 (J.A.B.).

Statement of Interest

None.