For the RNA-seq study, male Sprague-Dawley rats (Harlan, Houston, TX, USA) arrived at 21 days of age and were enriched or isolated for 30 days before behavioral testing. EC rats (n = 20) were group-housed in a large metal cage (70 cm × cm 70 × 70 cm) with 14 hard children’s plastic toys changed and rearranged daily. This density (245 cm²/rat) was higher than a majority of environmental enrichment studies, but was still more than double that required by NIH. EC rats were split into two cages after 50 days of age. IC rats (n = 20) were single housed in standard polycarbonate cages. These conditions produce a resistant (i.e., EC) and susceptible (i.e., IC) behavioral addiction phenotype (Green et al., 2002, 2003, 2010). While it is true that short term isolation is a stressor, enriched rats (even at lower density) show greater signs of chronic stress, even though these rats do not show outward signs of stress (Crofton et al., 2015). Rats remained in these homecage conditions throughout all behavioral tests, except during testing. For vector injection and behavioral tests, male Sprague-Dawley rats were obtained at 225–249 g. Rats were pair-housed and maintained in a controlled environment (temperature, 22°C; relative humidity, 50%; and 12 h light/dark cycle, lights on 0600 h) in an Association for Assessment and Accreditation of Laboratory Animal Care (AAALAC) approved colony and procedures were approved by the UTMB Institutional Animal Care and Use Committee and conform to the NIH Guide for the Care and Use of Laboratory Animals.

Rats were anesthetized with ketamine (100 mg/kg, IP) and xylazine (10 mg/kg, IP), and implanted with an indwelling Silastic catheter (0.2 mm I.D.; Fisher Scientific, Pittsburgh, PA, USA) into the jugular vein. The catheter passed under the skin to exit on the rat’s back. The catheters were infused with 0.1 ml of a sterile saline solution containing heparin (30.0 U/ml), ticarcillin (250,000 U/ml) and streptokinase (8000 IU/ml) daily, to prevent infection and maintain catheter patency throughout the duration of experiments.

One week after catheter surgery, rats were allowed to self-administer 0.5 mg/kg/infusion cocaine (National Institute on Drug Abuse, Bethesda, MD, USA) or saline under a fixed ratio 1 (FR1) schedule for 2 h/day for 14 days. The session terminated when the rat received 30 infusions to eliminate cocaine intake differences between EC and IC rats. Tissue was harvested 3 h after the beginning of the last session (Figure 1A).

Rat brains were harvested 3 h after the beginning of the last self-administration session and the left side of the NAc was dissected on an ice-cold platform for mRNA analysis. The right side was used for protein quantification (Lichti et al., 2014). The RNA was extracted and purified with the RNeasy mini kit (Qiagen, Valencia, CA, USA). cDNA libraries were created by reverse transcribing the RNA and creating the second strand. Blunt ends were phosphorylated and “a-tailed” so that adapters could be ligated to both ends. Adapters were individually “bar coded” and thus samples were not pooled despite having 4 samples per flow cell, yielding n = 7–8 for each condition (total N = 30). RNA was sequenced with a HiSeq 1000 system from Illumina. cDNA was amplified using “bridge” amplification. Base calls were made using fluorescently labeled nucleotides. More than 100 million reads with 50 bp (paired-end reads) were mapped for each rat and quality was checked with FastQC (v0.9.1) (Andrew, 2010). Reads were mapped to the rat reference genome (RN4) using Tophat2 (v2.0.4) (Kim et al., 2013) and Bowtie2 (v2.0.0.6) (Langmead and Salzberg, 2012) software packages. The R package EdgeR (v.3.0.8) (Anders and Huber, 2010; Robinson et al., 2010) was then used for analysis using the log-transformed “trimmed mean for M-values” (TMM) method for normalization and tag-wise dispersion using “count” data. A likelihood ratio F-test was used for generating p-values to compare EC vs. IC rats and cocaine vs. saline. Cross-validation of RNA-seq results was achieved by looking at correlation of expression between qPCR and RNA-seq for Fabp5 (Forward: 5′ CTTGCACCTTGGGAGAGAAG 3′, Reverse: 5′ CATCTTCCCGTCCTTCAGTT 3′) and Hspa5 (Forward: 5′ AACCAAGGATGCTGGCACTA 3′; Reverse: 5′ ATGACCCGCTGATCAAAGTC 3′), with the qPCR being normalized to Reep5 (Forward: 5′ GGTTCCTGCACGAGAAGAACT 3′; Reverse: 5′ GAGAGAGGCTCCATAACCGAA 3′) (Supplementary Figure S6). The Fabp5 was chosen because it is in the RA pathway and selectively expressed in the NAc shell, both aspects central to this paper. Hspa5 (Bip) was chosen because it has extremely high expression levels, and because we have studied this transcript with cocaine in the past (Pavlovsky et al., 2013). Using qPCR to cross-validate the RNA-seq data is somewhat problematic. Our conclusion is that using a less accurate and less precise method to validate a more accurate and precise method is difficult at best. The problem encountered is that qPCR needs to be normalized to a control or “housekeeping” transcript. The variance from the transcript of interest gets compounded with the variance inherent in the housekeeping gene, increasing variance. Multiple transcripts were assessed for normalization and the end results were quite different depending solely upon which normalization gene was chosen. Many traditional “housekeeping” transcripts were regulated (e.g., gapdh) or trended toward regulation (e.g., β-actin) by enrichment and/or cocaine. In the end, Reep5 was chosen because it was a housekeeping gene not regulated by cocaine or enrichment. However, the variability of the Reep5 expression among samples washed out main effects.

The quantitative transcriptomic data have been deposited in NCBI’s Gene Expression Omnibus (Edgar et al., 2002) and are accessible through GEO Series accession number GSE88736².

In order to study the biological functions and pathways regulated by cocaine and environmental enrichment, transcripts significantly regulated (p < 0.05) were analyzed with IPA. Canonical Pathways, Upstream Regulators, Diseases and Biological Functions, and Networks were used to identify significantly regulated transcript sets. Because regulation of any given gene could be a statistical anomaly (i.e., false positive), bioinformatic analyses have been developed under the assumption that regulation important for function will occur in a coordinated fashion at multiple targets within a given pathway. Thus, the IPA analysis assesses over-representation of multiple targets within known pathways. For this analysis, the list of significantly regulated genes is further analyzed for significantly regulated pathways, thus minimizing the effect of individual false positives. Preliminary analysis of ribosomal proteins, tyrosine phosphatases, de-ubiquitinating enzymes and proteasomal proteins, all of which show highly coordinated regulation (see Supplementary Figures S4A,B) demonstrate that the p < 0.05 cutoff does not produce an overabundance of false positives.

To complement the IPA analysis and avoid the problems of Type I vs. II error, normalized expression intensity (normalized by EdgeR using TMM) of all identified transcripts were analyzed by GSEA. The normalized enrichment score (NES) indicates the degree to which this gene set is overrepresented at the top or bottom of the ranked list of genes in the expression dataset (Subramanian et al., 2005). In IPA analyses, a p-value cutoff is required to decide whether transcription is significantly regulated or not. The problem of setting a cutoff is that high stringency introduces false negative results (i.e., high Type II error), whereas low stringent cutoffs introduce false positive results (i.e., high Type I error). To complement the IPA analysis and avoid the problem of Type I vs. II error, normalized expression intensity of all identified transcripts were analyzed by GSEA. In the GSEA analysis, all transcripts are ranked by a signal2noise metric (the difference of means scaled by the standard deviation), and the significance of a given set of genes is determined using a running-sum statistic to determine rank-order over-representation. Thus, this analysis produces a statistic at the gene set level without the need for a p-value cutoff. Gene sets were from the Broad Institute‘s set (v.4.0) of curated gene sets [C2.all], gene ontology gene sets [C5.all] and transcription factor target gene sets [C3.tft]. In GSEA, the Enrichment Score (ES) for the gene set is the degree to which this gene set is overrepresented (i.e., “enriched”) at the top or bottom of the ranked list of genes in the expression dataset. NES indicates the enrichment score for the gene set after it has been normalized to adjust for size of the gene set.

A secondary analysis of our previously published liquid chromatography tandem mass spectrometry (LC-MS/MS) protein data from these same rats (Lichti et al., 2014) was used to corroborate RNA results where appropriate.

The regional specificity of gene expression within the brain strongly suggests that genes are not expressed where they are not needed. Thus, in situ visualization of regional expression patterns in brain slices can provide a topographical map of enhanced gene expression for each specific brain region. One could then hypothesize that genes with regionally enhanced expression patterns in the NAc shell would likely be important for addiction-related behavior. Accordingly, the Allen Brain Institute’s Anatomic Gene Expression Atlas Gene Finder³ (Lein et al., 2007) algorithm was used with coronal image sets, seed coordinates of 4200, 5400, 6400 (right side) and 4200, 5600, 5000 (left side) with an expression threshold of 1. This Atlas defines mouse brain expression patterns; however, there is good concordance in expression patterns among mice, rats and humans (Zetterstrom et al., 1999; Strand et al., 2007). The genes with a regional enrichment score of 1.25 fold or higher (>250 genes) were then culled by individual manual inspection to remove genes whose enrichment fold change was clearly driven by experimental artifact such as bubbles, debris, misalignment and high expression in nearby regions (e.g., Islands of Calleja, NAc core, olfactory tubercles, etc.) to yield 178 genes. These 178 genes were submitted to IPA analysis. Once the RA signaling pathway was identified, manual curation of sagittal sets for RA signaling genes without coronal sets was used to identify additional targets (Rbp1 and Rdh10) with shell-specific expression. Images for Figures 6A–D were taken from the Allen Brain Atlas:

STRA6: http://mouse.brain-map.org/experiment/siv?id=75041492&imageId=74948838&initImage=ish&coordSystem=pixel&x=5288.5&y=3704.5&z=1

FABP5: http://mouse.brain-map.org/experiment/siv?id=70634396&imageId=70561807&initImage=ish&coordSystem=pixel&x=4104.5&y=3104.5&z=0

RBP1: http://mouse.brain-map.org/experiment/siv?id=68076899&imageId=68005433&initImage=ish&coordSyste&imgeId=68005433&initImage=ish&coordSystem=pixel&x=7448.5&y=4096.5&z=1

RDH10: http://mouse.brain-map.org/experiment/siv?id=75831736&imageId=75791261&initImage=ish&coordSystem=pixel&x=7472.5&y=4040.5&z=1

ALDH1A3: http://mouse.brain-map.org/experiment/siv?id=75861799&imageId=75806893&initImage=ish&coordSystem=pixel&x=6872.5&y=3496.5&z=1

CYP26B1: http://mouse.brain-map.org/experiment/siv?id=72081548&imageId=71940737&initImage=ish&coordSystem=pixel&x=4280.5&y=3160.5&z=1

RARβ: http://mouse.brain-map.org/experiment/siv?id=75038442&imageId=74930074&initImage=expression&colormap=0.5,1,0,256,4&coordSystem=pixel&x=4416.5&y=2952.5&z=1

Finally, a literature search of published papers with in situ and immunohistochemical images was used to confirm Allen Brain Atlas expression and identify additional shell-specific expression components of the RA signaling pathway (McCaffery and Drager, 1994; Zetterstrom et al., 1999).

In order to knock down CYP26b1 expression, five 24-nucleotide sequences were identified within the CYP26b1 mRNA sequence (Ensembl transcript ID: ENSRNOT00000020505) using the criteria previously described (Hommel et al., 2003; Benzon et al., 2014) (Supplementary Table S1). The oligonucleotide sequences were synthesized and the annealed hairpin oligonucleotides were cloned into pAAV-shRNA plasmids (pAAV-Cyp26b1shRNA). Hairpin expression from these plasmids was driven by the mouse U6 promoter using a pol-III mechanism. In addition to the hairpin, enhanced green fluorescent protein (eGFP) was expressed from a separate expression cassette driven by a pol-II promoter (CMV).

In order to determine the most effective hairpin, all five hairpins were screened in vitro. Since HEK293 does not express rat Cyp26b1, a Cyp26b1 overexpression plasmid was constructed. To create the Cyp26b1 overexpression plasmid to test knockdown efficiency, the rat CYP26b1 gene sequence was amplified from rat genome cDNA using polymerase chain reaction (forward primer: TAGGAATTCCTCCTGGGTTTCTTCGAGGG; Reverse: TAGGTCGACATCCAAGAGGGTGGGAGTCA) and cloned into the pAAV-IRES-hrGFP plasmid (Agilent Technologies, Santa Clara, CA, USA). The various pAAV-Cyp26b1-shRNA plasmids or pAAV-Control shRNA plasmid was co-transfected with pAAV-Cyp26b1-IRES-hrGFP plasmid into HEK-293 cells using FuGENE^® 6 Transfection Reagent (Promega)/Lipofectamine 2000 (Life Technologies, Grand Island, NY, USA).

The cells were harvested 24–48 h later, followed by RNA extraction and reverse transcription to cDNA. The RNA was extracted using RNeasy Mini Kit (Cat No. 74104). Contaminating DNA was removed (TURBO DNA- Free, Life Technologies, Carlsbad, CA, USA) and 5 μg total RNA was reverse transcribed into cDNA (SuperScript III First Strand Synthesis: Invitrogen catalog # 18080051). Relative knockdown was measured with Real-time PCR (SYBR Green: Applied Biosystems, Foster City, CA, USA) on an Applied Biosystems 7500 fast thermocycler with Cyp26b1 qPCR primers (forward: CCAGCAGTTTGTGGAGAATG; Reverse: GTCCAGGGCGTCTGAGTAGT). The results were normalized to Gapdh (forward: AACGACCCCTTCATTGAC; reverse: TCCACGACATACTCAGCAC). All primers were validated and analyzed for specificity and linearity prior to experiments (Alibhai et al., 2007).

Efficiency of hairpin was further validated by western blot. HEK293 cells were homogenized in a buffer containing sucrose, Hepes buffer, sodium fluoride, 10% SDS, and protease and phosphatase inhibitors (Sigma–Aldrich: P-8340, P-2850, P-5726). Protein concentration was assessed using the Pierce BCA Protein Assay Kit (Thermo Fisher Scientific, Waltham, MA, USA). Protein samples were denatured at 95° for 10 min and run on a 12% gel (Criterion TGX, Bio-Rad Laboratories, Hercules, CA, USA) then transferred to a polyvinylidene fluoride (PVDF) membrane (Millipore, Billerica, MA, USA). The membrane was blocked with blotting-grade blocker (non-fat dry milk), incubated with Cyp26b1 primary antibody (rabbit, 1:500, 21555-1-AP, Proteintech, Rosemont, IL, USA) and GAPDH primary antibody (mouse, 1:10000, Abcam, Cambridge, MA, USA), washed with TBST and then incubated with fluorescent secondary antibodies (donkey anti-rabbit (780 nm), donkey anti-mouse (680 nm), 1:15000, Li-Cor Biosciences, Lincoln, NE, USA). Western blots were then imaged (Odyssey, Li-Cor Biosciences, Lincoln, NE, USA) and protein levels quantified with the Odyssey software. The hairpin plasmid with the highest knockdown efficacy (5′.AGTTCTTTGGTCTAGACTCCAATC.3′) was packaged into Adeno-Associated Virus2 (AAV2) by the University of North Carolina Gene Therapy Core Facility and used in subsequent behavioral tests. Control vector expressed a previously validated control hairpin not targeted to any gene (Hommel et al., 2003; Benzon et al., 2014).

An AAV2-based vector that expresses Cyp26b1 shRNA and eGFP or a non-targeted hairpin control vector (n = 10–11) was injected bilaterally into the rat NAc shell (1 μl/side over 10 min) using coordinates AP = 1.7, L = 2.2, D = -6.7. To validate the knockdown efficiency of AAV in vivo, AAV-control shRNA or AAV-Cyp26b1 shRNA (n = 6) was injected in the nucleus accumbens shell (NAcSh) in rats. Two microliter of AAV was injected per side of the NAc to increase the number of infected neurons. NAc regions with eGFP fluorescence were collected and tissues from two rats were pooled together to increase the yield of protein concentration. The expression of CYP26B1 protein level was detected using western blot described above. For behavioral tests, an AAV2-based vector that expresses Cyp26b1 shRNA and eGFP, or control vector (n = 10–11 each) was injected bilaterally with 1 μl/side into the rat NAc. Pair-housed rats were used instead of isolated rats for this study to increase relevance to the scientific community by demonstrating the effects of Cyp26b1 independent of the EC/IC procedure. Behavioral tests started 5 weeks after stereotaxic surgery (Figure 7). Accurate placement was verified immunohistochemically after the conclusion of behavioral testing.

For Figure 7D, the placement of AAV-Cyp26b1shRNA expression in vivo was validated by immunofluorescence staining with eGFP. The brains were extracted, post fixed, cryoprotected and sectioned into 40 μm slices containing the NAc on a sliding freezing microtome (Leica Biosystems, Richmond, IL, USA). The slices remained floating and were rinsed with 1xPBS prior to blocking with 3% normal donkey serum (Jackson ImmunoResearch, West Grove, PA, USA) with 0.3% triton. NAc slices were incubated with eGFP primary antibody overnight (1:500, chicken, Aves labs, Tigard, OR, USA) with 3% donkey serum, 0.3% triton in 1xPBS. After washing, slices were incubated with secondary Alexa 488 donkey anti chicken antibody (Jackson ImmunoResearch, West Grove, PA, USA) in 1xPBS. Finally, slices were mounted, dehydrated using ethanol and CitriSolv (Fischer Scientific, Waltham, MA, USA) and coverslipped with DPX (Fisher Scientific).

Two-factor analyses of variance (ANOVAs) and two-factor repeated-measures ANOVAs were performed to compare four treatment groups. Significance between only two conditions was analyzed using a Student’s t-test. All t-test data passed the Shapiro–Wilk test of normality. All data are expressed as mean ± SEM. Statistical significance was set at p < 0.05.

Raw and processed RNA sequencing data can be found in the Gene Expression Omnibus database with the project number GSE88736.

After the primary data alignment and analysis, 14,309 transcripts were quantified as the result of RNA-seq. The first step of our analysis was to investigate the regulation of individual transcripts. Based on the likelihood F-tests, the Venn diagram shows 106 transcripts significantly regulated (p < 0.001) by cocaine, 683 transcripts significantly regulated by environmental enrichment and 64 transcripts significant for the interaction (Figure 1B). In addition, the Venn diagram also displays the number of transcripts common among the effects, which indicates the overlapping effects of cocaine and environmental enrichment. Note that there are more transcripts upregulated than downregulated by cocaine, while there are many more downregulated transcripts than upregulated by environmental enrichment (Figure 1C).

In the significantly regulated gene lists, we first looked at the top 50 upregulated and downregulated transcripts from cocaine and environmental enrichment. These transcripts give the highest confidence of regulation. For cocaine (Figure 1D), immediate early genes such as EGR4, NR4A3, FOS, EGR2 were induced, which agrees with previous publications (Hope et al., 1992; Berke et al., 1998; Werme et al., 2000; Guez-Barber et al., 2011). For enrichment, the top regulated transcripts include transcription factors, such as NR4A1, EGR3, ARC, EGR2 and NR4A3, etc. (Figure 1E). In our search for a therapeutic target, we moved beyond individual transcripts to explore molecular pathways.

To investigate which biological functions and cellular pathways are regulated by cocaine, significantly regulated transcripts were analyzed with IPA. Figure 2A lists some top-ranked diseases and biological functions of interest for the cocaine main effect. Complete IPA data can be found in the supplemental information.

To explore the molecular mechanism of environmental enrichment, IPA and GSEA were used to analyze the biological functions and pathways. For the environmental enrichment main effect, many mental disorder-related functions and diseases were significantly regulated (Figure 3A). The top-ranked canonical pathways involved Protein Translation-Related EIF2 Signaling, PKA Signaling, Mitochondrial Dysfunction, Kinase Signaling, etc. (Figure 3B). Complete IPA data can be found in the supplemental information.

The transcription regulated by the interaction indicates that EC and IC rats respond differently to cocaine (Figure 4). Studying this differential response helps to identify the molecular mechanisms of the protective EC phenotype. For the interaction, Drug Dependence (p = 3.38E-6; Supplementary Figure S5A) and Release of Dopamine (p = 5.41E-7; Supplementary Figure S5B) were top-regulated diseases and biological functions in IPA. Regulated transcripts in Drug Dependence were dominated by ion channels and G-protein coupled receptors (GPCRs). Release of Dopamine was also dominated by GPCRs.

To confirm the validity of the quantitative RNA sequencing technique, real-time PCR was used to quantify the mRNA expression of Fabp5 and Hspa5 from the same rats. mRNA fold change results from RNA sequencing and qPCR of Fabp5 (R² = 0.5155, p < 0.0001) and Hspa5 (R² = 0.4301, p = 0.0003) was compared for every rat. These results indicate qPCR and RNA sequencing results are well correlated in addition to the orthogonal validation of comparing the current results against previous cocaine and enrichment findings.

A topographic transcriptomic analysis for genes with selectively enhanced gene expression in the NAc shell from the Allen Brain Atlas identified 178 transcripts with selective expression in the NAc shell ≥ 1.25 fold over surrounding regions. An IPA analysis of these 178 genes revealed the central RA pathway as being significant with Stimulated by RA 6 (Stra6), Retinoic acid receptor β (Rarb), Fatty Acid Binding Protein 5 (Fabp5) and Cytochrome P450, Family 26, Subfamily B, Polypeptide 1 (Cyp26b1) as shell-selective genes (p = 1.67E-5; Figure 6). A manual search of NAc shell-selective RA gene expression in sagittal sections identified Rbp1. Rdh10 and Aldh1a3 as additional selective genes. A search of the literature confirmed Rbp1 and Rarb selective expression, along with Aldh1a1 and Rxrg (McCaffery and Drager, 1994; Zetterstrom, 1999). These nine NAc shell-enhanced genes are depicted in the pathway for Figure 6H.

Two strategies for changing concentrations of RA are to either alter synthesis or degradation. However, because there are several different subtypes of Rdhs and Raldhs, knocking down any one could be compensated by the other RA synthases. Therefore, we decided to alter RA concentration by knocking down the degradation enzyme, Cyp26b1, since RA signaling was regulated by cocaine, enrichment, and their interaction, and many genes showed selectively enhanced expression in the NAc shell. With decreased expression of Cyp26b1, RA subsequently builds up in neurons, enhancing RA downstream signaling (Kim et al., 2014). To knock down the expression of Cyp26b1, an shRNA targeting the Cyp26b1 coding sequence was designed and knockdown efficiency was examined in vitro and in vivo. Compared with a non-targeted control shRNA, Cyp26b1 shRNA significantly decreased expression of Cyp26b1 in HEK293 cells in both mRNA and protein (Figure 7A) and in rat NAc shell at the protein level (Figure 7B). Figure 7C shows the schematic diagram of the experimental timeline for behavioral tests. For behavioral testing, AAV-Cyp26b1shRNA or control vector was injected into the NAc shell of rats (Figure 7D, atlas comparison Figure 7E). For acquisition of cocaine self-administration, results demonstrate significant acquisition across sessions [F(4,80) = 3.855, p = 0.006; Figure 7F] with a trend for increased acquisition in Cyp26b1 shRNA rats [interaction F(4,80) = 2.142, p = 0.083]. A t-test showed that Cyp26b1 shRNA rats responded significantly more for Sessions 3 and 4. For maintenance responding (Figure 7G), knocking down Cyp26b1 significantly increased responding for cocaine at low unit doses, resulting in significant main effects of Dose [F(7,26) = 39.239, p < 0.001], Vector [F(1,18) = 7.938, p = 0.011] and a significant interaction [F(7,126) = 4.871, p < 0.001]. Knockdown significantly elevated the responding for cues in a cue responding test [Figure 7H; t(20) = -2.572, p = 0.018]. Finally, in a within-session extinction procedure (Figure 7I), rats with Cyp26b1 knockdown exhibited increased responding compared to control rats, with significant main effects of Session [F(2,36) = 15.317, p < 0.001] and the Vector main effect is at the threshold of the p-value cutoff for significance [F(1,18) = 4.433, p = 0.05], indicating that knockdown of Cyp26b1 in the NAc shell enhances drug-seeking behavior. In cocaine-induced reinstatement, high variance prevented detection of a difference in reinstatement between the two groups of rats at any dose.

In the upstream analysis of cocaine-regulated transcripts, cocaine itself and CREB1, an important mediator of the effects of psychostimulants (Carlezon et al., 1998; Pliakas et al., 2001), ranked at the top of the list as upstream regulators, strongly supporting that cocaine-regulated transcription seen here agrees with previous studies. Our prior research demonstrated that enriched rats have less phospho-CREB in the NAc and that decreasing CREB function in the accumbens shell produces a behavioral phenotype identical to that of environmental enrichment (Bowling and Bardo, 1994; Green et al., 2002, 2010), an interesting behavioral phenotype marked by increased sensitivity to the rewarding effects of stimulants (as measured by CPP) coupled with decreased self-administration (Pliakas et al., 2001; Larson et al., 2011).

Compared to the cocaine main effect, there were approximately 5X more transcripts significantly regulated by environmental enrichment, revealing that environment has a much more extensive impact on gene expression than cocaine exposure. Transcription factors were the most regulated gene sets by environmental enrichment. Another impressive difference between EC and IC rats is the regulation of transcripts involved with EIF2 signaling. Even though Eif2 itself is not regulated at the mRNA level, decreased upstream inhibitors of Eif2 and upregulated downstream ribosomal subunits suggests regulation of the protein translation process. In addition to protein synthesis, the protein degradation system is also altered by enrichment. Prior research from this laboratory demonstrated that expression of ubiquitin target proteins is different in EC and IC rats (Fan et al., 2013a,b). Ubiquitination is also important in differential expression of proteins from the current rats (Lichti et al., 2014). The current mRNA data revealed increased transcription of ubiquitin and proteasomal subunits, but reduced mRNA expression for deubiquitinating enzymes in the NAc, possibly suggesting enhanced protein degradation in EC rats. Taken together, enhanced protein translation and degradation likely indicate more rapid protein turnover in EC rats compared with IC rats.

Retinoic acid genes are selectively expressed in the NAc shell, as shown by a topographic transcriptomic analysis of the Allen Brain Atlas and the published literature (McCaffery and Drager, 1994; Zetterstrom, 1999). Given the importance of the NAc shell to addiction, the RA signaling pathway offers promising targets for novel therapeutic development for cocaine addiction. One previous report found that constitutive whole-body RARβ/RXRβ or RARβ/RXRγ double null mice had a selective decrease of dopaminergic D2 receptors in the shell of the NAc, with concomitant decrements in locomotor and rotarod performance (Krezel et al., 1998). Thus far, however, there have not been any other systematic studies aimed at understanding RA signaling and addiction.

The current report provides converging evidence from the Upstream Regulator analysis (predicting function) of environmental enrichment and the vector knockdown of Cyp26b1 suggesting that RA signaling activity in the NAc shell increases susceptibility to drug taking. Given that every core component of the RA signaling pathway involves protein interactions with small molecules (i.e., retinoids) this pathway is a prime candidate for the development of selective small molecule inhibitors as possible pharmacotherapeutics for cocaine addiction. One advantage of choosing targets in this pathway is that nine components of this pathway have some enhancement of expression in the NAc shell (Figure 6), providing some degree of regional selectivity and thereby decreasing the likelihood of unwanted side effects. These regionally-enhanced components include the binding proteins Stra6. Rbp1, and Fabp5, the synthesis enzymes Adh10. Aldh1a1, and Aldh1a3, the degradation enzyme Cyp26b1, and the RA receptors Rarβ and Rxrγ. Ongoing experiments are investigating which of these targets would be most suitable for pharmacotherapeutic development.