Environmental enrichment during forced abstinence from cocaine self-administration opposes gene network expression changes associated with the incubation effect

Environmental enrichment (EE) is a robust intervention for reducing cocaine-seeking behaviors in animals when given during forced abstinence. However, the mechanisms that underlie these effects are not well-established. We investigated the adult male rat transcriptome using RNA-sequencing (RNA-seq) following differential housing during forced abstinence from cocaine self-administration for either 1 or 21 days. Enriched, 21-day forced abstinence rats displayed a significant reduction in cocaine-seeking behavior compared to rats housed in isolation. RNA-seq of the nucleus accumbens shell revealed hundreds of differentially regulated transcripts between rats of different forced abstinence length and housing environment, as well as within specific contrasts such as enrichment (isolated 21 days vs. enriched 21 days) or incubation (isolated 1 day vs. isolated 21 days). Ingenuity Pathway Analysis affirmed several pathways as differentially enriched based on housing condition and forced abstinence length including RELN, the Eif2 signaling pathway, synaptogenesis and neurogenesis pathways. Numerous pathways showed upregulation with incubation, but downregulation with EE, suggesting that EE may prevent or reverse changes in gene expression associated with protracted forced abstinence. The findings reveal novel candidate mechanisms involved in the protective effects of EE against cocaine seeking, which may inform efforts to develop pharmacological and gene therapies for treating cocaine use disorders. Furthermore, the finding that EE opposes multiple pathway changes associated with incubation of cocaine seeking strongly supports EE as a therapeutic intervention and suggests EE is capable of preventing or reversing the widespread dysregulation of signaling pathways that occurs during cocaine forced abstinence.

www.nature.com/scientificreports/ Insight into novel targets may be gained from examining transcriptomic changes in rodent models that manipulate the degree of motivation for drugs. For instance, motivation for drug is affected by length of forced abstinence. In humans, after a cocaine binge there is initially little if any craving for drug; however, craving emerges within hours to days of abstinence and is often triggered by exposure to drug-associated cues 7 . In parallel, we found that cocaine-seeking behavior elicited by drug-associated cues in rats becomes increasingly stronger over 3-4 weeks of forced abstinence 8,9 . Grimm et al. 11 coined the term "incubation" to describe this effect and found that it persists for up to 2 months of forced abstinence 10 . Mechanisms contributing to the incubation effect include plasticity within brain regions involved in reward learning and memory, which play a critical role in the development of CUDs 8,9,[11][12][13][14][15] .
Environmental enrichment (EE) is another robust manipulation for reducing motivation to seek and take drugs in rodents as demonstrated for psychostimulants [16][17][18][19] , heroin 20 , and nicotine [21][22][23][24][25][26] . EE typically involves housing animals with social companions along with the opportunity for exercise and exploration of novel items. At the opposite end of the spectrum of housing conditions is isolation, which typically enhances motivation for drug. Compared to standard group housing for instance, isolation enhances cocaine self-administration 27 , whereas EE reduces cocaine-induced hyperactivity [28][29][30] . Housing conditions contribute to individual differences in cocaine self-administration, exemplified by Puhl et al., 31 who found that 96% of rats living in EE qualified as "low drug-takers" compared to 75% of rats living in isolation. Furthermore, EE given as an intervention during forced abstinence from cocaine self-administration reduces cue reinstatement of operant cocaine-seeking behavior compared to both group and isolation housing 32,33 .
Previous research using RNA-seq to examine transcriptomic changes related to the protective effects of EE on reducing cocaine intake in rats during self-administration identified novel pathways that had not been previously implicated in the drug abuse literature 34,35 , including the retinoic acid pathway. However, transcriptomic changes related to the protective effects of EE as an intervention during forced abstinence to reduce motivation to seek cocaine have not been investigated. The present study aimed to address this question by using RNA-seq in the nucleus accumbens shell (NAcsh) of male rats to examine transcriptomic differences associated with varying degrees of cocaine-seeking behavior. We focused on the NAcsh because this striatal region is critically involved with motivation to seek cocaine 36,37 and shows correlations between changes in addiction-related genes and cocaine-seeking behavior [38][39][40] . We manipulated motivation levels utilizing isolation and EE housing conditions during forced abstinence for either 1 or 21 days. We expected variation in cocaine-seeking behavior across groups during a test for cue reactivity, with 21 days of forced abstinence in isolation resulting in robust cocaine seeking and brief forced abstinence in EE resulting in the least cocaine seeking.

Results
Self-administration and cue reactivity. Following a minimum of 21, 2-h sessions of cocaine selfadministration (0.75 mg/kg, IV per infusion), assignment to housing and forced abstinence length conditions was counterbalanced to equate previous consumption and active lever presses ( Figure S1) across groups. Rats either remained in isolation (IC) or were placed into EE for either 1 (1D) or 21 days (21D) of forced abstinence (n = 9/housing condition for 1D; n = 15-16/housing condition for 21D), after which all rats underwent a 1-h cue reactivity test. During the test, light/tone cues previously paired with cocaine delivery were presented responsecontingently but cocaine was not available (Fig. 1A). A two-factor ANOVA of the number of active lever presses during the test revealed a main effect of forced abstinence length [F 1,45 = 12.13, p < 0.05] with rats abstinent for 21 days exhibiting more active lever presses than those abstinent for 1 day. In addition, there was a main effect of housing [F 1,45 = 6.67, p < 0.05] with IC rats exhibiting more active lever presses than EE rats. There was no housing × forced abstinence length interaction; however, pairwise comparisons using Bonferroni-corrected t-tests revealed that in animals abstinent for 21 days, cue reactivity was greater when they were housed in isolation (IC21D) than when housed in enrichment (EE21D) [t 30 = 3.60, p < 0.0167], whereas there was no difference between housing conditions in animals abstinent for only 1 day (IC1D vs. EE1D).

RnA-Seq.
Following the 1-h cue reactivity test, bilateral micropunches of the NAcsh were pooled from each animal from a subset of animals [N = 12 assigned to one of 4 groups: IC1D, IC21D, EE1D, and EE21D (n = 3/group)] and analyzed for changes in transcript expression. An average of 52 (± 3.3 SEM) million reads were mapped per sample. Of the 41,064 transcripts in the ENSEMBL rat database, we consistently detected 27,942 transcripts in our samples using a cutoff of 1 read across all samples. Based on the Wald test statistic, 135 transcripts were differentially regulated between abstinence lengths regardless of housing and 150 transcripts were differentially regulated between the housing conditions during abstinence prior to testing regardless of abstinence length (Fig. 1B). Among the 285 total transcripts differentially regulated by these factors, 75 were shared between housing conditions and forced abstinence length (Fig. 1C, D). Among the transcripts regulated between forced abstinence lengths (1 vs. 21 days ignoring housing; n = 6/condition), more were downregulated than upregulated, whereas the opposite pattern was observed among those regulated between housing conditions (Isolation vs. EE ignoring abstinence length; n = 6/condition). Using the contrast function of 'DESeq2′ in R, the number of differentially expressed transcripts in the isolation vs. enrichment groups following 21 days of forced abstinence were quantified; this contrast focused on enrichment as an intervention (Fig. 1B, Enrichment contrast; IC21D vs. EE21D; n = 3/group). We also quantified the number of differentially expressed transcripts in the 1 vs. 21 days of forced abstinence groups that remained isolated; this contrast focused on the incubation of motivation for cocaine across time during abstinence (Fig. 1B, Incubation contrast; IC1D vs. IC21D; n = 3/group). 185 transcripts were differentially regulated in the Enrichment contrast and 218 were differentially regulated in the Incubation contrast. Among these 403 differentially-expressed (DE) transcripts, 99 were shared among Enrichment and Incubation (Fig. 1E) www.nature.com/scientificreports/ ment contrast, more were downregulated than upregulated, whereas the opposite pattern was observed among those regulated within the Incubation contrast (Fig. 1F).
Forced abstinence effect. Of the 135 transcripts regulated within the forced abstinence contrast (1D vs. 21D), the top 30 up-or down-regulated transcripts by log2 fold change can be seen in Fig. 2A. An additional approach to examining the forced abstinence contrast was to use Ingenuity Pathway Analysis (IPA) to investigate transcripts with p ≤ 0.05 from the DESeq2 analysis described above and the effect of changes in these transcripts in pathway activation or inhibition is shown as positive or negative Z scores, respectively. A total of 693 transcripts met this criterion for the "Forced Abstinence" effect; following removal of duplicate genes by IPA a total of 586 transcripts remained. Selected pathways differentially regulated in forced abstinence are shown in Fig. 2B, Cocaine-seeking behavior shown as the number of lever presses (+ SEM) on the cocaineassociated, active lever during cue reactivity testing following abstinence. Isolated animals pressed the active lever more than enriched animals (a, main effect of housing); animals abstinent for 21 days pressed the active lever more than animals abstinent for 1 day (b, main effect of abstinence length); and animals isolated for 21 days pressed more than animals in an enriched environment for 21 days (c, Bonferroni t-test, p < 0.025 www.nature.com/scientificreports/ with the "threshold" representing the negative log of the alpha-level (0.05) and "ratio" representing the number of differentially expressed transcripts in each canonical pathway vs. total number of transcripts in each category (i.e., a ratio of 0.10 is equivalent to 10% genes in the pathway). The top-ranked canonical pathways included Insulin Receptor Signaling, Reelin Signaling in Neurons, and Synaptogenesis Signaling in Neurons. All output data from IPA analysis of the abstinence effect can be found in the Table S1.
Synaptogenesis signaling in neurons. One significant canonical pathway within the forced abstinence effect, Synaptogenesis Signaling in Neurons, was of particular interest. The differentially regulated molecules www.nature.com/scientificreports/ within the pathway that are driving this effect (activation Z-score: 1.213) are predicted to be spread across the synaptic cleft. This activation Z-score indicates that the pathway is predicted to have increased activity in animals that experienced prolonged forced abstinence. In the presynaptic bouton, calcium/calmodulin-dependent protein kinase II gamma (Camk2g), adenylate cyclase 6 (Adcy6), synaptotagmin 17 (Syt17), adaptor related protein complex 1 subunit beta 1 (Ap1b1), and heat shock protein family A (Hsp70) member 8 (Hspa8) are all upregulated (Table S1). At the postsynaptic membrane, cadherin 13 (Cdh13) and neuroligin 2 (Nlgn2), both involved with scaffolding of the postsynaptic region, are downregulated. Additionally, the mammalian target of rapamycin kinase mTor (Mtor) is upregulated in abstinence, is part of the Synaptogenesis Signaling in Neurons pathway, and is known to influence synaptic spine development and maturation as well as local protein translation in dendrites 41,42 . other functions and pathways related to forced abstinence. In addition to canonical pathways, IPA predicts upstream regulators of the molecules and expression levels that are input. Among the top upstream regulators, the dopamine precursor levodopa (L-dopa) was predicted to be in an inhibited state (Z-score = − 3.023, p = 3.16E−03; Fig. 2C). Additionally, Reelin signaling in neurons (p = 6.15E−04, ratio = 0.103) was among the topranked canonical pathways regulated by forced abstinence length, although reelin (Reln) itself was unchanged due to forced abstinence. Lastly, Figure S2  Incubation effect. We used the "contrasts" feature of "DESeq2" in R to compare expression data from animals that experienced isolation housing during forced abstinence for 1 or 21 days (IC1D vs. IC21D). The top 30 (of 218 total differentially expressed) up-or down-regulated transcripts can be seen in Fig. 4A   All output data from IPA analysis of the enrichment effect can be found in the Table S4.  www.nature.com/scientificreports/ to be inhibited, suggesting phosphorylation of eIF-2-α is down when translation and cocaine-seeking are up. However, normalized expression counts from the enrichment comparison indicate that another eIF-2-α kinase, Eif2ak2 (also known as Pkr, Ppp1r83, or Prkr), showed increased expression in EE rats, as does Eif4e2. Thus, overall phosphorylation of Eif signaling is potentially increased in enrichment, which would lead to a reduction in protein translation according to the established mechanisms 45 . other functions and pathways related to the enrichment contrast. In addition to the IPA path-  (Fig. 5B). Many of these pathways were positively activated with EE compared to isolation, the opposite of that seen with the incubation effect above. Finally, IPA analysis indicates that Diseases and Functions significantly regulated by the Enrichment effect include similar groupings as seen above for the incubation effect, including Nervous System Development and Function, Neurological Disease, Cellular Assembly and Organization, Cellular Function and Maintenance, and Cell Death and Survival, as well as Post-Translational Modification and Protein Synthesis ( Figure S5).

Rt-qpcR results.
To validate the effects seen using RNA-seq, a selection of transcripts from a subset of the rats (21 days of forced abstinence; isolation: n = 6, enrichment: n = 7) was screened using RT-qPCR. RNA-seq results for Camk2d (ENSRNOT00000016026) showed a 9.01-log2 fold increase that was validated as a 1.30-fold increase using RT-qPCR ( Figure S6A). One of the Eif4e2 transcript isoforms (ENSRNOT00000026646) showed a log2 fold increase of 7.62 in the RNA-seq analysis that was validated as 1.69-fold increase with RT-qPCR (Figure S6B). It should be noted that RT-qPCR primers were designed from transcript-specific exon sequences. For example, the Eif4e2 transcript chosen (ENSRNOT00000026646) represents 1 of 6 splice variants in the Ensembl Rattus norvegicus library.

Discussion
The present study identified transcriptome differences in male rats with a history of cocaine self-administration that were housed in isolation or in an enriched environment during forced abstinence from cocaine for either 1 or 21 days. As expected from previous research 32, 33 , isolation housing during the 21-day forced abstinence period from cocaine self-administration resulted in robust cocaine-seeking behavior during a cue reactivity test, whereas EE housing, which included social, exercise, and novelty components, blunted cocaine-seeking behavior to levels similar to that observed with only 1 day of forced abstinence (Fig. 1A). A lead hypothesis of the mechanisms underlying the robust effects of EE on cocaine-related behaviors is stress innoculation 46 . Based on this theory, EE presents mild stress that fine tunes stress reactivity systems such that EE rats are better able to cope with the stress. In the present study, EE rats may better cope with stress associated with the cue reactivity test than rats housed in isolation, thereby reducing incentive to seek cocaine. Another feature of the present experiment that introduces differential stress across groups is the switch in environments in EE rats, but not isolated rats, during forced abstinence. Previous research from our lab employing a similar experimental design and timeline showed no differences in plasma corticosterone levels between enriched and isolated saline-yoked animals 47 . Furthermore, a decrease in plasma corticosterone levels was observed in enriched rats compared to isolated rats with a history of cocaine self-administration 47 . These previous findings suggest that EE rats adapt rapidly to potential stress effects of the switch from isolation to EE and suggest that the reduction of cocaine seeking in EE rats may involve stress inoculation. Given that RNA-seq identifies hundreds of regulated transcripts, it is important to validate findings via other measures of RNA or comparison to previous literature. In this study, the Signaling by Rho Family GTPases pathway was downregulated by 21 days of forced abstinence, similar to the results of a recent RNA-seq study of cocaine self-administration where this pathway was reported as the top-ranked canonical pathway 34 . Furthermore, this pathway has been shown to be downregulated by experimenter-delivered cocaine administration 48 . Together, these results suggest that the downregulation of the Rho associated pathway is in part due to cocaine history but potentially amplified by protracted, forced abstinence. Additionally, the retinoic acid receptor activation pathway was differentially regulated by cocaine across the two housing conditions, with upregulation in animals housed in enrichment and downregulation in animals housed in isolation. Similarly, Zhang and colleagues 34 previously found significant upregulation of the retinoic acid pathway due to housing in EE prior to cocaine self-administration. In addition, Camk2d and Eif4e2 transcripts were chosen for validation from a cohort of animals that experienced the same cocaine self-administration protocol and 21 days of abstinence in either isolation or enrichment. Both of these transcripts demonstrated fold increases in accordance with the RNA-seq data ( Figure S6). These findings together with the consistency of the present RNA-seq data with the previous literature offers validation of our analytical approach.
An exciting result from the RNA-seq analysis is that the robust transcriptomic differences observed with incubation (i.e., 1 vs. 21 days forced abstinence contrast in isolated rats) showed largely opposite changes with enrichment (i.e., EE versus isolation contrast in 21-day forced abstinent rats) within several pathways. Of the 155 regulated pathways in either the incubation or enrichment contrasts, 66 are shared between the two conditions. Furthermore, 49 of those 66 shared pathways show reverse activity levels between incubation and enrichment. For instance, Synaptic Long Term Potentiation has significantly upregulated activity in enrichment but downregulated activity in incubation. Additionally, Synaptogenesis Signaling Pathway has the identical patternupregulated activity in enrichment and downregulated in incubation. Additional  www.nature.com/scientificreports/ Calcium Signaling, and Glutamate Receptor Signaling, which have been implicated in cocaine-seeking behavior previously [49][50][51][52][53] . The Opioid Signaling Pathway was also differentially regulated in the incubation contrast, which is a pathway that has received relatively little attention as a mechanism of cocaine-seeking behavior but includes molecules such as Camk2d, Camk2g, Grin2d, Th, and Nfkb1 as contributing toward its significant regulation. The enrichment contrast implicated the Wnt/Ca ++ Pathway, Dendritic Cell Maturation, PI3K/AKT Signaling, and Dopamine-DARPP32 Feedback in cAMP Signaling pathways in the blunting effect of EE on cocaine-seeking behavior. Thus, environmental enrichment may prevent or reverse multiple gene network changes that take place during abstinence and that presumably contribute to incentive motivation for cocaine, which manifests as cocaine craving in individuals with CUD and is thought to be a major factor in relapse. The IPA analysis suggested additional pathways that exhibit differential regulation across abstinence period regardless of housing conditions and across housing conditions regardless of forced abstinence length. The abstinence effects were observed for Huntington's Disease Signaling and Reelin Signaling in Neurons pathways. The reelin pathway is a key contributor to the development of schizophrenia and other neuropsychiatric disorders such as depression and psychosis 54,55 , and these psychiatric conditions are often comorbid with substance use disorders 56,57 . Additionally, the reelin pathway modulates gamma-aminobutyric acid (GABA) transmission and neuroplasticity, two mechanisms thought to be involved in cocaine use disorders 58,59 . Housing condition effects were observed for ILK Signaling, a pathway known to be linked to cocaine sensitization and structural plasticity 60 , and Tcf7l2, a transcription factor commonly associated with diabetes but recently linked to nicotine consumption 61 .
The top pathways exhibiting whole transcriptome changes related to Diseases and Biological Functions highly overlapped between the incubation and EE contrasts, although the changes were in opposite directions. These include nervous system development, neurological disease, and several cellular functions, such as neuronal cell death (increased activity) and proliferation of neuronal cells (increased activity) in the enrichment contrast. By contrast, in the incubation contrast (Isolated 1 day vs. Isolated 21 days), the proliferation of neuronal cells pathway shows significantly reduced activity with prolonged withdrawal as well as a reduction in the neuronal cell death category. Under some experimental manipulations, the ventral striatum acquires new neurons (see review by Gould, 2007). For example, chronic infusion of the dopamine D3 receptor agonist, 7-OH-DPAT, into the subventricular zone leads to increased proliferation of new cells within the striatum of adult female rats 63 . Furthermore, intracerebroventricular delivery of platelet-derived growth factors and brain-derived neurotrophic factor (BDNF) in adult female rats leads to an increase in the number of newly generated neurons within the striatum 64 . While little attention has been given to adult neurogenesis within the nucleus accumbens with drugs of abuse, considerable research has shown that cocaine exposure reduces neurogenesis in the adult hippocampus, including in self-administration studies 65,66 . Interestingly, hippocampal neurogenesis functions have previously been implicated in post-mortem tissues of human cocaine users, with 3 molecules-liprin beta 1 (PPFIBP1), oligophrenin 1 (OPHN1), axotrophin (AXOT)-upregulated by cocaine, and 2 molecules-SH3 and multiple ankyrin repeat domains 2 (SHANK2) and semaphorin 6A (SEMA6A)-downregulated by cocaine 67 . In contrast, EE increases neurogenesis in the dentate gyrus of adult rats as well as a corresponding improvement in spatial memory in a water maze testing paradigm 68 . These findings allude to a loss of plasticity within NAcsh circuits involved in incubation, resulting in enhanced expression of the prepotent drug seeking response, and a "reboot" of plasticity by enrichment, which weakens the prepotency of drug seeking behavior.
An interesting overlap between the present study and previous research examining EE effects on cocaine selfadministration is the upregulation of transcripts in the eukaryotic initiation factor (EIF) signaling pathway 34 . Three EIF transcript variants were differentially expressed among our conditions: (1) Eif1b-201 is increased in prolonged forced abstinence; (2) Eif4e2-201 is increased in EE; and, (3) Eif2b2-201 is increased in incubation. Eif2 signaling pathway genes were also upregulated in isolated animals abstinent from cocaine for 21 days compared to 1 day, suggesting increased pathway activation.  previously observed that the top canonical pathway altered by EE housing prior to cocaine access was Eif2 signaling, which regulates mRNA translation. Recent studies in mice indicate that when expression levels of phosphorylated eIF-2-α, the protein product of Eif2s1, in the ventral tegmental area are reduced through mutation or blocked pharmacologically, vulnerability to nicotine 69 and cocaine 70 increases. Furthermore, when tobacco smokers express a single nucleotide polymorphism in the orthologous gene that encodes eIF-2-α in humans (which shares > 99% of the sequence with mice and rats), a reduction in reward-induced activity within the caudate and putamen nuclei is observed 69 . Control of translation by eIF-2-α phosphorylation is critical for regulation of synaptic strength and memory (see review by Costa-Mattioli et al. 71 ). These same processes in the amygdala have been shown to mediate reconsolidation of drug-conditioned place preference memory for morphine and cocaine as well as for heroin-associated cue memories following establishment of self-administration 72 . Furthermore, Werner et al. (2018) found that eIF-2-α is dephosphorylated when rats experience a test for cocaine-seeking following withdrawal from cocaine and thus translation is increased; however, when they then blocked dephosphorylation of eIF-2-α to restore inhibition of translation, cocaine seeking was reduced 73 . These findings combined with the effects observed on the Eif2 signaling pathway here highlight the importance of cocaine on translational control within the nucleus accumbens and that EE influences those mechanisms in a restorative manner.
Surprisingly, none of our contrasts included a significant portion of the ERK/MAPK Signaling pathway even though there are converging lines of evidence that this pathway is involved in cocaine-seeking behavior (see reviews [74][75][76] ). There are several differences between the previous studies and the present study that may account for this discrepancy, including differences in the location of brain sample 47 , measures of protein vs. transcript 35,[77][78][79] , and differences in experimental design regarding the control comparison (saline control or other cocaine-experienced groups) 34,35 and when EE was utilized (i.e., prior to self-administration or during abstinence) 34,35,77 . Other studies of ERK changes with drugs of abuse have focused on the amygdala or hippocampus, two regions known to be crucial for the development and persistence of CUD but not tested here (see www.nature.com/scientificreports/ reviews 80,81 ). Additional investigation in these other brain regions of the complex interaction between housing environment and duration of forced abstinence could potentially reveal region-specific differences in the ERK/ MAPK Signaling pathway crucial to our understanding of CUD.
In conclusion, EE provided as an intervention during forced abstinence from cocaine self-administration leads to numerous changes in mRNA expression in the NAcsh that impact multiple signaling pathways and neuronal functions, including synaptic and neuronal plasticity as well as neuroinflammation. Additionally, the EE intervention regulates the reelin pathway, similar to previous studies using EE prior to cocaine exposure. Importantly, the present findings suggest two novel pathways-translational control and neurogenesis-within the striatum that are ripe for further exploration using established techniques in drug addiction research. For example, neurogenesis and similar annotations within the Cellular Growth and Proliferation category are downregulated in animals with increased motivation, suggesting that treatments that promote neurogenesis given during forced abstinence may help prevent reinstatement. Additionally, the work discussed here illustrates that environmental enrichment reverses multiple dysregulated pathways that underlie drug seeking behavior, particularly the pathways that switched directions of expression between the incubation (isolated 1 day vs. isolated 21 days) and enrichment (isolated 21 days vs. enriched 21 days) groups. Together, these findings suggest that therapies, either behavioral, genetic, or pharmacologic, that target these pathways may be useful treatments for CUDs.

Methods
Animals. 62 male Sprague-Dawley rats (225-250 g on arrival; Charles River Laboratories, Wilmington, Massachusetts USA) were separated into standard, individual housing upon arrival (21.6 × 45.7 × 17.8 cm) with ad libitum access to food and water. Animals were handled daily for 5 days prior to surgery. All procedures were approved by the Institutional Animal Care and Use Committee of Arizona State University and were in accordance with institutional and NIH guidelines.
Surgery. Jugular vein catheters were implanted in all rats as described previously 82 under 2-3% isoflurane anesthesia. Rats were given meloxicam (1 mg/kg/ml, SC) and buprenorphine (0.05 mg/kg/ml, SC) analgesics, and cefazolin (30 mg/kg/ml, SC) antibiotic. The free end of the catheter was tunneled under the skin and cemented to the skull using dental acrylic. Post-operational care continued for a minimum of 5 days and included handling and administration of cefazolin (100 mg/kg/ml, IV).

Self-administration.
Animals initially were food restricted to 16 g standard chow per day with ad libitum access to water. Rats (n = 62) received 2-h cocaine (0.75 mg/kg/0.1 ml, IV) self-administration sessions 6 days per week at the same time of day. Upon completion of operant schedule on the active lever, the light and tone stimuli were activated followed 1-s later by infusion of cocaine administered over 6-s. An inactive lever was also available and presses were recorded but produced no consequences. A 20-s timeout with a lit house light followed termination of the infusion and compound conditioning stimuli. All rats began self-administration on a fixed-ratio 1 (FR1) schedule with in-session progression to a higher schedule if 7 infusions were achieved within 1-h. Schedule progression went to variable-ratio 2 (VR2), VR3, and VR5. After 3 consecutive sessions ending on VR5, rats started the next session at the next highest schedule (e.g., moving from FR1 to VR2 to start the session). The first day of self-administration was capped at 50 infusions to prevent overdose. After all selfadministration sessions, animals were fed an amount to maintain 90% free-feeding body weight, and administered IV cefazolin to ensure patency. As animals progressed from an FR1 to a VR5 schedule, food restriction was gradually removed, with eventual ad libitum access to food upon reaching stability on a VR5 schedule of reinforcement. Training continued on free feeding until rats achieved stability on a VR5 schedule across three consecutive sessions within 15% coefficient of variability for infusions. forced abstinence and enrichment. Animals remained in standard isolation housing (21.6 × 45.7 × 17.8 cm) or were transferred to an enriched environment for the duration of forced abstinence. EE housing consisted of 3-6 rats in a (74 × 91 × 36 cm) cage and standard bedding material. Additionally, a running wheel, tubes, toys, and nesting material was provided. Toys were exchanged every 3 days throughout forced abstinence to enhance novelty. Group assignment was counterbalanced for cocaine consumption. All rats were handled every 2 days during abstinence, including rats in isolation. The length of forced abstinence was either 1 or 21 days. cue reactivity testing. After forced abstinence concluded, animals were returned to their self-administration chambers for a 1-h cue reactivity test where the previously paired lever light and tone conditioned stimuli were presented for 7-s upon a press of the active lever, but no drug was available. The timeout and house light illumination previously paired with drug infusion were present. tissue collection and analysis. Immediately after the cue reactivity test, animals were anesthetized with isoflurane, decapitated, and the brain was removed and flash-frozen in methylbutane before storage in a − 70 °C freezer. Brains were dissected using a brain matrix. Based on anatomical landmarks on the ventral surface of the brain, sections 2 mm thick containing the striatum were taken with guidance from a rat brain atlas 83 . A 2.0 mm diameter punch (Harris Unicore™) of the NAcsh was then taken. RNA was isolated using standard Tri-zol® (#15596026; Invitrogen Corp., Carlsbad, California USA) extraction, as performed previously 38 . Samples were analyzed at the Arizona State University Genomics Core facility. RNA was prepared using the Nugen Ovation RNA-seq system (#7102-32; Tecan Genomics, Inc., Redwood City, California USA). cDNA was quantified Scientific RepoRtS | (2020) 10:11291 | https://doi.org/10.1038/s41598-020-67966-8 www.nature.com/scientificreports/ using Nanodrop and sheared to approximately 300 base pair fragments. Libraries were generated using the Kapa Biosystem library preparation kit. Fragments were then end repaired and A-tailed. Sequencing was performed on a 1 × 75 flow cell using an Illumina NextSeq500 instrument.
Data analysis. Fastq read files were uploaded to the public server at usegalaxy.org to analyze the data 84 .
Access to the workflow can be found here: https ://usega laxy.org/u/g.powel l/w/powel l-et-al---envir onmen talenric hment -rna-seq-workfl ow. Quality control analysis was performed for each file using FastQC, Galaxy Version 0.69 85 and aggregated using MultiQC, Galaxy Version 1.5.0 86 . An average of 58.43 million total reads were obtained per sample (± 3.78 million reads, SEM). Reads were aligned using HISAT2, Galaxy Version 2.1.0, a splice-aware alignment program 87 , and the Ensembl release 92 Rattus norvegicus reference genome 88 with average mapping at 89.28%. Transcript assembly and quantification were performed using StringTie, Galaxy Version 1.3.3.1 89 , and an assembly reference guide acquired from Ensembl. Only alignments that matched reference transcripts were processed. All output files were then merged into a new reference transcript, keeping the original reference annotations from Ensembl, using the Stringtie merge function Galaxy Version 1.3.3 before reassembly and quantification of aligned reads with the new merged reference guide. Differential expression analysis was performed using the DESeq2 package [90][91][92] in R 93 , with additional contrasts probed using the "contrasts" feature, when necessary.
ingenuity pathway analysis. Lists of differentially expressed transcripts with a p value ≤ 0.05 from DESeq2 output for each comparison were analyzed using Ingenuity Pathway Analysis (IPA, version 46901286; Qiagen, Hilden, Germany). Expression analysis settings included all available data sources using human, mouse, or rat species and the confidence level for analyses included experimentally observed outcomes and those predicted with high confidence. Measurement values for assigning directionality of expression were set to log fold change.
RT-qPCR. For mRNA analysis, cDNA was prepared with the SuperScript II First-Strand Synthesis system (Thermo Fisher Scientific, Waltham, Massachusetts USA). Quantification was performed using iTaq™ Universal SYBR® Green Supermix (Bio-Rad Laboratories, Hercules, California USA). Primers were designed against selected, splice variant-specific targets using Ensembl transcript sequences and the National Center for Biotechnology Information's Primer-Blast (Eif4e2 [ENSRNOT00000026646] Forward: GCA GAA CAT CCC CTG CAG TA, Reverse: GAC GGA CCA TGT GGC TGT AA; Camk2d [ENSRNOT00000016026] Forward: GTT GAA GAA ACC GGA TGG GG, Reverse: AGC GCC GGG GTA GGA ATA ) and compared against the housekeeping transcript, Gapdh (Forward: GTG CCA GCC TCG TCT CAT AG, Reverse: AAG AGA AGG CAG CCC TGG TA). Relative expression for each transcript was determined using the 2 -ΔΔCt method 94 , modified according to Pfaffl (2001) to incorporate real-time PCR efficiency for each gene target. With each test, a no reverse transcriptase (RT) reaction was included to act as a negative control; zero no-RT controls. Samples were run in triplicate for each transcript.
Statistical analysis. Analysis of lever presses and cocaine infusions was performed using ANOVAs in SPSS 25 (IBM Corp., Armonk, New York USA) and Prism 8.0 (Graphpad Software, San Diego, California USA). Significance for behavioral analysis was set at 0.05. All significant results have been reported here.

Data availability
All RNA-seq data have been deposited into the Gene Expression Omnibus with accession number GSE144606.