Abstract
Cocaine use disorder is a significant public health issue without an effective pharmacological treatment. Successful treatments are hindered in part by an incomplete understanding of the molecular mechanisms that underlie long-lasting maladaptive plasticity and addiction-like behaviors. Here, we leverage a large RNA sequencing dataset to generate gene coexpression networks across six interconnected regions of the brain's reward circuitry from mice that underwent saline or cocaine self-administration. We identify phosphodiesterase 1b (Pde1b), a Ca2+/calmodulin-dependent enzyme that increases cAMP and cGMP hydrolysis, as a central hub gene within a nucleus accumbens (NAc) gene module that was bioinformatically associated with addiction-like behavior. Chronic cocaine exposure increases Pde1b expression in NAc D2 medium spiny neurons (MSNs) in male but not female mice. Viral-mediated Pde1b overexpression in NAc reduces cocaine self-administration in female rats but increases seeking in both sexes. In female mice, overexpressing Pde1b in D1 MSNs attenuates the locomotor response to cocaine, with the opposite effect in D2 MSNs. Overexpressing Pde1b in D1/D2 MSNs had no effect on the locomotor response to cocaine in male mice. At the electrophysiological level, Pde1b overexpression reduces sEPSC frequency in D1 MSNs and regulates the excitability of NAc MSNs. Lastly, Pde1b overexpression significantly reduced the number of differentially expressed genes (DEGs) in NAc following chronic cocaine, with discordant effects on gene transcription between sexes. Together, we identify novel gene modules across the brain's reward circuitry associated with addiction-like behavior and explore the role of Pde1b in regulating the molecular, cellular, and behavioral responses to cocaine.
Significance Statement
Cocaine use disorder (CUD) is a major public health challenge without an effective pharmacological treatment. Here, we leverage a combination of genome-wide RNA sequencing, gene coexpression network analysis, and bioinformatic analyses of cocaine self-administration behavior to identify a role for phosphodiesterase 1b (Pde1b) in regulating maladaptive, addiction-like behavior. Our studies reveal cell-type– and sex-specific roles for Pde1b in regulating the molecular, cellular, and behavioral responses to cocaine, yielding insight into the molecular mechanisms by which cocaine induces maladaptive plasticity in the brain's reward circuity to drive addiction-like behavior. These discoveries guide directions for future research investigating the molecular basis of cocaine action and provide a pathway for therapeutic development for CUD.
Introduction
Cocaine use disorder (CUD) is characterized by compulsive drug seeking and taking, loss of control over drug intake, and a negative emotional state in the absence of drug (Koob and Volkow, 2010). CUD represents a significant public health challenge in the United States, with an estimated one million patients diagnosed with CUD and ∼20,000 overdose deaths involving cocaine annually. There remain no effective pharmacological treatments for CUD, highlighting the importance of identifying novel targets for therapeutic development (Kampman, 2019).
Gene network analyses have yielded insight into the underlying neurobiology of numerous brain disorders, including Parkinson's disease (Wang et al., 2019, 2021), Alzheimer's disease (Neff et al., 2021; Horgusluoglu et al., 2022), depression (Bagot et al., 2016; Labonté et al., 2017; Lorsch et al., 2019), and schizophrenia (Kim et al., 2021). Given the growing number of genome-wide transcriptomic analyses of the brain tissue from animal models (Walker et al., 2018; Mews et al., 2022; Browne et al., 2023) and human patients with substance use disorders (SUDs; Ribeiro et al., 2017; Seney et al., 2021; Mews et al., 2023), transcriptomic network analyses are poised for applications to addiction neuroscience.
In one study, RNA sequencing (RNA-seq) was performed on mice that self-administered saline or cocaine, followed by a 24 h or 30 d forced abstinence period and a saline or cocaine challenge (Walker et al., 2018). Six brain regions implicated in the pathophysiology of addiction (Koob and Volkow, 2010) were collected for this study: prefrontal cortex (PFC), ventral hippocampus (vHip), caudate/putamen (CPU), nucleus accumbens (NAc), ventral tegmental area (VTA), and basolateral amygdala (BLA). In addition, this study found three factors most closely associated with cocaine self-administration behavior using exploratory factor analysis: (1) cocaine intake, (2) discrimination between the active and inactive lever, and (3) compensatory active lever responses at higher fixed-ratio schedules. These factors were linearly transformed resulting in values between 0 and 1. The product of these values was calculated to generate an “addiction index (AI)”—a composite score estimating the severity of addiction-like behavior for each mouse, with higher AI values representing increased addiction-like behavior.
Here, we apply a multiscale embedded gene coexpression network analysis (MEGENA; Song and Zhang, 2015) to this published RNA-seq dataset and generate gene coexpression networks across the brain's reward circuitry. We analyzed brain region-specific modules by quantifying enrichment in either (1) differentially expressed genes (DEGs) or (2) genes significantly correlated with the AI. We identified NAc module 41 (M41) as having the highest enrichment in AI-associated genes of all modules across all six brain regions. Within NAc M41, phosphodiesterase 1b (Pde1b) is a central hub gene that shows a strong negative correlation with the AI. Therefore, we hypothesized that upregulating Pde1b expression in NAc would reduce addiction-like behaviors and may guide therapeutic development for CUD.
Phosphodiesterase enzymes (PDEs) catalyze the hydrolysis of the 3′-cyclic phosphate bond of cAMP and/or cGMP, terminating the intracellular signaling cascade (Bender and Beavo, 2006). There are 11 families of PDEs distinguished by their protein composition, pharmacological properties, tissue-specific expression, and affinity for cAMP versus cGMP. All type 1 PDEs (PDE1A, PDE1B, and PDE1C) are expressed in the brain and hydrolyze cAMP and cGMP in a Ca2+/calmodulin-dependent manner. PDE1B is highly expressed in brain regions that receive dense dopaminergic input from the VTA, suggesting that PDE1B is involved in terminating cAMP/cGMP signaling in response to dopaminergic neurotransmission.
Previous studies investigating the physiological functions of PDE1B within the brain have shown that PDE1B is involved in regulating diverse processes such as learning and memory (McQuown et al., 2019), anxiety- and depression-like behaviors (Hufgard et al., 2017a,b), and locomotor responses to psychostimulants (Reed et al., 2002; Ehrman et al., 2006; Siuciak et al., 2007). In this study, we focus on the cell-type– and sex-specific actions of Pde1b in NAc in regulating the molecular, cellular, and behavioral response to cocaine.
Materials and Methods
MEGENA
As described previously (Song and Zhang, 2015), gene coexpression networks were constructed for each brain region using the R package MEGENA from published RNA-seq data (Walker et al., 2018). Each node within a network represents a specific RNA transcript. A subset of nodes within given network are identified as hub genes, which are characterized by a high node strength and node degree with a 5% false discovery rate. Node degree is determined by a gene's total number of connections, referred to as “edges.” Node strength represents the sum of the absolute correlation coefficients of each edge (edge weights).
Modules with <50 genes were excluded from the downstream analyses. To measure the enrichment of DEGs and AI-associated genes across gene modules, we ranked the modules by the log10(adj. p value) with a fold enrichment cutoff of >5. Enrichment p values were calculated using Fisher's exact test and adjusted using the Benjamini–Hochberg procedure (Benjamini and Hochberg, 1995) for multiple testing. Module subnetworks were visualized using the software Cytoscape_v3.7 (Shannon et al., 2003).
Gene ontology
Gene ontology (GO) analysis was performed on all genes in NAc M41 using Enrichr (Xie et al., 2021). Biological process terms were filtered (adj. p value, ≤0.05) and ranked by −log10(adj. p value). The size of each data point is scaled by the number of genes assigned to that GO term.
Cell-type enrichment
Cell-type enrichment analysis was performed on all genes in the NAc M41 module using a previously published single-nuclei RNA–seq dataset from NAc of male and female rats (Savell et al., 2020). Feature matrices were obtained from Gene Expression Omnibus (GEO), and the R notebook Adult_NAc_Analysis.Rmd from https://github.com/Jeremy-Day-Lab/Science_Advances_2020 was used to process the datasets. The DoHeatmap function in the Seurat R package was used to generate a feature expression heatmap to show the expression of NAc M41 AI-associated genes in the single-nuclei dataset across all clusters (Hao et al., 2021).
Animals
Rodent experiments were performed in accordance with the Icahn School of Medicine at Mount Sinai's Institutional Animal Care and Use Committee. Male and female C57BL/6 mice were purchased from Jackson Laboratories. Male and female D1-Cre × CAG-Rpl10a-eGFP and D2-Cre × CAG-Rpl10a-eGFP mice were bred and maintained at Mount Sinai. Mice were group-housed in a 12 h light/dark cycle, with food and water available ad libitum. Mice were 8–12 weeks old during experimentation. Mouse experiments were performed during the light cycle. Male and female Sprague Dawley rats were purchased from Envigo. Rats were housed in a 12 h reverse light/dark cycle, with food and water available ad libitum. Rats were 11–15 weeks old during experimentation. Rats were group-housed until after the jugular vein catheterization (JVC), at which point they were single-housed. Treatment assignments were randomized, and the experimenter was blinded to the experimental group during all behavior experiments.
Stereotaxic surgery and viral constructs
Stereotaxic surgeries were performed as described previously (Hamilton et al., 2018). Briefly, mice were anesthetized with a mixture of ketamine (100 mg/kg) and xylazine (5 mg/kg) in sterile saline, while rats were anesthetized with inhaled isoflurane. Stereotaxic coordinates targeting the NAc were calculated relative to the bregma (mouse, +1.6 mm A/P, +1.4 mm M/L, −4.5 mm D/V, 10° angle; rat, +1.7 mm A/P, +2.3 mm M/L, −6.5 mm D/V, 10° angle). Viral constructs (AAV9-hSyn-mCherry-WPRE-pA; AAV9-hSyn-PDE1B-P2a-mCherry-WPRE-pA; AAV9-hSyn-DIO-mCherry-WPRE-pA; AAV9-hSyn-DIO-PDE1B-P2a-mCherry-WPRE-pA) were provided by the Duke Viral Vector Core and administered at a volume of 1 µl per brain hemisphere for mice and 1.2 l per brain hemisphere for rats (1 × 109 vg/µl) using a Hamilton syringe. The virus was administered at a rate of 0.1 µl per minute, and the virus was allowed to diffuse for 10 min prior to the removal of the Hamilton syringe. Experiments were conducted 3–4 weeks after stereotaxic surgery to allow for viral transduction.
Cocaine self-administration
Rats were anesthetized with isoflurane using a precision vaporizer, and a catheter was inserted into the jugular vein. Catheters were flushed daily with 0.1 ml of heparin dissolved in sterile saline (30 U/ml). Rats recovered from the JVC surgery for 5 d before beginning cocaine self-administration. Rat weights were measured daily, and the dose of cocaine was adjusted as necessary. Rats underwent cocaine self-administration on a fixed-ratio 1 (FR1) schedule with a dose of 0.8 mg/kg/inf. Each session was performed during the dark cycle and was 3 h in duration. Pressing the active lever illuminated a cue light (20 s) in the operant chamber signaling delivery of an infusion of cocaine and was followed by a time-out period (20 s) in which additional active lever responses were recorded but did not result in a programmed response. Inactive lever presses also did not result in a programmed response. Following acquisition of cocaine self-administration, rats underwent two sessions of a previously established behavioral economics paradigm (Oleson and Roberts, 2009; Calipari et al., 2018, 2019). In this paradigm, rats undergo a 110 min cocaine self-administration session with a descending series of cocaine doses (421, 237, 133, 75, 41, 24, 13, 7.5, 4.1, 2.4, 1.3 µg/inf), in which each dose is available for 10 min. Cocaine infusions during the behavioral economics task are on an FR1 schedule without a time-out period. Each behavioral economics session was followed by an FR1 self-administration session as described in Figure 2b. To measure relapse-like behavior, rats underwent a 30 min seeking task in which responding on the active lever triggered the cue light (20 s), but did not result in a cocaine infusion. The drug-free–seeking tests were performed 24 h and 30 d following the last FR1 session.
Drugs
Cocaine hydrochloride was obtained from the National Institute on Drug Abuse and dissolved in sterile 0.9% NaCl (saline). Cocaine solutions were prepared daily prior to experimentation.
RNA isolation and RT-qPCR
Mice were euthanized via cervical dislocation, and the brain was rapidly extracted and placed in chilled PBS. Coronal brain sections were performed at 1 mm thickness, and the NAc was collected using a 14 gauge needle and placed on dry ice. RNA isolation was performed using the Qiagen RNeasy Micro Kit according to manufacturer's instructions. cDNA conversion was performed using the Bio-Rad iScript cDNA Synthesis Kit. qPCR reactions were performed in triplicate using the TaqMan Fast Advanced Master Mix (Applied Biosystems) with TaqMan probes directed toward Pde1b (catalog #4331182). Expression levels were calculated using the ΔΔCt method relative to Hprt1 (catalog #4331182).
RNA-seq
RNA-seq libraries were generated using the SMARTer Stranded Total RNA-Seq Kit v2 (Takara Bio; catalog #634417) and the Indexing Primer Set HT for Illumina v2 (Takara Bio; catalog #634420). The quality of the libraries was assessed by Azenta Life Sciences using TapeStation and Qubit analysis prior to sequencing. Library sequencing was performed by Azenta Life Sciences with >20 million reads per sample. Raw sequencing reads were mapped to the mm10 genome using HISAT2 (Kim et al., 2015). Read counts mapping to genes were obtained using the featureCounts software of the Subread package against Ensembl v90 annotation (Liao et al., 2019). Differential expression analysis was performed using the DESeq2 package (Love et al., 2014). DEGs were determined using a p value of <0.05 and log2(fold change) >0.3. The ggplot2 R package was used to generate heatmaps comparing the various differential lists. Rank–rank hypergeometric overlap (RRHO) plots were generated using the RRHO2 package in R (Cahill et al., 2018). The raw RNA-seq data was deposited in the GEO (accession ID, GSE244767).
Fluorescence in situ hybridization
Mice were deeply anesthetized with a mixture of ketamine (100 mg/kg) and xylazine (5 mg/kg) in sterile saline. Transcardial perfusions were performed with chilled PBS followed by 4% PFA in 1× PBS. The brain was rapidly dissected and underwent fixation for 24 h in 4% PFA at 4°C. Overnight cryoprotection steps were performed by incubating the brains in 15 and 30% sucrose at 4°C. Cryostat tissue sections (25 μm) were collected and stored at −20°C. Fluorescent in situ hybridization was performed using the RNAscope Multiplex Fluorescent v2 assay kit with probes targeting Pde1b (catalog #435571), Drd2 (catalog #406501-C2), and Drd1 (catalog #461901-C3), paired with Opal (Akoya Biosciences) 520, 570, and 690, respectively. Images were collected using a laser scanning confocal microscope (LSM 900, Zeiss) with an oil objective at 40× magnification. Images were collected from three sections per mouse of the NAc core and shell and analyzed using Fiji (Version 2.3.0) and CellProfiler (Version 4.2.1; Stirling et al., 2021). Fluorescence intensity was recorded from individual nuclei within a section. The robust regression followed by outlier identification method (Q = 0.1%) was used to remove fluorescence intensity measurements from individual nuclei that were found to be statistical outliers.
Conditioned place preference
Conditioned place preference (CPP) was performed as described previously (Calipari et al., 2018). Briefly, the CPP apparatus consists of three contextually distinct chambers. During the pretest, mice were placed in the center of the box and allowed free access to all three chambers for 20 min. The side of the box that the mouse spent less time in was paired with cocaine in the conditioning sessions. During the 2 d of conditioning, mice received an intraperitoneal injection of saline in the morning and were allowed to explore the saline-paired chamber for 20 min. In the afternoon, mice received an intraperitoneal injection of cocaine and were allowed to explore the cocaine-paired chamber for 20 min. Throughout both conditioning days, locomotor activity was measured by quantifying the number of horizontal beam breaks that occurred during the 20 min session. During the post-test, mice were placed in the center of the box and allowed access to all three chambers of the box. CPP score is reported as the time spent in the cocaine-paired chamber minus the time spent in the saline-paired chamber during the post-test.
Locomotor activity
Mice underwent habituation for 2 d by administering an intraperitoneal injection of saline and placing a mouse in the locomotor box for 20 min before being returned to their home cage. For the next 5 d, mice were given an intraperitoneal injection of cocaine (20 mg/kg) and placed in the locomotor box for 20 min. Throughout the experiment, locomotor activity was measured by quantifying the number of horizontal beam breaks.
Electrophysiological recordings
Virally infected D1 or D2 medium spiny neurons (MSNs) in NAc core and shell were distinguished from noninfected cells based on Cre-dependent mCherry expression identified using epifluorescence microscopy. Brain slices were superfused with ACSF at a temperature between 29 and 31°C. For sEPSC and intrinsic membrane excitability recordings, electrodes (3–5 MΩ) were filled with a potassium-based internal solution (in mM: 130 KmeSO3, 10 KCl, 0.4 EGTA, 10 HEPES, 2.5 Mg-ATP, 0.25 Na-GTP, 2 MgCl2-6H2O; pH 7.3). To measure sEPSCs, we allowed NAc MSN cells 2–5 min to reach a stable baseline before a 3 min; a gap-free recording was performed at a membrane potential of −70 mV. To measure intrinsic membrane excitability, we held the cells at a membrane potential of −80 mV and administered a current-step protocol from −200 to 200 pA with a 50 pA increment. sEPSC and membrane excitability recordings were analyzed off-line via Clampfit.
Experimental design and statistical analyses
Statistical analyses were performed using GraphPad Prism (Version 9.4.0). Pairwise analyses were performed using a two-sided unpaired t test, nested t test, or a one-way ANOVA with Dunnett's multiple-comparison correction. If the variation of the within-subject means were found to be less than random error, an unpaired t test was performed instead of a nested t test. Experiments were analyzed using two- or three-way ANOVA and Sidak's post hoc test as appropriate. Statistical analyses were performed collapsed across sexes when there was no statistically significant effect of sex. Data are presented disaggregated across sexes as recommended when considering sex as a biological variable (Shansky, 2019). Sample sizes were based on previously published research, rather than a statistical power analysis.
Results
Circuit-wide transcriptomic network analysis
To identify gene networks across the brain's reward circuitry associated with addiction-like behavior, we generated gene coexpression modules using transcriptomic data from a previously published study comprised of six interconnected brain regions of male mice that underwent saline or cocaine self-administration followed by a 24 h or 30 d forced abstinence (Fig. 1a; Walker et al., 2018). The mice that underwent a 30 d forced abstinence were challenged with either a saline or cocaine injection and returned to the self-administration chambers 1 h prior to tissue collection.
MEGENA identifies gene networks associated with addiction-like behavior across the brain's reward circuitry in male mice. a, An overview of self-administration paradigm, RNA-seq across six brain regions (PFC, NAc, CPU, BLA, VTA, vHip), and MEGENA analysis to identify gene networks associated with addiction-like behavior. MEGENA modules were ranked by enrichment in genes associated with the AI or DEGs (Extended Data Tables 1-1, 1-2). b, Network structure of NAc M41. Each node within the network represents a gene, and the color of the node represents the gene's correlation with the AI (yellow, positive correlation; blue, negative correlation; gray, nonsignificant correlation). Pde1b, a NAc M41 hub gene, has a negative correlation with the AI (Extended Data Fig. 1-1). The lines connecting the nodes, commonly referred to as “edges,” indicate a statistically significant coexpression relationship. Hub genes (Extended Data Table 1-3) are scaled in size by node strength (see Materials and Methods, MEGENA). c, A dot plot depicting the normalized expression levels of the top 10 NAc M41 hub genes across all clusters within a single RNA-seq dataset (n = 15,631) from the NAc of adult male and female rats (Savell et al., 2020). The size of the dot represents the percentage of nuclei expressing a given gene. GO analysis reveals the known physiological functions of the genes in NAc M41 (Extended Data Table 1-4).
Fig 1-1
Pde1b expression in the NAc is negatively correlated with the addiction index in male mice that self-administered cocaine. Correlation between Pde1b expression and the addiction index in mice that self-administered cocaine (Pearson’s r = -0.7035, p = 0.0005) or saline (Pearson’s r = -6.690 × 10−2, p = 0.7855). Download Fig 1-1, TIF file.
Table 1-1
Top 10 MEGENA modules ranked by enrichment in genes associated with the addiction index. Modules are ranked by adjusted p-value (FE > 5). Download Table 1-1, XLSX file.
Table 1-2
Top 10 MEGENA modules ranked by enrichment in differentially expressed genes. Modules are ranked by adjusted p-value (FE > 5). Download Table 1-2, XLSX file.
Table 1-3
Top 10 hub genes in NAc M41. Download Table 1-3, XLSX file.
Table 1-4
Gene Ontology analysis of NAc M41 using Enrichr. Download Table 1-4, XLSX file.
We constructed gene coexpression modules for each brain region individually using MEGENA with the transcriptomic data from all self-administration cohorts. We ranked the gene coexpression modules across all six brain regions based on their enrichment in either (1) DEGs or (2) genes with a statistically significant correlation with the AI (Extended Data Tables 1-1, 1-2; Walker et al., 2018).
Among all the gene networks across all six brain regions, NAc M41 shows the highest enrichment in genes associated with the AI (p = 3.56 × 10−47; Fig. 1b, Extended Data Table 1-1). NAc M41 contains 181 genes in total, with 77 (42.5%) of those genes showing a statistically significant negative correlation with the AI. A subset of genes in NAc M41 met the criteria of hub genes (Extended Data Table 1-3), which are characterized by dense connections with other nodes in the network and high node strength (see Materials and Methods, MEGENA). To assess physiological functions of genes in NAc M41, we performed a GO enrichment analysis. We identified numerous processes relevant to neuronal function including the regulation of neurotransmitter signaling, synaptic plasticity, gene transcription, and neuronal morphology (Extended Data Table 1-4). In addition, we investigated the NAc cell types that express the genes in NAc M41 to gain insight into the underlying cellular mechanisms by which these genes could be contributing to addiction-like behaviors. Using a single-nuclei RNA–seq dataset from the NAc of adult male and female rats (Savell et al., 2020), we found that M41 hub genes show the highest levels of expression in D1 and D2 MSNs (Fig. 1c).
Pde1b overexpression in NAc neurons regulates cocaine self-administration and relapse-like behavior in male and female rats
Pde1b is one of the top hub genes in NAc M41, and levels of its expression in this brain region show a strong negative correlation with the AI in male mice (Extended Data Fig. 1-1, Extended Data Table 1-3). Pde1b shows statistically significant coexpression relationships with 23 other genes in NAc M41, most of which are also negatively correlated with the AI.
To investigate the role of Pde1b in regulating addiction-like behavior, we delivered a Pde1b or mCherry overexpression construct into the NAc of male and female rats (Fig. 2a) and performed cocaine self-administration as described in Figure 2b. We validated that the Pde1b overexpression construct significantly increases Pde1b expression in NAc neurons (Extended Data Fig. 2-1). Rats acquired cocaine self-administration behavior on an FR1 schedule. We found that Pde1b overexpression significantly decreases cocaine intake during self-administration in female rats (Fig. 2f–h), with no effect observed in male rats (Fig. 2c–e). The observations in female rats are consistent with our hypothesis that Pde1b expression in the NAc is associated with reduced addiction-like behavior (Extended Data Fig. 1-1).
Pde1b overexpression in the NAc reduces cocaine self-administration in female rats but elevates relapse-like behavior after prolonged abstinence in both sexes. a, Representative confocal microscopy image of viral targeting to the NAc (10× magnification). b, Experimental timeline of cocaine self-administration in male (n = 6 mCherry; n = 6 PDE1B) and female (n = 7 mCherry; n = 8 PDE1B) rats. We validated that Pde1b overexpression in the NAc significantly increases Pde1b expression (Extended Data Fig. 2-1). c–h, Combined across sexes, the overexpression of Pde1b in the NAc reduces cocaine intake [repeated measures (RM) ANOVA; virus, F(1,23) = 4.774; p = 0.0393; sex, F(1,23) = 0.1590; p = 0.6938] but has no significant effect on active (RM ANOVA; virus, F(1,23) = 0.4654; p = 0.5019; sex, F(1,23) = 0.02499; p = 0.8758) or inactive lever presses (mixed-effect model; virus, F(1,23) = 0.5298; p = 0.4740; sex, F(1,23) = 0.4213; p = 0.5227). Given the intrinsic sex differences in cocaine self-administration behavior and that the effect of Pde1b overexpression on infusions is driven primarily by female rats, we present the data and statistical analysis disaggregated by sex. In male rats, the overexpression of PDE1B in the NAc has no effect on active lever presses (RM ANOVA; virus, F(1,10) = 0.2473; p = 0.6298), inactive lever presses (mixed-effect model; virus, p = 0.3427), or cocaine intake (RM ANOVA; virus, F(1,10) = 0.5653; p = 0.4695) during the acquisition of cocaine self-administration behavior. In female rats, the overexpression of PDE1B in the NAc reduced cocaine intake (RM ANOVA; virus, F(1,13) = 11.37; p = 0.0050) in female rats. No significant differences were observed in active (RM ANOVA; virus, F(1,13) = 0.1907; p = 0.6695) or inactive (RM ANOVA; virus, F(1,13) = 1.318; p = 0.2717) lever presses in female rats. i, j, Pde1b overexpression decreases active lever responses during a thresholding task in female rats (RM ANOVA; virus, F(1,132) = 5.756; p = 0.0178) but has no effect in male rats (RM ANOVA; virus, F(1,88) = 3.497 × 10−2; p = 0.8521). K, L, PDE1B overexpression in the NAc potentiates active lever responses in a drug-free–seeking test after a 30 d abstinence period in male and female rats [RM ANOVA; virus, F(1,22) = 16.36; p = 0.0005; sex, F(1,22) = 10.12; p = 0.0043; Sidak's post hoc test (mCherry males vs PDE1B males), p = 0.0171; Sidak's post hoc test (mCherry females vs PDE1B females), p = 0.0186]. No significant effects of Pde1b overexpression were observed after 24 h forced abstinence (RM ANOVA; virus, F(1,22) = 8.922 × 10−4; p = 0.9764; sex, F(1,22) = 5.183; p = 0.0329). Data are presented as the mean ± SEM.
Fig 2-1
Pde1b overexpression in the NAc has no effect on conditioned place preference for cocaine in WT male mice. A) Overview of the experimental design for the conditioned place preference paradigm. B) Depiction of the neuron-specific control and Pde1b overexpression plasmid constructs packaged into AAV9 for intra-NAc delivery. C) Overexpression of Pde1b in the NAc significantly increases Pde1b RNA expression (Two-tailed t-test; p < 0.0001; n = 12 per group). D) Overexpression of Pde1b in the NAc has no effect on conditioned place preference for cocaine (two-way ANOVA; Virus: F1,44 = 1.218, p = 0.2758). Data are presented as the mean ± SEM. Download Fig 2-1, TIF file.
Following the acquisition of cocaine self-administration behavior, rats underwent a previously established behavioral economics paradigm (Fig. 2b; Oleson and Roberts, 2009; Calipari et al., 2018, 2019). This model involves a 110 min cocaine self-administration session in which the dose of each cocaine infusion decreases every 10 min. Consequently, the “price” of cocaine (no. of responses/milligram) increases over time. Rats increase active lever responses to maintain stable cocaine intake until reaching a threshold, at which point responding declines. Behavioral economics is used to measure motivation for drug seeking and taking (Venniro et al., 2020), a defining characteristic of SUDs (Koob and Volkow, 2010). Here, we demonstrate that Pde1b overexpression in NAc neurons significantly reduces active lever responses during the behavioral economic task in female rats (Fig. 2j), particularly at low doses of cocaine, with no effect in male rats (Fig. 2i). In addition, male rats exhibited a lower threshold for responding than female rats during the behavioral economics task, which may reflect intrinsic sex differences in the motivation for cocaine.
Lastly, we measured susceptibility to relapse-like behavior by performing a 30 min drug-free–seeking test 24 h and 30 d following the final FR1 cocaine self-administration session (Fig. 2b). During this task, responses on the active lever would trigger the cue light (20 s), but would not result in drug infusion. The overexpression of Pde1b in NAc neurons significantly increased active lever responses in this drug-free–seeking test in both sexes following the 30 d home-cage forced abstinence (Fig. 2l), but had no effect after 24 h (Fig. 2k).
Cocaine regulates the expression of Pde1b in NAc D2 MSNs in male mice
After establishing a role for Pde1b in the NAc in regulating addiction-like behavior, we investigated how cocaine influences endogenous levels of Pde1b in the brain's reward circuitry of male mice. Under normal physiological conditions, Pde1b is expressed predominantly in brain regions involved in dopaminergic neurotransmission including the striatum, hippocampus, and pallidum. Following cocaine self-administration, Pde1b expression levels remain unchanged in bulk tissue samples from the NAc, CPU, PFC, BLA, and VTA of male mice (Extended Data Fig. 3-1; Walker et al., 2018). In the vHip of male mice, Pde1b RNA expression is downregulated after acute cocaine and after 30 d of forced abstinence from cocaine self-administration.
To investigate transient regulation of Pde1b, we quantified Pde1b mRNA expression in the NAc tissue from male mice treated with acute or chronic cocaine 6 h after the last injection (Fig. 3a). Consistent with our previous observations, Pde1b mRNA expression remained unchanged in the NAc (Fig. 3b). However, Pde1b is expressed predominantly in D1 and D2 MSNs, compared with all other cell types, in the NAc of rats (Savell et al., 2020) and humans (Tran et al., 2021). Some genes are oppositely regulated by cocaine in D1 versus D2 MSNs (Mews et al., 2022), which may be obscured by measuring Pde1b RNA expression in bulk NAc tissue. To investigate the cell-type–specific regulation of Pde1b in NAc MSN subtypes, we performed RNA fluorescence in situ hybridization on NAc tissue from male and female mice treated chronically with saline or cocaine (Fig. 3c,d). We found that chronic cocaine significantly increases the percentage of Drd2 + nuclei that express Pde1b in the NAc of male mice, with no significant effects observed in female mice (Fig. 3e–h).
Chronic cocaine increases Pde1b expression in D2 MSNs of NAc in male mice. a, Experimental timeline of saline or cocaine injections in WT male mice (saline, n = 6; acute cocaine, n = 7; chronic cocaine, n = 9). b, Acute or chronic injections of cocaine (20 mg/kg) had no effect on Pde1b expression in bulk NAc tissue (one-way ANOVA; Dunnett's multiple-comparison test, saline vs acute cocaine, adj. p = 0.7445; saline vs chronic cocaine, adj. p = 0.9992). Pde1b expression values were normalized to Hprt1. Similarly, cocaine self-administration has no effect on NAc Pde1b expression (Extended Data Fig. 3-1). c, Experimental timeline of chronic saline or cocaine injections in WT mice (n = 4 per group). d, Representative confocal microscopy (40× objective) images representing DAPI (blue), Pde1b (green), Drd1 (red), and Drd2 (yellow) RNAscope in situ hybridization staining in the NAc. The fluorescence intensity of Pde1b was measured in Drd1+ and Drd2+ nuclei from male (saline Drd1+ nuclei, n = 291; cocaine Drd1+ nuclei, n = 435; saline Drd2+ nuclei, n = 304; cocaine Drd2+ nuclei, n = 409) and female (saline Drd1+ nuclei, n = 492; cocaine Drd1+ nuclei, n = 481; saline Drd2+ nuclei, n = 545; cocaine Drd2+ nuclei, n = 628) mice treated with chronic saline or cocaine (n = 4 mice per group). We present the data and statistical analysis disaggregated by sex due to observed sex differences in Pde1b intensity measurements (ANOVA; sex, F(1,24) = 29.00; p < 0.0001). e, In male mice, chronic cocaine has no effect on Pde1b intensity in Drd1+ and Drd2+ nuclei (two-way ANOVA; cocaine, F(1,12) = 1.509; p = 0.2428; cell type, F(1,12) = 0.9592; p = 0.3467). f, Chronic cocaine increases the percentage of Drd2+ nuclei expressing Pde1b (two-way ANOVA; cocaine, F(1,12) = 11.21; p = 0.0058; Drd1+ nuclei Sidak's post hoc test; p = 0.1546; Drd2+ nuclei Sidak's post hoc test, p = 0.0303). g–h, In female mice, chronic cocaine has no effect on Pde1b intensity (two-way ANOVA; cocaine, F(1,12) = 1.044; p = 0.3270) or the percentage of nuclei expressing Pde1b (two-way ANOVA; cocaine, F(1,12) = 0.1656; p = 0.6912). Data are presented as the mean ± SEM.
Fig 3-1
Cocaine regulates the expression of Pde1b mRNA in the vHip of male mice. A) Overview of self-administration conditions and brain regions collected from male mice for RNA-seq in a previously published study (Walker et al., 2018). B-G) Cocaine self-administration has no effect on Pde1b mRNA expression in the BLA, CPU, NAc, PFC, or VTA. In the vHip, Pde1b levels are decreased 1 hour after acute cocaine administration and following 30 days of forced abstinence from cocaine self-administration (One-way ANOVA; Dunnett’s multiple comparisons test: SC vs. SS, adj. p = 0.0650; CS v SS, adj. p = 0.0311). Data are presented as the mean ± SEM. Download Fig 3-1, TIF file.
Pde1b oppositely regulates locomotor responses to cocaine in NAc MSN subtypes in female mice
D1 and D2 MSNs comprise ∼95% of the neuronal population of the NAc. These cell types differ in their intracellular responses to dopaminergic signaling, transcriptomic profile (Kronman et al., 2019; Mews et al., 2022; Teague and Nestler, 2022), afferent and efferent projections (Zinsmaier et al., 2022), and effects on addiction-like behavior (Lobo et al., 2010; Kravitz et al., 2012; Calipari et al., 2016; Soares-Cunha et al., 2016, 2019; Cole et al., 2018; Pardo-Garcia et al., 2019). Therefore, we hypothesized that the effects of Pde1b overexpression on addiction-like behavior differ between MSN subtypes. To investigate the cell-type–specific functions of Pde1b, we designed a Cre-dependent Pde1b overexpression construct and delivered it into the NAc of D1- or D2-Cre male and female mice (Fig. 4b).
The overexpression of Pde1b in D1 or D2 MSNs in NAc of male and female mice oppositely regulates the acute locomotor response to cocaine, but not CPP. a, A timeline of CPP in D1-Cre (n = 11 mCherry males; n = 12 mCherry females; n = 12 PDE1B males; n = 13 PDE1B females) and D2-Cre (n = 9 mCherry males; n = 13 mCherry females; n = 12 PDE1B males; n = 10 PDE1B females) mice. b, Cre-dependent mCherry and Pde1b overexpression plasmids. c, d, The overexpression of Pde1b has no effect on CPP for cocaine in D1 (two-way ANOVA; virus, F(1,44) = 1.118; p = 0.2960; sex, F(1,44) = 2.038; p = 0.1605) or D2 (two-way ANOVA; virus, F(1,38) = 1.163; p = 0.2877; sex, F(1,38) = 0.6587; p = 0.4221) MSNs. e, f, During Day 1 of conditioning, Pde1b overexpression reduces the cocaine-induced locomotor response in D1 MSNs (ANOVA; virus, F(1,45) = 5.092; p = 0.0289; sex, F(1,45) = 5.108; p = 0.0287; virus:sex interaction, F(1,45) = 0.7589; p = 0.3883; virus:cocaine interaction, F(1,45) = 3.654; p = 0.0623) but increases the cocaine-induced locomotor response in D2 MSNs (ANOVA; virus, F(1,38) = 7.288; p = 0.0103; sex, F(1,38) = 0.3286; p = 0.5699; virus:sex interaction, F(1,38) = 2.240; p = 0.1428; virus:cocaine interaction, F(1,38) = 5.705; p = 0.0220). The overexpression of Pde1b had no effect on the locomotor response to saline administration in D1 or D2 MSNs. Data are presented as the mean ± SEM collapsed across sexes.
First, we applied this approach to the CPP assay (Fig. 4a) in male and female mice to determine how Pde1b overexpression influences reward processing in response to cocaine. Previous work has implicated the cAMP-PKA-CREB signaling pathway in regulating cocaine CPP behavior (Carlezon et al., 1998; McClung and Nestler, 2003; Teague and Nestler, 2022). For example, the overexpression of CREB in all NAc neurons reduces CPP behavior for cocaine, suggesting that dampening the cAMP-PKA-CREB pathway by overexpressing Pde1b would increase CPP for cocaine. However, we observed that Pde1b overexpression in NAc D1 or D2 MSNs had no effect on cocaine CPP behavior in male or female mice (Fig. 4c,d). Similarly, Pde1b overexpression in all NAc neurons of WT male mice had no effect on CPP for cocaine (Extended Data Fig. 2-1). These observations may be due to the combinatorial effect of Pde1b overexpression in terminating cGMP as well as cAMP signaling and downstream effects on CREB phosphorylation (Mattson et al., 2005). In future studies, manipulating the expression of a PDE with single-substrate specificity (cAMP or cGMP) would be useful in delineating the contribution of these signaling cascades to the development of CPP for cocaine. Although Pde1b overexpression had no effect on CPP for cocaine, we observed that Pde1b overexpression in NAc D1 versus D2 MSNs oppositely regulated the acute locomotor response to cocaine during Day 1 of CPP conditioning in male and female mice (Fig. 4e,f).
It is well established that in response to the repeated administration of cocaine or many other classes of drugs of abuse, rodents show elevated locomotor responses over time. This process, commonly referred to as locomotor sensitization, involves remodeling of synaptic inputs onto NAc neurons (Kalivas, 2009; Brown et al., 2011). Given the effect of Pde1b overexpression on the acute locomotor response to cocaine, we overexpressed Pde1b in NAc D1 or D2 MSNs and analyzed locomotor responses to repeated cocaine doses in male and female mice (Fig. 5a,b). In female mice, Pde1b overexpression in D1 MSNs blunted locomotor responses to repeated cocaine, with the opposite effect seen for D2 MSNs (Fig. 5d,f). By contrast, overexpressing Pde1b in D1 or D2 MSNs in male mice had no effect on locomotor sensitization (Fig. 5c,e). Collectively, these data indicate that Pde1b overexpression in NAc neurons exerts cell-type– and sex-specific effects on locomotor responses to cocaine, which may reflect underlying sex differences in the molecular mechanisms that lead to escalating locomotor activation over time in male and female mice.
Pde1b overexpression oppositely regulates cocaine-induced locomotor responses in NAc D1 and D2 MSNs female mice, but not male mice. a, A timeline for locomotor activity measures in D1-cre male mice (n = 9 mCherry; n = 9 Pde1b), D1-cre female mice (n = 6 mCherry; n = 6 Pde1b), D2-cre male mice (n = 6 mCherry; n = 7 Pde1b), and D2-cre female mice (n = 9 mCherry; n = 8 Pde1b). b, Cre-dependent mCherry and Pde1b overexpression plasmids. c–f, Combined across sexes, PDE1B overexpression reduces the locomotor response to chronic cocaine in D1 MSNs (ANOVA; virus, F(1,26) = 5.165; p = 0.0316; sex, F(1,26) = 4.734; p = 0.0388; virus:sex interaction, F(1,26) = 2.354; p = 0.1371), with the opposite effect observed in D2 MSNs (ANOVA; virus, F(1,26) = 5.021; p = 0.0338; sex, F(1,26) = 1.775, p = 0.1943; virus:sex interaction, F(1,26) = 8.137; p = 0.0084). Given the observed sex differences in locomotor response following Pde1b overexpression and cocaine administration, we present the data and statistical analysis disaggregated by sex. In D1 MSNs, Pde1b overexpression decreased cocaine-induced locomotor responses in female mice (RM ANOVA; virus, F(1,10) = 6.663; p = 0.0274), but not male mice (RM ANOVA; virus, F(1,16) = 0.3009; p = 0.5909). In D2 MSNs, Pde1b overexpression increased cocaine-induced locomotor responses in female mice (RM ANOVA; main effect of virus, F(1,15) = 20.54; p = 0.0004), with no effect in male mice (RM ANOVA; main effect of virus, F(1,11) = 0.2514; p = 0.6260). Data are presented as the mean ± SEM.
Pde1b overexpression alters the physiological properties of NAc MSNs in male and female mice
We then explored the physiological alterations correlated to the behavioral effects of Pde1b overexpression in NAc neurons. We overexpressed Pde1b selectively in NAc D1 or D2 MSNs in male and female mice and made whole-cell recordings of NAc MSNs after >3 weeks of viral expression. The overexpression of Pde1b reduced the sEPSC frequency in NAc D1 MSNs (Fig. 6a) without affecting this parameter in D2 MSNs (Fig. 6c). Pde1b overexpression did not change the amplitude of sEPSCs (Fig. 6b,d) or passive membrane properties (Fig. 6g,i) of D1 or D2 MSNs. The intrinsic membrane excitability of MSNs was assessed by measuring the frequency of action potentials evoked by current injections. Pde1b overexpression reduced the membrane excitability of D1 MSNs in female mice (Fig. 6j), with no effect in male mice (Fig. 6i). By contrast, Pde1b overexpression increased the membrane excitability of D2 MSNs in male mice (Fig. 6k), with no significant effect in female mice (Fig. 6l).
Pde1b overexpression regulates the electrophysiological activity of D1 and D2 MSNs in the NAc of male and female mice. a, b, Pde1b overexpression reduces sEPSC frequency in D1 MSNs (nested t test, t(9) = 2.453; p = 0.0366; mCherry group, n = 22 cells; n = 6 mice; Pde1b group, n = 15 cells; n = 6 mice) but has no effect on sEPSC amplitude in D1 MSNs (nested t test, t(9) = 0.8943; p = 0.3945). Sex differences in D1 MSN sEPSC frequency (two-way ANOVA; sex, F(1,33) = 1.310; p = 0.2606; sex:virus interaction, F(1,33) = 0.9009; p = 0.3494) and amplitude (two-way ANOVA; sex, F(1,35) = 1.434; p = 0.2392; sex:virus interaction, F(1,35) = 0.7673; p = 0.3870) were not observed. c, d, Pde1b overexpression has no effect on sEPSC frequency (nested t test, t(10) = 0.1824; p = 0.8589; mCherry group, n = 25 cells; n = 6 mice; Pde1b group, n = 20 cells; n = 6 mice) or amplitude (unpaired t test, t(43) = 1.177; p = 0.2458) in D2 MSNs. Sex differences in D2 MSN sEPSC frequency (two-way ANOVA; sex, F(1,41) = 3.486; p = 0.0691; sex:virus interaction, F(1,41) = 1.866; p = 0.1794) and amplitude (two-way ANOVA; sex, F(1,41) = 0.5838; p = 0.4492; sex:virus interaction, F(1,41) = 0.4599; p = 0.5015) were not observed. e, f, Representative sEPSC traces from D1 (left) and D2 (right) MSNs. g, h, Pde1b overexpression has no effect on passive membrane properties in D1 (ANOVA; virus, F(1,35) = 0.7135; p = 0.4040; sex, F(1,35) = 4.508; p = 0.0409; sex:virus interaction, F(1,35) = 2.802; p = 0.1030) or D2 (ANOVA; virus, F(1,40) = 1.048; p = 0.3121; sex, F(1,40) = 3.113 × 10−3; p = 0.9558; virus:sex interaction, F(1,40) = 1.990; p = 0.1661) MSNs. i, j, We observed sex differences in the effect of Pde1b overexpression on the excitability of D1 (ANOVA; virus, F(1,28) = 4.446; p = 0.0441; sex, F(1,28) = 4.987; p = 0.0337; virus:sex interaction, F(1,28) = 13.58; p = 0.0010) and D2 (ANOVA; virus, F(1,32) = 170.2; p < 0.0001; sex, F(1,32) = 1.084; p = 0.3056; virus:sex interaction, F(1,32) = 15.97; p = 0.0004) MSNs. Pde1b overexpression reduces excitability in D1 MSNs of female mice (two-way ANOVA; virus, F(1,12) = 24.23; p = 0.0004), but not male mice (two-way ANOVA; virus, F(1,16) = 1.090; p = 0.3120). k, l, Pde1b overexpression increases excitability in D2 MSNs of male mice (two-way ANOVA; virus, F(1,16) = 22.69; p = 0.0002), but not female mice (two-way ANOVA; virus, F(1,16) = 3.530; p = 0.0786). m, n, Representative action potentials evoked by 200 pA injection currents. Data are presented as the mean ± SEM.
Pde1b overexpression in NAc neurons regulates transcriptomic responses to cocaine in male and female mice
To investigate the role of Pde1b in regulating cocaine-induced transcriptomic adaptations, we overexpressed Pde1b in NAc neurons, treated male and female mice with 10 d of saline or cocaine (20 mg/kg), and collected NAc tissue for RNA-seq 24 h after the last injection (Fig. 7a). Using RNA-seq RPKM values, we validated successful Pde1b overexpression in mice treated with saline or cocaine (Fig. 7b). We further validated that the degree of Pde1b overexpression did not differ significantly between sexes. Male and female mice in the mCherry group showed a similar pattern of DEGs after chronic cocaine, albeit with fewer DEGs observed in female mice (Fig. 7c, Extended Data Tables 7-1, 7-2). Pde1b overexpression significantly reduced the number of DEGs observed after chronic cocaine in both sexes (Extended Data Tables 7-3, 7-4). A subset of the genes that are differentially expressed following Pde1b overexpression are contained in NAc M41, such as Slc8a2, Glra2, and Rbms2. In addition, we observed a minimal overlap in the cocaine-induced DEGs between male and female mice following Pde1b overexpression (Fig. 7d–e; Extended Data Tables 7-5, 7-6). These data suggest that Pde1b induces distinct transcriptional changes in the NAc of male versus female mice, which may contribute to the sex differences observed in cocaine self-administration and locomotor responses.
Pde1b overexpression regulates the transcriptomic response to cocaine in the NAc of male and female mice. a, Experimental timeline (mCherry males, n = 5; mCherry females, n = 5; PDE1B males, n = 5; PDE1B females, n = 5). b, Validation of PDE1B overexpression (two-way ANOVA, F(1,36) = 59.75; p < 0.0001) in mice treated with saline (Sidak's post hoc test, p < 0.0001) or cocaine (Sidak's post hoc test, p < 0.0001). These data were analyzed collapsed across sexes, given that sex differences in Pde1b expression levels following Pde1b overexpression were not observed (ANOVA; virus, F(1,32) = 54.89, p < 0.0001; sex, F(1,32) = 0.1875; p = 0.6679; virus:sex interaction, F(1,32) = 0.2082; p = 0.6521). c, Heatmaps comparing DEGs (Extended Data Tables 7-1, 7-2, 7-3, 7-4) after chronic cocaine across viral treatment groups and sexes [p value <0.05; log2(fold change) >0.3]. d–e, RRHO plots comparing the overlap of cocaine-induced changes in gene expression between sexes in mCherry and PDE1B mice (Extended Data Tables 7-5, 7-6). Data are presented as the mean ± SEM.
Table 7-1
DEGs in mCherry male mice following chronic cocaine. Download Table 7-1, XLSX file.
Table 7-2
DEGs in mCherry female mice following chronic cocaine. Download Table 7-2, XLSX file.
Table 7-3
DEGs in PDE1B male mice following chronic cocaine. Download Table 7-3, XLSX file.
Table 7-4
DEGs in PDE1B female mice following chronic cocaine. Download Table 7-4, XLSX file.
Table 7-5
RRHO p values comparing DEGs after chronic cocaine in mCherry male vs mCherry female mice. As described previously (Cahill et al., 2018), each cell represents the hypergeometric p-value calculated by overlapping subsets of genes in mCherry male mice (rows) and mCherry female mice (columns). The subsets of genes are obtained by first sorting all genes from their respective dataset by the log-fold change and the p-value and using a sliding-window approach to select the genes for each subset. Download Table 7-5, XLS file.
Table 7-6
RRHO p values comparing DEGs after chronic cocaine in PDE1B male vs PDE1B female mice. As described previously (Cahill et al., 2018), each cell represents the hypergeometric p-value calculated by overlapping subsets of genes in PDE1B male mice (rows) and PDE1B female mice (columns). The subsets of genes are obtained by first sorting all genes from their respective dataset by the log-fold change and the p-value and using a sliding-window approach to select the genes for each subset. Download Table 7-6, XLS file.
Discussion
In this study, we constructed gene coexpression networks across six interconnected regions of the brain's reward circuitry from male mice that underwent saline or cocaine self-administration using RNA-seq data from a previously published study (Walker et al., 2018). To investigate gene coexpression networks involved in addiction-like behavior, we quantified module enrichment in genes associated with the AI. We identified PDE1B as a prominent hub gene within NAc M41, the gene network with the highest enrichment in AI-associated genes among all gene modules.
First, we examined the functional characteristics of NAc M41 using a combination of differential expression, GO, and cell-type enrichment analyses (Fig. 1). Surprisingly, although ∼43% of the 181 genes in NAc M41 show a statistically significant negative correlation with the AI, few of these genes show evidence of differential regulation after cocaine self-administration (Walker et al., 2018). Therefore, a sole focus on gene networks enriched in DEGs would have overlooked NAc M41 as being involved in the development of addiction-like behaviors. Using published single-nuclei RNA–seq data from the NAc of male and female rats (Savell et al., 2020), we showed that NAc M41 hub genes are primarily enriched in D1 and D2 MSNs. GO analysis demonstrated that the genes in this network regulate numerous physiological functions within the nervous system.
The top three hub genes in NAc M41 are Pde1b, Camk2b, and Ptk2b, all of which have been linked to SUDs. Pde1b and Ptk2b regulate acute locomotor responses to psychostimulants (Ehrman et al., 2006; de Pins et al., 2020), while a single-nucleotide polymorphism in Camk2b was linked to opioid dependence (Gelernter et al., 2014). Furthermore, a recent large-scale GWAS linked SNPs in PDE4B and PDE1C to greater susceptibility to developing SUDs (Hatoum et al., 2023). Other genes in NAc M41 include Drd2, members of transcription factor families implicated in addiction-like behavior (Teague and Nestler, 2022), intracellular messengers downstream of GPCR signaling, and nucleosome-remodeling complexes. These observations suggest that pairing gene coexpression network analysis with the AI was effective in identifying a subset of genes previously linked to addiction-like behavior, as well as additional targets for future investigation.
Consistent with our hypothesis that increased Pde1b expression is associated with reduced addiction-like behavior, we found that Pde1b overexpression in NAc neurons significantly reduced cocaine self-administration and active lever responses in a behavioral economics task in female rats, with no effect in male rats. These results suggest that the overexpression of Pde1b in NAc neurons reduces the motivation for cocaine in female rats, a hallmark characteristic of SUDs (Koob and Volkow, 2010). However, these sex differences were surprising considering that the data used to generate the MEGENA modules and AI analyses were derived exclusively from male mice (Walker et al., 2018). Nonetheless, we opted to include both sexes in all of the behavioral paradigms used in this study (Shansky, 2019) and identified a prominent role for Pde1b in attenuating cocaine self-administration in female rats.
Male and female rats subsequently underwent a drug-free–seeking test 24 h and 30 d after the last FR1 cocaine self-administration session. This translationally relevant approach is thought to model relapse susceptibility by measuring active lever responses after a home-cage forced abstinence period (Marchant et al., 2013; Venniro et al., 2016). Pde1b overexpression in NAc neurons significantly increased relapse-like behavior after a 30 d forced abstinence period, with no effect after 24 h. Thus, while Pde1b in NAc neurons is associated with a reduction in cocaine self-administration behavior in female rats, it is also related to heightened relapse-like behavior following prolonged abstinence in both sexes.
Within the NAc, Pde1b is expressed most abundantly in D1 and D2 MSNs in both rats and humans (Savell et al., 2020; Tran et al., 2021). While prior studies have not found evidence of differential expression of Pde1b in bulk NAc samples of male mice (Walker et al., 2018), we measured the cell-type–specific expression of Pde1b RNA using RNAscope and found that chronic cocaine increases Pde1b expression in D2 MSNs of male mice but not female mice (Fig. 3). These data add to growing evidence that NAc Pde1b expression is influenced by numerous factors. For example, acute cocaine reduces Pde1b expression in bulk NAc of socially isolated male mice, while chronic cocaine shows a trending increase (p = 0.084) in Pde1b expression in bulk NAc of socially isolated female mice (Walker et al., 2022). By contrast, Pde1b expression in D1/D2 MSNs is not altered 1 h following acute cocaine in male and female rats (Savell et al., 2020).
Given the distinct and often opposing actions of D1 and D2 MSNs in addiction-related behaviors, we evaluated the effects of Pde1b overexpression in NAc D1 versus D2 MSNs in male and female mice using CPP and locomotor activity paradigms. Our results show that Pde1b overexpression had no effect on CPP behavior in D1- or D2-Cre male or female mice but significantly regulated acute locomotor responses during Day 1 of CPP conditioning. We further investigated these cell-type–specific effects with repeated cocaine exposure. In female mice, Pde1b overexpression in D1 MSNs significantly decreased locomotor responses, with the opposite effect observed for D2 MSNs. By contrast, overexpressing Pde1b in D1 or D2 MSNs in male mice had no effect on locomotor responses. Opposing roles of D1 and D2 MSNs in regulating locomotor behavior have been observed previously (van Zessen et al., 2021) and may relate to differences in the efferent projections and electrophysiological responses of these neurons (Creed et al., 2016). However, distinct roles for NAc cAMP/cGMP signaling between male and female mice remain understudied to date. One possibility is that the transcription factors downstream of cAMP/cGMP signaling are differentially regulated or have different effects on gene transcription between male and female mice, ultimately leading to differences in cellular and behavioral responses to drugs of abuse.
We also investigated the cell-type–specific effects of Pde1b overexpression on the electrophysiological properties of NAc D1 and D2 MSNs in male and female mice. In both sexes, Pde1b overexpression reduced the sEPSC frequency in D1 MSNs. Pde1b overexpression reduces the membrane excitability of D1 MSNs in female mice while increasing the membrane excitability of D2 MSNs in male mice. Previous studies have shown that inhibiting D1 MSNs or activating D2 MSNs reduces behavioral responses to cocaine (Lobo et al., 2010; Pascoli et al., 2011; Creed et al., 2016; van Zessen et al., 2021). Thus, the reduction in D1 excitability in female mice and increase in D2 excitability in male mice offers a potential mechanistic link by which Pde1b overexpression influences the locomotor response to cocaine in a sex-specific manner by tilting the functional balance between D1 and D2 MSNs. Given that chronic cocaine influences the physiological properties of NAc MSNs (Zinsmaier et al., 2022), investigating the cell-type– and sex-specific effects of Pde1b overexpression on D1 and D2 MSN physiology in rodents exposed to chronic cocaine is an important area for future research.
To evaluate the role of Pde1b in regulating cocaine-induced transcriptomic adaptations, we overexpressed Pde1b in NAc neurons of male and female mice, administered saline or cocaine (20 mg/kg) for 10 d, and performed RNA-seq on NAc tissue. In the mCherry groups, we observed a similar pattern of DEGs across male and female mice, albeit with fewer DEGs in female mice. This finding confirms sex differences in transcriptional responses to cocaine (Walker et al., 2022), an important area for future investigation given that most studies investigating cocaine-induced transcriptomic changes in the NAc have focused on male rodents.
Pde1b overexpression in the NAc altered the transcriptional profile of both male and female mice treated with saline or cocaine. Several of the DEGs following Pde1b overexpression are in NAc M41, consistent with previous reports that altering hub gene expression influences the expression of other genes in a module (Bagot et al., 2016; Labonté et al., 2017; Lorsch et al., 2019). Pde1b overexpression in the NAc significantly reduced the number of cocaine-induced DEGs across both sexes. This effect is presumably due to dampening of cocaine-induced cAMP/cGMP intracellular signaling cascades by Pde1b and their downstream effectors. However, the cocaine-induced DEGs observed after Pde1b overexpression showed minimal overlap between sexes, indicating that Pde1b induces distinct transcriptional profiles in male versus female mice exposed to cocaine. These data suggest that the sex differences we observed in cocaine self-administration and locomotor responses may be due in part to the sex-specific effects of Pde1b on cocaine-induced transcriptomic adaptations. Further work is needed to understand the molecular mechanisms downstream of cAMP/cGMP signaling that mediate these sex differences.
Together, our study identified gene coexpression networks associated with addiction-like behavior and reveals cell-type– and sex-specific roles for Pde1b in regulating the molecular, cellular, and behavioral response to cocaine. These findings highlight the importance of future studies leveraging PDEs to probe the contributions of cAMP/cGMP signaling within specific cell types to addiction-related behaviors. Considering that PDE1B terminates both cAMP and cGMP signaling, the downstream behavioral effects of selectively manipulating one of these cascades within NAc MSN subtypes are an important focus for future investigations.
Footnotes
This work was supported by National Institute on Drug Abuse Grants to E.J.N. (R01DA007359 and P01DA008227). We thank the Microscopy and Advanced Bioimaging CoRE at Mount Sinai, Min Li at the University of Pittsburg, and James Callens at the Icahn School of Medicine at Mount Sinai for their technical assistance with experiments.
The authors declare no competing financial interests.
- Correspondence should be addressed to Eric J. Nestler at eric.nestler{at}mssm.edu.
