Abstract Background Opioid use disorder (OUD) has emerged as a severe, ongoing public health emergency. Current treatments for OUD are unsuccessful in leading to lasting abstinence in most users. This underscores the lasting effects of chronic opioid use and emphasizes the need to understand the molecular mechanisms of drug seeking and taking and how those alterations persist through acute and protracted withdrawal. Methods Here, we used RNA sequencing in postmortem human tissue from males (n = 10) and females (n = 10) with OUD and age- and sex-matched control subjects. We compared molecular alterations associated with human OUD in the nucleus accumbens (NAc) to mouse and rat models of nonvolitional (n = 4–5 per group per sex) and volitional (n = 5–6 per group per sex) exposure to opioids across distinct stages of opioid use and withdrawal (acute and prolonged). Results We found that the molecular signature in the NAc of females with OUD mirrored effects seen in the NAc of female rodents in a nonvolitional paradigm at all stages of exposure. Conversely, males with OUD showed an expression profile similar to that of rodents with volitional exposure but only during the acute withdrawal phase. Shared coexpression networks were involved in posttranscriptional modification of RNA and epigenetic modification of chromatin state. Conclusions Our results provide fundamental insight into the conserved molecular pathways altered by opioids across species, with evidence suggesting that alterations in females with OUD may be driven by drug exposure, while alterations in males with OUD may be driven by volitional intake. Keywords: Human, Morphine, Mouse, Opioids, Opioid use disorder, Oxycodone Plain Language Summary This study highlights the persistent molecular changes in the brain due to opioid use disorder (OUD). Using RNA sequencing on postmortem human tissue and comparing it to rodent models, researchers found sex-specific molecular signatures in a key reward area of the brain. Females with OUD showed similarities to nonvolitional rodent models, while males resembled volitional models during acute withdrawal. These findings suggest that different mechanisms drive OUD in males and females. Plain Language Summary This study highlights the persistent molecular changes in the brain due to opioid use disorder (OUD). Using RNA sequencing on postmortem human tissue and comparing it to rodent models, researchers found sex-specific molecular signatures in a key reward area of the brain. Females with OUD showed similarities to nonvolitional rodent models, while males resembled volitional models during acute withdrawal. These findings suggest that different mechanisms drive OUD in males and females. __________________________________________________________________ Opioid use disorder (OUD) has emerged as a severe, ongoing public health emergency, with rapidly increasing prevalence rates driven by prescription opioids and recently by a surge in the availability of synthetic opioids, such as fentanyl ([60]1). The rate of overdose deaths due to any opioids has seen a parallel increase over the same period, reaching more than 80,000 overdose deaths per year in the United States alone ([61]2,[62]3). Current treatment strategies have proven inadequate for most users, and rates of relapse are extremely high, underscoring the chronic nature of OUD ([63]4, [64]5, [65]6). This relapse propensity emphasizes the need to understand not only the mechanisms of drug seeking and taking but also what occurs in the brain during acute and protracted withdrawal. Disentangling withdrawal mechanisms may allow for the development of novel therapies to reduce the negative consequences of withdrawal and promote long-term recovery. Drug-induced neuroplasticity is thought to involve molecular adaptations to mesolimbic reward regions such as the nucleus accumbens (NAc) and higher-order executive regions such as the dorsolateral prefrontal cortex (DLPFC) ([66]7, [67]8, [68]9). Chronic opioid use leads to a decrease in drug response. For example, activation of opioid receptors leads to molecular events that promote desensitization and the development of tolerance, including a cascade of receptor adaptations that involve phosphorylation, internalization, and a reduction in signal transduction ([69]10). Opioid exposure also produces an early epigenetic response. The promoter for OPRM1 is hypermethylated in response to short-term therapeutic use ([70]11). Long-term opioid exposure leads to profound alterations in DNA methylation and gene expression ([71]12), microRNA–messenger RNA network dysregulation ([72]13), DNA damage ([73]14), and dysregulation of angiogenic and cytokine gene networks ([74]15). These epigenetic changes persist long after the drug has been metabolized. However, our knowledge of how these changes evolve from intoxication through withdrawal and into long-term abstinence in human OUD is incomplete. Previously, we found enrichment of differentially expressed (DE) transcripts involving neuroinflammation and extracellular matrix (ECM) remodeling in the DLPFC and NAc of people with OUD. Using cell-type deconvolution and weighted gene coexpression network analysis (WGCNA), we identified an accompanying upregulation of microglia markers and increased connectivity of neuroinflammatory and ECM signaling ([75]16). However, all the people in that study had opiates on board at the time of their death, including people who died by overdose. Similarly, preclinical models of molecular mechanisms are typically produced using tissue taken from rodents with drugs on board at the time of sacrifice. Thus, it is difficult to disentangle whether transcriptional changes are a product of acute or chronic drug exposure, and few studies have focused on molecular changes during withdrawal. Furthermore, it is challenging to determine whether alterations in humans with OUD are associated with drug exposure versus volitional intake. Here, we used a multispecies translational approach to compare transcriptional alterations in the NAc of humans with OUD to rodent models of opioid addiction. We leveraged resources available from the rodent biobank (National Institute on Drug Abuse Center of Excellence for Genetics, Genomics, and Epigenetics of Substance Use Disorders in Outbred Rats) and collaborators to collect and investigate tissues from a rodent model of nonvolitional opioid exposure ([76]17) and a rodent model of volitional exposure ([77]18). Both rodent models resulted in robust behavioral and physiological signs of opioid withdrawal. Rodent models were assessed during distinct stages of opioid use and withdrawal (acute and prolonged) (see [78]Table 1 for details on datasets used for cross-species analyses). This study provides fundamental insight into the converging molecular pathways altered by opioids across species. Furthermore, this work helps to disentangle which alterations observed in humans with OUD are driven by acute versus chronic drug effects, with insights into which are associated with drug exposure versus volitional intake. Table 1. Details Associated With Datasets Used in the Cross-Species Analyses Study Groups Sex __________________________________________________________________ Male Female Human: Volitional, Seney et al. ([79]16) Control 10 10 Opioid use disorder 10 10 Rodent: Nonvolitional, Eacret et al. ([80]17) Control 4 4 Intoxication 5 4 1-day withdrawal 5 4 14-day withdrawal 5 5 Rodent: Volitional, Carrette et al. ([81]18) Control 6 6 Intoxicated 5 6 1-day withdrawal 6 6 28-day withdrawal 6 6 [82]Open in a new tab Methods and Materials Animals All procedures were performed in accordance with the National Institutes of Health guidelines and approved by individual Institutional Animal Care and Use Committees at the University of Pennsylvania (C57BL/6J mice) and University of California San Diego (Heterogeneous Stock [HS] rats). Humans Consent for the use of human tissue and data collection were obtained from next of kin, and procedures were approved by the University of Pittsburgh’s committee for Oversight of Research and Clinical Training Involving Decedents and Institutional Review Board for Biomedical Research. Details on the postmortem OUD cohort are provided in [83]Figure S1. Transcriptomics Datasets We previously published the human OUD data used in the current cross-species comparisons (GEO Accession: [84]GSE174409 , ID: 200174409) ([85]16). Details of mouse and HS rat opioid exposure paradigms are included in the [86]Supplement. Behavioral results are reported elsewhere for the mouse nonvolitional exposure ([87]17). Tissue was obtained from the rat oxycodone biobank for the volitional intake dataset ([88]18). In collaboration with the National Institute on Drug Abuse Center of Excellence for Genetics, Genomics, and Epigenetics of Substance Use Disorders in Outbred Rats, we obtained brain samples for RNA sequencing collected from male and female HS rats at different stages of oxycodone self-administration (SA). Differential Expression Analysis Differential expression was assessed using limma ([89]19). For humans, we compared control subjects to people with OUD, with every individual having opioids on board at the time of their death. Briefly, for each transcript, we first adjusted the expression differences with covariates (sex, age, race, postmortem interval, pH, and RNA integrity number) by including each in the regression model; at most, 2 covariates were selected to be included in the full model using Bayesian information criteria for each individual transcript. For the mouse nonvolitional dataset, we examined DE in 3 groups compared with controls: 1) chronic forced morphine drinking and intoxicated at the time of sacrifice (intoxication), 2) chronic forced drinking and in withdrawal for 24 hours (1-day withdrawal), and 3) chronic forced drinking and in withdrawal for 14 days (14-day withdrawal). For the rat volitional dataset, we examined DE in 3 groups compared with controls: 1) intravenous SA (IVSA) of oxycodone and intoxicated at the time of sacrifice (intoxication), 2) IVSA and in withdrawal for 24 hours (1-day withdrawal), and 3) IVSA and in withdrawal for 28 days (28-day withdrawal). Transcripts with p < .01 and log[2] fold change > ±0.26 were considered DE. We also used rank-rank hypergeometric overlap (RRHO) to assess the overlap of differential expression patterns between humans and rodent models at each phase of opioid exposure using RRHO2 ([90]20,[91]21). RRHO2 is a threshold-free method that identifies overlap between 2 ranked lists of DE genes, where the ranking is based on the −log[10](p value) multiplied by the effect size direction. Identification of Coexpression Networks Using the human dataset, we performed WGCNA to identify gene modules ([92]22,[93]23) (threshold power = 0.95, soft power = 30, cut height = 0.25, minimum module size = 30, network type = signed). Fisher’s exact tests were used to determine whether DE transcripts were enriched within WGCNA modules. ARACNe (Algorithm for the Reconstruction of Accurate Cellular Networks) was used to identify hub and OUD-specific hub genes for network analysis ([94]24), and Cytoscape was used to visualize networks. Pathway overrepresentation categories for each module were assessed using Metascape, with the 5000 WGNCA-analyzed genes as background. Results Differential Expression Across Human OUD and Rodent Models of Nonvolitional and Volitional Opioid Exposure Because corticostriatal circuit dysfunction contributes to the impairments that are a hallmark of substance use disorders ([95]25), we examined the impact of OUD on transcriptional differences in the human NAc. Consistent with our previous work, we found that many more NAc DE transcripts were downregulated than upregulated in OUD when we examined samples from humans ([96]Figure S2A and [97]Table S1). However, sex differences have been described across several domains in OUD including comorbidity with psychiatric symptoms ([98]26), cue-induced craving/neural activity ([99]27), and DNA methylation of OUD-related genes ([100]28). To identify potential sex-specific transcriptional signatures of OUD, we examined males and females separately. Notably, there was a low level of overlap in the patterns of upregulation and downregulation between males and females, with most DE transcripts exhibiting distinct patterns in males and females with OUD ([101]Figure 1A and [102]Table S1). Across analyses, most DE transcripts were protein coding, with a much smaller fraction representing long noncoding RNAs (lncRNAs) and pseudogenes. Figure 1. [103]Figure 1 [104]Open in a new tab Sex-specific transcriptomic changes in the NAc from human subjects with OUD and 2 rodent models of opioid exposure. (A) Alluvial plots showing transcripts that exhibit shared or unique patterns of transcripts that were upregulated, downregulated, or not altered in males and females with OUD. (B) Alluvial plots showing transcripts that exhibit shared or unique patterns of transcripts that were upregulated, downregulated, or not altered in male and female mice across 3 stages of opioid exposure in a nonvolitional paradigm. (C, D) Alluvial plots showing transcripts exhibiting shared or unique patterns of transcripts that were upregulated, downregulated, or not altered in male and female rats in the NAc core (C) and shell (D) across 3 stages of opioid exposure in a volitional paradigm. NAc, nucleus accumbens; OUD, opioid use disorder. Then, we investigated the effect of opioid administration on the NAc transcriptome in a nonvolitional exposure paradigm (mouse; entire NAc) and a volitional exposure paradigm (rat; core and shell separately) at 3 stages of opioid use and withdrawal. For the nonvolitional paradigm ([105]17), we compared controls (not exposed to drug) with 1) 11 days forced administration in drinking water (intoxicated), 2) 1-day withdrawal, and 3) 14-day withdrawal rodents. For the nonvolitional paradigm, opioid exposure was associated with more downregulated than upregulated genes in all 3 opioid-exposed groups ([106]Figure S2B and [107]Table S2). Similar to the sex-specific patterns of OUD effects in humans, male and female mice exhibited distinct alterations for all 3 stages of exposure, with very few genes altered in the same direction in males and females ([108]Figure 1B). Most of the DE transcripts were protein coding genes followed by lncRNAs across the 3 stages of opioid exposure ([109]Figure S2B). In the volitional paradigm ([110]18), we compared controls (not exposed to drug) with 1) 6 weeks IVSA (intoxicated), 2) 1-day withdrawal, and 3) 28-day withdrawal rodents. More DE transcripts were upregulated than downregulated in the NAc core ([111]Figure S2C) and shell ([112]Figure S2D) across the 3 opioid-exposed groups. When we examined the results separately by sex, we again observed distinct patterns across the 3 stages of exposure for both core ([113]Figure 1C and [114]Table S3) and shell ([115]Figure 1D and [116]Table S4). Transcriptional Concordance Between Human OUD and the Rodent Model of Nonvolitional Opioid Exposure Next, we interrogated whether the pattern of gene expression in human OUD mapped onto the signature from our nonvolitional exposure groups using threshold-free RRHO2 analysis. These analyses have the potential to identify which transcriptional changes in human OUD are driven by exposure to opioids that are independent of volitional intake. These comparisons also allow us to determine which alterations in human OUD are a product of the acute effects of opioid intoxication versus the long-term effects of chronic exposure. When the sexes were pooled, there was substantial overlap in both upregulated and downregulated transcripts between humans and intoxicated rodents. However, when the sample was split by sex, OUD females but not males showed strong concordance with sex-matched intoxicated rodents ([117]Figure 2A). The same pattern was true for 1-day ([118]Figure 2C) and 14-day ([119]Figure 2E) withdrawal. Pathway enrichment was performed on shared transcripts between females with OUD and female rodents across the 3 exposure groups ([120]Figure 2B, D, F and [121]Table S5). Top shared pathways included Wnt/β-catenin, myelination, CLEAR (coordinated lysosomal expression and regulation), PDGF-mediated signaling, and synaptogenesis signaling. In intoxicated and 1-day withdrawal rodents, we identified levodopa (L-DOPA), the central precursor in the synthesis of the catecholamine neurotransmitters dopamine, norepinephrine, and epinephrine, as an upstream pathway regulator ([122]Figure 2B, D). We also identified the gene for estrogen receptor alpha (ERα), ESR1, as a potential upstream regulator in intoxicated and 1-day withdrawal rodents ([123]Figure 2B, D). When activated by its ligands, ERα acts as a transcription factor with roles across the organism ([124]29). The HNF4A gene, which also encodes a transcription factor, emerged as an upstream regulator across all 3 opioid-exposed groups. Figure 2. [125]Figure 2 [126]Open in a new tab Transcriptional changes in the NAc from female human subjects with OUD overlap with rodent nonvolitional exposure. (A) RRHO2 plots showing concordant gene expression between human subjects with OUD pooled by sex (left column), human female OUD (center column), human male OUD (right column), and mice intoxicated ATOD. Increasing color warmth indicates increasing −log[10]p value. (B) Pathway enrichment analysis of transcripts concordant between females with OUD and female mice intoxicated ATOD (left). Circle size corresponds to gene number, and increasing color warmth indicates enrichment of −log[10]p value. The bar graph displays potential upstream regulators with −log[10]p value on the x-axis (right). (C) RRHO2 plots showing concordant gene expression between human subjects with OUD pooled by sex (left column), human female OUD (center column), human male OUD (right column), and mice at 1-day withdrawal. (D) Pathway enrichment analysis of transcripts concordant between females with OUD and female mice at 1-day withdrawal (left). The bar graph displays potential upstream regulators (right). (E) RRHO2 plots showing concordant gene expression between human subjects with OUD pooled by sex (left column), human female OUD (center column), human male OUD (right column), and mice at 14-day withdrawal. (F) Pathway enrichment analysis of transcripts concordant between females with OUD and female mice at 14-day withdrawal (left). The bar graph displays potential upstream regulators (right). ATOD, at time of death; CLEAR, coordinated lysosomal expression and regulation; NAc, nucleus accumbens; OUD, opioid use disorder; RRHO, rank-rank hypergeometric overlap. Transcriptional Concordance Between Human OUD and the Rodent Model of Volitional Opioid Exposure Then, we turned to our volitional exposure rodent model and compared the transcriptional signature in the NAc shell and core with the NAc of humans with OUD. The pattern of gene expression was discordant between humans and intoxicated rodents when the sexes were pooled for both the shell ([127]Figure 3A) and core ([128]Figure 3E). When the sexes were examined separately, gene expression was still discordant but more strongly in females than in males ([129]Figure 3A, E). Gene expression was concordant between humans and the 1-day withdrawal rodents in both shell ([130]Figure 3B) and core ([131]Figure 3F). However, this effect was driven by males; when the sexes were examined separately, only males with OUD showed concordance with male 1-day withdrawal rodents. The same overall patterns were observed between humans and 28-day withdrawal rodents. Gene expression was weakly concordant between all humans and 28-day withdrawal rodents, but when separated by sex, males were weakly concordant, and females were weakly discordant ([132]Figure 3D, H and [133]Table S5). In the shell, top pathways associated with shared transcripts between OUD males and 28-day withdrawal rodent males included protein ubiquitination and mTOR (mechanistic target of rapamycin) signaling ([134]Figure 3C), with ESR1 and HNF4A predicted as upstream regulators. In the core, the 1-day withdrawal comparison in males was associated with estrogen receptor signaling and HIF1a ([135]Figure 3G), while the 28-day withdrawal comparison in males was associated with synaptogenesis and CLEAR signaling ([136]Figure 3I). Predicted upstream regulators for both the 1-day and 28-day withdrawal comparisons were HNF4A, TP53, and β-estradiol ([137]Figure 3G, I). Figure 3. [138]Figure 3 [139]Open in a new tab Transcriptional changes in the NAc from male humans with OUD overlap with acute and prolonged withdrawal in a rodent volitional exposure paradigm. (A) RRHO2 plots showing concordant gene expression between human subjects with OUD pooled by sex (left column), human female OUD (center column), human male OUD (right column), and the NAc shell of rats intoxicated ATOD. (B) RRHO2 plots showing concordant gene expression between human subjects with OUD pooled by sex (left column), human female OUD (center column), human male OUD (right column), and the NAc shell of rats at 1-day withdrawal. (C) Pathway enrichment analysis of transcripts concordant between males with OUD and the NAc shell of male rats at 1-day withdrawal (left). The bar graph displays potential upstream regulators (right). (D) RRHO2 plots showing concordant gene expression between human subjects with OUD pooled by sex (left column), human female OUD (center column), human male OUD (right column), and the NAc shell of rats at 28-day withdrawal. (E) RRHO2 plots showing concordant gene expression between human subjects with OUD pooled by sex (left column), human female OUD (center column), human male OUD (right column), and the NAc core of rats intoxicated ATOD. (F) RRHO2 plots showing concordant gene expression between human subjects with OUD pooled by sex (left column), human female OUD (center column), human male OUD (right column), and the NAc core of rats at 1-day withdrawal. (G) Pathway enrichment analysis of transcripts concordant between males with OUD and the NAc core of male rats at 1-day withdrawal (left). The bar graph displays potential upstream regulators (right). (H) RRHO2 plots showing concordant gene expression between human subjects with OUD pooled by sex (left column), human female OUD (center column), human male OUD (right column), and the NAc core of rats at 28-day withdrawal. (I) Pathway enrichment analysis of transcripts concordant between males with OUD and the NAc core of male rats at 28-day withdrawal (left). The bar graph displays potential upstream regulators (right). ATOD, at time of death; CLEAR, coordinated lysosomal expression and regulation; mTOR, mechanistic target of rapamycin; NAc, nucleus accumbens; OUD, opioid use disorder; RRHO, rank-rank hypergeometric overlap. Gene Network Analysis Identifies Sex-Specific Hub Genes Associated With Nonvolitional and Volitional Opioid Exposure To investigate gene correlations in human OUD and our rodent models of nonvolitional and volitional opioid intake, we interrogated gene networks using WGCNA. First, we built the WGCNA modules using the human OUD data and then looked for enrichment in OUD-associated DE transcripts as well as cross-species concordance patterns within each module (see module membership in [140]Table S6). There was strong enrichment of DE transcripts across all human OUD subject groups in the darkorange, yellowgreen ([141]Figure S3), and lightgreen modules ([142]Figure 4A). In addition to being enriched for DE transcripts in all our human OUD subject groups, the lightgreen module showed a high prevalence of DE transcripts that were concordant between human female OUD and rodent female nonvolitional intake as well as human male OUD and rodent male volitional intake groups ([143]Figure 4A). Because this module showed the strongest enrichment for DE transcripts, we focused our attention there. We visualized the pathway network associated with the lightgreen module and identified clusters with significantly enriched terms based on pathways ([144]Figure 4B). The 2 largest clusters were categorized as RNA splicing/spliceosome and acetylation lysine/acetyltransferase. Other network clusters included DNA transcription preinitiation, regulation gene heterochromatin, cytosolic ribosome structural, and transcription coactivator activity. Together, this network analysis underscores the profound effect that opioid exposure has on the regulatory mechanisms of DNA transcription from epigenetic modifiers through posttranscriptional regulation of RNA splicing. Other notable network clusters involved synaptic activity: synaptic vesicle localization and signal release synapse. Figure 4. [145]Figure 4 [146]Open in a new tab OUD-associated gene networks in the NAc. (A) Coexpression modules and network structure of transcription in the NAc were generated using weighted gene coexpression network analysis. Circos plot identified by module names and colors. Enrichment for transcripts DE in OUD, concordant human female and mouse, and concordant human male and rat are indicated by semicircle colors within each module, with increasingly warm colors indicating increasing −log[10]p value. (B) Left: Network visualization generated using Cytoscape with categorical labels for multiple clusters. Right: Hub gene coexpression networks of the NAc in the lightgreen module. Node size indicates the degree of connectivity for that gene. Blue nodes indicate DE hub genes. Gray nodes indicate hub genes. Red halos indicate OUD-specific hub genes. Yellow halos indicate sex-specific hub genes. DE, differentially expressed; NAc, nucleus accumbens; OUD, opioid use disorder. To distinguish potential regulators of this coexpression network, we identified highly connected hub genes within the lightgreen module that were predicted to drive the expression of other genes within the module ([147]Figure 4B). Of the 11 identified hub genes, 5 were specific to OUD: ANKRD11, CAMTA2, IQSEC1, RIPOR1, and SKI. We also identified sex-specific hub genes as part of our network analysis. Three of those genes overlapped with OUD hub genes: CAMTA2, IQSEC1, and ANKRD11, while the other 2, SEC16A and WNK2, were uniquely sex specific. The central hub gene in our network is KMT2D, a histone methyltransferase with a direct relationship with the estrogen receptor ([148]30). This is particularly interesting given that one of the largest clusters in the lightgreen module is acetylation lysine/acetyltransferase, further suggesting that opioids lead to heterochromatin modifications ([149]16,[150]31). The lightgreen module represents a potential interacting, correlative gene network that is integrated across species and drug administration paradigms. To further deconvolve the cell types represented by the genes in the lightgreen module, we assessed whether module genes were enriched for specific cell types in human striatum ([151]32,[152]33). Overall, the lightgreen module was enriched for genes in neuron subtypes compared with glial cells ([153]Figure 5A), suggesting that the module is composed mainly of genes active in neurons across species in response to opioids. In OUD, module genes were significantly reduced in enrichment within medium spiny neurons relative to interneurons and glial cells ([154]Figure 5B). Enrichment changes were independent of sex ([155]Figure 5B), occurring in both females and males with OUD. Thus, module genes were decreased in expression in medium spiny neurons in OUD, reflecting similar genes and consistent correlative expression across the gene network in medium spiny neurons in rodent models of opioids and individuals with OUD. Figure 5. [156]Figure 5 [157]Open in a new tab Cell-type enrichments identify opioid-related gene network alterations in striatal medium spiny neurons. (A) AUCell of the lightgreen module identified by weighted gene coexpression network analysis derived from human striatum with the y-axis representing the relative expression of module gene score ([158]32). Each major neuronal (left) and glial (right) cell subtype is represented. Higher module scores in neurons reflect preferential enrichment of genes in the lightgreen module within neurons of human striatum relative to glial cells. (B) AUCell of the lightgreen module genes are decreased in expression primarily in medium spiny neuron subtypes (matrix and striosome neurons that express dopamine D[1] or D[2] receptor) in individuals with OUD. The y-axis represents the relative expression of module gene score. ∗∗∗Significantly different between the UC and OUD cohorts at p < .001. Astro, astrocyte; AUCell, area under the curve for cell types; D1/D2, D[1]/D[2] receptor-expressing neurons; F, female; M, male; NS, nonsignificant; NUDAP, neurochemically unique domains in the accumbens and putamen; Oligo, oligodendrocyte; OPC, oligodendrocyte precursor cell; OUD, opioid use disorder; UC, unaffected control individual. Discussion Here, we demonstrated sex specificity in concordance of the molecular signature of OUD in the human NAc and the molecular signature of opioid exposure in preclinical models. Females with OUD mirrored effects seen in the NAc of female rodents in a nonvolitional paradigm at all stages of opioid exposure, while males with OUD showed overlap in expression profile with male rodents in acute withdrawal (NAc shell, core) and prolonged withdrawal (NAc core) in a volitional paradigm. The top 2 shared pathways, synaptogenesis and CLEAR signaling, were consistently identified in concordant genes across species and opioid exposure, suggesting that opioid addiction is closely linked to synaptic alterations and regulation of lysosomal degradation. WGCNA identified gene coexpression networks involved in posttranscriptional modification of RNA and chromatin state, underscoring the effect of opioid exposure on transcriptional regulation via epigenetic modification. While epidemiological studies have demonstrated a higher prevalence of illicit drug use in males, evidence suggests that drug use in females transitions more quickly from casual, recreational use to dependence ([159]34, [160]35, [161]36). Females report higher levels of baseline and cue-induced craving ([162]27,[163]37,[164]38), more physical withdrawal symptoms, greater functional impairment, and higher levels of comorbid psychological suffering ([165]39, [166]40, [167]41). Preclinical studies have shown that females acquire opioid SA more quickly, self-administer more, reach higher breakpoints during fixed ratio, and exert more behavioral effort during extinction and cue-induced reinstatement ([168]42, [169]43, [170]44, [171]45, [172]46, [173]47, [174]48). Our work provides added nuance to understanding sex differences in OUD. When we compared samples from humans to our rodent models of disease, females with OUD were concordant with female rodents that were nonvolitionally exposed to opioids, while males with OUD were concordant with male rodents that were volitionally exposed to opioids. These results help provide potential insight into mechanisms that underlie the OUD-associated alterations that we observed. For example, our results suggest that the pattern of alterations in females with OUD may be due to opioid exposure itself and not necessarily volitional intake, while the pattern in males with OUD is driven more by volitional opioid intake. Notably, several consistent patterns emerged with regards to top shared pathways in concordant genes and their upstream regulators. Estrogen receptor signaling was the strongest pathway finding in the rat NAc core, and both ESR1, the gene for ERα, and a ligand for this receptor, estradiol, were identified as upstream regulators across sex and exposure paradigm (nonvolitional and volitional). ERα is the predominant estrogen receptor subtype, and estradiol is the primary circulating estrogen hormone. When bound, the receptor undergoes a conformational change, homodimerizes, and together estradiol and ERα form a complex that binds the promoter region of target genes to regulate their expression ([175]29,[176]49). Evidence suggests that gonadal hormones generally and estradiol specifically play a role in guiding behavior related to substance use ([177]50, [178]51, [179]52, [180]53, [181]54). Female rats take a similar number of heroin infusions during estrus, metestrus, and diestrus but self-administer significantly less during proestrus when estradiol is at its peak ([182]46). Ovariectomy (OVX) removes the primary source for estrogens and, in mice, increases morphine conditioned place preference (CPP) at a 10-mg/kg dose, which is reversed by estradiol replacement ([183]55). These findings suggest that estradiol leads to an overall reduction in drug seeking/reward. However, the data are complex. At smaller doses of morphine (2, 5 mg/kg), OVX in female mice decreases CPP, and estradiol replacement increases CPP ([184]55). Following estradiol replacement in OVX rats, animals acquired opioid SA more quickly, self-administered higher levels of opioids, and showed an enhanced sensitivity to opioid-related cues compared with vehicle-treated OVX controls ([185]56,[186]57), which is consistent with estradiol’s effect in studies that have used psychostimulants and alcohol ([187]58, [188]59, [189]60, [190]61, [191]62, [192]63). However, other studies have found no effect of OVX in rat opioid SA models ([193]64,[194]65). Together, this suggests that gonadal hormones play some role in the rewarding effects of drugs of abuse, but the relationship to opioids needs further study. One potential mechanism is through regulation of synaptic density. Estradiol and the primary male gonadal hormone testosterone can modulate spine plasticity in both the NAc core and shell ([195]35,[196]66, [197]67, [198]68, [199]69). Notably, synaptogenesis signaling emerged as a consistent pathway finding across sex, species, and opioid exposure group in our analyses. It is interesting to speculate how gonadal hormones may modulate the rewarding effect of drugs of abuse through altering baseline spine density in reward-related cortical regions. Consistent with the idea that the long-lasting effects of opioid exposure are mediated through epigenetic modification, the central hub gene that emerged in our network is KMT2D, a histone methyltransferase with a direct relationship with ERα. As discussed above, activated ER forms a complex with other transcription factors and binds chromatin at specific sequences. KMT2D actively modulates chromatin, creating a more open chromatin state so that the activated ER complex can be recruited to ER response elements in the genome and promote ER-dependent transcription ([200]30,[201]70). This is interesting because it is a potential link between epigenetic regulation on the one hand and sex specificity on the other. Sex differences have previously been described in the epigenetic regulation of opioid-related genes. In a retrospective study of 250 outpatients who were experiencing chronic noncancer pain, there were sex differences in methylation of the promoter region of OPRM1, which was significantly correlated with lower rates of OUD in females ([202]28). We observed a variety of alterations in opioid receptor expression across different patterns of opioid exposure between males and females in rodent models and in individuals with OUD ([203]Table S7). Unpacking the mechanisms by which opioid exposure modulates the ability of transcriptional regulators to interact with chromatin is integral for understanding how environment and experience shape gene expression in the context of addiction. Together, our findings demonstrate how the relationship between opioid use, gene expression, and its regulation through epigenetic modifications is informed by input from gonadal hormones with consequences for synaptic modeling in corticostriatal circuitry. Regardless of sex or exposure paradigm, HNF4A was identified as an upstream regulator across all concordant pathways. HNF4A is a nuclear transcription factor that binds DNA as a homodimer and plays a role in the development of several major organ systems including the liver, kidneys, and intestines ([204]71, [205]72, [206]73, [207]74). Converging lines of evidence implicate HNF4A in the development of several kinds of cancer ([208]75, [209]76, [210]77). Specifically, HNF4A interacts with the Wnt/β-catenin pathway to guide cell adhesion, cell proliferation, apoptosis, and tumorigenesis ([211]78, [212]79, [213]80). Interestingly, Wnt/β-catenin was the strongest pathway finding in our female with OUD versus nonvolitional rodent female intoxication comparison, and molecular mechanisms of cancer emerged in several of our concordance findings. It is currently unclear what this means for the development of dependence, but the consistency of this finding suggests that HNF4A should be a focus of study in future OUD work. We identified highly connected hub genes within the lightgreen module that were specific to OUD: ANKRD11, CAMTA2, IQSEC1, RIPOR1, and SKI. Previous evidence has linked the activity and regulation of CAMTA2, IQSEC1, and RIPOR1 to OUD. CAMTA2 is a member of the ubiquitous CaMKs (Ca^2+/calmodulin-dependent protein kinase family), which are crucial players in producing the synaptic changes that underlie the formation of memory ([214]81, [215]82, [216]83). Although CAMTA2 has not been directly linked to opioid addiction, this protein family generally, and 1 member, CaMKII specifically, have been linked to the development of opioid tolerance and dependence ([217]84, [218]85, [219]86, [220]87, [221]88). CaMKII activity is increased by morphine treatment, and inhibiting CaMKII in rat hippocampus attenuates morphine tolerance and dependence ([222]89,[223]90). IQSEC1 has been identified as a potential regulator for receptor endocytosis by Kyoto Encyclopedia of Genes and Genomes (KEGG) analysis of putative targets for responsive NAc microRNAs after morphine SA ([224]91), which has direct relevance to mu opioid receptor desensitization, a molecular correlate of the development of dependence. Lastly, RIPOR1 is a member of the Rho family of GTPases, which regulate cellular processes such as cytoskeleton remodeling and synapse maintenance ([225]92, [226]93, [227]94). Drugs of abuse, including morphine and heroin, alter the activity of these proteins ([228]95). In addition, we compared the genes from our lightgreen module with genes linked to OUD by existing genome-wide association study (GWAS) work. Two genes stood out as particularly interesting, GABRG2 and FURIN. GABRG2 encodes a GABA (gamma-aminobutyric acid) receptor subunit. Not only has GABRG2 been linked to both alcohol use disorder and OUD ([229]96,[230]97), but we previously also demonstrated a microglia-specific upregulation of GABRG2 in OUD using single-nuclei transcriptomics in human OUD ([231]33). FURIN is a proteolytic enzyme and notably has a single variant association with OUD as identified through GWAS ([232]98, [233]99, [234]100, [235]101). Our data provide crucial nuance to understanding sex-specific mechanisms of OUD in human subjects. However, we acknowledge several limitations to our work. First, due to the nature of human postmortem work, we cannot accurately compare the effects, whether additive or synergistic, of polysubstance use on molecular alterations in the brain. Second, each of the individuals in our human cohort used opioids for years and terminating from drug toxicity and overdose. Thus, it remains difficult to definitively parse the acute versus chronic opioid effects on molecular alterations in human brains. However, the overlap between molecular alterations in humans and rodent models provides some insights. Third, the duration of opioid administration, the type and dose of opioids administered (oxycodone vs. morphine), and the pattern of opioid use, including the duration of withdrawal from opioids (28 vs. 14 days), were different in rats and mice. Nevertheless, our findings support the notion that opioid addiction is driven by large-scale gene alterations in the NAc in both humans and rodents. Furthermore, these results suggest that researchers may consider the relationship between biological sex and exposure paradigm when designing preclinical studies to investigate mechanisms with relevance to human OUD (e.g., use of a nonvolitional paradigm to model female OUD; use of a volitional paradigm to model male OUD). This work deepens our understanding of sex specificity in OUD and opens the door for future research into putative therapies. Acknowledgments and Disclosures