Cocaine addiction inflicts enormous health and economic costs to individuals, families, and society (Reid et al., 2012; United Nations Office on Drugs and Crime, 2020). Recently, significantly increased studies have focused on the field of neuroscience of cocaine use disorder, but its neurobiological mechanism is still unclear, and there is no effective clinical treatment for cocaine use disorder (Gawin and Ellinwood, 1989; Majewska, 1996a,b; Heal et al., 2014).

Epigenetic mechanisms can integrate both genetic and diverse environmental stimuli to exert potent and often long-lasting changes in gene expression (Jaenisch and Bird, 2003). Accumulating research has found epigenetic mechanism plays an important role in the drug addiction (Robison and Nestler, 2011; Nestler, 2014; Nestler and Lüscher, 2019). Non-coding RNAs, specifically long non-coding RNAs, circular RNAs (circRNAs), and small non-coding RNAs, are one type of common epigenetic regulators that play a vital role in many biological processes associated with diseases (Amin et al., 2019; Mehta et al., 2020).

Circular RNAs are vastly conserved non-coding RNAs formed by back-splicing and covalent fusion of RNA free ends into natural circles (Vicens and Westhof, 2014; Szabo and Salzman, 2016; Greene et al., 2017; Li et al., 2018). Because circRNAs lack poly(A) tails and cap structure, they are not affected by RNA exonuclease (Vicens and Westhof, 2014; Szabo and Salzman, 2016; Greene et al., 2017; Li et al., 2018). CircRNAs usually exert their functions as transcriptional and post-transcriptional regulators through various functional mechanisms, such as RNA binding protein (RBP) “sponges” (Du et al., 2016; Holdt et al., 2016), translated proteins (Legnini et al., 2017; Pamudurti et al., 2017), and RNA–RNA interaction (Li et al., 2015). At present, circRNAs function mainly by absorbing microRNAs (miRNAs) as competing endogenous RNAs (ceRNAs) regulatory network to regulate their target genes expression, which construct a functional circRNA–miRNA–mRNA regulation network (Hansen et al., 2013; Vicens and Westhof, 2014; Rybak-Wolf et al., 2015; Du et al., 2017; Greene et al., 2017; Li et al., 2018; Mehta et al., 2020). For example, knockdown of circHIPK2 expression significantly inhibited astrocyte activation induced by methamphetamine through the targeting of miR124 and SIGMAR1 (Huang et al., 2017). Another study reported that circTmeff-1 promotes incubation of context-induced morphine craving by sponging miR-541/miR-6934 in the nucleus accumbens (Yu et al., 2021).

Although several circRNAs have been identified as participating in cocaine addiction, the regulatory networks in patients with cocaine use disorder are still unknown. It is necessary to conduct the circRNA–miRNA–mRNA regulatory networks in patients with cocaine use disorder to help to advance our understanding of the molecular mechanism of cocaine use disorder. Dorsolateral prefrontal cortex (dlPFC), similar role to medial PFC in rodents (Seamans et al., 2008), is a crucial component brain region of inhibitory control (Gass and Chandler, 2013; Moeller et al., 2014), which undergoes significant changes after long-term cocaine use (Matochik et al., 2003; Moreno-Lopez et al., 2012) and is involved in compulsive drug-seeking behaviors, increasing drug intake and addiction severity (Chen B. T. et al., 2013; Conti and Nakamura-Palacios, 2014; Terraneo et al., 2016). In this study, we aimed to investigate the functional circRNA–miRNA–mRNA regulatory networks in the dlPFC of patients with cocaine use disorder. Lastly, we constructed a circRNA–miRNA–mRNA regulation network including 24 circRNAs, 43 miRNAs, and 82 mRNAs, which may reveal a novel molecular mechanism in pathogenesis of patients with cocaine use disorder.

Most of previous research on the mechanism of addiction was based on animal models, or the peripheral blood of patients with substance use disorders; however, the studies do not truly portray the changes that occur in the brains of patients with substance use disorders, which may be an important obstruction to the study of drugs for the treatment of substance use disorders. The circRNA–miRNA–mRNA regulatory network we constructed will enhance the understanding of the addiction mechanism in the brain of patients with cocaine use disorder.

Through the functional enrichment analysis of the mRNAs in the network, we found that the up-regulated mRNAs were mainly involved in regulation of axonogenesis and cell junction, which suggests that our up-regulated mRNA may have a close connection with synaptic transmission, and previous studies have confirmed that abnormal synaptic transmission is a very critical factor for cocaine addiction (Khibnik et al., 2016; Martínez-Rivera et al., 2017; Li et al., 2021). Down-regulated mRNAs were found to be involved in many biological processes, including cellular response to hormone stimulus, response to unfolded protein, cellular response to heat, intracellular receptor signaling pathway, myeloid cell differentiation, calcium-ion regulated exocytosis, calcium-mediated signaling, and autophagy. All the biological processes related to down-regulated mRNAs in the circRNA–miRNA–mRNA regulatory network have been involved with cocaine addiction. For example, clinical trials have investigated that cocaine associated cues could significantly increase adrenocorticotropic hormone and cortisol (Berger et al., 1996). A single dose of cocaine can cause the accumulation of different heat shock proteins (Salminen et al., 1997), which leads to blood-brain barrier breakdown and brain edema formation thereby promoting cocaine intoxication (Sharma et al., 2009). Apart from this, Cocaine has the propensity to cause hyperthermia which increases the mortality rates to cocaine (Crandall et al., 2002). These are also evidences reported that various intracellular receptors, especially dopamine receptors and glutamate receptors, are all critical for cocaine addiction (Ellenbroek, 2013; Howell and Cunningham, 2015; Smaga et al., 2019). In addition, brain myeloid cells, particularly microglia, presented in the brain parenchyma, serve as a surveillance function for neuroinflammation and neurodegeneration in the central nervous system (Ransohoff and Cardona, 2010; Ajami et al., 2018; Jordão et al., 2019). Addictive drugs, especially cocaine, have been consistently shown to activate microglia both in vitro and in vivo (Guo et al., 2015; Liao et al., 2016). In rodents, inhibiting glial cell activation was shown to block cocaine-mediated behavioral changes (Chen et al., 2009). In humans, cocaine exposure can reduce microglial cells viability and inhibit the expression of extracellular vesicle-associated proteins disrupting cellular signaling and cell-to-cell communication (Kumar et al., 2020). Therefore, it can be considered that circRNAs were thought to play an important role in the multiple addiction-related networks in dlPFC of patients with cocaine use disorder.

Circular RNAs are endogenous non-coding RNAs with widespread distribution and various cellular function (Hansen et al., 2013; Vicens and Westhof, 2014; Rybak-Wolf et al., 2015; Du et al., 2017; Greene et al., 2017; Li et al., 2018; Mehta et al., 2020). Numerous studies have shown that circRNAs have an important influence on many complicated neuropsychiatric disorders (Cui et al., 2016; Zhang et al., 2018; An et al., 2019; Liu et al., 2019; Mahmoudi et al., 2019; Huang et al., 2020; Zhang Y. et al., 2020; Zimmerman et al., 2020), including drug addiction (Huang et al., 2017; Bu et al., 2019; Li et al., 2019, 2020; Zhang H. et al., 2020). Knockdown of circHomer1 ameliorates methamphetamine-induced neuronal injury through inhibiting Bbc3 expression (Li et al., 2020). The abnormal expression of mmu_circRNA_002381 in striatum was induced by cocaine self-administration and cocaine-induced locomotor activity model (Bu et al., 2019). Interestingly, siRNA-mediated mmu_circRNA_002381 down-regulation increased the expressions of limk1 and bdnf, which are the targets of miR-138 associated with synaptic plasticity. Additionally, some studies predicted that circRNAs are involved in the progress and development of many addictive drug models by sponging miRNA to regulate downstream targets (Li et al., 2017, 2020; Bu et al., 2019; Zhang H. et al., 2020). In our study, 24 circRNAs were identified to be involved in the circRNA–miRNA–mRNA regulatory network. Among these, 15 were identified previously in the mammalian brain as being dysregulated during neuronal differentiation and highly enriched in synapses (Rybak-Wolf et al., 2015). It was suggested that circRNAs in the circRNA–miRNA–mRNA regulatory network may play their regulatory functions in the neurons or synapses of patients with cocaine use disorder.

Several studies have revealed that circRNAs contain multiple miRNA response elements and can bind to miRNAs, often termed as “miRNA sponges,” decreasing cytoplasmic levels of miRNAs and liberating their respective downstream target mRNAs (Memczak et al., 2013; Piwecka et al., 2017; Kleaveland et al., 2018; Zhang Y. et al., 2020). Here, we constructed a circRNA–miRNA–mRNA regulation network involved in cocaine use disorder, including 24 circRNAs, 42 miRNAs, and 82 mRNAs. CircSLC30A6 was down-regulated in the dlPFC of patients with cocaine use disorder. Based on the analysis of circRNA–miRNA–mRNA network, we found that circSLC30A6 interacts with hsa-miR-9-3p. Interestingly, it has been known that hsa-miR-9-3p mediates the dynamic regulation of neural progenitor proliferation during neurogenesis (Pascale et al., 2020). Moreover, hsa-miR-9-3p is significantly increased in serums of patients with methamphetamine use disorder compared with normal controls (Gu et al., 2020). Therefore, we assume that circSLC30A6 up-regulation induced by the cocaine may be involved in cocaine effect through interacting with hsa-miR-9-3p. CircRASA1 and circMNAT1 were down-regulated in the dlPFC of patients with cocaine use disorder. Through the analysis of circRNA–miRNA–mRNA network, we found that circRASA1 interacts with hsa-miR-26b-3p and circMNAT1 interacts with hsa-miR-22-5p. miR-26b was found to be up-regulated in hippocampus following the acquisition and extinction but miR-22 was only up-regulated during extinction of cocaine-induced conditioned place preference in rats (Chen C. L. et al., 2013). Therefore, it is conceivable that circSLC30A6, circRASA1, and circMNAT1 might play an important role in cocaine use disorder through modulating their target miRNA.

To further identify the key circRNAs participating in the regulatory network, we established a PPI network and screened four hub genes, including HSP90AA1, HSPA1B, YWHAG, and RAB8A. Accordingly, we constructed a circRNA–miRNA-hub gene subnetwork. In the cocaine self-administration model, HSP90AA1 had a significant decrease in PFC after withdrawal of 15 days (Bhattacherjee et al., 2019), however, the expression of HSP90AA1 in posterior hippocampus increased significantly after 28 days of withdrawal (García-Fuster et al., 2012). These delayed neurobiological effects of HSP90AA1 likely contribute to sustained vulnerability to cocaine relapse, which may be regulated by circMTHFD2L. HSPA1B gene is one of heat shock protein 70 (HSP70)-encoding transcripts, and it is reported that HSPA1B expression was increased in the postmortem brains of patients with cocaine use disorder exhibiting excited delirium in comparison with other (non-excited delirium) cocaine-related deaths and drug-free controls, concluding that elevated HSPA1B provides a reliable forensic biomarker for the identification of excited delirium (Mash et al., 2009; Johnson et al., 2012). The structure of circRNA makes them more stable and has a longer half-life, so it is considered to be a more ideal marker (Enuka et al., 2016). Previous studies reported that cocaine exposure dysregulated the expression of YWHAG (Bhattacherjee et al., 2019), and reduced YWHAG can lead to neuronal hyperexcitability, and normalization of hyperexcitability can rescue memory deficits (Roy et al., 2021). RAB8A is a member of the RAS superfamily, which are key regulators of intracellular membrane trafficking from the formation of transport vesicles to their fusion with membranes, and involve in polarized vesicular trafficking, and neurotransmitter release (Núñez et al., 2009; Esseltine et al., 2012; Sellier et al., 2016; Nüchel et al., 2018). Although no studies have shown that RAB8A has a direct effect on cocaine addiction, the biological process in which it participates is very important in cocaine addiction (Periyasamy et al., 2016; Harraz et al., 2021). Here, we identified five circRNA–miRNA-hub gene axes, indicating competitive regulatory relationships of five circRNAs with the four genes in cocaine use disorder. Nevertheless, the expression of downstream genes may be regulated by multiple circRNAs and miRNAs, the expression of the five key circRNAs may be not significantly correlated with its potential downstream gene targets in the sequencing data. Cocaine use disorder is a complex brain disease in which many factors, such as cell subtype specificity, synaptic plasticity, and neural circuit, can influence genes expression. The bioinformatics analysis, which integrated several datasets, could only provide a possible research direction, how the circRNAs contributes to the specific mechanism of cocaine use disorder requires more in-depth studies.

In conclusion, our research is the first to use dlPFC circRNAs and mRNA of patients with cocaine use disorder via bioinformatic tools to identify a circRNA–miRNA–mRNA regulatory network in the patients with cocaine use disorder. The circRNA–miRNA-hub genes regulatory sub-network uncovered five important circRNAs that might be involved in cocaine use disorder, providing new insight into the pathogenesis of cocaine use disorder and suggesting potential therapeutic targets that warrant further investigation.

Publicly available datasets were analyzed in this study. This data can be found here: http://www.ncbi.nlm.nih.gov/geo/, GSE99349, GSE124952, and GSE89572.

YC, JS, and SC contributed toward conception and design of research. YC and XL analyzed the data and wrote the manuscript together. YC, JS, SM, SC, XL, and SH interpreted the results and revised the manuscript. All authors have read and approved the final version of the manuscript.

The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.

All claims expressed in this article are solely those of the authors and do not necessarily represent those of their affiliated organizations, or those of the publisher, the editors and the reviewers. Any product that may be evaluated in this article, or claim that may be made by its manufacturer, is not guaranteed or endorsed by the publisher.