Background

Pain is a debilitating problem that cancer patients face, impairing their quality of life. Pain may be related to multiple factors, including radiotherapy, chemotherapy, surgery, and cancer progression. In order to mitigate therapy-related pain or cancer-related pain, physicians often prescribe opioid analgesics to cancer patients [1, 2]. The prescription of opioids for pain carries risk for opioid abuse and addiction. Because of the increased survival rate in cancer patients, their exposure to prescriptions of opioids are also prolonged, further increasing their risk for opioid abuse and addiction [3–5].

Studies showed that opioid abuse was associated with past histories of drug and alcohol abuse in patients treated for cancer-related pain with opioid analgesics [6, 7]. Several clinical trials also found that patients with a history of cigarette smoking and illicit drug abuse had a significantly higher risk for opioid addiction than those without the history [8–11]. Taken together, these studies suggest that past addictive behaviors to various substances may predict opioid addiction in cancer patients with opioid prescriptions for pain. However, very few studies have explored whether there exists a genetic basis and common pathways to the relationship between smoking, alcohol, and opioid addiction.

Bioinformatics uses methods and software tools to organize and analyze biological data [12]. Specifically, gene network analyses have been used frequently to identify genes associated with drug abuse and addiction [13–15]. However, there has been limited application of bioinformatics in understanding multiple addiction phenotypes, specifically, smoking, alcohol and opioid addiction. We hypothesize that the associations between alcohol, smoking, and opioid addiction observed in the clinical setting have a genetic basis.

The goal of the current study is to use bioinformatics tools to determine whether there exists a genetic basis and common pathways to the relationship between smoking, alcohol, and opioid addiction and identify new candidate target genes. Understanding the genetic bases of addiction will underscore the importance of integrating genetic studies into the process of drug administration, as well as allow clinicians to more accurately tailor a patient’s drugs and dosage based on medical history and genetic risk factors [16].

Methods

With the goal of identifying commonly shared genes for alcohol, smoking and opioid addiction we conducted a literature search as described below. Subsequently, using genes pooled from literature as a starting point, we performed gene network analyses: a) specific to each phenotype (Phenotype Specific Biological Network) and b) commonly shared between alcohol, smoking and opioid addiction (Common Biological Network). Finally, we used the Connect function from IPA My Pathway toolbox to connect the commonly shared genes of the three phenotypes to the signaling network involved in neuronal adaptation/plasticity in substance addiction [17, 18].

Literature search

Each substance of abuse was searched on the PubMed database using the keywords “addiction” and “SNPs” in July 2014. Specifically, we used the term “alcohol addiction SNPs” for alcohol addiction, “smoking/nicotine/tobacco addiction SNPs” for smoking addiction, and “opioid addiction SNPs” for opioid addiction (Fig. 1). No limitations were placed on the year of publication. Non-human trials, literature reviews, and meta-analyses were excluded. Articles about treatment of drug addiction and drug addiction in patients with mental illnesses were also excluded. The genes reported in the literature to be statistically significantly associated with one of the addiction phenotypes were included in the pathway analysis and are called focus genes. The genes that were not replicated in an independent study were excluded. Figure 1 shows the criteria of the literature search.

Fig. 1
figure 1

Literature search flowchart. *Subset after using the following Exclusion criteria: Literature review/meta-analysis, non-human experiments, other mental disorders, recovery/withdrawal, unrelated to phenotype, genes that were not replicated in or confirmed by at least one independent study. **Some overlaps between phenotypes for articles and genes

Full size image

Ingenuity pathway analysis

Ingenuity Pathway Analysis (IPA) was used to produce a comprehensive analysis of the genes commonly shared in these addiction pathways. IPA is a software used to connect molecules based on the Ingenuity Knowledge Base, its database of information on biomolecules and their relationships [19]. The Core Analysis function was used to compare genes pooled from literature for each phenotype of addiction with the genes and other molecules in IPA’s database and generates gene networks based on their interactions.

We first designated a set of criteria for the molecules included in the Core Analysis. The following criteria were used: genes and endogenous chemicals, maximum molecules per network (140) and networks per analysis (25), humans, tissues and primary cells. Figure 2 illustrates the steps of the network generation process [20]. The resulting networks are then scored based on the negative base 10 logarithm of the p-value obtained using the Fisher’s exact test (i.e., -log10(p-value)), with the null hypothesis being that the molecules within the networks were connected based on chance.

Fig. 2
figure 2

IPA network generation process

Full size image

Phenotype specific biological network

Gene networks were created for each addiction phenotype. Only the networks with a p-score of 5 or higher were considered significant (i.e., p-value ≤ 10−5), a nominal significance used in previous studies [21]. The genes in each network were ranked based on number of edges, or interactions with other genes in the network.

Common biological network

In order to identify the shared genes, the opioid addiction network was compared with alcohol and smoking addiction networks. In addition, another network was generated by combining all 56 focus genes for all three addiction phenotypes (Fig. 3). In these analyses, only the network with a score ≥ 5 was considered significant [21]. Supplementary to the gene network, IPA also provides a list of top canonical pathways associated with the focus genes, along with a Fisher’s exact test p-value and the ratio between the number of focus genes in the canonical pathway and the total number of molecules in the canonical pathway. In this study, we also reported the top canonical pathways associated with all 56 focus genes for all three addiction phenotypes.

Fig. 3
figure 3

Network generated by pooling all 56 focus genes for alcohol, nicotine and opioid addiction (p-score = 45)

Full size image

Finally, in order to understand the biological context of the gene network (association of genetic variations with addiction to opioids, alcohol and nicotine), we used the Connect function from IPA My Pathway toolbox to connect the commonly shared genes of three phenotypes to the signaling network involved in neuronal adaptation/plasticity in substance addiction [17, 18]. The Connect function adds specific interactions between molecules. While performing this analysis, we limited the interactions from only human studies. All results were generated through the use of Ingenuity® iReport [19].

Results

Literature search

A total of 73 unique articles were extracted based on the PubMed search for a thorough review. Figure 1 illustrates how the PubMed search produced this final list of articles for literature review. The articles associated with the corresponding type of addiction were summarized in tabular format (Tables 1, 2, 3). This resulted in a list of 56 focus genes total (Fig. 1), and each of these genes was used in the IPA Core Analysis. Opioid receptor genes [22] were frequently studied for alcohol and opioid addiction [22–28]. Nicotinic acetylcholine receptor genes were widely explored for alcohol and nicotine addiction [29–46]. Dopamine receptor genes were frequently explored in all three phenotypes [5, 27, 47–53]. Several overlapping focus genes across the three addiction phenotypes were observed, including DRD2 and CRHR1 for all three phenotypes, OPRM1 for alcohol and opioid addiction network, and BDNF and CNR1 for nicotine and opioid addiction network (Table 4). The 56 focus genes were subsequently used as seed genes in Ingenuity Pathway Analysis.

Table 1 Summary of literature search - alcohol addiction
Full size table
Table 2 Summary of literature search - smoking addiction
Full size table
Table 3 Summary of literature search - opioid addiction
Full size table
Table 4 Overlapping genes for networks of nicotine, alcohol and opioid addiction; focus genes from literature are bolded
Full size table

IPA – Phenotype-specific biological network

Individual gene networks were generated through IPA’s Core Analysis for each addiction phenotype (Additional file 1: Figures S1-S3). TNF, NF-κB, and ERK1/2 were present as highly interconnected genes for alcohol addiction (103, 86, and 62 edges, respectively). For nicotine addiction, TNF, ERK1/2 and Akt had the most edges (85, 76, and 53, respectively). NF-κB, RELA, and ERK1/2 were most interconnected for opioid addiction (112, 92, and 74 edges respectively).

IPA – Common biological network

Table 4 lists overlapping genes for alcohol and opioids (A), smoking and opioids (B), and all three addiction phenotypes (C). Genes were ranked by the number of edges within the opioid network. The network for opioid addiction was found to have the most number of genes that overlap with the network for alcohol addiction relative to the smoking addiction genes. ERK1/2 was found to be very strongly interconnected across all three addiction networks with 74 edges in opioid network, 62 edges in alcohol network and 76 edges in nicotine network (Table 4, panel C). ERK1/2 also shows with highest number of edges in opioid and nicotine network (Table 4, panel B) and second highest edges in opioid and alcohol network (Table 4, panel A). We also noticed that some commonly shared genes are involved in the immune response. Specifically, the immune response genes that were common in the three networks (panel C) were: corticotropin-releasing hormone receptor 1 (CRHR1), chemokine ligand 21 (CCL21), chemokine ligand 3 (CXCL3), chemokine ligand 5 (CXCL5) and toll-like receptor 6 (TLR6). In addition to the above genes, the following immune response genes were also found in opioid and alcohol genes networks (panel A): beta-defensin 103 (DEFB103A/DEFB103B), beta-defensin 2 (DEFB4A/DEFB4B), elastase neutrophil expressed (ELANE), protease activated receptor 2 (F2RL1), lactoferrin (LTF), nuclear factor kappa-light-chain-enhancer of activated B cells (NF-kappa B), toll-like receptor 1 (TLR1), TSC22 domain family protein 3 (TSC22D3), chemokine ligand 22 (CCL22), chemokine ligand 2 (CXCL2), interleukin 1 receptor type 1 (IL1R1), tumor necrosis factor ligand superfamily member 11 and 15 (TNFSF11 and TNFSF15).

By pooling all 56 focus genes from three addiction phenotypes, a total of 8 networks were generated by using IPA Core Analysis. Figure 3 shows the network with the highest statistical significance (p-value = 10−45). Figure 4 shows the top canonical pathways for the combined focus genes, including calcium signaling, GPCR signaling, cAMP-mediated signaling, GABA receptor signaling, and Gαi signaling (p-values = 1.26E-12, 4.45E-12, 1.71E-11, 6.3E-10, 4.29E-8).

Fig. 4
figure 4

Top canonical pathways obtained by pooling all 56 focus genes for alcohol, nicotine and opioid addiction. Blue bars: p-score for each of the canonical pathways. Yellow lines: ratio for each of the canonical pathways, calculated as the number of focus genes included in the canonical pathway divided by the total number of genes that constitute the canonical pathway

Full size image

Biological context

Finally, we used the “Connect” function from IPA My Pathway toolbox to connect the commonly shared genes (i.e., overlapping genes) related to addiction to opioids, alcohol and nicotine (Table 4, panel C) to the signaling network involved in neuronal adaptation/plasticity in substance addiction (Fig. 5) [17, 18]. Particularly, DRD2 is the gene common to both the list of genetic variations associated with substance addiction and the components of the brain neuronal signaling network involved in substance addiction. IPA identified multiple links between components of these 2 lists of genes. ERK1/2 was directly connected to DRD1 and indirectly connected to RAC1, FOS, ERK, Creb, PI3K, BDNF and Pka in the signaling network in neuronal adaptation/plasticity in substance addiction (i.e., reward circuit). All the commonly shared immune response genes for the three addiction phenotypes, including TLR6, CXCL5, CXCL3, CRHR1 and CCL21, were indirectly linked to NFkB in the reward circuit. Gene CCL21 was also indirectly linked to Akt and ERK in the reward circuit.

Fig. 5
figure 5

The links of genes associated with addiction to opioids, tobacco and alcohol to components of the brain “reward circuit”

Full size image

Discussion

One of the most challenging areas of oncologic medicine is the management and treatment of severe, chronic pain that arises from cancer therapies, including surgery, chemotherapy, and radiation, as well as cancer itself. Opioids remain the drugs of choice for cancer pain management [54], however, the use of opioids for treatment of chronic pain in cancer patients remains debatable. An increasing concern is the potential rise in aberrant drug-taking behaviors of cancer patients undergoing treatment for chronic pain [3, 55]. Given that addictions to alcohol and tobacco are known risk factors for cancer, exploring genetic markers of risk for these addiction phenotypes in cancer patients may help in risk stratification. Indeed, studies have begun to show that genetic vulnerability to different substances of addiction may partly overlap [56]. The primary aims of this study were to determine whether there exists a genetic basis to the relationship between smoking, alcohol, and opioid addiction, and to identify candidate genes associated with the three phenotypes for further study.

We used IPA, a bioinformatics tool, to identify commonly shared genes for alcohol, smoking, and opioid addiction. Of the 20 genes commonly shared across the alcohol, smoking and opioid addiction phenotypes, extracellular-signal-regulated kinases 1 and 2 (ERK1/2) was found to have the most interconnections across all three addiction networks as indicated by the number of edges (biological interactions; Table 4). Recent studies suggest the relevance of ERK pathway in drug addiction. Several studies have cited the role of ERK in brain’s response to drugs of abuse [57–59]. Specifically, Valjent et al. [59] demonstrated that multiple drugs of abuse increased activation of ERK1/2. Molecular mechanisms underlying ERK1/2 activation by drugs of abuse and the role of ERK1/2 signaling in long-term neuronal plasticity in the striatum may provide novel targets for therapeutic intervention in addiction [60]. Moreover, studies exploiting ERK activation for cancer therapy have been promising, including the use of MEK inhibitors to block ERK activation in acute lymphoblastic leukemia for instance [61]. Future studies are needed to assess the potential clinical relevance of ERK1/2 for addiction, e.g., to genotype ERK1/2 and stratify patients for prompt intervention, or to determine appropriate dosage of opioid analgesics to patients with specific genotypes.

Of note, the identified shared genes for the three addiction phenotypes are involved in immune response. This is consistent with recent research that implicates immune signaling in drug addiction. Dafney et al. demonstrated that certain immunosuppressive treatments controlled morphine withdrawal in rats [62, 63]. More recent studies demonstrated that blocking pro-inflammatory glial activation could block the elevation of dopamine induced by opioid receptor activity [64, 65]. Hutchinson et al. have also found evidence that toll-like receptors (TLRs), a class of innate immune receptors, interact with opioids and glial cells, contributing to opioid reward behaviors [65]. Our recent studies also showed that cytokine genes are implicated in pain, depressed mood, and fatigue in cancer patients [66–68], and these cytokines may serve as biomarkers of risk for persistent pain in cancer patients.

Furthermore, it is also speculated that synaptic plasticity induced by substances of abuse in the neuronal circuits of reward may underlie behavioral changes that characterize addiction. Importantly, NF-kappa B may be the link between inflammation and neuronal/synaptic plasticity involved in behavioral changes in addiction, as we have shown that all the commonly shared immune response genes of three addiction phenotypes were linked to NF-kappa B in the reward circuit (Fig. 5). NF-kappa B is one of several transcription factors present at the synapse, and it is activated by brain-specific activators such as glutamate (via AMPA/KA and NMDA receptors) and neurotrophins [69]. To date, there are currently no pharmacotherapies for drug addiction targeting immune signaling.

Our results also showed the top canonical pathways associated with all the 56 focus genes of three addiction phenotypes were: 1) calcium signaling, 2) GPCR signaling, 3) cAMP-mediated signaling, 4) GABA receptor signaling, and 5) Gαi signaling. These pathways have been confirmed to be associated with substance addiction in the literature [70–74]. They are the post-receptor signaling pathways for the glutaminergic, dopaiminergic and GABAergic neurons involved in the “reward circuitry” in mammalian brains [75]. Whether these pathways can be used as targets for drug addiction therapy needs to be explored. Our approach of identifying genetic variations associated with addiction to multiple substances and linking to known the neural signaling network involved in substance addiction in the brain has clarified the functional significance of many of the genetic associations to substance addiction. This bioinformatics approach has also identified signaling pathways that may be targeted by drugs. Promising research has shown that allosteric modulators of GPCRs can be used to treat addiction by altering the affinity of the GPCR to its ligand or impacting its downstream signaling responses [72]. Other studies have also suggested positive allosteric modulation of GABAB as a therapeutic strategy for treatment of addiction [71, 76].

Among the limitations of this study is that edges are simplified in the IPA designates only a single edge between each pair of molecules in a network regardless of the number of interactions the two molecules share. Furthermore, this bioinformatics analysis is hypothesis-generating, and the findings must be further investigated and validated experimentally.

Conclusions

Studying smoking, alcohol, and opioid addiction phenotypes in conjunction allowed us to identify molecules and pathways involved in multiple types of drug addiction. IPA is able to use large-scale information to produce comprehensive networks of genes and underlying biological pathways implicated in a phenotype [19]. Most of the current literature on addiction genes focuses on genes specific to each type of addiction, while in this study we studied genes relating to multiple addiction phenotypes. Our findings show immune signaling and ERK1/2 as novel genetic markers for multiple addiction phenotypes including alcohol, smoking and opioid addiction. Future studies are needed to validate our findings in large cohorts of patients.