Multiomic Network Analysis Identifies Dysregulated Neurobiological Pathways in Opioid Addiction

Abstract

Background: Opioid addiction is a worldwide public health crisis. In the United States, for example, opioids cause more drug overdose deaths than any other substance. However, opioid addiction treatments have limited efficacy, meaning that additional treatments are needed.

Methods: To help address this problem, we used network-based machine learning techniques to integrate results from genome-wide association studies of opioid use disorder and problematic prescription opioid misuse with transcriptomic, proteomic, and epigenetic data from the dorsolateral prefrontal cortex of people who died of opioid overdose and control individuals.

Results: We identified 211 highly interrelated genes identified by genome-wide association studies or dysregulation in the dorsolateral prefrontal cortex of people who died of opioid overdose that implicated the Akt, BDNF (brain-derived neurotrophic factor), and ERK (extracellular signal-regulated kinase) pathways, identifying 414 drugs targeting 48 of these opioid addiction-associated genes. Some of the identified drugs are approved to treat other substance use disorders or depression.

Conclusions: Our synthesis of multiomics using a systems biology approach revealed key gene targets that could contribute to drug repurposing, genetics-informed addiction treatment, and future discovery.

Keywords: Addiction; Bioinformatics; Multiomic; Networks; Opioids; Systems biology.

Figure 1.

Multiomic integration of OA phenotypes. Beginning with multiple datasets collected from OA cases and controls, OA genes were assembled from H3K27ac ChIP-seq (peaks assigned to nearest gene), DNAm (CpG methylation site to nearest gene), GWASs (SNP-to-gene assignment), differentially abundant proteins by LC/MS proteomics (protein-coding genes), and differential gene expression by RNA-seq. After integrating the overlapping and distinct genes identified by each omics data type, a biological multiplex network consisting of networks from multiple lines of biological evidence were constructed using data sources separate from any of the OA omics datasets. Network traversal algorithms were used to identify mechanistic connections among the multiomic genes and identify dysregulated pathways in the dlPFC. ChIP-seq, chromatin immunoprecipitation sequencing; dlPFC, dorsolateral prefrontal cortex; DNAm, DNA methylation; FC, fold change; GWAS, genome-wide association studies; LC/MS, liquid chromatography-mass spectrometry; OA, opioid addiction; RNA-seq, RNA sequencing; SNP, single nucleotide polymorphism. (Figure created with BioRender.com.)

Figure 2.

Opioid addiction genes are retained by GRIN from multiple omics types and result in more GO enrichments compared with the unfiltered gene set. (A) Percentages of genes retained (orange) or removed (blue) by GRIN by each omics type, including genes shared across omics types (ChIP-seq_RNA-seq, GWAS_MTAG_GWAS, pcHiC-VELs_ChIPseq, and pcHiC-VELs_ChIPseq_RNA-seq). (B) From the 404 original opioid addiction genes (Pre-GRIN), 31 GO biological process terms were significantly enriched. The 211 genes retained by GRIN (post-GRIN) were enriched for 28 of the same GO biological processes but were uniquely enriched for 105 additional terms. Only 3 terms were significantly enriched in the set of 404 genes prior to GRIN that were not significantly enriched in the post-GRIN set of 211 genes. ChIP-seq, chromatin immunoprecipitation sequencing; DNAm, DNA methylation; GO, Gene Ontology; GRIN, Gene set Refinement through Interacting Networks; GWAS, genome-wide association studies; pcHIC, promotor capture HIC; RNA-seq, RNA sequencing.

Figure 3.

Multiomic OA genes are tightly interconnected as demonstrated by network biology. (A) Network visualization of the shortest pathways between all pairs of 50 OA-associated genes from GWAS and dorsolateral prefrontal cortex omics. Forty-three genes were directly connected to at least 1 other gene by the networks (direct neighbor OA genes), and only 127 additional genes (network-connecting genes) were necessary to connect the other 7 genes (one neighbor OA genes) from 10 network layers. Notably, 3 network-connecting genes (SERPINB1, SORCS1, and SORL1) were members of the larger 211 GRIN-retained OA gene set. Gene legend indicates gene color from which omics data type or whether it is a network-connecting gene; network connections legend indicates network layer used to connect gene pairs. (B) Using random walk with restart to explore the biological networks starting from 50 OA-associated genes, 5-fold cross-validation exhibits high recall based upon a mean AUROC value of 0.94. AUROC, area under the receiver operating characteristic curve; ChIP-seq, chromatin immunoprecipitation sequencing; GRIN, Gene set Refinement through Interacting Networks; GTEx, Genotype-Tissue Expression Project; GWAS, genome-wide association studies; LC/MS, liquid chromatography-mass spectrometry; OA, opioid addiction; RNA-seq, RNA sequencing.

Figure 4.

Conceptual model of opioid addiction pathways implicated by multiomic integration. Conceptual model of 45 opioid addiction genes identified via multiple omics data types and 26 additional genes, proteins, or molecules in associated pathways. The mu opioid receptor (OPRM1) and GABA_B receptor (GABBR2) inhibit downstream adenylyl cyclase/protein kinase A (PKA) signaling. PKA can phosphorylate ERK1, which is bound by the scaffolding protein PEA-15, and ERK can also be activated by upstream netrin (NTN1) and BDNF signaling molecules that were implicated in opioid addiction (BDNF, NTRK2, RASGRF1). The proteins encoded by DUSP2, DUSP4, DUSP6, DUSP10, and PPP6C all function as ERK phosphatases, and phosphodiesterase 4B (PDE4B) can reduce PKA activation, while Akt signaling (implicated by PIK3RA and GSK3B) and ERK signaling can activate transcription factors encoded by NPAS4 and CREB5 to activate transcription of NTN1 (identified by DNA hypermethylation) and synaptic plasticity and immediate early genes (ARC, BDNF, EGR1, EGR2, EGR4, ETS1, ETV5, FOS, MYC, NPAS4, SST). ERK can also activate the transcription factor encoded by RORA to promote transcription of ASTN2 and galectin-3 (LGALS3), which is important in microglial inflammatory processes. OPRM1 (chaperoned to the cell membrane by RTP4) and FURIN share a common scaffolding protein (filamin A; FLNA) with the glutamatergic kainate receptor subunit GRIK3, and SLC1A2 is an important glutamate transporter in astrocytes. Additional potassium and calcium channel subunits (ABCC8, KCNMA1, KCNN1) were implicated together with multiple ionotropic GABA_A receptor subunits (GABRE, GABRG3) and cell adhesion molecules (NCAM1 and NRXN3). The color of gene text indicates which opioid addiction omics dataset the gene originated from, and the shading of the gene indicates the logFC of expression or histone acetylation state (not applicable for pcHi-C VELs or GWAS genes). NTN1 was hypermethylated by a mean difference of 0.29 rather than a logFC difference in methylation state. Gray text genes indicate genes or molecules that are involved in pathways but were not implicated by an omics study. BDNF, brain-derived neurotrophic factor; ChIP-seq, chromatin immunoprecipitation sequencing; ERK, extracellular signal-regulated kinase; FC, fold change; GABA, gamma-aminobutyric acid; GWAS, genome-wide association studies; LC/MS, liquid chromatography-mass spectrometry; pcHi-C VELs, promoter capture–Hi-C variant enhancer loci; RNA-seq, RNA sequencing; TLR, toll-like receptor. (Figure created with BioRender.com.)

Figure 5.

Drug-gene target network includes current opioid treatments and putative candidate drugs for treatment repurposing. Using DrugBank and Cytoscape, a network of 414 drugs and 48 target genes (orange) was implicated in multiomic opioid addiction studies. These drug interactions included Food and Drug Administration–approved, investigational, and experimental uses of drugs targeting these genes, resulting in 48 gene targets and 414 total drugs. Current opioid addiction/OUD treatments (light yellow-green) that target the μ opioid receptor (OPRM1) include buprenorphine, methadone, and naltrexone, as well as nalmefene and naloxone to prevent overdose. In addition to nalmefene and naltrexone, which are used to treat OUD and AUD, 4 approved or experimental AUD treatments (light green) were present in the network: acamprosate (targeting GABRE and GABRG3), baclofen (targeting GABBR2), ibudilast (targeting PDE4B), and topiramate (targeting CACNB2, GRIK3, and SCN8A). Drugs with known psychiatric effects (teal) targeting opioid addiction genes include spironolactone (also investigated for AUD) and antidepressants (e.g., amitriptyline, amoxapine, and esketamine, targeting 6 unique genes). Some drugs in the network with known misuse potential (dark blue) include antipsychotic drugs (e.g., aripiprazole, quetiapine) and many benzodiazepines that act as anxiolytics (e.g., alprazolam, lorazepam). Multiple ion channel receptor subunits (15 total) are also known drug targets, and 9 genes are known to be targeted by fostamatinib. Other drugs with yet unknown psychiatric effects that target opioid addiction genes are shown in light blue together with a number of opioids (gray). This drug-gene target network may guide additional hypotheses and follow-up experiments to test the efficacy of these drugs in combating opioid addiction processes. AUD, alcohol use disorder; OUD, opioid use disorder.