Pleiotropy and genetically inferred causality linking multisite chronic pain to substance use disorders.

The SNP-based h 2 of each trait was statistically significant ( Supplemental Table 1 ). Genetic correlations between MCP and SUDs ranged from MCP-OUD r g =0.20 (se=0.063) to MCP-CanUD r g =0.37 (se=0.029), there was also substantial r g between the SUDs, ranging from PTU-OUD and AUD-OUD r g =0.24 to PTU-CanUD r g =0.84 ( Figure 1 , Supplemental Table 2 ). Comparing MCP-SUD r g estimates using a z-test ( Supplemental Table 2 ), we found that MCP genetic correlation with CanUD was statistically higher than that observed for OUD (z-score=2.40, p-value=0.017) and AUD (z-score=2.68, p-value=0.007). We assessed bi-directional genetically inferred causal relationships between MCP and SUDs using the two-sample MR approach. Although the initial MR analyses showed consistent MCP-SUD effects across IVW, weighted median, simple mode, and weighted mode, we observed evidence of horizontal pleiotropy (significant MR-Egger intercept) and heterogeneity among the genetic instruments ( Supplemental Table 3 ). After removing the outliers contributing to heterogeneity and horizontal pleiotropy ( Supplemental Table 3 ), MCP showed a positive, bi-directional relationship with AUD, CanUD, and PTU ( Figure 2 ). However, the effect of MCP on PTU (beta=0.434) and CaUD (beta=0.575) was statistically larger than the effect of PTU (beta=0.085) and CaUD (beta=0.051) on MCP (difference-p-value=1.52×10 −5 and 1.36×10 −6 , respectively). Additionally, in line with the genetic correlation results, the MCP→CanUD effect was statistically higher than the one observed for MCP→AUD (z-score=3.37, p-value=7.64×10 −4 ), while there was no difference between CanUD→MCP and AUD→MCP (p-value=0.572). The Steiger directionality test showed inaccurate estimates only for outlier variants removed during the MR-PRESSO analysis. Using PolarMorphism 31 to test simultaneously the five traits of interest, eight independent SNPs had significant theta q-value. Among them, IHO1 rs7652746 (chr3:49224874; Z MCP =3.81, Z AUD =3.98, Z CanUD =5.74, Z OUD =3.54, Z PTU =8.74, q theta =2.46×10 −3 ) and CADM2 rs1248857 (chr3:84969461; Z MCP =3.57, Z AUD =−3.90, Z CanUD =−2.80, Z OUD =−3.91, Z PTU =3.76, q theta =0.042) showed significant pleiotropy (q theta 2.73; Supplemental Table 4 ). Rs7652746 (beta= 0.014, p-value=1.19×10 −8 ) and rs1248857 (beta=−0.01, p-value=0.001) were also associated with the general addiction risk factor defined by a previous multivariable GWAS 18 . Testing MCP pleiotropy with each SUD separately, we identified additional pleiotropic variants (q theta 3.41; Supplemental Table 5 ; top-result: SLC39A8 rs13107325 q theta =0.001), 22 MCP-CanUD loci (|z-score|> 3.37; Supplemental Table 6 ; top-result: rs61771920 q theta =0.027), and four MCP-OUD loci (|z-score|> 2.50; Supplemental Table 7 ; top-result: NCAM1 rs4444109 q theta =0.009). No pleiotropic locus was found between MCP and PTU. The gene-based and pathway analyses using pleiotropic SNPs as inputs did not reveal associations surviving FDR multiple testing correction ( Supplemental Tables 8 – 15 ). Because of the sensitivity analyses performed, the MR analyses described above did not include any of the variants identified by the PolarMorphism approach as presenting horizontal pleiotropy between MCP and SUD. We performed a BrainWAS (3,146 brain imaging phenotypes) to characterize MCP-SUD pleiotropic loci. Considering a Bonferroni correction (p-value<7.95×10 −6 ) accounting for the two independent MCP-SUD pleiotropic variants (i.e., IHO1 rs7652746 and CADM2 rs1248857) and 3,146 brain imaging phenotypes, we identified 12 statistically significant associations related to rs7652746 ( Figure 3 , Supplemental Table 16 ). The associated imaging phenotypes included seven regional tissue volumes (top result: volume of grey matter in Left IX Cerebellum beta=0.05, p-value=6.41×10 −11 ), a tfMRI activation measurement (90 th percentile of the blood-oxygen-level dependent (BOLD) effect in group-defined amygdala activation mask for faces-shapes contrast, beta=0.05, p-value=3.65×10 −9 ), a white matter tract fractional anisotropy measurement (mean fractional anisotropy in right cerebral peduncle on fractional anisotropy skeleton, beta=0.04, p-value=5.77×10 −8 ), two rfMRI connectivity measures (top-result: NET100 0072 beta=−0.04, p-value=5.29×10 −6 ), and a rfMRI node amplitude measure (NODEamps100 021 beta=0.04, p-value=7.64×10 −6 ). With respect to MCP pleiotropic loci identified by analyzing each SUD individually, we observed many brain-imaging associations: 379 for MCP-AUD variants ( Supplemental Table 17 ), 427 for MCP-CanUD variants ( Supplemental Table 18 ), and 253 for MCP-OUD variants ( Supplemental Table 19 ). Most of these brain-imaging associations were related to variants located in SLC39A8 gene (MCP-AUD rs13107325, MCP-CanUD rs13105581, and MCP-OUD rs13135092) that all showed the strongest association with the volume of grey matter in ventral striatum (p-value<10 −85 ). Testing the colocalization of MCP-SUD pleiotropic loci (i.e., rs7652746 and rs1248857) with respect to the transcriptomic regulation of the surrounding genes, we did not observe evidence of shared mechanisms linking MCP-SUD pleiotropy to genetic control of gene transcription in the brain tissues investigated.

Genome-wide association statistics investigated in the present study were generated previously ( Table 1 ). MCP GWAS was performed in the UKB cohort 13 . MCP was defined as the sum of body sites where chronic pain was recorded (0–7 sites: head, face, neck/shoulder, back, stomach/abdomen, hip, knee). MCP phenotype definition can be considered as a proxy of the degree of pain severity experienced. Indeed, UKB MCP is highly genetically correlated with pain intensity assessed in the MVP cohort with the NRS (r g =0.79) 14 . GWAS statistics for SUD were generated by the MVP 24 and the PGC 25 . In the AUD GWAS 19 , AUD and alcohol dependence (AD) measurements were combined due to the high genetic correlation between them (r g = 0.98). CanUD meta-analysis combined information from six cohorts 16 , including International Classification of Disease (ICD) codes for cannabis dependence or cannabis abuse. OUD GWAS based on a stringent definition combined data from four cohorts and required at least one inpatient or two outpatient visits 17 . The PTU GWAS 18 was performed by combining Fagerström Test for Nicotine Dependence (FTND) and cigarettes per day (CPD) GWAS data using the Multi-Trait Analysis of GWAS (MTAG) approach 26 . Due to the lack of large-scale MCP and SUD GWAS in diverse ancestries, the data described above refer only to individuals of European descent. Similarly, sex-stratified information is not available for most of the traits investigated. Linkage disequilibrium score regression 27 was used to calculate SNP-based heritability and genetic correlation with genome-wide association statistics for MCP, AUD, CanUD, OUD, and PTU. The 1000 Genomes Project European populations were used as linkage disequilibrium reference panel. We performed a bi-directional two-sample Mendelian randomization (MR) analysis using the TwoSampleMR R package (v0.5.6, https://mrcieu.github.io/TwoSampleMR/index.html ) 28 to test causal genetic effects between MCP and SUDs (i.e., AUD, CanUD, OUD, and PTU). We defined genetic instruments with independent genome-wide significant variants (P < 5×10 −8 ; linkage disequilibrium clumping criteria: r 2 =0.001 and a 10,000-kb window). For OUD as exposure, we used a suggestive threshold of P < 5×10 −6 as only four variants were genome-wide significant. We only included variants present in both the exposure and outcome datasets. Variant ID, effect size, standard error, effect allele, other allele, effect allele frequency, and p-value were used in each dataset. Five MR methods were used: inverse variance weighted (IVW), MR-Egger, weighted median, simple mode, and weighted mode 28 . The IVW method was the primary approach due to its higher statistical power 29 . The other tests were assessed to evaluate the concordance of the direction of effects. To verify the reliability of the MR analyses, MR-Egger regression intercept and MR heterogeneity tests were performed. When significant heterogeneity was observed, the MR-PRESSO (Pleiotropy RESidual Sum and Outlier) test was used to remove outliers 30 . The Steiger filtering directionality test was also applied to examine if the results followed the direction in our hypothesis 28 . The PolarMorphism R package 31 was used to investigate individual loci responsible for the horizontal pleiotropy between MCP and SUDs (AUD, CanUD, OUD, and PTU). This method analyzes genome-wide data to identify SNPs associated with multiple traits due to horizontal pleiotropy, attenuating the effect of vertical pleiotropy observed in the genetic correlation via a decorrelating transform 31 . A comparison analysis showed that PolarMorphism approach is more efficient and powerful in detecting pleiotropic variants than other methods currently available 31 . Variants with significant pleiotropy (theta q-value < 0.05; false discovery rate (FDR)-corrected p values related to theta statistics describing the extent to which a SNP is shared) were clumped using the following parameters in PLINK 2.0 32 : –clump-kb 5000000, –clump-p1 0.05, –clump-p2 0.05 and –clump-r2 0.2. Pleiotropy was assessed between all five traits of interest simultaneously and also between MCP and the four SUDs individually. A gene-based analysis was performed using the Versatile Gene-based Association Study 2 tool 33 “best-SNP test” using variants with nominally significant pleiotropy between the traits. A pathway enrichment analysis was performed using the gene sets to investigate gene ontology, biological pathways, and molecular functions. FDR correction (FDR q<0.05) was applied to account for multiple testing. To investigate further MCP-SUD pleiotropic variants, we analyzed their association with brain imaging measures. This analysis was limited only to independent variants identified by PolarMorphism analysis as characterized by MCP-SUD horizontal pleiotropy, because MCP-SUD vertical pleiotropy (assessed in the MR analysis) is expected to be related to direct effects between the traits investigated rather than shared brain mechanisms. Accordingly, we performed a brain-wide association study (BrainWAS) of PolarMorphism-identified variants using 3,146 brain imaging genome-wide association statistics assessed in up to 33,224 individuals from the UKB 34 . Briefly, six types of data were collected for each participant: a T1-weighted structural image, resting-state functional MRI time-series, task functional MRI time-series, T2-weighted FLAIR structural imaging, diffusion MRI, and susceptibility-weighted imaging. The complete brain imaging documentation is available at https://biobank.ctsu.ox.ac.uk/crystal/crystal/docs/brain_mri.pdf . Bonferroni correction was applied to define the significance threshold accounting for the number of MCP-SUD pleiotropic variants and brain imaging phenotypes tested. Considering cis expression quantitative trait loci (i.e., variants located ±1 Mb from the transcription site start of the gene tested) available from GTEx v8 35 , we performed a colocalization analysis with respect to IHO1 and WDR6 genes for rs7652746, and CADM2 for rs1248857 and gene expression in brain tissues using the hyprcoloc R package 36 . Hyprcoloc identifies colocalized traits and candidate causal SNPs using a Bayesian clustering algorithm. Loci with posterior probability >70% were considered as colocalized between all traits, as recommended by the developers 36 . GTEx data permitted us to investigate MCP-SUD pleiotropy in the context of transcriptomic regulation across multiple brain regions. Specifically, restricted our colocalization analysis only to brain regions presenting statistically significant expression quantitative trait loci for rs1248857 (i.e., putamen, caudate, cortex, spinal cord, substantia nigra, hypothalamus, and hippocampus; Supplemental Figure 1 ) and rs7652746 (cortex, substantia nigra, hippocampus, amygdala, putamen, nucleus accumbens, frontal cortex, spinal cord, cerebellar hemisphere, cerebellum, caudate, and anterior cingulate cortex; Supplemental Figure 2 ).

In the present study, we systematically investigated genetic pleiotropy linking MCP to multiple SUDs. Applying multiple complementary approaches, we provide evidence that both cause-effect relationships and shared molecular pathways contribute to MCP-SUD comorbidities. Because the analyses were performed with respect to MCP phenotype, the results should be interpreted in the context of chronic pain itself rather than associations related to chronic pain. Similarly, because of the well-established difference in the genetics of SUDs vs. substance use 37 , 38 , our findings should be considered in the context of the relationship of MCP with addiction rather than substance consumption and use. Since genetic correlation can be considered an estimate of the overall pleiotropy linking two traits of interest 39 , we can derive information regarding the degree of MCP pleiotropy across multiple SUDs by comparing their genetic correlation estimates. While previous studies reported genetic overlap of pain-related phenotypes with several addiction-related outcomes 14 , 16 – 19 , 40 , we observed that MCP-CanUD genetic correlation was statistically higher than those estimated for OUD and AUD. A recent 9-month follow-up of a randomized clinical trial showed increased CanUD risk among individuals using medical cannabis to alleviate pain and other adverse health outcomes 41 . The interplay between pain management and misuse of alcohol, opioids, and tobacco has also been reported in cohorts representing different race-ethnicity groups 42 – 44 . Considering patterns reported by previous studies 13 , 14 , 38 , SUDs, MCP, and pain intensity showed consistent genetic correlations across several traits, such as college completion (r g from −0.20 to −0.54), neuroticism (r g from 0.22 to 0.28), risk taking (r g from 0.22 to 0.37), and depression (r g from 0.32 to 0.4). SUD-MCP genetic correlations contribute to our appreciation of the role of genetic factors in the comorbidity of these outcomes. However, it is important to distinguish between pleiotropy driven by cause-effect relationships and pleiotropy driven by shared molecular pathways. The MR analysis identified possible causal relationships between MCP and SUDs, in both directions. Initially, the MR estimates showed a high degree of heterogeneity, indicating that the genetic instruments used may include loci implicated in both MCP and SUD pathogenesis. After removing the genetic variants responsible for the heterogeneity within the MR instrumental variables, we still observed a bidirectional relationship between MCP and SUDs. However, the effect of MCP on AUD, CanUD, and PTU was much stronger than the one observed in the opposite direction (i.e., AUD/CanUD/PTU→MCP). Additionally, the MCP→CanUD effect was stronger than the MCP→AUD effect in line with the difference observed in the genetic correlation analysis, and with observational studies 45 . The potential role of chronic pain in increasing SUD risk is in line with the well-known use of psychoactive substances to alleviate somatic symptoms. Chronic pain is one of the most common reasons for unprescribed drug use 46 . This result may have important implications, because of the increasing number of individuals using medical cannabis to relieve pain symptoms due to increased accessibility 47 , 48 . The effect of MCP on AUD has also important health implications due to alcohol’s legal status worldwide. In community settings, using alcohol to attempt to temporarily alleviate pain symptoms is common 42 . More severe pain is associated with higher frequency and quantity of alcohol intake, especially in patients with pre-existing alcohol drinking problems 49 . The MR analysis also highlighted a potential causal effect of MCP on PTU, which is consistent with the analgesic properties of nicotine 50 and smoking cigarettes as a coping strategy to reduce emotional distress due to pain 51 . Most of the previous studies we referred to included substance use, rather than SUDs. In the case of the latter, our MR results could have the potential to be used for prevention and intervention practices. In the last few decades, several efforts have been ongoing to monitor SUDs and provide resources to individuals suffering from chronic pain 52 – 54 , with significant improvements in addiction severity scores and reduced mortality risk. However, treating chronic pain in these patients still remains challenging 55 , 56 . Although smaller than the reverse direction, we also observed genetically inferred effects of AUD, CanUD, and PTU on MCP. After removing genetic variants contributing to the heterogeneity within the MR instrumental variables, these SUD→MCP effects were still significant. This could be in line with experimental and epidemiological findings highlighting the potential role of alcohol, cannabis, and tobacco in increasing pain sensitivity and chronic pain risk among substance users 57 – 60 . In line with this evidence, a recent study found that substance use, such as smoking and alcohol intake increased the risk of chronic back, neck/shoulder, hip, knee, abdominal, facial, and widespread pain 20 . With respect to tobacco smoking, pain interference appears to be due to nicotine-induced acute anti-nociceptive effects 61 . Among the SUDs investigated, we did not see any evidence of a cause-effect relationship between MCP and OUD. Despite opioids being the most commonly used substances for pain management 62 , our analyses identified few pleiotropic mechanisms linking MCP to OUD due to the limited power of the GWAS of the latter outcome. Indeed, the null MR results, marginal statistical significance of the r g estimate, and the limited loci identified by the PolarMophism approach suggest that OUD GWAS can provide limited information regarding the pleiotropy of this trait, independently of the method applied. The source of the MCP GWAS could also contribute to limiting the OUD analysis. Indeed, while SUD GWAS were derived from cohorts recruited in different countries (mainly the United States and Europe), the MCP GWAS was based only on UKB participants. Because opioids are less prescribed for pain management in Western Europe than in the United States 63 , UKB may be less informative to investigate the relationship between MCP and OUD. The MR sensitivity analyses showed horizontal pleiotropy and heterogeneity among the genetic instruments. Indeed, we had to remove these outlier variants to obtain robust estimates of the possible causal effects between MCP and SUDs. However, the presence of cause-effect relationships does not exclude the contribution of molecular pathways shared between MCP and SUDs. Applying the PolarMorphism approach that is specifically designed to investigate horizontal pleiotropy among complex traits 31 , we identified rs7652746 and rs1248857 variants as pleiotropic among all five traits of interest (i.e., MCP, AUD, CanUD, OUD, and PTU). Rs7652746 is in high linkage disequilibrium (R 2 >0.5) with genome-wide significant loci associated with depressive symptoms 64 , life satisfaction 64 , coffee consumption 65 , and smoking initiation 66 . Our brain-wide analysis highlighted that rs7652746 is associated with 12 brain brain imaging phenotypes. Seven of them were related to cerebellum volumetric changes, where the effect allele (rs7652746*A) increasing SUDs and MCP was associated with increased cerebellar volume. While it is known that cerebellum is linked to pain perception 67 , with respect to addiction, the cerebellum could play a role in controlling motivation and saliency due to the widespread presence of opioid receptors in this brain region 68 . Rs7652746 was also associated with the white matter tract fractional anisotropy measurement in the cerebral peduncle, and while the relationship between SUD and white matter alterations is still unclear 69 , changes in the cerebral peduncle may be relevant in the context of the role of cerebellum among SUD-MCP shared pathogenetic pathways. Beyond cerebellum, rs7652746 variant was also associated with amygdala activation in response to observing faces and shapes 70 , 71 . Although, to our knowledge, there is no information regarding this imaging phenotype in the context of pain and SUDs, the amygdala is linked to addiction-related emotion dysregulation 72 and is a center for pain modulation 73 . Conversely, rs7652746 association with two rfMRI phenotypes is in line with evidence of abnormal brain functional connectivity in SUD- and pain-affected individuals 74 , 75 . Rs7652746 maps to an intron in the IHO1 gene, which encodes a protein involved in meiotic recombination 76 . Additionally, it is an expression trait locus for 14 genes in multiple brain tissues ( Supplemental Table 20 ). However, our colocalization analysis did not identify a causal link between MCP-SUD rs7652746 pleiotropy and transcriptomic regulation, suggesting that the effect of this locus may be related to cell-specific or other types of regulatory mechanisms. The other variant showing MCP-SUD pleiotropic effects was CADM2 rs1248857, which is in linkage disequilibrium (R 2 >0.20) with genome-wide significant loci associated with addiction-related traits such as problematic alcohol use 77 , 78 , alcohol consumption 79 , smoking initiation 80 , and cannabis use 81 . A previous phenome-wide study conducted across multiple cohorts showed that CADM2 locus is associated with both behavioral traits (e.g., neuroticism, mood instability, and risk-taking) and somatic outcomes (body mass index, blood pressure, obesity, and C-reactive protein levels) 82 . CADM2 encodes a synaptic cell adhesion protein important for synapse organization axon guidance, and neuron myelination 83 . This gene appears to play a role in impulsivity and personality traits that can be associated with an increased risk of substance use disorders and other comorbid conditions 84 . Impulsivity was also reported as genetically correlated with chronic pain 84 . Accordingly, impulsivity may be the link responsible for CADM2 pleiotropic effect on MCP and SUDs. However, rs1248857 showed no association with brain imaging phenotypes and no brain-specific transcriptomic colocalization with MCP-SUD pleiotropy. This highlights the need for further studies to understand the brain mechanisms connecting this locus MCP and SUDs. When testing MCP pleiotropy with each SUD separately, we identified multiple loci for AUD (n=25), CanUD (n=22), and OUD (n=4). Many of these loci showed robust associations with multiple imaging phenotypes. Variants located in the SLC39A8 gene (MCP-AUD rs13107325, MCP-CanUD rs13105581, and MCP-OUD rs13135092) accounted for the majority of the brain-imaging associations with the strongest being with the volume of grey matter in ventral striatum. SLC39A8 rs13135092 (linkage disequilibrium R 2 = 0.92 with rs13107325 and 0.49 with rs13105581) is a schizophrenia-associated variant that appears to affect zinc concentration and dendritic spine density 85 . In the context of MCP-SUD pleiotropy, dopaminergic innervation in the ventral striatum appears to mediate the effect of psychoactive substances 86 and to be altered by chronic pain 87 . Taken together, these findings suggest both substance-shared and substance-specific MCP pleiotropic loci. However, we did not identify any pleiotropic locus linking MCP to PTU. This null observation does not appear to be due to a lack of power, but rather we hypothesize that MCP-PTU genetic correlation is primarily driven by possible causal effects rather than shared molecular pathways. Overall, the key findings of the present study have substantial translational implications. The potential effect of MCP on SUDs observed in our MR analyses highlighted that screening programs among individuals affected by MCP could contribute to reducing SUD risk among them. With respect to pleiotropy analysis, the loci identified may be further investigated to refine potential molecular targets to develop novel compounds and/or to repurpose existing drugs to specifically address MCP-SUD comorbidity. Our study has several limitations. Because the MCP GWAS was defined as the sum of body sites where chronic pain was recorded, we could not focus on SUD association with chronic pain in specific body locations, caused by certain illnesses, or related to other external factors. Furthermore, chronic pain is a highly heterogeneous trait that can vary based on the patients’ pain tolerance, the scale used to record the pain, and specific treatments 88 . As the International Association for the Study of Pain (IASP) is proposing novel classifications to distinguish chronic pain as a disease from chronic pain as a symptom 89 , future studies may able to rely on more detailed information to characterize accurately chronic pain spectrum. We also recognize that participation and self-reported biases could affect our results, especially with respect to the UKB MCP GWAS. However, selection bias appears to not affect strongly MR analyses 90 . Additionally, MCP GWAS based on UKB cohort and SUD GWAS based on meta-analyses excluding UKB are likely affected by different dynamics, decreasing the probability of shared biases that generate false positive results. The null results observed in some OUD analyses are likely driven by the small sample size of the GWAS available to date. The analyses were conducted only using GWAS data generated from individuals of European descent, because of the lack of large-scale datasets informative of other human populations. Similarly, sex-stratified GWAS are not available for SUD. Accordingly, further studies will be needed to investigate whether the pleiotropic mechanisms identified in the present study are generalizable to cohorts more representative of different demographic groups. Furthermore, there may be possible bias in the MR analysis due to assortative mating, dynastic effects, and population structure 91 . Finally, we could not perform multivariable MR analyses testing whether MCP-SUD relationships are partially due to other socioeconomic factors due to the lack of datasets independent from the ones investigated in the present study ( Table 1 ). To our knowledge, this is the first large-scale study to investigate MCP with respect to multiple SUDs. The results demonstrated that the MCP-SUD association is due to different pleiotropic mechanisms. Among them, we observed that genetic factors might contribute more to MCP-CanUD relationship than to MCP-AUD and MCP-OUD associations. Our genetically informed causal inference analysis highlighted a strong MCP effect on AUD, CanUD, and PTU. In addition to possible causal effects, we also identified loci with genome-wide significant evidence of MCP pleiotropy with multiple SUDs and with specific SUDs. Finally, this investigation contributes to dissecting the complex dynamics linking chronic pain to SUDs, highlighting the need for complementary approaches to address these debilitating health conditions.

In ancient Greece, adequate pain management was already understood as a medical obligation and responsibility 1 . Nevertheless, patients still regularly experience uncontrolled pain 1 . Chronic pain denotes persistent and enduring pain that extends beyond the regular recovery timeframe – typically longer than three months -, or is experienced along with a long-term chronic health condition, such as arthritis or endometriosis 2 . Approximately 10% of the global population experience persistent pain with different prevalence between countries and regions 3 . A recent National Health Interview Survey conducted in the United States reported 50.2 million adults (20.5% of the population) with pain on most days or every day 4 . Individuals living with chronic pain develop substance use disorders (SUDs) more often than others due to frequent misuse of substances to alleviate pain 5 . Assessing and treating patients with chronic pain and a substance use history can be challenging for healthcare professionals 6 . Clinicians often face ethical, legal, and practical concerns when treating patients with addiction 6 . Co-occurring of other mental illnesses, such as anxiety and mood disorders, contribute to the difficulty of treating patients with SUDs and chronic pain 7 . During the last few decades, the prescription of opioids for treating chronic pain not related to cancer has led to an opioid epidemic, and consequently other substance use disorders 8 . Approximately 22.6 million, or 8.9% of all population over the age of 12 in the United States were either current or past month illicit drug users in 2012 8 , and prescription-opioid use caused the deaths of nearly 500,000 people between 1996 and 2019 9 . Currently available treatments offer only modest pain relief, and minimal improvements in physical and emotional well-being, which could have contributed to the opioid epidemic 10 . Understanding the biological basis of the comorbidity between chronic pain and SUDs could help developing early intervention tools for patients in need. A previous study including 7,644 individuals found significant family-based heritability ( h 2 ) for “any chronic pain” ( h 2 = 0.16; i.e., mild or severe pain) and “severe chronic pain” ( h 2 = 0.30) while accounting for age, body mass index, sex, household income, occupation, and physical activity 11 . Although chronic pain shows significant h 2 , and is one of the most prevalent issues in healthcare, studying chronic pain genetics is difficult partly due to the heterogeneity in pain assessment and experience 12 . A genome-wide association study (GWAS) of multisite chronic pain (MCP; a measure of the number of sites of chronic pain, ranging from 0 to 7) in ~380,000 UK Biobank (UKB) participants reported a single nucleotide polymorphism (SNP)-based h 2 of 10.2% and identified 75 associated variants in 39 independent loci 13 . A polygenic risk score (PRS) constructed for MVP was significantly associated with having chronic pain all over the body 13 . A subsequent GWAS of 598,339 participants in the Million Veteran Program (MVP) identified 125 independent loci associated with pain intensity and reported a SNP- h 2 of 8% 14 . The pain intensity of veterans was measured with the question “Are you in pain?”, followed by a response on a scale from 0 (“no pain”) to 10 (“unbearable pain”) using the Numerical Rating Scale (NRS) 15 . Pain intensity had a positive genetic correlation (rg) with other pain phenotypes, including MCP (r g =0.79), a variety of smoking-related measures (e.g., smoking initiation, r g =0.33), cannabis use disorder (CanUD; r g =0.35), alcohol dependence (r g =0.45), and opioid use disorder (OUD; r g =0.37). Previous research also reported a genetic correlation between MCP and CanUD (r g =0.35) 16 . Additionally, several pain phenotypes from the UKB, including low back pain were genetically correlated with OUD (r g =0.44) 17 . Chronic pain was associated with the PRS for addiction risk factor 18 , while back pain was associated with the PRS for problematic alcohol use 19 . Using latent class analysis, a previous study also found that substance abuse/addiction-associated traits (e.g., smoking, alcohol intake) increased the risk of chronic back, neck/shoulder, hip, knee, abdominal, facial, and widespread pain 20 . These previous findings suggest a complex genetic link between chronic pain and SUDs. However, to our knowledge, no study has systematically investigated the pleiotropy underlying the association between chronic pain and different SUDs. Additionally, while some imaging studies highlighted which could be the brain mechanisms linking SUDs to pain-related outcomes 21 – 23 , no large-scale study integrated brain imaging data to understand the genetic mechanisms related to the comorbidity between SUD and chronic pain. In the present study, we evaluated the genetic basis of the comorbidity observed between MCP and SUDs using large-scale genome-wide datasets from UKB, MVP, and the Psychiatric Genomics Consortium (PGC). We tested genetic correlation, possible cause-effect relationship, and SNP-level pleiotropy between these disorders. Additionally, we conducted a brain-wide association analysis, testing more than 3,000 brain imaging phenotypes to link MCP-SUD pleiotropy with changes in the central nervous system. The results of these analyses provided insights into the brain mechanisms contributing to the relationship between SUDs and MCP.