Chronic overlapping pain conditions and nociplastic pain

We fitted a common factor GWAS model (including individual SNP effects), using data for six COPC traits (CWP, low-back pain, endometriosis, TMJ, IBS, and broad headache). This model estimates the size of association between SNPs and a latent common factor (nociplastic pain), producing results effectively equivalent to a standard GWASs of this unmeasured latent factor ( Figures 3 A and 3B). We found a total of 663 SNPs across 15 GWAS genomic risk loci significantly associated with nociplastic-type pain ( p  < 5 × 10 −8 ), consisting of 24 independent SNPs. The majority of these SNPs (18 of 24) have not been previously associated with pain-related traits ( Table S2 ). Figure 3 Manhattan plot of nociplastic pain common-factor GWAS and TWAS (A) Manhattan plot of common-factor GWAS output. Orange dotted line = genome-wide p value significance threshold (-log 10 (5 x 10 -8 )). (B) Quantile-quantile plot of GWAS output. (C) Manhattan plot of common-factor TWAS output. Manhattan plot of nociplastic pain common-factor GWAS and TWAS (A) Manhattan plot of common-factor GWAS output. Orange dotted line = genome-wide p value significance threshold (-log 10 (5 x 10 -8 )). (B) Quantile-quantile plot of GWAS output. (C) Manhattan plot of common-factor TWAS output. We estimated the SNP-based heritability of the nociplastic pain trait using output from our GenomicSEM common-factor GWAS. We also estimated genetic correlation between nociplastic pain and another complex chronic pain trait, MCP, between nociplastic pain and RA, a disease where chronic pain is a common symptom but is likely associated with nociceptive rather than nociplastic pain mechanisms, 34 , 35 and between nociplastic-type pain and neuropathic pain. Nociplastic pain was found to be significantly heritable (liability scale SNP-h2 = 0.025, SE = 0.0014), and was significantly genetically correlated with MCP (rg = 0.92, SE = 0.04) and to a much lesser degree with rheumatoid arthritis (rg = 0.18, SE = 0.04). Neuropathic and nociplastic-type pain were also significantly genetically correlated (rg = 0.79, SE = 0.099). The LDSR intercept values when calculating these genetic correlations was close to 1 (0.97–1.03), suggesting the majority of genomic inflation captured by lambda GC (estimates ranging 1.0025–1.3) is due to polygenicity rather than population structure. To explore nociplastic pain at the transcriptomic level and uncover tissue-level gene expression relevant to this trait, we carried out a common-factor TWAS analysis of six COPC traits using GenomicSEM. We found 819 tissue-wide significant gene-tissue associations consisting of 127 unique genes, across all 49 tested GTEx tissues ( Figure 3 C). As part of multivariate GWAS and TWAS analyses with GenomicSEM, a Q (heterogeneity) value is calculated per association. This value indicates degree of heterogeneity (i.e., the proportion of gene expression or SNP association effect that is mediated through pathways other than the shared common factor, across the traits included in the model that load onto the common factor). Specific genes were identified through subsetting multivariate TWAS output to include genes in significant (tissue-wide) gene-tissue associations and with non-significant Q p values (i.e., non-significant heterogeneity). Non-specific genes were also identified as above but with significant Q p values. Q p values were Bonferroni-corrected for multiple testing within tissue (Q p adj = Q p /N tests in that tissue). A total of 450 of the 819 tissue-wide significant gene-tissue associations showed significant heterogeneity (Q p adj 0.05). Twenty-eight unique genes showed significant heterogeneity across all tested tissues (Q p adj 0.05). Five genes showed both significant and non-significant heterogeneity depending on tissue, and were excluded from gene set enrichment tests of specific/heterogeneous genes. To determine whether specific tissues showed more or fewer significant TWAS results than expected by chance, we carried out a series of binomial tests for enrichment of significant TWAS results within each tested tissue. We found that nine of the 49 tested tissues showed a different proportion of significant gene-tissue TWAS findings than expected by chance. Cultured fibroblasts, atrial appendage of the heart, tibial nerve, and thyroid were enriched for significant associations, whereas amygdala, substantia nigra, Epstein-Barr-virus-transformed lymphocytes, and terminal ileum of the small intestine and vagina showed significantly fewer significant associations than expected. At the nominal significance level (unadjusted p  < 0.05), two tissues showed a different proportion of associations than expected by chance—whole blood and skeletal muscle were enriched for significant (unadjusted p  < 0.05) associations. To explore potential mechanisms in nociplastic pain that are shared with other complex traits of interest, we performed three sets of gene set enrichment tests using FUMA. First, we used all unique tissue-wide significant gene findings from multivariate TWAS, then a subset of those findings that showed significant heterogeneity (non-specific genes) and finally a subset with non-significant heterogeneity (specific genes). Of 127 unique genes, 120 had a recognized ensemble gene ID within FUMA and were included in analyses. Eight positional gene sets were enriched ( p adj <0.05) for nociplastic-type pain-associated genes; chr17q21, chr3p21, chr9q33, chr16q22, chr12q13, chr2q34, chr4q33, and chr1q21. Kyoto Encyclopedia of Genes and Genomes (KEGG) pathway gene set nitrogen metabolism was also enriched for nociplastic pain genes, as was chemical and genetic perturbation gene set “SU_LIVER” (genes specifically upregulated in the liver). A total of 44 GWAS Catalog traits including cognitive function (adjusted p  = 1.76 × 10 −10 ), extremely high intelligence (adjusted p  = 3.3 × 10 −9 ), sleep duration (adjusted p  = 5.95 × 10 −9 ), and headache (adjusted p  = 1.10 × 10 −8 ) were enriched for nociplastic-type pain genes ( Table S3 ; Figure 4 A). Additional gene sets enriched for nociplastic pain genes including microRNA target sets, transcription factor target sets, computational gene sets, and cancer gene modules can be found in Table S4 . Figure 4 GWAS Catalog trait gene set enrichments (A) Gene sets enriched for nociplastic-type pain genes found in multivariate TWAS analysis. (B) Gene sets enriched for specific nociplastic pain genes. (C) Gene sets enriched for non-specific nociplastic pain genes. GWAS Catalog trait gene set enrichments (A) Gene sets enriched for nociplastic-type pain genes found in multivariate TWAS analysis. (B) Gene sets enriched for specific nociplastic pain genes. (C) Gene sets enriched for non-specific nociplastic pain genes. Ninety of 94 tissue-wide significant genes without significant heterogeneity across all tissues where a significant association was found, and with an ensembl gene ID recognized within FUMA were included in this analysis. We found seven positional gene sets to be enriched for specific nociplastic pain genes, including chr3p21, chr9q33, chr16q22, chr2q34, chr4q33, chr12q13, and chr1q21. Chemical and genetic perturbation gene set “SU_LIVER” was enriched for specific nociplastic pain genes, as was nitrogen metabolism, and the hallmark gene set fatty acid metabolism. Additional gene sets enriched for specific nociplastic pain genes are listed at Table S5 . We found 28 GWAS Catalog traits enriched for specific nociplastic pain genes ( Table S6 ; Figure 4 B), including extremely high intelligence (adjusted p  = 8.02 × 10 −10 ), sleep duration (adjusted p  = 1.09 × 10 −9 ), regular attendance at a religious group (adjusted p  = 2.77 × 10 −9 ), and cognitive function (adjusted p  = 3.08 × 10 −7 ). Twenty-five of 28 tissue-wide significant genes with significant heterogeneity across tissues where a significant association was found, and with an ensembl gene ID recognized within FUMA were included in this analysis. Positional gene sets chr17q21 and chr6q16 were enriched for these non-specific nociplastic pain genes, as were several cancer gene neighborhoods ( Table S7 ) and nine GWAS Catalog traits including migraine without aura (adjusted p  = 9.31 × 10 −7 ), headache (adjusted p  = 1.79 × 10 −6 ), and migraine (adjusted p  = 7.63 × 10 −5 ) ( Table S8 ; Figure 4 C). There was no overlap in gene sets enriched for specific and non-specific genes, and migraine and headache gene sets were enriched for non-specific nociplastic-type pain genes only.

See Figure 1 for workflow diagram of all analyses. Figure 1 Workflow diagram of analyses Workflow diagram of analyses GWAS summary statistics for six COPC traits were obtained through publicly available downloads, requests made directly to study authors, or through data request to FinnGenn (release R9, May 2023). Traits, sample sizes, and sources are summarized in Table 1 , and include chronic widespread pain, 12 low-back pain, 18 broad headache, 19 temporomandibular joint (TMJ) disorder, 20 and IBS. 21 We opted not to include migraine or ME/CFS summary statistics. With ME/CFS 22 this is due to sample size being too small for successful completion of the multivariable linkage disequilibrium-score regression (LDSR) step of GenomicSEM common-factor GWAS analysis (Ncase = 427, Ncontrol = 972), and because genotyping in this study was performed using the Illumina Immunochip array (i.e., not whole-genome genotyping, but a specialized assay focused on immune-relevant SNPs). For migraine, a significant proportion of the “broad headache” GWAS case participants are likely migraineurs and/or have both tension-type headache and migraine, therefore including the broad headache GWAS but not migraine GWAS allows for capturing a fuller spectrum of headache and migraine-associated genetic variation, without overrepresentation of migraine. Table 1 Sources and sample sizes for individual GWASs used in common-factor GWAS and TWAS analyses Trait Acronym Source Cases Controls Sample prevalence Population prevalence Chronic widespread pain CWP Rahman et al. 2021 12 6,914 242,929 0.029 0.01 23 Endometriosis NA FinnGenn (R9, May 2023) 15,088 107,564 0.123 0.1 24 Low-back pain LBP Suri et al. 2021 18 49,182 51,629 0.488 0.075 25 Broad headache NA Meng et al. 2018 19 74,461 149,312 0.5 0.158 26 Temporomandibular joint disorder TMJ Jiang et al. 2021 20 (GWAS Catalog) 217 456,131 0.0005 0.3 27 Irritable bowel syndrome IBS Eijsbouts et al. 2021 21 (GWAS Catalog) 53,400 433,201 0.12 0.1 28 All six traits are considered chronic overlapping pain conditions (COPCs). Sources and sample sizes for individual GWASs used in common-factor GWAS and TWAS analyses All six traits are considered chronic overlapping pain conditions (COPCs). For each trait excluding endometriosis, GWAS samples are at least partially, and in some cases completely, composed of UK Biobank participants and samples may overlap—sample overlap is permitted and does not statistically bias GenomicSEM analyses. 29 , 30 Sample size of individual GWASs varies, but this is accounted for within GenomicSEM through provision of sample and population prevalence estimates during the multivariable LDSR step. 29 GenomicSEM was used to carry out a common-factor GWAS. First, we prepared (munged) GWAS summary statistics using “munge” function included in GenomicSEM. Sum of effective sample sizes was calculated for low-back pain and IBS, as these represented GWAS meta-analyses of case-control phenotypes. 31 Briefly, variants with info values ≤0.9, minor allele frequency (MAF) ≤0.01, that were not SNPs with missing values, that were strand ambiguous, those with duplicated rsIDs, those without a match in the HapMap 3 SNP file used for quality control, and those with mismatched allele labels compared with the HapMap 3 SNP file were removed. Next, multivariable LDSR was carried out to produce matrices used in the common-factor GWAS step. GenomicSEM “sumstats” function was used for a final preparation step, jointly processing the GWAS summary statistics files for each of the six traits included in the analysis. This function merges across all summary statistics using listwise deletion, performs quality control (including backing out logistic betas, checking for allele mismatches, missing data, and duplicate variants), and merges with the reference SNP file. Output from multivariable LDSR and from GenomicSEM “sumstats” is taken forward to common factor GWAS using GenomicSEM “commonFactorGWAS” function. Our chosen common-factor model, where the single latent factor represents “nociplastic pain,” is shown in Figure 2 . Figure 2 Path diagram of common factor GWAS GenomicSEM model Values = Standardized Estimate (Standard Error). Path diagram of common factor GWAS GenomicSEM model Values = Standardized Estimate (Standard Error). First, we fitted the common-factor GWAS model without SNP effects and noted loadings of each trait onto the singular factor ( Figure 2 ). These were significant ( Table 2 , p  < 0.05) and model fitting was successful, so we proceeded to fitting the model including SNP effects with “smooth_check = T.” We note that TMJ variance standard error is very large ( Figure 2 , SE = 3.8), likely due to the small case number in the TMJ GWAS, but factor loading was significant ( Table 2 , 0.69, p  < 0.05) so we elected to include TMJ in this model. Output produced from the above model is equivalent to GWAS summary statistics (magnitude of association between each SNP and a trait of interest) for the latent factor onto which the six traits load, which we describe as nociplastic-type pain. Next, we calculate an estimated sample size for the “GWAS” of nociplastic-type pain according to instruction provided on the GenomicSEM github, first reserving SNP results for SNPs with MAF ≤0.4 and ≥0.1, and then calculating N ˆ according to the formula for each SNP: (Equation 1) 1 ( 2 ∗ M A F ∗ ( 1 − M A F ) ) ∗ S E 2 Table 2 Standardized common-factor GWAS model factor loadings and p values Trait Standardized estimate Standardized SE p value CWP 0.76 0.06 1.78 × 10 −41 IBS 0.62 0.03 3.87 × 10 −73 Broad Headache 0.73 0.04 4.56 × 10 −81 LBP 0.39 0.04 5.53 × 10 −28 TMJ 0.69 0.29 0.0188 Endometriosis 0.57 0.06 3.34 × 10 −20 SE = standard error. Standardized common-factor GWAS model factor loadings and p values SE = standard error. Equation 1 : Formula to estimate per-SNP sample size for SNPs included in common-factor GWASs. Taking the mean of this set of values then gives N ˆ . For nociplastic-type pain N ˆ  = 578,561.3 (where this sample size is necessary in calculations, e.g., LD-score regression analyses for genetic correlation, we use the value 578,561). Common-factor GWAS output (nociplastic pain summary statistics) was then taken forward and analyzed within FUMA (Functional Mapping and Annotation of Genome-Wide Association Studies), 32 a web-based suite of tools for downstream GWAS analyses. Genome-wide significant, independent SNPs were defined using FUMA as SNPs associated with nociplastic pain factor ( p  < 5 × 10 −8 ) and independent from one another (r 2  < 0.6). We used “ldsc” 30 , 33 to estimate SNP heritability of nociplastic-type pain, and to estimate genetic correlation between nociplastic-type pain and three pain-related phenotypes: multisite chronic pain (MCP), a measure of number of chronic pain sites, rheumatoid arthritis, a chronic pain condition that is not usually considered a COPC and where pain for a majority of individuals is likely mainly nociceptive/inflammatory, 34 , 35 and a neuropathic pain phenotype. 36 We performed additional genetic correlation with major depressive disorder (MDD), 37 post-traumatic stress disorder (PTSD), 38 and type 2 diabetes (T2D) 39 using European ancestry GWAS output for each trait (see Table S1 ). We obtained summary statistics for GWASs of multisite chronic pain (MCP), 13 a general chronic pain phenotype previously found to be heritable, polygenic, and significantly associated with gene expression changes in the brain through download from University of Glasgow Enlighten (research data repository). MCP summary statistics were munged as previously described, using “ldsc” package, as was common factor GWAS output for nociplastic-type pain. We then carried out LDSR, estimating genetic correlation between MCP and nociplastic-type pain. We obtained rheumatoid arthritis GWAS summary statistics from a recent large GWAS ( N  = 311,292) 40 from the GWAS Catalog. 41 As previously described and again using “ldsc,” we calculated genetic correlation between nociplastic pain and rheumatoid arthritis. We obtained GWAS summary statistics for a study on neuropathic pain susceptibility (N effective  = 16,311.72) by request to the study authors, 36 using “ldsc” as previously described to calculate genetic correlation between nociplastic-type pain and neuropathic pain. An extension of GenomicSEM common-factor GWAS is multivariate (common-factor) TWAS. Here, FUSION TWAS 42 output for each of the six COPCs serves as input (as GWAS summary statistics did for common-factor GWAS in GenomicSEM common-factor GWAS). We performed TWAS in each of the six COPC traits using GWAS summary statistics and the FUSION package and scripts, and using pre-computed predictive models for all 49 Genotype-Tissue Expression project (GTEx) 43 v8 tissues. 44 We used models including genes with significant heritability and with weights calculated using all genetic ancestries, as recommended for typical analyses and to increase sensitivity. FUSION output (TWAS summary statistics for each of the six COPCs) was prepared using the GenomicSEM “read_fusion” function, and then taken forward with multivariable LDSR output from previous common-factor GWAS analysis to perform common-factor TWAS. Significant gene-tissue association findings were defined at the tissue-wide and experiment-wide level, through Bonferroni multiple testing correction across all tests within a tissue and across all genes tested across all tissues, respectively. As part of multivariate GWAS and TWAS analyses, a Q (heterogeneity) value is calculated per SNP-trait association (or per gene-tissue-trait association), indicating degree of heterogeneity (i.e., the proportion of gene expression or SNP association effect that is mediated through pathways other than the shared common factor). Genes specific to nociplastic-type pain were identified through subsetting multivariate TWAS output to include genes in significant (tissue-wide) gene-tissue associations and with non-significant Q p values (i.e., non-significant heterogeneity). Non-specific genes were also identified as above but with significant Q p  values. Q p values were Bonferroni-corrected for multiple testing within tissue (Qp adj = Qp/N tests in that tissue). We carried out binomial tests of enrichment within our TWAS results to investigate whether certain tissues showed a higher proportion of tissue-wide significant ( p bonf <0.05) associations results than expected by chance, and whether certain tissues showed higher proportion of nominally significant ( p  < 0.05) associations than expected by chance. Genes tested per tissue are available from FUSION predictor models. 42 Gene set enrichment analyses on all tissue-wide significant genes, on tissue-wide significant and specific genes (Q p adj > 0.05), and on tissue-wide significant non-specific genes (Q p adj < 0.05) was carried out using FUMA, with all genes tested in multivariate TWAS and with recognized Ensembl gene ID as background ( N  = 26,455 genes).

We carried out common-factor GWAS and TWAS analyses incorporating six COPCs to investigate genetic variation associated with nociplastic pain, a mechanistic pain descriptor and type of pain likely important across COPCs. This method allows us to find genetic variation associated with nociplastic pain in the absence of a dataset where nociplastic pain is directly assessed. This method also allows for assessment of “heterogeneity,” i.e., the ability to test whether certain SNPs primarily influence a subset or even only one of the traits included in analyses (high heterogeneity), and which are most relevant, in our analyses, to the latent factor (nociplastic pain), and to this end we identify “specific” and “non-specific” SNPs and genes associated with nociplastic pain. We observed extremely high genetic correlation between nociplastic pain and MCP (rg = 0.92). Individuals with a non-zero trait value for MCP could as a group be majority composed of individuals with COPCs—to assess this we carried out a series of Fisher’s exact tests on counts of COPC ICD-10 code occurrences within and outside of MCP “cases” (MCP trait value ≥1) in the UK Biobank (see Table S9 ). We found that for ICD-10 category codes for COPCs that were available in the UK Biobank (back pain, post-viral fatigue (ME/CFS), migraine, IBS, endometriosis, and tension headache), all were significantly overrepresented in MCP cases compared with those without chronic pain (MCP trait value 0), with the most overrepresented COPC being fibromyalgia (odds ratio [OR] 4.48). However, we also tested for enrichment of rheumatoid arthritis (a non-COPC) cases, and again found significant enrichment ( Table S9 ). This suggests high phenotypic overlap in COPCs and MCP in the UK Biobank cannot fully explain high genetic correlation values, although most significant enrichment of a COPC in MCP was with fibromyalgia (OR 4.48), considered a prototypical nociplastic pain condition. Another explanation may be that a large amount of genetic variation associated with the main characteristic of MCP (increasing number of sites of chronic pain) is also captured in a GWAS/GWAS-equivalent analysis of nociplastic pain—a recent paper outlining clinical criteria in assessing and grading nociplastic pain lists regional (as opposed to discrete) location and spread of pain as a key characteristic of nociplastic pain. 45 Neuropathic pain was also relatively highly genetically correlated with nociplastic-type pain (rg = 0.79); however, this rg value is significantly less than 1, indicating a significant portion of genetic variation is distinct between these two pain traits. In addition, a high degree of genetic correlation between nociplastic and neuropathic pain, to a greater extent than rg with nociceptive pain types, could be expected. In their recent review on nociplastic pain, Fitzcharles et al. emphasize not only how common mixed pain states are (pain with nociceptive/neuropathic/nociplastic features), but that neuropathic features in hip and knee arthritis and low-back pain observed in studies predating the concept of nociplastic pain likely represent nociplastic etiology. 8 Neuropathic features of pain in various rheumatic diseases have also been suggested to actually represent nociplastic pain. 46 Gene set enrichment results showed certain genes significantly contribute to variation in nociplastic pain phenotype, but that a large amount of their influence is not truly multivariate and is instead relevant to a subset, or even one, of the contributing COPCs (non-specific genes, i.e., genes with significant heterogeneity in genomicSEM analyses). GWAS trait gene sets enriched for such genes included migraine and headache. Non-nociplastic-type-pain-specific genes were also enriched in trait gene sets for male-pattern baldness (androgenic alopecia)—alopecia, including androgenic alopecia, has been previously associated with drugs that block CGRP (calcitonin gene–related peptide), a treatment for acute migraine and migraine prevention. 47 GWAS trait gene sets enriched for specific nociplastic pain genes included amino acid and acylcarnitine levels. Changes in amino acid levels are associated with fibromyalgia, migraine, osteoarthritis, and complex regional pain syndrome. 48 High acylcarnitine levels can indicate disorders in fatty acid metabolism, and diets high in certain fatty acids have been associated with increased allodynia in rodent models, and associated with human chronic pain conditions. 49 Lipids generally are also involved in acute and chronic inflammation 50 and changes to acylcarnitine metabolism are observed in dementia, certain cancers, heart failure, and coronary artery disease. 51 Changes to circulating lipids have also been observed in COPCs including fibromyalgia, headache, migraine, temporomandibular disorder, low-back pain, and IBS, 52 and cholesterol metabolism in microglia has been linked to neuropathic pain in a rodent model. 53 In addition, certain drug therapies used in treatment for systemic lupus erythematosus and rheumatoid arthritis can disrupt lipid metabolism, 54 and experiencing chronic pain is associated with changes in diet that can result in changes in lipid profiles. 55 Genes specifically upregulated in liver tissue were also enriched for specific nociplastic pain genes. Gene findings from TWAS in this study in theory most likely represent gene expression mediating the relationship between genotype and trait—in other words, gene expression that is likely genetically regulated and that occurs before development of nociplastic pain. Therefore, changes in the liver associated with nociplastic pain may develop prior to nociplastic pain development. Curcuminoids (components of turmeric) have been previously investigated in treatment of neuropathic pain, and previous studies found that a possible mechanism of action in alleviating neuropathic pain was via modulating nitrogen metabolism. 56 Although ME/CFS summary statistics were not used in this analysis, changes in nitrogen metabolism have been implicated in this condition 57 ; our findings suggest these metabolic changes may be shared across COPCs through their association with nociplastic pain. Those with ME/CFS have also been found to have altered lipid, acylcarnitine, and amino acid levels. 58 In line with extensive previous studies linking pain and chronic pain to the immune system, 59 , 60 , 61 , 62 several positional gene sets highlighted in our findings suggest immune-related factor involvement in nociplastic pain. Positional gene sets associated with specific nociplastic pain genes have been previously linked to COVID-19 susceptibility and severity (chr3p21 63 ), schizophrenia and bipolar disorder (chr3p21 64 ), and Alzheimer disease in an African American cohort (chr3p21 17 ). Regions in chr9q33 are frequently deleted in certain cancers, 65 , 66 and chr16q22 has been previously associated with a rare duplication syndrome accompanied by varying psychiatric disorder symptoms, 67 and with schizophrenia. 68 A type of syndactyly has been associated with chr2q34, 69 along with autoimmunity, amyotrophic lateral sclerosis, and schizophrenia, 70 and age-related degeneration in the lumbar spine. 71 Variants in the chr12q13 region have been associated with childhood obesity 72 and asthma. 73 , 74 Finally, chr1q21 has been previously implicated in GWASs of circulating interleukin 6 levels. 75 These findings suggest shared immune and musculoskeletal-related etiology in these phenotypes and nociplastic pain. Other recent studies in the context of pain and long COVID have found that infection may trigger and/or exacerbate existing painful conditions, and that those with an existing COPC were more likely to develop long COVID. 76 In contrast, positional gene sets enriched for non-specific nociplastic pain genes included chr17q21, where duplication and deletion have previously been associated with syndromes involving distinctive craniofacial features, developmental delay, and cardiac symptoms (Online Mendelian Inheritance in Man OMIM: 610443 , 613533 ), and genes at chr6q16 with cardiac phenotypes 77 and cluster headache and migraine. 78 Both chr6q16 and chr17q21 harbor known structural variants, including those associated with smooth muscle involvement in vascular disease. 77 , 79 Autoimmune disorders including autoimmune hypothyroidism may also be misdiagnosed as fibromyalgia, and hypothyroidism may both cause pain and worsen pain associated with pre-existing conditions. 80 , 81 , 82 Enrichment of nociplastic pain gene associations in thyroid may therefore indicate high levels of thyroid involvement in COPCs, presence of individuals with comorbid (or misdiagnosed) hypothyroidism among COPC GWAS participants, or both. Atrial fibrillation, the most common abnormal heart rhythm in adults, can lead to formation of clots (most commonly in the left atrial appendage) and subsequent stroke. 83 , 84 , 85 Electrocardiogram abnormalities, particularly atrial fibrillation, have been linked to chronic pain, 86 , 87 and with physiological stress associated with chronic illness and major surgery. 88 , 89 Chronic pain has also been associated with higher risk of myocardial infarction, death due to cardiovascular event, heart failure, and stroke. 90 Our findings may suggest this atrial fibrillation in particular could be commonly associated with nociplastic pain, increasing risk for stroke in these patient populations. We also found enrichment of nociplastic pain genes in tibial nerve. An ultrasound study comparing participants with fibromyalgia and controls found significant increased cross-sectional area in several nerves, including tibial nerve. 91 Tarsal tunnel syndrome, a nerve entrapment syndrome analogous to carpal tunnel syndrome in the wrist, was also found to be more common in fibromyalgia patients 92 ; while causes of tarsal tunnel syndrome are likely multifactorial, enlarged tibial nerve diameter may contribute to this nerve entrapment. Neuromodulation involving the tibial nerve (e.g., through percutaneous tibial nerve stimulation) has also been investigated in the treatment of a range of pelvic pain disorders, including IBS, dysmenorrhea, and bladder pain syndrome. 93 One caveat is that the tibial nerve is the only peripheral nerve tissue sampled in GTEx; other peripheral nerve tissues, potentially also representing therapeutic targets in neuromodulation for chronic pain, could potentially be enriched for nociplastic pain gene expression. There are several general caveats that should be considered when interpreting these results. There is a lack of availability of suitable large GWASs for several COPC traits. While there are epidemiological studies of pelvic pain conditions (e.g., MAPP 94 , 95 ), there are no GWASs of vulvodynia, chronic pelvic pain, or bladder pain syndrome/IC, and existing GWASs of ME/CFS are underpowered for multivariate GWAS analyses. Discussion of phenotyping, assessing, and understanding nociplastic pain in clinical and population cohorts is ongoing in the field, 8 , 96 , 45 , 97 and there are to date no sufficiently large studies with both genotyping and questionnaire data designed specifically to ascertain nociplastic pain (such as the Central Sensitization Inventory 98 or Nociplastic-Based Fibromyalgia Features tool 99 ). Re-analysis including these GWASs when available would positively impact multivariate GWAS analyses of nociplastic pain. Finally, a major limitation is lack of available GWASs in non-European ancestry populations—it will be key to expand these analyses to include diverse genetic ancestry populations, and this requires availability of large-scale GWASs in other genetic ancestry populations for all individual COPC traits. A third mechanistic pain descriptor, nociplastic pain, may best represent the pain experience of those with COPCs, where tissue and/or nerve damage is often not present. Using existing COPC GWAS data and a network-informed genomics approach, GenomicSEM, we found genetic variation at the SNP, gene expression, and gene set level associated with nociplastic pain. Our findings indicate distinct pathology in migraine and headache compared with other COPCs and link this distinct pathology with traits such as Parkinson disease as well as provide unique genes associated with this pathology. We demonstrate various degrees of genetic overlap between nociplastic pain and various pain traits, showing highest overlap with a trait capturing an increasing number of pain sites on the body (a key characteristic of nociplastic pain in the literature), and lowest with rheumatoid arthritis, a chronic pain condition where pain mechanisms are considered mostly nociceptive. We also find tissue enrichment relevant to chronic pain comorbidities, including stroke and therapies such as peripheral nerve stimulation. Chronic pain conditions can be highly stigmatized 100 , 101 , 102 , 103 , 104 ; this is likely even more common in COPCs due to a lack of or disproportionate-to-pain-level presence of tissue or nerve damage, 105 , 106 , 107 and higher prevalence of COPCs in women. 100 , 108 Understanding mechanisms of chronic pain that are not due to, or cannot be fully explained by, nervous system damage or dysfunction or tissue damage (i.e., nociplastic pain) will contribute to legitimizing the pain experience in COPCs. In addition, finding genetic variation associated with nociplastic pain could inform new treatment approaches in COPCs. Our findings contribute to further understanding mechanisms of nociplastic pain and suggest this type of pain is important across COPCs.

Chronic pain can be defined as pain that persists 3+ months 1 and is a main symptom of many conditions as well as being associated with injury and surgery. Recently, the International Association for the Study of Pain (IASP) also redefined pain and “chronic primary pain” codes for the International Classification of Diseases, 11 th Edition (ICD-11). 2 , 3 , 4 More than 1 in 5 US adults experience chronic pain, 5 and chronic pain is associated with high socioeconomic and quality-of-life burden 1 , 6 Mechanistic pain descriptors, neuropathic and nociceptive, 7 are used to categorize pain and chronic pain according to suspected or confirmed underlying mechanism(s). Neuropathic pain is defined as being caused by lesions or disease in the somatosensory nervous system, and nociceptive pain, designed to directly contrast neuropathic pain, is defined as pain arising from actual or threatened damage to non-neural tissue. However, pain experienced in the context of many different chronic pain conditions may not fit with these two descriptors; therefore, a proposed third mechanistic pain descriptor, “nociplastic pain,” was added to IASP terminology 7 in 2017. Nociplastic pain is defined as pain arising from altered nociception in the absence of clear lesion/disease of the somatosensory nervous system and/or absence of actual or threatened tissue damage causing activation of peripheral nociceptors, and is associated, compared with nociceptive or neuropathic pain, with greater risk of CNS-related symptoms such as fatigue, changes in cognition and memory, depression, and anxiety. The pain experience in a particular subset of chronic pain conditions, chronic overlapping pain conditions (COPCs), may be better captured by this new mechanistic pain descriptor. 8 COPCs 8 , 9 , 10 are commonly comorbid with one another and more prevalent in people assigned female at birth (AFAB). The US Congress and the National Institutes of Health (NIH) listed 10 conditions as COPCs: myalgic encephalitis/chronic fatigue syndrome (ME/CFS), vulvodynia, temporomandibular disorders, irritable bowel syndrome (IBS), interstitial cystitis/painful bladder syndrome, fibromyalgia, endometriosis, chronic tension-type headache, chronic migraine headache, and chronic low-back pain. In these conditions, there may not be actual or threatened tissue damage or lesion/disease at the somatosensory nervous system (e.g., fibromyalgia). If features of nociceptive/neuropathic pain are present they do not fully capture the pain experience (e.g., chronic low-back pain), and if disease/tissue damage is present, pain is often non-proportional and can be diffuse beyond affected tissue sites (e.g., endometriosis, chronic low-back pain). Previous studies also suggest COPCs could be viewed as a single “lifelong” disease that “manifests in different bodily regions over time.” 11 Taken together, these characteristics suggest nociplastic pain could be an underlying common factor across COPC diagnostic boundaries. In addition, while chronic pain conditions can separately be the subject of genetic study, chronic pain can also be studied as a complex disease trait as evidenced by large recent genome-wide association studies (GWASs) 12 , 13 , 14 , 15 , 16 , 17 where chronic pain traits are agnostic of diagnosis. In this study we have therefore taken a multivariate GWAS approach across COPCs to uncover genetic variation associated with nociplastic pain as an unmeasured latent common factor across COPCs.

The authors declare no competing interests.

K.J.A.J. is supported by 10.13039/100000025 NIMH ( R01MH118278 ; R01MH124839 ). R.S. and L.M.H. are supported by 10.13039/100000025 NIMH ( R01MH118278 ; R01MH124839 ) and 10.13039/100000066 NIEHS ( R01ES033630 ). Data associated with the UK Biobank–approved application number 18177 was used in this study. We thank the study authors who provided their GWAS summary statistics directly to us.

• Individual-level data for the UK Biobank are available upon approved application to the UK Biobank. Code and resources to implement FUSION, GenomicSEM, and ldsc are available at the provided web addresses. • Common-factor GWAS and multivariate TWAS summary statistics generated as part of this study are publicly available for download https://doi.org/10.5281/zenodo.8117582 Accession Number: 8117583. • GWAS summary statistics for broad headache, low-back pain, and neuropathic pain may be available upon request to the respective study authors. • GWAS summary statistics for endometriosis are available from https://www.finngen.fi/en/access_results . • GWAS summary statistics for CWP are available from https://zenodo.org/record/4459546 . • GWAS summary statistics for IBS are available from http://ftp.ebi.ac.uk/pub/databases/gwas/summary_statistics/GCST90044001-GCST90045000/GCST90044107/ . • GWAS summary statistics for TMJ are available from http://ftp.ebi.ac.uk/pub/databases/gwas/summary_statistics/GCST90016001-GCST90017000/GCST90016564/ . Individual-level data for the UK Biobank are available upon approved application to the UK Biobank. Code and resources to implement FUSION, GenomicSEM, and ldsc are available at the provided web addresses. Common-factor GWAS and multivariate TWAS summary statistics generated as part of this study are publicly available for download https://doi.org/10.5281/zenodo.8117582 Accession Number: 8117583. GWAS summary statistics for broad headache, low-back pain, and neuropathic pain may be available upon request to the respective study authors. GWAS summary statistics for endometriosis are available from https://www.finngen.fi/en/access_results . GWAS summary statistics for CWP are available from https://zenodo.org/record/4459546 . GWAS summary statistics for IBS are available from http://ftp.ebi.ac.uk/pub/databases/gwas/summary_statistics/GCST90044001-GCST90045000/GCST90044107/ . GWAS summary statistics for TMJ are available from http://ftp.ebi.ac.uk/pub/databases/gwas/summary_statistics/GCST90016001-GCST90017000/GCST90016564/ .