Browse the corpus

Causal evidence linking chronic pain genetics to late-onset asthma via the nervous system. BACKGROUND: Chronic pain and asthma are associated, but the direction and basis of their genetic and biological relationship remain unclear. METHODS: We conducted genome-wide association studies (GWAS), multi-trait analysis of GWAS (MTAG), polygenic risk score (PRS) prediction, bivariate causal modelling, and Mendelian randomisation (MR) across nine chronic pain traits and three asthma age-of-onset strata (<18, 18-40, and >40 yr for childhood-, adult-, and late-onset asthma, respectively) in 456 958 UK Biobank and 25 275 Canadian Longitudinal Study on Aging participants of European descent. We analysed shared and distinct genetic architecture using gene-, pathway-, tissue-, and cell-type-based enrichment analyses. RESULTS: Multisite chronic pain (MCP) showed the strongest and most consistent genetic overlap with asthma, with genetic correlation increasing from childhood (rg=0.01) to late-onset asthma (rg=0.40). Estimated causal variants for late-onset asthma (∼1.8 K), and fewer for childhood asthma (∼0.2 K), were nested within a broader MCP profile (∼9.4 K). Using PRS, MR, and longitudinal analyses, we found that MCP contributes causally to late-onset asthma. Top causal variants from MR mapped to GMPPB-RNF123, DCC, and FOXP2. Conditioning by MCP amplified late-onset asthma variant effect sizes using MTAG, and uncovered genes enriched for immune and CNS function across pathways, tissues, and cell types. In contrast, childhood asthma showed immune-specific enrichment alone. CONCLUSIONS: These findings reveal neurological function linking chronic pain to late-onset asthma, distinct from childhood asthma, and highlight a CNS contribution to asthma emerging later in life.

This study leveraged data from the UK Biobank, a large-scale population cohort of individuals aged 37–73 yr across the UK.i Participants were recruited voluntarily between 2006 and 2010, with baseline and first follow-up data collected through questionnaires (sociodemographic, lifestyle, and health), physical measurements, and blood samples.26 Chronic pain phenotypes were determined based on pain reports at specific body sites—back, neck and shoulder, hip, knee, stomach and abdomen, headache, facial, and pain over body (widespread). A musculoskeletal pain phenotype was defined based on back, neck and shoulder, hip, and knee pain. MCP, a quantitative trait, reflects the count of body sites with a report of chronic pain. Doctor-diagnosed asthma was identified via self-reported non-cancer illnesses and stratified by age of onset; <18 yr (childhood, N=15 063),18–40 yr (adult, N=13 970), and >40 yr (late, N=19 764) to capture heterogeneity in disease presentation. We drew on UKB-linked hospital admission records (ICD-10 codes J45 for asthma, J40-J44 for chronic obstructive pulmonary disease [COPD]), lung function measures, forced expiratory volume in 1 second (FEV1), and forced vital capacity (FVC), and smoking status in sensitivity analyses (Supplementary Methods). To identify individuals of European ancestry, we applied a standard genetic clustering method, expanding beyond UKB’s ‘White British’ label.27 Participants with sex discrepancies or who opted out of the study were excluded, resulting in 456 958 individuals (Supplementary Methods).

Chronic pain phenotypes were determined based on pain reports at specific body sites—back, neck and shoulder, hip, knee, stomach and abdomen, headache, facial, and pain over body (widespread). A musculoskeletal pain phenotype was defined based on back, neck and shoulder, hip, and knee pain. MCP, a quantitative trait, reflects the count of body sites with a report of chronic pain. Doctor-diagnosed asthma was identified via self-reported non-cancer illnesses and stratified by age of onset; <18 yr (childhood, N=15 063),18–40 yr (adult, N=13 970), and >40 yr (late, N=19 764) to capture heterogeneity in disease presentation. We drew on UKB-linked hospital admission records (ICD-10 codes J45 for asthma, J40-J44 for chronic obstructive pulmonary disease [COPD]), lung function measures, forced expiratory volume in 1 second (FEV1), and forced vital capacity (FVC), and smoking status in sensitivity analyses (Supplementary Methods). To identify individuals of European ancestry, we applied a standard genetic clustering method, expanding beyond UKB’s ‘White British’ label.27 Participants with sex discrepancies or who opted out of the study were excluded, resulting in 456 958 individuals (Supplementary Methods). Phenotypic associations between age-of-onset-stratified asthma and chronic pain traits were assessed using generalised linear models (glm) in R (v4.2.2),ii adjusting for sex and age, within a random subset of UKB participants (80% of cases, 90% of controls) used for polygenic risk score (PRS) training; the remainder were used for PRS testing. We evaluated baseline characteristics across asthma strata and chronic pain categories. To assess robustness of association findings, we conducted sensitivity analyses of varying asthma definitions, sample restrictions (including COPD exclusion), and covariate adjustments (lung function, smoking). Similar association analyses were performed in the Canadian Longitudinal Study on Aging (CLSA; Supplementary Methods)iii.

bustness of association findings, we conducted sensitivity analyses of varying asthma definitions, sample restrictions (including COPD exclusion), and covariate adjustments (lung function, smoking). Similar association analyses were performed in the Canadian Longitudinal Study on Aging (CLSA; Supplementary Methods)iii. The GWA scans were conducted using REGENIE (v3.1.3).28,iv Covariates included age, sex, and the top 40 principal components. Post-GWA scan filtering excluded variants with low minor allele frequency (MAF <0.01), low imputation quality (Info score ≤ 0.3), and Hardy–Weinberg disequilibrium (P<10−6). A total of 9.4 million variants were retained for analysis. Lead SNPs were identified through the online platform for functional mapping and annotation of GWAS (FUMAv), using linkage disequilibrium (LD) thresholds to define independent loci. Functional annotations and gene mappings were performed using ANNOVAR and Ensemble data29 (Supplementary Methods). Heritability and genetic correlations were estimated using LD score regression (LDSC).30,vi Genetic architecture was examined with MiXeR.vii MiXeR modelled discoverability (proportion of variance detectable at genome-wide significance) and polygenicity (number of variants explaining 90% of heritability) under a bivariate model across asthma and pain traits to estimate number of trait-specific or overlapping causal genetic variants31,32 (Supplementary Methods).

eR modelled discoverability (proportion of variance detectable at genome-wide significance) and polygenicity (number of variants explaining 90% of heritability) under a bivariate model across asthma and pain traits to estimate number of trait-specific or overlapping causal genetic variants31,32 (Supplementary Methods). To increase statistical power, we conducted trait-specific meta-analyses with MTAGviii using asthma or chronic pain traits as the primary trait, and traits from the other category as the secondary trait, which accounts for overlapping samples and adjusts for LD structure. MTAG-enhanced summary statistics were then used to refine gene prioritisation33 (Supplementary Methods). Polygenic scores were constructed for MCP and asthma traits using PRSice-2.34,ix Scores were validated in independent UKB samples (20% of cases and 10% of controls who were not included in primary GWAS) and externally in 25 275 participants of European ancestry in the CLSA,35,36 testing cross-phenotype prediction (Supplementary Methods). Mendelian randomisation (MR) using Causal Analysis Using Summary Effect estimates (CAUSE)x was conducted to assess causal and pleiotropic relationships between traits, accounting for LD and sample overlap by incorporating all variants.37 To support the MR findings, longitudinal analyses using logistic regression were performed to examine directional associations between asthma by age of onset and chronic pain across baseline and follow-up data (Supplementary Methods).

between traits, accounting for LD and sample overlap by incorporating all variants.37 To support the MR findings, longitudinal analyses using logistic regression were performed to examine directional associations between asthma by age of onset and chronic pain across baseline and follow-up data (Supplementary Methods). We conducted gene-, gene-set–, and gene-property–level analyses using MAGMA (v1.08)xi on both primary and MTAG-boosted GWAS results.38 Gene-level associations were tested using SNP-wise models to compute gene-based P-values. Tissue and cell-type enrichment analyses were performed using expression data from GTEx (v8) and the Human Protein Atlas,39, 40, 41, 42 and cell-type specificity was assessed via cell-type-specific LDSC (ct-LDSC)xii across multiple transcriptomic datasets.43,44 Additionally, CAUSE-derived expected log pointwise posterior density (ELPD) values were used to identify causal and pleiotropic pathways through SNP-to-gene mapping and enrichment analyses with permutation testing. To enhance the interpretability of our gene-set analysis results, we clustered significant Gene Ontology (GO) terms with a nominal P-value threshold ≤0.05 based on their semantic similarity using the rrvgo R package45,xiii (Supplementary Methods).

gh SNP-to-gene mapping and enrichment analyses with permutation testing. To enhance the interpretability of our gene-set analysis results, we clustered significant Gene Ontology (GO) terms with a nominal P-value threshold ≤0.05 based on their semantic similarity using the rrvgo R package45,xiii (Supplementary Methods). Ethical approval was granted by the National Information Governance Board for Health and Social Care and North West Multicentre Research Ethics Committee for the UKB (11/NW/0382) and CLSA (190213), with institutional approvals from McGill University (A03-M20-21B, A05-M25-19A). This study was conducted under UKB’s application number 20802. All data provided to us were anonymised.

This study leveraged data from the UK Biobank, a large-scale population cohort of individuals aged 37–73 yr across the UK.i Participants were recruited voluntarily between 2006 and 2010, with baseline and first follow-up data collected through questionnaires (sociodemographic, lifestyle, and health), physical measurements, and blood samples.26

Chronic pain phenotypes were determined based on pain reports at specific body sites—back, neck and shoulder, hip, knee, stomach and abdomen, headache, facial, and pain over body (widespread). A musculoskeletal pain phenotype was defined based on back, neck and shoulder, hip, and knee pain. MCP, a quantitative trait, reflects the count of body sites with a report of chronic pain. Doctor-diagnosed asthma was identified via self-reported non-cancer illnesses and stratified by age of onset; <18 yr (childhood, N=15 063),18–40 yr (adult, N=13 970), and >40 yr (late, N=19 764) to capture heterogeneity in disease presentation. We drew on UKB-linked hospital admission records (ICD-10 codes J45 for asthma, J40-J44 for chronic obstructive pulmonary disease [COPD]), lung function measures, forced expiratory volume in 1 second (FEV1), and forced vital capacity (FVC), and smoking status in sensitivity analyses (Supplementary Methods).

To identify individuals of European ancestry, we applied a standard genetic clustering method, expanding beyond UKB’s ‘White British’ label.27 Participants with sex discrepancies or who opted out of the study were excluded, resulting in 456 958 individuals (Supplementary Methods).

Phenotypic associations between age-of-onset-stratified asthma and chronic pain traits were assessed using generalised linear models (glm) in R (v4.2.2),ii adjusting for sex and age, within a random subset of UKB participants (80% of cases, 90% of controls) used for polygenic risk score (PRS) training; the remainder were used for PRS testing. We evaluated baseline characteristics across asthma strata and chronic pain categories. To assess robustness of association findings, we conducted sensitivity analyses of varying asthma definitions, sample restrictions (including COPD exclusion), and covariate adjustments (lung function, smoking). Similar association analyses were performed in the Canadian Longitudinal Study on Aging (CLSA; Supplementary Methods)iii.

The GWA scans were conducted using REGENIE (v3.1.3).28,iv Covariates included age, sex, and the top 40 principal components. Post-GWA scan filtering excluded variants with low minor allele frequency (MAF <0.01), low imputation quality (Info score ≤ 0.3), and Hardy–Weinberg disequilibrium (P<10−6). A total of 9.4 million variants were retained for analysis. Lead SNPs were identified through the online platform for functional mapping and annotation of GWAS (FUMAv), using linkage disequilibrium (LD) thresholds to define independent loci. Functional annotations and gene mappings were performed using ANNOVAR and Ensemble data29 (Supplementary Methods).

Heritability and genetic correlations were estimated using LD score regression (LDSC).30,vi Genetic architecture was examined with MiXeR.vii MiXeR modelled discoverability (proportion of variance detectable at genome-wide significance) and polygenicity (number of variants explaining 90% of heritability) under a bivariate model across asthma and pain traits to estimate number of trait-specific or overlapping causal genetic variants31,32 (Supplementary Methods).

To increase statistical power, we conducted trait-specific meta-analyses with MTAGviii using asthma or chronic pain traits as the primary trait, and traits from the other category as the secondary trait, which accounts for overlapping samples and adjusts for LD structure. MTAG-enhanced summary statistics were then used to refine gene prioritisation33 (Supplementary Methods).

Polygenic scores were constructed for MCP and asthma traits using PRSice-2.34,ix Scores were validated in independent UKB samples (20% of cases and 10% of controls who were not included in primary GWAS) and externally in 25 275 participants of European ancestry in the CLSA,35,36 testing cross-phenotype prediction (Supplementary Methods).

Mendelian randomisation (MR) using Causal Analysis Using Summary Effect estimates (CAUSE)x was conducted to assess causal and pleiotropic relationships between traits, accounting for LD and sample overlap by incorporating all variants.37 To support the MR findings, longitudinal analyses using logistic regression were performed to examine directional associations between asthma by age of onset and chronic pain across baseline and follow-up data (Supplementary Methods).

We conducted gene-, gene-set–, and gene-property–level analyses using MAGMA (v1.08)xi on both primary and MTAG-boosted GWAS results.38 Gene-level associations were tested using SNP-wise models to compute gene-based P-values. Tissue and cell-type enrichment analyses were performed using expression data from GTEx (v8) and the Human Protein Atlas,39, 40, 41, 42 and cell-type specificity was assessed via cell-type-specific LDSC (ct-LDSC)xii across multiple transcriptomic datasets.43,44 Additionally, CAUSE-derived expected log pointwise posterior density (ELPD) values were used to identify causal and pleiotropic pathways through SNP-to-gene mapping and enrichment analyses with permutation testing. To enhance the interpretability of our gene-set analysis results, we clustered significant Gene Ontology (GO) terms with a nominal P-value threshold ≤0.05 based on their semantic similarity using the rrvgo R package45,xiii (Supplementary Methods).

Ethical approval was granted by the National Information Governance Board for Health and Social Care and North West Multicentre Research Ethics Committee for the UKB (11/NW/0382) and CLSA (190213), with institutional approvals from McGill University (A03-M20-21B, A05-M25-19A). This study was conducted under UKB’s application number 20802. All data provided to us were anonymised.

Figure 1 presents a multi-step analytical pipeline to investigate the genetic basis and biological mechanisms underlying chronic pain traits and asthma considering age-of-onset strata. Our integrative framework combines genetic association, risk and causal modelling, and functional exploration to disentangle the shared and distinct genetic mechanisms linking the two sets of traits (Fig. 1). We first assessed the co-occurrence of chronic pain and asthma in 456 958 UKB participants of European descent (Fig. 1a-i). We identified significant associations across 39 pairwise combinations of chronic pain and asthma phenotypes (Fig. 2a and b). Asthma risk was highest for late and lowest for childhood asthma. Among pain traits, widespread pain showed the highest asthma risk (three-fold increase with late asthma). For late asthma, risk increased with number of pain sites, peaking in the top five-to-seven sites stratum (odds ratio [OR] 3.02, 95% confidence interval [CI] 2.65–3.46), with similar patterns for adult and childhood asthma (Fig. 2b). Chronic pain risk estimates increased markedly from childhood to adult asthma and moderately from adult to late asthma, with non-overlapping 95% CIs between childhood and late categories (Supplementary Result 1, Supplementary Tables 1 and 2). Late asthma cases showed distinct characteristics including higher associations with COPD comorbidity, smoking, and reduced lung function compared with earlier-onset asthma (Supplementary Result 1, Supplementary Table 3). Given these associations, we conducted extensive sensitivity analyses (Supplementary Result 2, Supplementary Tables 4 and 5, Supplementary Fig. 1) showing robust association between asthma and chronic pain, including with asthma hospitalisation, and after removing individuals with COPD and adjusting for lung function and smoking. Nonetheless, COPD comorbidity contributed substantially to the particularly strong associations with late asthma.

Supplementary Fig. 1) showing robust association between asthma and chronic pain, including with asthma hospitalisation, and after removing individuals with COPD and adjusting for lung function and smoking. Nonetheless, COPD comorbidity contributed substantially to the particularly strong associations with late asthma. We confirmed these chronic pain-asthma associations in an independent Canadian cohort (CLSA, N=25 261), with similar demographic patterns and association magnitudes across asthma strata (Supplementary Result 3, Supplementary Tables 6 and 7).Fig 1Overview of the study. (a) Data sources used for statistical genetic analyses, including chronic pain traits, asthma stratified by age of onset, and COPD. (i) Phenotypes were collected at baseline and imaging visits. (ii) Genotype and imputed data were used for GWA scans, and significant SNPs from trait-specific GWA scans and meta-analyses were functionally annotated. (b) Genetic pleiotropy between chronic pain and asthma was examined using (i) genetic correlation, (ii and iii) univariate and bivariate MiXeR models, and (iv) cross-trait/cross-population polygenic risk score (PRS) analyses. (c) The causal modelling framework included (i) GW Mendelian randomisation using CAUSE, (ii) longitudinal analysis of late asthma incidence at follow-up based on baseline chronic pain status, and (iii) a dose–response relationship between the number of chronic pain sites and late asthma risk. Phenotypes were collected at baseline and first follow-up visits. (d) Biological interpretation involved (i and ii) gene-set and tissue-specific enrichment analyses, gene-property, and cell-type heritability enrichment.

nd (iii) a dose–response relationship between the number of chronic pain sites and late asthma risk. Phenotypes were collected at baseline and first follow-up visits. (d) Biological interpretation involved (i and ii) gene-set and tissue-specific enrichment analyses, gene-property, and cell-type heritability enrichment. CAUSE, Causal Analysis Using Summary Effect estimates; Child, childhood; CLSA, Canadian Longitudinal Study on Aging; COPD, chronic obstructive pulmonary disease; ct-LDSC, cell-type-specific LD score regression; FDR, false discovery rate; FUMA, Functional Mapping and Annotation; GO, Gene Ontology; GTEx, Genotype-Tissue Expression; GW, genome-wide; GWA, genome-wide association; HPO, Human Phenotype Ontology; KEGG, Kyoto Encyclopedia of Genes and Genomes; Log₂Fold, log base 2 fold change; LDSC, LD score regression; MAGMA, Multi-marker Analysis of GenoMic Annotation; MTAG, Multi-Trait Analysis of GWAS; Num., number; PRS, polygenic risk score; PRSice-2, Polygenic Risk Score software, version 2; REGENIE, tool for regression with genetic data; R glm, R statistical software generalized linear model; SNP, single nucleotide polymorphism.Fig 1Fig 2Phenotypic and genome-wide genetic parameters across chronic pain and asthma traits. (a) Associations between asthma stratified by age of onset (childhood, adult, late) and chronic pain traits, including body site-specific pain and widespread pain. (b) Associations between asthma age-of-onset groups and the number of chronic pain sites. Red vertical lines in panels (a) and (b) indicate no association (odds ratio [OR] of 1). Circles represent estimated ORs, with horizontal black lines indicating 95% CIs. (c) Narrow-sense heritability estimates for childhood asthma (light orange), adult asthma (pink), late asthma (light green), and chronic pain traits (blue). (d) Genetic correlations between asthma (stratified by age of onset) and chronic pain traits. The heatmap shows all pairwise correlations, with red indicating positive and blue indicating negative correlations; darker hues represent stronger effects. Child, childhood; 95% CI, 95% Confidence Interval; MCP, multisite chronic pain.Fig 2

relations between asthma (stratified by age of onset) and chronic pain traits. The heatmap shows all pairwise correlations, with red indicating positive and blue indicating negative correlations; darker hues represent stronger effects. Child, childhood; 95% CI, 95% Confidence Interval; MCP, multisite chronic pain.Fig 2 Overview of the study. (a) Data sources used for statistical genetic analyses, including chronic pain traits, asthma stratified by age of onset, and COPD. (i) Phenotypes were collected at baseline and imaging visits. (ii) Genotype and imputed data were used for GWA scans, and significant SNPs from trait-specific GWA scans and meta-analyses were functionally annotated. (b) Genetic pleiotropy between chronic pain and asthma was examined using (i) genetic correlation, (ii and iii) univariate and bivariate MiXeR models, and (iv) cross-trait/cross-population polygenic risk score (PRS) analyses. (c) The causal modelling framework included (i) GW Mendelian randomisation using CAUSE, (ii) longitudinal analysis of late asthma incidence at follow-up based on baseline chronic pain status, and (iii) a dose–response relationship between the number of chronic pain sites and late asthma risk. Phenotypes were collected at baseline and first follow-up visits. (d) Biological interpretation involved (i and ii) gene-set and tissue-specific enrichment analyses, gene-property, and cell-type heritability enrichment. CAUSE, Causal Analysis Using Summary Effect estimates; Child, childhood; CLSA, Canadian Longitudinal Study on Aging; COPD, chronic obstructive pulmonary disease; ct-LDSC, cell-type-specific LD score regression; FDR, false discovery rate; FUMA, Functional Mapping and Annotation; GO, Gene Ontology; GTEx, Genotype-Tissue Expression; GW, genome-wide; GWA, genome-wide association; HPO, Human Phenotype Ontology; KEGG, Kyoto Encyclopedia of Genes and Genomes; Log₂Fold, log base 2 fold change; LDSC, LD score regression; MAGMA, Multi-marker Analysis of GenoMic Annotation; MTAG, Multi-Trait Analysis of GWAS; Num., number; PRS, polygenic risk score; PRSice-2, Polygenic Risk Score software, version 2; REGENIE, tool for regression with genetic data; R glm, R statistical software generalized linear model; SNP, single nucleotide polymorphism.

AGMA, Multi-marker Analysis of GenoMic Annotation; MTAG, Multi-Trait Analysis of GWAS; Num., number; PRS, polygenic risk score; PRSice-2, Polygenic Risk Score software, version 2; REGENIE, tool for regression with genetic data; R glm, R statistical software generalized linear model; SNP, single nucleotide polymorphism. Phenotypic and genome-wide genetic parameters across chronic pain and asthma traits. (a) Associations between asthma stratified by age of onset (childhood, adult, late) and chronic pain traits, including body site-specific pain and widespread pain. (b) Associations between asthma age-of-onset groups and the number of chronic pain sites. Red vertical lines in panels (a) and (b) indicate no association (odds ratio [OR] of 1). Circles represent estimated ORs, with horizontal black lines indicating 95% CIs. (c) Narrow-sense heritability estimates for childhood asthma (light orange), adult asthma (pink), late asthma (light green), and chronic pain traits (blue). (d) Genetic correlations between asthma (stratified by age of onset) and chronic pain traits. The heatmap shows all pairwise correlations, with red indicating positive and blue indicating negative correlations; darker hues represent stronger effects. Child, childhood; 95% CI, 95% Confidence Interval; MCP, multisite chronic pain.

c correlations between asthma (stratified by age of onset) and chronic pain traits. The heatmap shows all pairwise correlations, with red indicating positive and blue indicating negative correlations; darker hues represent stronger effects. Child, childhood; 95% CI, 95% Confidence Interval; MCP, multisite chronic pain. To characterise the shared and distinct genetic architecture across the nine chronic pain and three asthma phenotypes, we first performed GWA analyses, and identified 179 distinct loci (Fig. 1a-ii, Supplementary Table 8). Five loci overlapped between at least one chronic pain and one asthma phenotype. All five included childhood asthma, one included adult asthma and headache (with nearest gene to lead SNP SDR9C7-TMEM194A), three included headache, one included knee and back pain (HSD17B8-IP6K3) (Supplementary Result 4, Supplementary Fig. 2).

loci overlapped between at least one chronic pain and one asthma phenotype. All five included childhood asthma, one included adult asthma and headache (with nearest gene to lead SNP SDR9C7-TMEM194A), three included headache, one included knee and back pain (HSD17B8-IP6K3) (Supplementary Result 4, Supplementary Fig. 2). For chronic pain, SNP heritability ranged from 2.4% for facial pain to 11% for MCP, a quantitative trait counting the number of body sites with a report of chronic pain, with most musculoskeletal pain traits showing heritability at 6–8.5%. For asthma, heritability decreased with age of onset: 3.5% for childhood, 2.6% for adult, and 1.5% for late asthma (Figs 1b-i and 2c, Supplementary Table 9). Pain traits showed high genetic correlations with MCP (rg=0.81–0.96), demonstrating MCP's effectiveness at capturing features of the genetic architecture of individual pain sites (Fig. 2d, Supplementary Table 10). Adult asthma was more strongly genetically correlated with both late and childhood asthma (rg=0.78 and 0.83, respectively) than childhood and late asthma (rg=0.50). Pain traits showed the strongest genetic correlation with late asthma, particularly with widespread pain (rg=0.50) and MCP (rg=0.45), with correlations decreasing from late to childhood asthma.

ally correlated with both late and childhood asthma (rg=0.78 and 0.83, respectively) than childhood and late asthma (rg=0.50). Pain traits showed the strongest genetic correlation with late asthma, particularly with widespread pain (rg=0.50) and MCP (rg=0.45), with correlations decreasing from late to childhood asthma. We investigated the genetic architecture of chronic pain and asthma traits to estimate polygenicity—the number of causal variants contributing to trait heritability—and discoverability—the average strength of those variant effects—under univariate modelling assumptions (Fig. 1b-ii, iii). Chronic pain traits exhibited high polygenicity and low discoverability, an inverse trend compared with asthma traits (Fig. 3a and b, Supplementary Table 11). Late asthma resembled chronic pain in its polygenicity-to-discoverability ratio, whereas the ratio for childhood asthma reflected larger-effect-size variants.Fig 3Common polygenic architecture underlying chronic pain and asthma traits. (a) Estimated polygenicity (total number of causal variants) across chronic pain and asthma traits. (b) Discoverability estimates, reflecting the average effect sizes among causal variants for each trait. (c) Shared genetic architecture between trait pairs illustrated by genetic correlations, Venn diagrams, and bivariate effect size distributions. Each panel (i–xiv) displays three components for a given trait pair: (1) genetic correlation (ρg): estimated correlation of genetic effects between the two traits from the MiXeR model; (2) Venn diagram: estimated number of causal variants unique to each trait and shared between traits; numbers indicate estimated causal variant counts in thousands (k), with standard errors in parentheses; outer numbers show trait-specific variants; central numbers show shared variants; (3) bivariate effect size heatmaps: estimated bivariate density of causal additive allele substitution effects (β1, β2) inferred by the MiXeR model; colour intensity is scaled by log10(N), where N is the expected number of SNPs per bin. Panels (i–viii) show comparisons between multisite chronic pain (MCP) and site-specific or widespread chronic pain traits; (ix–xi) show comparisons across asthma subtypes; and (xii–xiv) compare MCP and asthma traits. Child, childhood; Num., number; SNP, single nucleotide polymorphism.Fig 3

ected number of SNPs per bin. Panels (i–viii) show comparisons between multisite chronic pain (MCP) and site-specific or widespread chronic pain traits; (ix–xi) show comparisons across asthma subtypes; and (xii–xiv) compare MCP and asthma traits. Child, childhood; Num., number; SNP, single nucleotide polymorphism.Fig 3 Common polygenic architecture underlying chronic pain and asthma traits. (a) Estimated polygenicity (total number of causal variants) across chronic pain and asthma traits. (b) Discoverability estimates, reflecting the average effect sizes among causal variants for each trait. (c) Shared genetic architecture between trait pairs illustrated by genetic correlations, Venn diagrams, and bivariate effect size distributions. Each panel (i–xiv) displays three components for a given trait pair: (1) genetic correlation (ρg): estimated correlation of genetic effects between the two traits from the MiXeR model; (2) Venn diagram: estimated number of causal variants unique to each trait and shared between traits; numbers indicate estimated causal variant counts in thousands (k), with standard errors in parentheses; outer numbers show trait-specific variants; central numbers show shared variants; (3) bivariate effect size heatmaps: estimated bivariate density of causal additive allele substitution effects (β1, β2) inferred by the MiXeR model; colour intensity is scaled by log10(N), where N is the expected number of SNPs per bin. Panels (i–viii) show comparisons between multisite chronic pain (MCP) and site-specific or widespread chronic pain traits; (ix–xi) show comparisons across asthma subtypes; and (xii–xiv) compare MCP and asthma traits. Child, childhood; Num., number; SNP, single nucleotide polymorphism.

e expected number of SNPs per bin. Panels (i–viii) show comparisons between multisite chronic pain (MCP) and site-specific or widespread chronic pain traits; (ix–xi) show comparisons across asthma subtypes; and (xii–xiv) compare MCP and asthma traits. Child, childhood; Num., number; SNP, single nucleotide polymorphism. The number of estimated causal variants ranged from 0.8 to 11.2 K for chronic pain traits. Most musculoskeletal pain and headache variants were contained within the MCP set, highlighting MCP’s broad genetic representation (Fig. 3c-i–viii). Childhood and adult asthma shared 90% of causal variants. Late asthma included all those variants plus 1.4 K trait-specific ones (>70%) (Fig. 3c-ix–xi). Moreover, all causal variants for late asthma (1.8 K) were nested within MCP’s larger set, whereas childhood asthma (0.4 K) shared only 66% of its variants with MCP (Fig. 3c-xii–xiv) under bivariate modelling assumptions. MCP–late asthma comparisons showed highly correlated observed and model-estimated GWA scan effect sizes, indicating shared genetic architecture, whereas MCP–childhood asthma comparisons showed no correlation. We confirmed the goodness of fit of the MiXeR model to our data in univariate and bivariate analyses (Supplementary Table 12).

comparisons showed highly correlated observed and model-estimated GWA scan effect sizes, indicating shared genetic architecture, whereas MCP–childhood asthma comparisons showed no correlation. We confirmed the goodness of fit of the MiXeR model to our data in univariate and bivariate analyses (Supplementary Table 12). We conducted pairwise meta-analyses using MTAG, enhancing statistical power for each chronic pain or asthma trait by conditioning on a trait from the other category. The conditional effect was most notable for late asthma when boosted by MCP, as measured by the difference in mean χ2 values between the boosted and straight GWA scans (delta χ2), whereas the magnitude was lowest for childhood asthma (Supplementary Fig. 3a–c). Mean χ2 values were higher for MCP than for other chronic pain traits. These values did not substantially increase through the inclusion of a secondary asthma trait, although late asthma showed a slight boosting effect on MCP (Supplementary Fig. 3d–f).

agnitude was lowest for childhood asthma (Supplementary Fig. 3a–c). Mean χ2 values were higher for MCP than for other chronic pain traits. These values did not substantially increase through the inclusion of a secondary asthma trait, although late asthma showed a slight boosting effect on MCP (Supplementary Fig. 3d–f). PRS aggregate genetic variants to estimate disease predisposition. Individuals were ranked into percentiles by their genetic risk, and we tested whether PRS for one trait predicted observed disease status for another. After validating UKB PRS in CLSA samples for chronic pain and asthma phenotypes (Figs 1b-iv and 4a and b, Supplementary Table 13), we tested whether MCP-based PRS could predict asthma. In the UKB, asthma risk increased from the 60th percentile of MCP PRS onward, with the strongest effects for late asthma and the weakest for childhood asthma (Fig. 4c). The CLSA showed a similar pattern: late asthma showed significant prediction above the 80th percentile, adult asthma above the 90th percentile, and childhood asthma showed no association (Fig. 4d). In the reverse direction, asthma PRS did not predict chronic pain traits, except for knee pain in the CLSA, where late asthma PRS was associated with increased risk above the 80th percentile (Supplementary Fig. 4a–c).Fig 4Predictive power of polygenic risk scores (PRS), causal modelling, and longitudinal analyses for chronic pain and asthma traits. Results from the UK Biobank (UKB; blue labels) and Canadian Longitudinal Study on Aging (CLSA; red labels). PRS percentiles rank individuals by genetic risk within each cohort, with the middle percentiles (45–55%) serving as the reference group. All error bars represent 95% confidence intervals (95% CI). (a–d) Risk of asthma or chronic pain across PRS percentile bins (<5% to >95%). Horizontal reference lines indicate no association (odds ratio [OR] of 1). (a) Risk of asthma strata in CLSA across asthma PRS percentiles. (b) Risk of chronic migraine, back pain, and knee pain in CLSA across MCP PRS percentiles. (c) Risk of asthma strata in UKB across MCP PRS percentiles. (d) Risk of asthma strata in CLSA across MCP PRS percentiles. (e and f) Bidirectional Mendelian randomisation (MR) using Causal Analysis Using Summary Effect estimates (CAUSE). (e) Causal effects of chronic pain traits on asthma strata. (f) Causal effects of asthma strata on chronic pain traits. Vertical red lines indicate no causal effect (γ=0); circles mark γ estimates. ∗P<0.0001.

and f) Bidirectional Mendelian randomisation (MR) using Causal Analysis Using Summary Effect estimates (CAUSE). (e) Causal effects of chronic pain traits on asthma strata. (f) Causal effects of asthma strata on chronic pain traits. Vertical red lines indicate no causal effect (γ=0); circles mark γ estimates. ∗P<0.0001. (g–j) Longitudinal associations. (g) Risk of late asthma at follow-up among individuals with vs without chronic pain at baseline. (h) Dose–response relationship between the number of chronic pain sites at baseline and risk of late asthma at follow-up. (i) Risk of chronic pain at follow-up among individuals with vs without late asthma at baseline. (j) Dose–response relationship between baseline late asthma and risk of chronic pain at follow-up. Vertical red lines indicate no association (risk ratio [RR] of 1). Child, childhood; γ, causal effect estimate; MCP, multisite chronic pain.Fig 4

ain at follow-up among individuals with vs without late asthma at baseline. (j) Dose–response relationship between baseline late asthma and risk of chronic pain at follow-up. Vertical red lines indicate no association (risk ratio [RR] of 1). Child, childhood; γ, causal effect estimate; MCP, multisite chronic pain.Fig 4 Predictive power of polygenic risk scores (PRS), causal modelling, and longitudinal analyses for chronic pain and asthma traits. Results from the UK Biobank (UKB; blue labels) and Canadian Longitudinal Study on Aging (CLSA; red labels). PRS percentiles rank individuals by genetic risk within each cohort, with the middle percentiles (45–55%) serving as the reference group. All error bars represent 95% confidence intervals (95% CI). (a–d) Risk of asthma or chronic pain across PRS percentile bins (<5% to >95%). Horizontal reference lines indicate no association (odds ratio [OR] of 1). (a) Risk of asthma strata in CLSA across asthma PRS percentiles. (b) Risk of chronic migraine, back pain, and knee pain in CLSA across MCP PRS percentiles. (c) Risk of asthma strata in UKB across MCP PRS percentiles. (d) Risk of asthma strata in CLSA across MCP PRS percentiles. (e and f) Bidirectional Mendelian randomisation (MR) using Causal Analysis Using Summary Effect estimates (CAUSE). (e) Causal effects of chronic pain traits on asthma strata. (f) Causal effects of asthma strata on chronic pain traits. Vertical red lines indicate no causal effect (γ=0); circles mark γ estimates. ∗P<0.0001. (g–j) Longitudinal associations. (g) Risk of late asthma at follow-up among individuals with vs without chronic pain at baseline. (h) Dose–response relationship between the number of chronic pain sites at baseline and risk of late asthma at follow-up. (i) Risk of chronic pain at follow-up among individuals with vs without late asthma at baseline. (j) Dose–response relationship between baseline late asthma and risk of chronic pain at follow-up. Vertical red lines indicate no association (risk ratio [RR] of 1). Child, childhood; γ, causal effect estimate; MCP, multisite chronic pain.

nic pain at follow-up among individuals with vs without late asthma at baseline. (j) Dose–response relationship between baseline late asthma and risk of chronic pain at follow-up. Vertical red lines indicate no association (risk ratio [RR] of 1). Child, childhood; γ, causal effect estimate; MCP, multisite chronic pain. We used bidirectional MR (CAUSE) to assess causal and pleiotropic effects between chronic pain and asthma traits (Fig. 1c-i). After Bonferroni correction, only MCP showed a significant causal effect on late asthma (γ=0.35, P-Bonf.=0.032; Fig. 4e, Supplementary Tables 14 and 15). No significant causal effects were found in the reverse direction (Fig. 4f). This directional relationship was supported by converging evidence from longitudinal analyses in the UKB and CLSA cohorts, reinforcing the MR finding that chronic pain precedes and contributes to late asthma (Fig. 1c-ii, iii). Among individuals initially free of asthma, baseline chronic pain status predicted late asthma development across all body sites, (relative risk [RR] 1.67–3.35) except for chronic widespread pain (RR 2.12, CI 0.77–5.86; Fig. 4g, Supplementary Table 16). This causal relationship was further supported by increasing risk observed with increasing numbers of pain sites at baseline (RR 1.57–3.35; Fig. 4h). In contrast, baseline late asthma showed weaker associations with subsequent chronic pain development, with the highest risks observed for neck/shoulder pain (RR 1.53, CI 1.07–2.19) and hip pain (RR 1.83, CI 1.19–2.82) in the UKB, and for migraine (RR 1.91, CI 1.40–2.60) in the CLSA (Fig. 4i). Baseline asthma did not predict the number of pain sites at follow-up (Fig. 4j), suggesting asymmetry in the relationship.

ment, with the highest risks observed for neck/shoulder pain (RR 1.53, CI 1.07–2.19) and hip pain (RR 1.83, CI 1.19–2.82) in the UKB, and for migraine (RR 1.91, CI 1.40–2.60) in the CLSA (Fig. 4i). Baseline asthma did not predict the number of pain sites at follow-up (Fig. 4j), suggesting asymmetry in the relationship. To biologically interpret the causal relationship between MCP and late asthma, we performed pathway, tissue, and cell-type enrichment analyses (Fig. 1d-i, ii). Gene-based tests (Fig. 5a) revealed that MCP boosting increased the number of significant asthma genes. For late asthma, 41.4% (141 genes) were significant (false discovery rate [FDR] ≤0.1), compared with only 7.7% (29 genes) for adult and 3.5% (39 genes) for childhood asthma, attributable to boosting by MCP (Fig. 5b-i–iii, Supplementary Table 17).Fig 5Functional characterisation of shared genetic architecture between chronic pain and asthma. (a) Input files and analysis workflow. Summary statistics from multiple complementary approaches: gene-based analyses aggregate single-nucleotide polymorphism (SNP) signals into gene-level associations; multi-trait analysis of GWAS (MTAG) meta-analyses leverage genetic correlations between related traits to increase statistical power; Causal Analysis Using Summary Effect estimates (CAUSE) distinguishes directional causal effects from shared pleiotropic effects across chronic pain and asthma trait pairs. Pathway analyses were performed using Human Phenotype Ontology (HPO), Reactomexiv, Gene Ontology (GO), and KEGG databases. (b) Multi-marker Analysis of Genomic Annotation (MAGMA) gene-based results comparing trait-specific GWA scans with MTAG summary statistics of asthma age-of-onset strata boosted by multisite chronic pain (MCP; i.e. MTAG analysis combining each asthma stratum with MCP). Quadrant plots display Z-statistics, with dashed vertical and horizontal lines indicating nominal significance thresholds (P=0.05). Panels: (i) late asthma, (ii) adult asthma, (iii) childhood asthma. (c) MAGMA gene-set enrichment analysis using MTAG results for childhood and late asthma, both boosted by MCP. Volcano plot shows log2 fold change in gene-set effect sizes (x-axis) vs –log10 false discovery rate (FDR)-adjusted P-values (y-axis); the horizontal line marks the FDR threshold (FDR<0.1). (d) MAGMA gene-property analysis of tissue enrichment using Genotype-Tissue Expression (GTEx) v8 and MTAG summary statistics for childhood and late asthma (boosted by MCP).

e-set effect sizes (x-axis) vs –log10 false discovery rate (FDR)-adjusted P-values (y-axis); the horizontal line marks the FDR threshold (FDR<0.1). (d) MAGMA gene-property analysis of tissue enrichment using Genotype-Tissue Expression (GTEx) v8 and MTAG summary statistics for childhood and late asthma (boosted by MCP). Quadrant plot shows Z-statistics, with dashed lines marking nominal significance thresholds (P=0.05). Ant., anterior; BA9, Brodmann Area 9; Child, childhood; EBV, Epstein-Barr Virus; GOBP, Gene Ontology Biological Process; GWAS, genome-wide association study; KEGG, Kyoto Encyclopedia of Genes and Genomes; No sig., no significant genes.Fig 5

statistics, with dashed lines marking nominal significance thresholds (P=0.05). Ant., anterior; BA9, Brodmann Area 9; Child, childhood; EBV, Epstein-Barr Virus; GOBP, Gene Ontology Biological Process; GWAS, genome-wide association study; KEGG, Kyoto Encyclopedia of Genes and Genomes; No sig., no significant genes.Fig 5 Functional characterisation of shared genetic architecture between chronic pain and asthma. (a) Input files and analysis workflow. Summary statistics from multiple complementary approaches: gene-based analyses aggregate single-nucleotide polymorphism (SNP) signals into gene-level associations; multi-trait analysis of GWAS (MTAG) meta-analyses leverage genetic correlations between related traits to increase statistical power; Causal Analysis Using Summary Effect estimates (CAUSE) distinguishes directional causal effects from shared pleiotropic effects across chronic pain and asthma trait pairs. Pathway analyses were performed using Human Phenotype Ontology (HPO), Reactomexiv, Gene Ontology (GO), and KEGG databases. (b) Multi-marker Analysis of Genomic Annotation (MAGMA) gene-based results comparing trait-specific GWA scans with MTAG summary statistics of asthma age-of-onset strata boosted by multisite chronic pain (MCP; i.e. MTAG analysis combining each asthma stratum with MCP). Quadrant plots display Z-statistics, with dashed vertical and horizontal lines indicating nominal significance thresholds (P=0.05). Panels: (i) late asthma, (ii) adult asthma, (iii) childhood asthma. (c) MAGMA gene-set enrichment analysis using MTAG results for childhood and late asthma, both boosted by MCP. Volcano plot shows log2 fold change in gene-set effect sizes (x-axis) vs –log10 false discovery rate (FDR)-adjusted P-values (y-axis); the horizontal line marks the FDR threshold (FDR<0.1). (d) MAGMA gene-property analysis of tissue enrichment using Genotype-Tissue Expression (GTEx) v8 and MTAG summary statistics for childhood and late asthma (boosted by MCP). Quadrant plot shows Z-statistics, with dashed lines marking nominal significance thresholds (P=0.05). Ant., anterior; BA9, Brodmann Area 9; Child, childhood; EBV, Epstein-Barr Virus; GOBP, Gene Ontology Biological Process; GWAS, genome-wide association study; KEGG, Kyoto Encyclopedia of Genes and Genomes; No sig., no significant genes.

ws Z-statistics, with dashed lines marking nominal significance thresholds (P=0.05). Ant., anterior; BA9, Brodmann Area 9; Child, childhood; EBV, Epstein-Barr Virus; GOBP, Gene Ontology Biological Process; GWAS, genome-wide association study; KEGG, Kyoto Encyclopedia of Genes and Genomes; No sig., no significant genes. Pathway enrichment analyses of GWA summary statistics identified 43 MCP-associated pathways (mostly neurological) and 27 shared asthma pathways (FDR ≤0.1; Supplementary Table 18). Shared asthma pathways were primarily immunological, including interleukin-1 receptor activity, the KEGGxv asthma pathway, and MHC class II function. Notably, the only neurological pathway—learned vocalisation behaviour—was shared between MCP and late asthma. To identify functional differences by asthma age of onset, we compared pathway enrichment beta values between MCP-boosted late and childhood asthma. Among 113 significantly enriched pathways, 99 were immunological and more strongly enriched in childhood asthma. In contrast, late asthma showed enrichment for two neurological pathways—vocalisation behaviour and vocal learning—and three additional pathways (Fig. 5c, Supplementary Table 19).

late and childhood asthma. Among 113 significantly enriched pathways, 99 were immunological and more strongly enriched in childhood asthma. In contrast, late asthma showed enrichment for two neurological pathways—vocalisation behaviour and vocal learning—and three additional pathways (Fig. 5c, Supplementary Table 19). We next used tag SNPs from CAUSE (delta ELPD <0, N=2743) to identify pathways enriched for causal effects of MCP on late asthma (Supplementary Tables 20 and 21). Gene Ontologyxvi and Human Phenotype Ontologyxvii analyses revealed 18 (four parent-node level clusters) and six significant pathways, respectively, all involving nervous system or musculoskeletal function (Supplementary Fig. 5a, Supplementary Tables 22 and 23). The top 12 causal genes (based on delta ELPD), including DCC, GMPPB-AMIGO3-RNF123, and FOXP2, were consistently represented across the enriched neurological pathways. Several of these genes were also highly boosted by MCP in the meta-analysis (Supplementary Fig. 5b). In contrast, no significant pathways were identified using pleiotropic variants (delta ELPD >0, N=1972).

ding DCC, GMPPB-AMIGO3-RNF123, and FOXP2, were consistently represented across the enriched neurological pathways. Several of these genes were also highly boosted by MCP in the meta-analysis (Supplementary Fig. 5b). In contrast, no significant pathways were identified using pleiotropic variants (delta ELPD >0, N=1972). Tissue enrichment analyses using GTEx highlighted 13 neural tissues including cerebellum, frontal cortex, and basal ganglia for both late asthma and MCP, whereas childhood and adult asthma were enriched in immune tissues such as whole blood, Epstein–Barr virus-transformed lymphocytes, and small intestine (Supplementary Tables 24 and 25). Comparing late and childhood asthma boosted by MCP, immune tissues were highly enriched in childhood asthma, whereas brain regions were significantly enriched for late asthma (FDR ≤0.1; Fig. 5d, Supplementary Table 25). MCP-boosted late asthma additionally showed enrichment in cerebellar substructures—cortex, vermis, and flocculonodular lobe—based on single-cell RNA data (Human Protein Atlas) (Fig. 6a-i, ii), whereas other asthma strata showed no such enrichment (Fig. 6a-iii–vi). Cell types enriched in late asthma through MCP boosting included microglia, astrocytes, excitatory and inhibitory neurones, and oligodendrocytes, with only microglia also enriched in MCP-boosted adult asthma (Supplementary Table 26).Fig 6Differential enrichment of immune and central nervous system signals in childhood and late asthma. (a) Brain tissue and cell-type enrichment using data from the Human Protein Atlas and both trait-specific genome-wide association (GWA) scans and multi-trait analysis of GWAS (MTAG) summary statistics of asthma age-of-onset strata boosted by multisite chronic pain (MCP; i.e. MTAG analysis combining each asthma stratum with MCP). Panels show results for asthma traits, with the top row showing tissue enrichment (red dots) and the bottom row (blue dots) showing cell-type enrichment. (i and ii) Late asthma. (iii and iv) Adult asthma. (v and vi) Childhood asthma. (b) Cell-type-specific partitioned heritability enrichment based on multi-tissue gene expression data using MTAG summary statistics for childhood and late asthma (both boosted by MCP). Quadrant plots show Z-statistics, with dashed vertical and horizontal lines indicating nominal significance thresholds (P=0.05). (c) Cell-type-specific heritability enrichment using the Gene Enrichment Profiler dataset.

e expression data using MTAG summary statistics for childhood and late asthma (both boosted by MCP). Quadrant plots show Z-statistics, with dashed vertical and horizontal lines indicating nominal significance thresholds (P=0.05). (c) Cell-type-specific heritability enrichment using the Gene Enrichment Profiler dataset. Pyramid plots compare trait-specific GWA scans (left) vs MTAG results (right). (i) Late asthma (boosted by MCP). (ii) MCP (boosted by late asthma). Child, childhood; CNS, central nervous system; GWAS, genome-wide association study; Lym. T, T lymphocytes; Lym. B, B lymphocytes; PNS, peripheral nervous system.Fig 6

e expression data using MTAG summary statistics for childhood and late asthma (both boosted by MCP). Quadrant plots show Z-statistics, with dashed vertical and horizontal lines indicating nominal significance thresholds (P=0.05). (c) Cell-type-specific heritability enrichment using the Gene Enrichment Profiler dataset. Pyramid plots compare trait-specific GWA scans (left) vs MTAG results (right). (i) Late asthma (boosted by MCP). (ii) MCP (boosted by late asthma). Child, childhood; CNS, central nervous system; GWAS, genome-wide association study; Lym. T, T lymphocytes; Lym. B, B lymphocytes; PNS, peripheral nervous system.Fig 6 Differential enrichment of immune and central nervous system signals in childhood and late asthma. (a) Brain tissue and cell-type enrichment using data from the Human Protein Atlas and both trait-specific genome-wide association (GWA) scans and multi-trait analysis of GWAS (MTAG) summary statistics of asthma age-of-onset strata boosted by multisite chronic pain (MCP; i.e. MTAG analysis combining each asthma stratum with MCP). Panels show results for asthma traits, with the top row showing tissue enrichment (red dots) and the bottom row (blue dots) showing cell-type enrichment. (i and ii) Late asthma. (iii and iv) Adult asthma. (v and vi) Childhood asthma. (b) Cell-type-specific partitioned heritability enrichment based on multi-tissue gene expression data using MTAG summary statistics for childhood and late asthma (both boosted by MCP). Quadrant plots show Z-statistics, with dashed vertical and horizontal lines indicating nominal significance thresholds (P=0.05). (c) Cell-type-specific heritability enrichment using the Gene Enrichment Profiler dataset. Pyramid plots compare trait-specific GWA scans (left) vs MTAG results (right). (i) Late asthma (boosted by MCP). (ii) MCP (boosted by late asthma). Child, childhood; CNS, central nervous system; GWAS, genome-wide association study; Lym. T, T lymphocytes; Lym. B, B lymphocytes; PNS, peripheral nervous system.

r dataset. Pyramid plots compare trait-specific GWA scans (left) vs MTAG results (right). (i) Late asthma (boosted by MCP). (ii) MCP (boosted by late asthma). Child, childhood; CNS, central nervous system; GWAS, genome-wide association study; Lym. T, T lymphocytes; Lym. B, B lymphocytes; PNS, peripheral nervous system. Cell-type heritability analysis using multiple-tissue analysis of gene expression revealed enrichment for both late and childhood asthma in broad lymphocytes and T lymphocytes (FDR ≤0.10; Fig. 6b, solid red dots), with lymphocyte-null cell types enriched only in childhood asthma (red ring). Late asthma showed unique enrichment in monocytes, phagocytes, the mononuclear-phagocyte system, and bone marrow cells (Fig. 6b, purple rings), and in CNS tissues, namely cerebellum, metencephalon, cerebellar hemisphere, limbic system, and brain stem cells (Fig. 6b, blue triangles, Supplementary Table 27). Gene Enrichment Profiler confirmed heritability enrichment for MCP-boosted late asthma in T and B lymphocytes, CNS regions (cerebellum, hypothalamus, olfactory bulb), and respiratory tissues (Fig. 6c-i). In contrast, MCP alone showed enrichment in 20 CNS tissues, but no additional enrichment when boosted by late asthma, reinforcing the asymmetry in causal direction (Fig. 6c-ii, Supplementary Table 28).

To characterise the shared and distinct genetic architecture across the nine chronic pain and three asthma phenotypes, we first performed GWA analyses, and identified 179 distinct loci (Fig. 1a-ii, Supplementary Table 8). Five loci overlapped between at least one chronic pain and one asthma phenotype. All five included childhood asthma, one included adult asthma and headache (with nearest gene to lead SNP SDR9C7-TMEM194A), three included headache, one included knee and back pain (HSD17B8-IP6K3) (Supplementary Result 4, Supplementary Fig. 2).

For chronic pain, SNP heritability ranged from 2.4% for facial pain to 11% for MCP, a quantitative trait counting the number of body sites with a report of chronic pain, with most musculoskeletal pain traits showing heritability at 6–8.5%. For asthma, heritability decreased with age of onset: 3.5% for childhood, 2.6% for adult, and 1.5% for late asthma (Figs 1b-i and 2c, Supplementary Table 9). Pain traits showed high genetic correlations with MCP (rg=0.81–0.96), demonstrating MCP's effectiveness at capturing features of the genetic architecture of individual pain sites (Fig. 2d, Supplementary Table 10). Adult asthma was more strongly genetically correlated with both late and childhood asthma (rg=0.78 and 0.83, respectively) than childhood and late asthma (rg=0.50). Pain traits showed the strongest genetic correlation with late asthma, particularly with widespread pain (rg=0.50) and MCP (rg=0.45), with correlations decreasing from late to childhood asthma.

We conducted pairwise meta-analyses using MTAG, enhancing statistical power for each chronic pain or asthma trait by conditioning on a trait from the other category. The conditional effect was most notable for late asthma when boosted by MCP, as measured by the difference in mean χ2 values between the boosted and straight GWA scans (delta χ2), whereas the magnitude was lowest for childhood asthma (Supplementary Fig. 3a–c). Mean χ2 values were higher for MCP than for other chronic pain traits. These values did not substantially increase through the inclusion of a secondary asthma trait, although late asthma showed a slight boosting effect on MCP (Supplementary Fig. 3d–f).

PRS aggregate genetic variants to estimate disease predisposition. Individuals were ranked into percentiles by their genetic risk, and we tested whether PRS for one trait predicted observed disease status for another. After validating UKB PRS in CLSA samples for chronic pain and asthma phenotypes (Figs 1b-iv and 4a and b, Supplementary Table 13), we tested whether MCP-based PRS could predict asthma. In the UKB, asthma risk increased from the 60th percentile of MCP PRS onward, with the strongest effects for late asthma and the weakest for childhood asthma (Fig. 4c). The CLSA showed a similar pattern: late asthma showed significant prediction above the 80th percentile, adult asthma above the 90th percentile, and childhood asthma showed no association (Fig. 4d). In the reverse direction, asthma PRS did not predict chronic pain traits, except for knee pain in the CLSA, where late asthma PRS was associated with increased risk above the 80th percentile (Supplementary Fig. 4a–c).Fig 4Predictive power of polygenic risk scores (PRS), causal modelling, and longitudinal analyses for chronic pain and asthma traits. Results from the UK Biobank (UKB; blue labels) and Canadian Longitudinal Study on Aging (CLSA; red labels). PRS percentiles rank individuals by genetic risk within each cohort, with the middle percentiles (45–55%) serving as the reference group. All error bars represent 95% confidence intervals (95% CI). (a–d) Risk of asthma or chronic pain across PRS percentile bins (<5% to >95%). Horizontal reference lines indicate no association (odds ratio [OR] of 1). (a) Risk of asthma strata in CLSA across asthma PRS percentiles. (b) Risk of chronic migraine, back pain, and knee pain in CLSA across MCP PRS percentiles. (c) Risk of asthma strata in UKB across MCP PRS percentiles. (d) Risk of asthma strata in CLSA across MCP PRS percentiles. (e and f) Bidirectional Mendelian randomisation (MR) using Causal Analysis Using Summary Effect estimates (CAUSE). (e) Causal effects of chronic pain traits on asthma strata. (f) Causal effects of asthma strata on chronic pain traits. Vertical red lines indicate no causal effect (γ=0); circles mark γ estimates. ∗P<0.0001.

We used bidirectional MR (CAUSE) to assess causal and pleiotropic effects between chronic pain and asthma traits (Fig. 1c-i). After Bonferroni correction, only MCP showed a significant causal effect on late asthma (γ=0.35, P-Bonf.=0.032; Fig. 4e, Supplementary Tables 14 and 15). No significant causal effects were found in the reverse direction (Fig. 4f). This directional relationship was supported by converging evidence from longitudinal analyses in the UKB and CLSA cohorts, reinforcing the MR finding that chronic pain precedes and contributes to late asthma (Fig. 1c-ii, iii). Among individuals initially free of asthma, baseline chronic pain status predicted late asthma development across all body sites, (relative risk [RR] 1.67–3.35) except for chronic widespread pain (RR 2.12, CI 0.77–5.86; Fig. 4g, Supplementary Table 16). This causal relationship was further supported by increasing risk observed with increasing numbers of pain sites at baseline (RR 1.57–3.35; Fig. 4h). In contrast, baseline late asthma showed weaker associations with subsequent chronic pain development, with the highest risks observed for neck/shoulder pain (RR 1.53, CI 1.07–2.19) and hip pain (RR 1.83, CI 1.19–2.82) in the UKB, and for migraine (RR 1.91, CI 1.40–2.60) in the CLSA (Fig. 4i). Baseline asthma did not predict the number of pain sites at follow-up (Fig. 4j), suggesting asymmetry in the relationship.

To biologically interpret the causal relationship between MCP and late asthma, we performed pathway, tissue, and cell-type enrichment analyses (Fig. 1d-i, ii). Gene-based tests (Fig. 5a) revealed that MCP boosting increased the number of significant asthma genes. For late asthma, 41.4% (141 genes) were significant (false discovery rate [FDR] ≤0.1), compared with only 7.7% (29 genes) for adult and 3.5% (39 genes) for childhood asthma, attributable to boosting by MCP (Fig. 5b-i–iii, Supplementary Table 17).Fig 5Functional characterisation of shared genetic architecture between chronic pain and asthma. (a) Input files and analysis workflow. Summary statistics from multiple complementary approaches: gene-based analyses aggregate single-nucleotide polymorphism (SNP) signals into gene-level associations; multi-trait analysis of GWAS (MTAG) meta-analyses leverage genetic correlations between related traits to increase statistical power; Causal Analysis Using Summary Effect estimates (CAUSE) distinguishes directional causal effects from shared pleiotropic effects across chronic pain and asthma trait pairs. Pathway analyses were performed using Human Phenotype Ontology (HPO), Reactomexiv, Gene Ontology (GO), and KEGG databases. (b) Multi-marker Analysis of Genomic Annotation (MAGMA) gene-based results comparing trait-specific GWA scans with MTAG summary statistics of asthma age-of-onset strata boosted by multisite chronic pain (MCP; i.e. MTAG analysis combining each asthma stratum with MCP). Quadrant plots display Z-statistics, with dashed vertical and horizontal lines indicating nominal significance thresholds (P=0.05). Panels: (i) late asthma, (ii) adult asthma, (iii) childhood asthma. (c) MAGMA gene-set enrichment analysis using MTAG results for childhood and late asthma, both boosted by MCP. Volcano plot shows log2 fold change in gene-set effect sizes (x-axis) vs –log10 false discovery rate (FDR)-adjusted P-values (y-axis); the horizontal line marks the FDR threshold (FDR<0.1). (d) MAGMA gene-property analysis of tissue enrichment using Genotype-Tissue Expression (GTEx) v8 and MTAG summary statistics for childhood and late asthma (boosted by MCP).

We provide the first comprehensive genetic dissection of the link between chronic pain and asthma, revealing a complex, age-dependent relationship driven primarily by shared architecture between MCP and late asthma. Multiple lines of evidence—including PRS in two large cohorts (UKB and CLSA), longitudinal prediction of asthma onset by baseline chronic pain, and MR—consistently supported a directional causal effect from MCP to late asthma. The substantial boosting effect through meta-analysis of late asthma by MCP highlights a shared polygenic component, characterised by increased late asthma effect sizes across the genome, as supported by bivariate causal mixture models. Pathway and cell-type heritability analyses indicated immunological function for childhood asthma while revealing both immunological and substantial neurological contributions for late asthma. Together, these findings suggest a biologically coherent and directionally causal relationship between chronic pain and late asthma, underpinned by shared neurological and immunological mechanisms that emerge in mid-to-late life.

e revealing both immunological and substantial neurological contributions for late asthma. Together, these findings suggest a biologically coherent and directionally causal relationship between chronic pain and late asthma, underpinned by shared neurological and immunological mechanisms that emerge in mid-to-late life. This work expands upon previous reports of a more modest genetic correlation (∼20%) between MCP and asthma as a broad phenotype17,24,25 by demonstrating a substantially higher genetic correlation between MCP and late asthma (rg=0.45), with correlations decreasing progressively from late to childhood asthma. These findings align with an earlier study showing genetic correlations between adult asthma and insomnia or depressive symptoms—traits that commonly co-occur with chronic pain—whereas no such correlation was observed for childhood asthma.46 Although a prior MR study suggested a horizontally pleiotropic relationship between MCP and asthma as a broad phenotype,25 our analyses provide converging evidence for a directional causal effect of chronic pain on late asthma (γ=0.35).

with chronic pain—whereas no such correlation was observed for childhood asthma.46 Although a prior MR study suggested a horizontally pleiotropic relationship between MCP and asthma as a broad phenotype,25 our analyses provide converging evidence for a directional causal effect of chronic pain on late asthma (γ=0.35). Previous UKB GWAS have independently examined asthma stratified by age of onset or MCP for chronic pain, but this is the first study to jointly analyse these traits while incorporating internal heterogeneity on both sides. A recent epidemiological survey stratified asthma by the same age of onset cut-offs used here.9 This study found strong differences in comorbidities, asthma control, and demographic features across the three strata, reinforcing the clinical and epidemiological relevance of our stratification scheme. Earlier studies concluded that childhood asthma has a distinct genetic profile, whereas adult asthma shares many of these signals but adds little unique signal of its own.46 Our findings extend this model by showing that late asthma not only shares genetic factors with earlier-onset forms but also carries additional trait-specific variation.

ed that childhood asthma has a distinct genetic profile, whereas adult asthma shares many of these signals but adds little unique signal of its own.46 Our findings extend this model by showing that late asthma not only shares genetic factors with earlier-onset forms but also carries additional trait-specific variation. Consistent with prior GWAS identifying lymphocyte-driven immune involvement in asthma,20,47,48 we observed strong enrichment for lymphocyte-specific genetic signal in both childhood and late asthma when boosted by MCP, suggesting a shared polygenic immune component. Childhood asthma showed stronger enrichment across immunological pathways and immune tissues, aligning with previous findings of its genetically driven, allergy-related architecture.19,49 In contrast, late asthma genetic signals uniquely demonstrated enrichment in neurological pathways and CNS tissues including the cerebellum, alongside broader leucocyte categories such as phagocytes and monocytes. This combined neurological and immunological involvement has not been prominent in prior asthma GWAS and highlights CNS involvement as a novel component of late asthma and its overlap with MCP, consistent with genetic studies of MCP showing enrichment in brain-specific tissues and structural features.50,51

d monocytes. This combined neurological and immunological involvement has not been prominent in prior asthma GWAS and highlights CNS involvement as a novel component of late asthma and its overlap with MCP, consistent with genetic studies of MCP showing enrichment in brain-specific tissues and structural features.50,51 Our sensitivity analyses revealed that COPD comorbidity substantially contributed to the particularly strong chronic pain associations with late asthma. Effect estimates were nearly unchanged after adjusting for lung function and smoking, indicating a low-level residual association. This pattern is consistent with a ‘pain–breathlessness cluster’—chronic pain co-occurring with greater reported breathlessness symptom burden beyond airflow limitation52,53—and is plausibly explained by the CNS pathways we identified for late asthma; moreover, experimental breathlessness engages an established pain neuromarker and is opioid-modulated.54 Although individuals with chronic pain have higher smoking prevalence and worse pain trajectories—and smoking precipitates COPD55—the pain–late asthma association in our data was essentially unchanged after adjustment for smoking and lung function, indicating that smoking does not account for this relationship. Further genetic and functional studies are needed to better characterise the phenotypic features associated with the CNS component of late asthma. These findings are consistent with a dose–response pattern in genetic architecture across asthma strata, and with a biologically grounded dimorphism: functional analyses revealed exclusively immunological pathways for childhood asthma and both immunological and neurological pathways for late asthma, whereas no clear enrichment emerged for intermediate adult asthma. The lack of pathway enrichment for intermediate-onset adult asthma may reflect a dilution effect, combining two distinct variant sets—some shared with childhood asthma, others with late asthma—without convergent biological evidence. In this context, the shared genetic risk may represent ‘poor aging” variation, consistent with the increasing prevalence and frequent co-occurrence of asthma and chronic pain in later life.56,57 More speculatively, our findings may reflect a common inflammaging-related latent phenotype—an underlying low-grade inflammatory state that manifests differently across individuals as either chronic pain, late asthma, or both.56,57

ing prevalence and frequent co-occurrence of asthma and chronic pain in later life.56,57 More speculatively, our findings may reflect a common inflammaging-related latent phenotype—an underlying low-grade inflammatory state that manifests differently across individuals as either chronic pain, late asthma, or both.56,57 Our study has some limitations. Firstly, as analyses focused on participants of European ancestry, results may not generalise to other populations. Secondly, UKB's nonspecific self-reported chronic pain classification fails to distinguish between nociceptive, neuropathic, and nociplastic mechanisms, constraining our causal interpretations, particularly across different body sites.58 This heterogeneity in pain mechanisms may explain the weak or non-causal effects observed for some of the body site-specific pain traits in our longitudinal analysis. Thirdly, age-of-onset stratification for asthma relied on self-reported diagnosis dates, which may introduce recall bias. A prior UKB study of childhood and adult-onset asthma demonstrated robust findings despite potential recall bias, and successfully replicated nearly all previously reported asthma signals.49 Moreover, our sensitivity analyses using concordant self-report and ICD-10 hospital codes yielded consistent associations. Despite these limitations, the multi-cohort design, integration of longitudinal and genetic data, and convergence across analytic approaches lend robustness to our findings.

orted asthma signals.49 Moreover, our sensitivity analyses using concordant self-report and ICD-10 hospital codes yielded consistent associations. Despite these limitations, the multi-cohort design, integration of longitudinal and genetic data, and convergence across analytic approaches lend robustness to our findings. In summary, our comprehensive analyses reveal a complex relationship between chronic pain and asthma that varies substantially by phenotypic subtype, with the strongest and most causally relevant link between MCP and late asthma. The complete overlap of causal variants, consistent directional evidence, and shared neurological pathways suggest that chronic pain may contribute to late asthma development via central neurological mechanisms, potentially involving neuroinflammatory processes that bridge cerebellar and immune systems. Our results may offer promising targets for interventions aimed not only at symptom management but at the prevention of both chronic pain and late asthma.

Conceived the study: AVG, LD Designed the analysis plan: GK Oversaw the analysis plan: AVG Advised on the analysis plan: LD Provided input and advice on statistical analysis using eQTL data and helped in creating a mathematical formal for causal pathway analysis: MP Performed GWA scans and bioinformatics analyses and generated results displays: GK Conducted sensitivity analyses: AVG, GK Interpretation of results: AVG, GK, MP, LD Participated in interpretation of brain and neurological data findings: MF Drafted and revised the manuscript: AVG, GK Participated in UKB data preparation and processing: ND, CC, and YC Participated in the CLSA phenotype classification and QC of CLSA genotyping data: ND, MZ, LH, PH, and SB Funding acquisition: LD, AVG Critically reviewed and approved the final version of the manuscript: all authors

The Louise and Alan Edwards Foundation c/o, The Jewish Community Foundation Montreal and Quebec Pain Research Network (QPRN) (to AVG). The Canadian Excellence Research Chair for Human Pain Genetics (CERC09), and NIH grant U54 DA049110 (held by LD). The Catherine Bushnell Pain Research Postdoctoral Fellowship, the International Association for the Study of Pain, and the Quebec Network of Junior Pain Investigators (to GK). This research was enabled in part by support provided by Calcul Québec (https://www.calculquebec.ca/) and the Digital Research Alliance of Canada (alliancecan.ca).

The authors declare no competing interests. SB is an employee of 5 Prime Sciences (www.5primesciences.com), which provides research services for biotech, pharma, and venture capital companies for projects unrelated to this research.