Genetic underpinning of the comorbidity between type 2 diabetes and osteoarthritis

Abstract

Multimorbidity is a rising public health challenge with important implications for health management and policy. The most common multimorbidity pattern is the combination of cardiometabolic and osteoarticular diseases. Here, we study the genetic underpinning of the comorbidity between type 2 diabetes and osteoarthritis. We find genome-wide genetic correlation between the two diseases and robust evidence for association-signal colocalization at 18 genomic regions. We integrate multi-omics and functional information to resolve the colocalizing signals and identify high-confidence effector genes, including FTO and IRX3, which provide proof-of-concept insights into the epidemiologic link between obesity and both diseases. We find enrichment for lipid metabolism and skeletal formation pathways for signals underpinning the knee and hip osteoarthritis comorbidities with type 2 diabetes, respectively. Causal inference analysis identifies complex effects of tissue-specific gene expression on comorbidity outcomes. Our findings provide insights into the biological basis for the type 2 diabetes-osteoarthritis disease co-occurrence.

Keywords: Mendelian randomization; colocalization analysis; multimorbidity; osteoarthritis; statistical genetics; type 2 diabetes.

Figure 1

Stronger evidence for a genetic correlation between type 2 diabetes and knee osteoarthritis than between type 2 diabetes and hip osteoarthritis (A) Genetic correlation (r_g) results between type 2 diabetes (T2D) and knee or hip osteoarthritis (OA). The error bars represent the standard error of the estimated genetic correlation. (B) Permutation-based testing results for knee OA and hip OA, respectively. The red line is the actual correlation.

Figure 2

Overview of colocalizing regions between type 2 diabetes and osteoarthritis The y axis depicts the posterior probability of a shared causal variant (PP4) and the x axis the number of variants in the 95% credible set for the causal variant. Each point represents a colocalized signal between type 2 diabetes and one osteoarthritis (OA) phenotype. Point size is proportional to the number of variants in the colocalization analysis 95% credible set. We find strong statistical evidence for colocalization (PP4 > 0.8) for 51 signals. Some of those 51 signals colocalize between type 2 diabetes and more than one OA phenotype, resulting in 18 unique colocalizing genomic loci.

Figure 3

Study design to derive a list of high-confidence effector genes for the type 2 diabetes and osteoarthritis comorbidity In each of the unique 18 genomic loci that colocalized between type 2 diabetes and osteoarthritis with a posterior probability of a single shared causal variant (PP4) ≥ 0.8, we explored all genes in a 1-Mb window on either side of the lead variants of the 95% credible set for the causal variant of the colocalization analysis. For each gene, we searched databases for knockout mice and rare and syndromic human diseases for pre-defined type 2 diabetes- and musculoskeletal-related phenotypes. We also examined differentially expressed genes (DEGs) in pancreatic islets of healthy versus diabetic individuals and of degraded versus intact osteoarthritis cartilage. We also assessed whether the genes were already previously defined as established effector genes for the individual diseases. We examined all variants in the 95% credible set for the causal variant of each colocalization locus for missense variants within genes located in the colocalized genomic loci. We performed regional multi-trait colocalization analyses between type 2 diabetes, each osteoarthritis phenotype, and molecular QTLs from disease-relevant tissues.

Figure 4

Overview of the 19 high-confidence effector genes for the type 2 diabetes and osteoarthritis comorbidity Genes are stratified based on the joint affected by osteoarthritis. The scoring of the six biological lines of evidence is depicted on the right (material and methods). OA = osteoarthritis; T2D = type 2 diabetes; molQTLs = molecular quantitative trait loci; DEG = differential expressed genes; KO mice = knockout mice; OMIM = Online Mendelian Inheritance in Man; HC = previously defined high-confidence effector genes; missense = missense variant.

Figure 5

Regional association plots of the highlighted colocalizing regions between type 2 diabetes and osteoarthritis (A–D) FTO and IRX3 region; (B) TCF7L2 region; (C) WSCD2 and TMEM119 region; (D) TMEM176A region. The plots are colored on the basis of linkage disequilibrium between the lead causal variant of the colocalization and all other variants in the region. The red dashed line represents the genome-wide significance threshold (p value = 5 × 10⁻⁸), and the blue dashed line represents a suggestive association threshold (p value = 10⁻⁶). PP4 = posterior probability of a single shared causal variant; OA = osteoarthritis; T2D = type 2 diabetes; all OA = osteoarthritis at any site.