A metabolomic signature of weight-loss and its associations with BMI-linked diseases: A post-hoc analysis of two intervention trials and a mendelian randomisation study
Nicholas Sunderland, Madeleine L. Smith, Alex McConnachie, Paul Welsh, Roy Taylor, Michael E. J. Lean, Chris A. Rogers, Jane M. Blazeby, Naveed Sattar, Lavinia Paternoster, R. Thomas Lumbers, Nicholas J. Timpson, Laura J. CorbinBackground
Obesity is a major risk factor for a broad spectrum of diseases, yet the molecular pathways underlying this risk remains poorly understood. Individual analytical approaches—interventional trials, observational data, and genetic methods—are each susceptible to distinct sources of bias. We aimed to identify a common metabolic perturbation signature of adiposity by triangulating evidence across multiple sources of body mass index (BMI) variation, and to characterise associations between this signature and contrasting adiposity-linked disease outcomes.
Methods and findings
We integrated mass spectrometry and nuclear magnetic resonance metabolomic profiling from two weight-loss interventions—a structured diet programme (DiRECT trial, ISRCTN03267836) and metabolic and bariatric surgery (By-Band-Sleeve trial, ISRCTN00786323)—with mendelian randomisation (MR) estimates of the effect of lifetime BMI exposure on metabolite levels, in order to identify a consistent BMI-associated metabolite signature across sources with distinct error structures. Weight loss-associated metabolites were then instrumented using genetic variants and their associations with heart failure (HF), osteoarthritis, and kidney cancer estimated in a two-sample MR framework, with instruments grouped into clusters based on genetic similarity to aid biological interpretation. 242 metabolites were consistently altered across weight loss interventions, predominantly comprising lipids, lipoproteins, and amino acids. The majority of these showed concordant associations with lifetime BMI in MR, consistent with regulation by adiposity. Differential components of this signature were associated with disease-specific outcomes. For HF, low-density lipoprotein (LDL) measures were strongly associated with increased risk, consistent with ischaemic aetiology enrichment, whilst high-density lipoprotein (HDL) cholesterol-enrichment traits were protective, and markers of dysfunctional HDL remodelling were associated with increased risk. For osteoarthritis, lipoprotein associations were again associated with disease risk but in the opposite direction to that seen with HF; additional signals were observed in glycine metabolism pathways. For kidney cancer, the adiposity metabolite signature translated poorly into disease associations, with only 5-methyluridine and ursodeoxycholate associations identified. The main limitations of our study include the potential for residual horizontal pleiotropy with metabolite instruments despite multiple sensitivity analyses and the rank-normalised scale of metabolite measures, which limits direct physiological interpretability.
Conclusions
Triangulating evidence across dietary intervention, metabolic and bariatric surgery, and MR of lifetime BMI identified a robust metabolic signature of adiposity. The association of components of this signature with obesity-related outcomes was disease-specific, supporting the notion that not all adiposity-associated metabolic perturbations are uniformly relevant across the spectrum of obesity-related diseases.