Genetically proxied growth differentiation factor 15 levels and body mass index

Growth-differentiation factor 15 (GDF15) is an inflammatory cytokine involved in energy homeostasis that is being pursued as a drug target for obesity. Its circulating levels are acutely increased by the type 2 diabetes medication metformin, resulting in reduced appetite and weight loss. We identified a genetic variant at the GDF15 gene to proxy a small, lifelong increase in circulating GDF15 levels, and leveraged it in colocalization and Mendelian randomization analyses to investigate the effects of chronically elevated GDF15 levels on body mass index (BMI) and type 2 diabetes liability. The results provide human genetic evidence supporting that chronically elevated GDF15 levels increase BMI. There was no genetic evidence to support bi-directional effects, or that chronically elevated GDF15 levels directly affect liability to type 2 diabetes. Our results contrast the BMI lowering effects of an acute increase in GDF15 levels observed after metformin use. These findings have direct implications for informing pharmacological strategies aimed at targeting GDF15 levels for weight loss.


Introduction
Growth-differentiation factor 15 (GDF15) is an inflammatory cytokine involved in energy homeostasis that is being pursued as a drug target for obesity 1 . Its circulating levels are acutely increased by the type 2 diabetes medication metformin, resulting in reduced appetite and weight loss 2 . However, the effect of chronically elevated circulating GDF15 levels on body weight is not known.
Here, we identified a genetic variant at the GDF15 gene to proxy a small, lifelong increase in circulating GDF15 levels, and leveraged it in genetic analyses to investigate the effects of chronically elevated GDF15 levels on body mass index (BMI) and type 2 diabetes liability.

Methods
Colocalization and Mendelian randomization analyses were performed. Briefly, colocalization compares the genetic associations for two traits within a given genetic locus, investigating whether the data supports a model with a shared causal variant for both traits 3 .
In Mendelian randomization, genetic variants that proxy the effect of varying the exposure are used to investigate its effect on an outcome 4 . Our choice of genetic proxy for the effect of GDF15 signalling was restricted to a variant at the GDF15 locus, to reduce the risk of bias related to pleiotropic associations unrelated to GDF15 5   This article is protected by copyright. All rights reserved.

Discussion
Using human data, we provide genetic evidence supporting the notion that chronically elevated GDF15 levels increase BMI. There was no genetic evidence to support bi-directional effects, or that chronically elevated GDF15 levels directly affect liability to type 2 diabetes.
Our results contrast the BMI lowering effects of an acute increase in GDF15 levels observed after metformin use 2 . There are a number of possible explanations for this discrepancy, including that metformin may be affecting BMI though mechanisms unrelated to GDF15.
Long-term elevation of circulating GDF15 levels may also contrast the effects of acute increases in GDF15 and lead to desensitization of the GDF15 receptor and reduced signalling 9 . Alternative possibilities include compensatory changes in other hormones and behaviour (including dietary intake) as result of elevated circulating GDF15 levels.
Furthermore, it is possible that GDF15 affects foetal and childhood development, thus impacting adult BMI through this mechanism.
The use of both colocalization and Mendelian randomization in this study provide complementary evidence supporting causal effects of chronically elevated GDF15 levels on BMI. As genetic variants are randomly allocated at conception, the Mendelian randomization paradigm is less susceptible to the confounding and reverse causation that that can hinder causal inference in observational studies. A previous Mendelian randomization analysis exploring the effect of circulating GDF15 levels on BMI used variants from throughout the genome in their main analysis 10 . While the greater number of variants generated through this strategy allows for the incorporation of Mendelian randomization statistical sensitivity analyses, this approach also risks including variants that have pleiotropic associations unrelated to GDF15, which can consequently bias the Mendelian randomization analysis 5 .
Previous Mendelian randomization studies exploring the effect of circulating GDF15 have This article is protected by copyright. All rights reserved. also used GDF15 genetic association estimates that were adjusted for systolic blood pressure, antihypertensive medication use, diabetes mellitus and smoking status 10 11 , which can introduce collider bias 12 . Another previous study used a genetic variant related to GDF15 gene expression in whole blood to proxy the effect of varying GDF15 circulating levels 13 .
However, colocalization analysis was not performed in this work, so it is unclear if genetic confounding was introducing bias into consequent Mendelian randomization analysis 13 . Such issues potentially explain the discrepancy in the results of these previous studies with our current work, which did not suffer from the same limitations.
As a limitation of our current work, the genetic associations were derived from individuals of European ancestries, and therefore our results may not generalize to other ethnic groups. A further limitation is that we used BMI as a measure of obesity. However, BMI may also be elevated in the context of increased lean mass and body water content, and is not always reflective of increased body fat.
In conclusion, this genetic analysis found robust evidence to support that, in contrast to acute elevations in GDF15 levels, chronically elevated GDF15 levels increase BMI. These findings may be used to inform the design of pharmacological strategies aimed at targeting GDF15 for weight loss.