A Mendelian Randomization Study of Metabolite Profiles, Fasting Glucose
Total Page:16
File Type:pdf, Size:1020Kb
Diabetes Volume 66, November 2017 2915 A Mendelian Randomization Study of Metabolite Profiles, Fasting Glucose, and Type 2 Diabetes Jun Liu,1 Jan Bert van Klinken,2 Sabina Semiz,3,4 Ko Willems van Dijk,2,5 Aswin Verhoeven,6 Thomas Hankemeier,1,7,8 Amy C. Harms,7,8 Eric Sijbrands,9 Nuala A. Sheehan,10 Cornelia M. van Duijn,1,6 and Aysxe Demirkan1,5 Diabetes 2017;66:2915–2926 | https://doi.org/10.2337/db17-0199 GENETICS/GENOMES/PROTEOMICS/METABOLOMICS Mendelian randomization (MR) provides us the opportunity paths to and from glucose and type 2 diabetes and underline to investigate the causal paths of metabolites in type 2 the value of additional information from high-resolution meta- diabetes and glucose homeostasis. We developed and bolomics over classic biochemistry. tested an MR approach based on genetic risk scoring for plasma metabolite levels, utilizing a pathway-based sensi- tivity analysis to control for nonspecificeffects.Wefocused Type 2 diabetes is a progressive metabolic disease character- on 124 circulating metabolites that correlate with fasting ized by hyperglycemia, initially as a result of insulin resistance glucose in the Erasmus Rucphen Family (ERF) study (n = and in later stages also as a result of insulin insufficiency. Type 2 2,564) and tested the possible causal effect of each me- diabetes is also associated with dyslipidemia, including higher tabolite with glucose and type 2 diabetes and vice versa. circulating concentrations of triglycerides and lower concen- We detected 14 paths with potential causal effects by MR, trations of HDL cholesterol. In addition, several circulating following pathway-based sensitivity analysis. Our results molecules have previously been shown to be dysregulated in suggest that elevated plasma triglycerides might be par- type 2 diabetes, including phospholipids, branched-chain amino tially responsible for increased glucose levels and type 2 acids, keto-acid metabolites, and other metabolites such as acyl- diabetes risk, which is consistent with previous reports. carnitines (1–3). However, the causal paths between these Additionally, elevated HDL components, i.e., small HDL triglycerides, might have a causal role of elevating glucose metabolites and glucose/type 2 diabetes in humans remain levels. In contrast, large (L) and extra large (XL) HDL lipid unclear from observational studies and require randomized fi components, i.e., XL-HDL cholesterol, XL-HDL–free cho- controlled trials that are dif cult to conduct. lesterol, XL-HDL phospholipids, L-HDL cholesterol, and As an alternative, Mendelian randomization (MR) is an L-HDL–free cholesterol, as well as HDL cholesterol seem instrumental variable method that has gained popularity over to be protective against increasing fasting glucose but not the last decade to investigate causal effects of traits using against type 2 diabetes. Finally, we demonstrate that genetic genetic predictors. MR uses the principle that the allocation predisposition to type 2 diabetes associates with increased of genetic variants that affect a specifictraitfromparentsto levels of alanine and decreased levels of phosphatidylcholine offspring is random and unrelated to factors other than the alkyl-acyl C42:5 and phosphatidylcholine alkyl-acyl C44:4. trait (4). Furthermore, associations between the genotype Our MR results provide novel insight into promising causal and the outcome will not be affected by reverse causation 1Department of Epidemiology, Erasmus Medical Center, Rotterdam, the Netherlands 9Department of Internal Medicine, Erasmus Medical Center, Rotterdam, the 2Department of Human Genetics, Leiden University Medical Center, Leiden, the Netherlands Netherlands 10Department of Health Sciences, University of Leicester, Leicester, U.K. 3 Genetics and Bioengineering Program, Faculty of Engineering and Natural Sci- Corresponding author: Aysxe Demirkan, [email protected]. ences, International University of Sarajevo, Sarajevo, Bosnia and Herzegovina Received 15 February 2017 and accepted 19 August 2017. 4Department of Biochemistry and Clinical Analysis, Faculty of Pharmacy, University of Sarajevo, Sarajevo, Bosnia and Herzegovina This article contains Supplementary Data online at http://diabetes 5Department of Endocrinology, Leiden University Medical Center, Leiden, the Netherlands .diabetesjournals.org/lookup/suppl/doi:10.2337/db17-0199/-/DC1. 6Center for Proteomics and Metabolomics, Leiden University Medical Center, Leiden, © 2017 by the American Diabetes Association. Readers may use this article as the Netherlands long as the work is properly cited, the use is educational and not for profit, and the 7Division of Analytical Biosciences, Leiden Academic Centre for Drug Research, work is not altered. More information is available at http://www.diabetesjournals Leiden University, Leiden, the Netherlands .org/content/license. 8Netherlands Metabolomics Centre, Leiden University, Leiden, the Netherlands 2916 Mendelian Randomization Study Diabetes Volume 66, November 2017 because disease will occur after the meiosis. MR has pre- markers including subfractions of lipoproteins, triglycerides, viously been used to determine whether metabolic markers phospholipids, ceramides, amino acids, acyl-carnitines, and such as classic blood lipids are causally involved in type 2 small intermediate compounds, which throughout this arti- diabetes (5–11) and has yielded contradicting results. One clewillbereferredas“metabolites,” were measured either reason for this could be that these studies are affected by by NMR spectrometry or by MS. The platforms used in this the heterogeneous nature of the metabolic markers chosen, research are the following: 1) liquid chromatography-MS such as in the example of total HDL cholesterol, which in (LC-MS) (116 positively charged lipids comprising 39 tri- reality consists of a collection of different-sized HDL parti- glycerides, 47 phosphatidylcholines, 8 phosphatidylethanol- cles possibly with different functions. This may dilute the amines, 20 sphingolipids, and 2 ceramides available in up to causal effects of single nucleotide polymorphisms (SNPs) 2,638 participants) measured in the Netherlands Meta- when only combined (total) HDL is considered. However, bolomics Centre, Leiden, using the method described previ- false signals may also be due to pleiotropic effects of the ously (18); 2) electrospray-ionization MS (ESI-MS) (in total, chosen genetic variants leading to possibly invalid instru- 148 phospholipids and sphingolipids comprising 16 plas- mental variables. As high-throughput analyses techniques mologens, 72 phosphatidylcholines, 27 phosphatidyletha- improve, the quantification of circulating molecules is becom- nolamines, 24 sphingolipids, and 9 ceramides available in ing ever more detailed and precise. For instance, instead of up to 878 participants) measured in the Institute for Clin- LDL cholesterol, HDL cholesterol, and total triglycerides de- ical Chemistry and Laboratory Medicine, University Hospi- termined by routine clinical biochemistry, lipoprotein particle tal Regensburg, Regensburg, Germany, using the method size distribution and content as well as tens of biochemical described previously (14); 3) small molecular compounds components can now be measured using nuclear magnetic res- window-based NMR spectroscopy (NMR-COMP) (41 mole- onance (NMR) spectroscopy– and mass spectrometry (MS)- cules comprising 29 low–molecular weight molecules and based approaches (12,13). These additional measures offer 12 amino acids available in up to 2,639 participants) mea- an opportunity to gain novel insight into the pathogenesis of sured in the Center for Proteomics and Metabolomics, Leiden diseases like type 2 diabetes. With the knowledge of genetic University Medical Center (19,20); 4) lipoprotein window- determinants of metabolites gained from genome-wide asso- based NMR spectroscopy (NMR-LIPO) (104 lipoprotein par- ciation studies (GWAS) (14–16), one can use MR for causal ticle subfractions comprising 28 VLDL components, 30 HDL inference given the specific conditions encoded in Fig. 1. In components, 35 LDL components, 5 IDL components, and the current study, with the aim of unraveling potentially 6 plasma totals available in up to 2,609 participants) mea- causal metabolic paths that underlie the observed associa- sured in the Center for Proteomics and Metabolomics, Leiden tions, we used genetic predictors from published metabolite University Medical Center (lipoprotein subfraction concen- GWAS, guided by pathway-based evidence to select instru- trations were determined by the Bruker algorithm [Bruker mental variables, and performed MR between selected met- BioSpin GmbH, Germany] [16]); and 5) AbsoluteIDQ p150 abolic markers and glucose/type 2 diabetes. Kit of Biocrates Life Sciences AG (153 molecules comprising 14 amino acids, 91 phospholipids, 14 sphingolipids, 33 acyl- RESEARCH DESIGN AND METHODS carnitines, and hexose available in up to 989 participants) Study Population measured with the experiments carried out at the Meta- The observational associations between metabolites and bolomics Platform of the Genome Analysis Center at the fasting glucose/type 2 diabetes were tested in the Erasmus Helmholtz Zentrum München, Germany, per manufacturer Rucphen Family (ERF) study, which is a prospective family- instructions (15). The laboratories had no access to pheno- based study with 3,465 individuals in the southwest of the type information. Netherlands. The study protocol for ERF was approved by the medical ethics board of the Erasmus