Author + information
- Peter Würtz, PhD†,‡,
- Antti J. Kangas, MSc†,
- Pasi Soininen, PhD†,§,
- Terho Lehtimäki, MD, PhD‖,
- Mika Kähönen, MD, PhD¶,
- Jorma S. Viikari, MD, PhD#,
- Olli T. Raitakari, MD, PhD∗∗,††,
- Marjo-Riitta Järvelin, MD, PhD‡‡,§§,‖‖,¶¶,
- George Davey Smith, MD, PhD##∗∗∗∗ ( and )
- Mika Ala-Korpela, PhD†,§,¶¶∗∗∗
- †Computational Medicine Research Group, Institute of Health Sciences, University of Oulu, Oulu, Finland
- ‡Institute for Molecular Medicine Finland, University of Helsinki, Helsinki, Finland
- §NMR Metabolomics Laboratory, School of Pharmacy, University of Eastern Finland, Kuopio, Finland
- ‖Department of Clinical Chemistry, Fimlab Laboratories, University of Tampere, Tampere, Finland
- ¶Department of Clinical Physiology, University of Tampere and Tampere University Hospital, Tampere, Finland
- #Department of Medicine, University of Turku and Turku University Hospital, Turku, Finland
- ∗∗Research Centre of Applied and Preventive Cardiovascular Medicine, University of Turku, Turku, Finland
- ††Department of Clinical Physiology and Nuclear Medicine, Turku University Hospital, Turku, Finland
- ‡‡Department of Epidemiology and Biostatistics, MRC-HPA Centre for Environment and Health, School of Public Health, Imperial College London, London, United Kingdom
- §§Department of Children and Young People and Families, National Institute for Health and Welfare, Oulu, Finland
- ‖‖Institute of Health Sciences and Biocenter Oulu, University of Oulu, Oulu, Finland
- ¶¶Oulu University Hospital, Oulu, Finland
- ##Medical Research Council Centre for Causal Analyses in Translational Epidemiology, University of Bristol, Bristol, United Kingdom
- ∗∗∗Computational Medicine, School of Social and Community Medicine, University of Bristol, Bristol, United Kingdom
- ↵∗MRC Integrative Epidemiology Unit, School of Social and Community Medicine, University of Bristol, Oakfield House, Oakfield Grove, Bristol, BS8 2BN, United Kingdom
To the Editor:
Mendelian randomization is increasingly used to infer causality in observational epidemiology (1). Confounding factors are then minimized through the random assignment of genetic predisposition. In recent extensive Mendelian randomization studies, remnant cholesterol (remnant-C) has been advanced as a causative risk factor for ischemic heart disease, independently of high-density lipoprotein (HDL) cholesterol (2,3). Remnant-C is understood as the cholesterol carried in the triglyceride-rich very-low-density lipoprotein (VLDL) and intermediate-density lipoprotein (IDL) particles together with chylomicrons in the nonfasting state. It is well established that small VLDL and IDL particles (i.e., those with the highest content of cholesterol in the remnant particle category) can penetrate the arterial wall and become retained in the intima. Thus, a causal role of remnant-C is physiologically plausible. However, implying causality by using Mendelian randomization is, in many cases, not appropriate (1). Genes that regulate lipoprotein metabolism are rarely specific for 1 lipid measure or a single lipoprotein (4).
We applied detailed lipoprotein subclass profiling to illustrate the inherent pleiotropy of the lipid-related genetic variants used to examine causality of remnant-C by Varbo et al. (2). We used nuclear magnetic resonance (NMR) spectroscopy to quantify 14 lipoprotein subclasses as well as multiple lipid measures and circulating metabolites (4) from 10,547 fasting serum samples from 3 population-based cohorts of adolescents and young adults in Finland. Figure 1 shows an association heat map of the lipid genes from the study by Varbo et al., with various lipoprotein and metabolite measures to clarify the specificity or pleiotropy of the genetic loci.
Three variants in the TRIB1 (tribbles homolog-1), GCKR (glucokinase regulatory protein), and APOA5 (apolipoprotein A-V) loci were incorporated as instruments for remnant-C in the Mendelian randomization analyses of causality for ischemic heart disease (2). However, the variant in TRIB1 was positively associated with the entire range of apolipoprotein B–carrying lipoprotein particles as well as HDL triglycerides and inversely associated with large HDL (Fig. 1). Although this genotype was also associated with remnant-C, it was not specific to this lipoprotein measure. Similarly, a common variant in GCKR was associated with all VLDL lipids at approximately equal magnitude as observed for remnant-C. The genetic variant in GCKR was also associated with multiple fatty acid measures and even certain glycolytic substrates and amino acids (4). In addition, a wide pleiotropic profile was observed for the variant in the APOA5 locus.
The genetic variants in the LPL locus, denoted as instruments for both remnant-C and HDL-C in the study by Varbo et al. (2), are indeed associated with both of these cholesterol fractions; however, the specific lipid associations do not suggest a stronger association with cholesterol than with other lipids in VLDL particles (Fig. 1). In addition, a polymorphism in the LIPC locus, considered to be an instrument specific for HDL-C in the study by Varbo et al. (2), was inversely associated not only with large HDL lipids but also with the smallest VLDL, IDL, and the largest low-density lipoprotein (LDL) lipids. The variant in PCSK9, used as a positive control for the LDL-C in their study, was strongly associated with LDL lipids, although similar associations were also observed for IDL and small VLDL lipids, including the summary measure of (non-HDL, non-LDL)-C. In general, the association profiles of the lipid loci extend beyond the composite lipids measures used in clinical medicine and epidemiology. Caution must therefore be exercised when addressing causality of lipid fractions by using these genotypes in instrumental analyses. More detailed phenotyping, such as lipoprotein subclass profiling, would therefore be advised in addressing the validity of the genetic instruments used in Mendelian randomization. This is increasingly important because Mendelian randomization is becoming a widespread tool for causal inference, and guidelines are placing evidence from such studies between observational data and randomized trials.
Regardless of the pleiotropic complications inherent in the lipid loci used as instruments in these studies (2,3), the causal role of remnant-C is physiologically supported, and its investigation is of high relevance. The definition advocated by some researchers is that remnant-C would be estimated as nonfasting total cholesterol minus HDL-C minus LDL-C, implying the use of the Friedewald approximation for LDL-C (2,3). This definition complicates the assessment of genetic variants supposedly specific to remnant-C, because the Friedewald LDL-C estimation does, in fact, include the IDL-C fraction as well (5). In contrast, lipoprotein lipid quantification by using NMR enables size-specific measurements of 14 lipoprotein subclasses and thereby permits a direct assessment of, for example, (non-HDL, non-LDL)-C and VLDL-C.
Overall, the concept of remnant-C as a causative risk factor for coronary heart disease deserves further attention and may constitute a prominent target for interventions to reduce vascular risk beyond that afforded by existing therapies. However, future studies on the role of triglyceride-rich lipoproteins should account for the complexity of the definition of remnant-C as well as the wide pleiotropy of the genes implicated in lipoprotein metabolism.
Please note: This study was supported by the Academy of Finland and the Strategic Research Funding from the University of Oulu, Finland. The authors have reported that they have no relationships relevant to the contents of this paper to disclose.
- American College of Cardiology Foundation
- Davey Smith G.,
- Ebrahim S.
- Varbo A.,
- Benn M.,
- Tybjærg-Hansen A.,
- Jørgensen A.B.,
- Frikke-Smith R.,
- Nordestgaard B.G.
- Jørgensen A.B.,
- Frikke-Schmidt R.,
- West A.S.,
- Grande P.,
- Nordestgaard B.G.,
- Tybjærg-Hansen A.
- Tukiainen T.,
- Kettunen J.,
- Kangas A.J.,
- et al.