Introduction
Cardiometabolic diseases are the predominant cause of mortality, morbidity and healthcare spending globally [
1,
2], and are believed to result in part from the combined additive and synergistic effects of genetic and environmental risk factors. Environmental exposures such as diet and physical activity have enormous potential for prevention and treatment of these diseases, but no single therapy works well in all individuals. Determining whether susceptibility to adverse environmental exposures is genetically determined (i.e. gene–environment interactions [
3]) and elucidating the specific nature of these interactions may facilitate the stratification of patient populations into subgroups that can be treated with optimal therapies.
In contemporary population genetics research, the heritability of a given trait is usually assessed by quantitative genetics approaches to make inferences about the extent to which polygenic variation influences the trait. Assessing heritability is usually done prior to embarking on studies that seek to discover specific loci influencing the trait. While it is equally logical to use quantitative genetics to determine whether traits are influenced by genotype–environment interactions as a prelude to studies focused on specific environmental exposures and genetic loci, this is rarely done in practice [
4‐
8]. The dearth of such studies may be because large, well-characterised cohorts including genealogies, which are necessary for genotype–environment quantitative genetic studies, are rare.
Here we sought to screen for genotype–environment interactions across a number of environmental exposures and cardiometabolic traits using quantitative genetic analyses in extended pedigrees. Accordingly, we characterised the genealogical structure of a large northern Swedish population, within which detailed measures of environmental exposures, cardiometabolic traits and other personal characteristics exist [
9].
Discussion
To our knowledge, this is the first compendium of genotype–environment interactions for cardiometabolic traits to be reported. The purpose of doing so is to provide a foundation for subsequent locus-specific analyses of interaction effects and to aid the interpretation of published locus-specific interaction studies. After accounting for multiple testing, we observed robust evidence of genotype–age interactions for body weight and SBP, genotype–sex interactions for BMI and triacylglycerol, and genotype–alcohol intake interaction for body weight.
There are many published reports concerning interactions of environmental exposures with genetic factors in cardiometabolic traits (reviewed in [
30‐
33]). Approaches include quantitative genetics studies, usually undertaken in twin or family-based cohorts [
4,
6,
7,
34‐
38] and candidate gene studies, focused on individual genetic variants, haplotypes, or genetic risk scores constructed from variants with high biological priors for interactions or those conveying genome-wide significant marginal effects [
39‐
47]. Several quantitative genetic studies have shown that physical activity attenuates the influence of genetic effects on cardiometabolic traits [
4,
6,
34,
35,
37,
38]. However, only
FTO–physical activity interactions in obesity [
39‐
42] have been adequately replicated in candidate gene studies. In the present study, we observed evidence of genotype–physical activity interactions for DBP, 2 h glucose (class 1) and triacylglycerol (class 2), but not for obesity-related traits. This may be because analyses of the kind reported here account for the overall modifying effect of genetic variation (polygenic interactions), whereas gene–physical activity interactions in obesity may be oligogenic in nature.
According to our analyses, variation in the intake of macronutrients (whether modelled together or separately) may interact with genetic variation to affect body composition and glycaemic control. Although several candidate gene studies have focused on gene–diet interactions (e.g. [
43‐
47]), there are few quantitative genetics studies on this topic, and these were restricted in scope and conducted in relatively small cohorts [
35]. On the other hand, family-based studies have reported class 2 genotype–smoking interactions with serum leptin levels (an important endophenotype of adiposity) [
7,
36], and those findings are consistent with the current analyses for body weight.
Although this is a hypothesis-generating study, and as such one might argue against multiple test adjustments owing to the risk of a type II error [
48], we adopted a conservative approach to minimise the number of false-positives reported. Nevertheless, as described in the Results section, many of the statistical models yielded nominal evidence of interactions for the environmental exposures and cardiometabolic traits assessed. We present those findings, as the approach used here is orthogonal to standard approaches used to model genotype–environment interactions; thus, the combination of these approaches may help verify the presence or absence of interaction effects. Despite the relatively large sample size used here, it is of course likely that some of the hypothesis tests were underpowered. Statistical power may be diminished by the imprecise nature of the self-reported methods used to assess many of the environmental exposures and the need to dichotomise some of these variables for analysis. Survival bias is a further possible limitation, as people with the most deleterious genetic and/or environmental risk characteristics might have been excluded from the cohort because of early mortality. Systematic error (bias), on the other hand, may lead to false-positive or false-negative conclusions: for example, if an environmental exposure is over-reported at high or low levels of the cardiometabolic trait, or a strong correlate [
49], an observed genotype–environment interaction may be false-positive. However, this limitation clearly does not impact our strongest findings (for age and sex), as these were objectively assessed. Additionally, as in other studies including genealogical information from registries (without genetic validation), the pedigrees are unlikely to be completely accurate due, for example, to false paternity. A further consideration is that some environmental exposures assessed here are to a limited extent influenced by genetic background [
50,
51]; hence, it is possible that what might on the surface appear to be a genotype–environment interaction reflects, at least in part, epistasis.
In conclusion, our results suggest that cardiometabolic traits are heavily influenced by the interactions between the genotype and environmental exposures. Our data indicate that future studies focused on identifying specific genetic variants underlying genotype–environment interactions should focus on the exposures of age, sex and alcohol intake on body composition. Numerous other exposures and outcomes defined here are also plausible candidates for genotype–environment interaction.
Acknowledgements
The authors thank the participants, the health professionals, data managers and other staff involved in the Västerbotten Health Survey. We also thank V. Diego, C. Peterson and J. Blangero (Department of Genetics, Texas Biomedical Research Institute, San Antonio, TX, USA) for sharing program scripts for interaction analyses in the SOLAR program, and J. Armelius Lindberg and Å. Ågren (Department of Biobank Research, Umeå University, Umeå, Sweden), M. Sandström and P. Vikström (Centre for Demographic and Ageing Research, Umeå University, Umeå, Sweden) for their work on data extraction. Some of the computations were performed on resources provided by the Swedish National Infrastructure for Computing at HPC2N in Umeå, Sweden.
Open Access This article is distributed under the terms of the Creative Commons Attribution 4.0 International License (http://creativecommons.org/licenses/by/4.0/), which permits unrestricted use, distribution, and reproduction in any medium, provided you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons license, and indicate if changes were made.