Gene-environment interaction study for BMI reveals interactions between genetic factors and physical activity, alcohol consumption and socioeconomic status

Abstract

Previous genome-wide association studies (GWAS) have identified hundreds of genetic loci to be associated with body mass index (BMI) and risk of obesity. Genetic effects can differ between individuals depending on lifestyle or environmental factors due to gene-environment interactions. In this study, we examine gene-environment interactions in 362,496 unrelated participants with Caucasian ancestry from the UK Biobank resource. A total of 94 BMI-associated SNPs, selected from a previous GWAS on BMI, were used to construct weighted genetic scores for BMI (GSBMI). Linear regression modeling was used to estimate the effect of gene-environment interactions on BMI for 131 lifestyle factors related to: dietary habits, smoking and alcohol consumption, physical activity, socioeconomic status, mental health, sleeping patterns, as well as female-specific factors such as menopause and childbirth. In total, 15 lifestyle factors were observed to interact with GSBMI, of which alcohol intake frequency, usual walking pace, and Townsend deprivation index, a measure of socioeconomic status, were all highly significant (p = 1.45*10-29, p = 3.83*10-26, p = 4.66*10-11, respectively). Interestingly, the frequency of alcohol consumption, rather than the total weekly amount resulted in a significant interaction. The FTO locus was the strongest single locus interacting with any of the lifestyle factors. However, 13 significant interactions were also observed after omitting the FTO locus from the genetic score. Our analyses indicate that many lifestyle factors modify the genetic effects on BMI with some groups of individuals having more than double the effect of the genetic score. However, the underlying causal mechanisms of gene-environmental interactions are difficult to deduce from cross-sectional data alone and controlled experiments are required to fully characterise the causal factors.

Fig 1. Distribution of weighted genetic scores for BMI (GS_BMI) in participants from the UK Biobank (left y-axis).

The average BMI in kg/m² (right y-axis) for participants in each bin of the histogram is plotted as black diamonds ± 95% confidence interval. The dotted line represents the regression of BMI across GS_BMI.

Fig 2

Interaction between GS_BMI genotype with frequency of alcohol consumption (A) and frequency of more than 10 minutes of walking per week (B). (A) Effect on BMI per GS_BMI by frequency of alcohol intake. The self-report questionnaire asked: “About how often do you drink alcohol?” The effect per GS_BMI is higher in low-frequency alcohol consumers compared to high-frequency consumers: (“Never”: N = 7,944; “Special occasions only”: N = 12,767; “once or twice a week”: N = 12,966; “One to three times a month”: N = 30,412; “Three or four times a week”: 27,250; “Daily or almost daily”: N = 24,424.) (B) Effect on BMI per GS_BMI by frequency of 10+ minutes of walking. The self-report questionnaire asked: “In a typical week, on how many days did you walk for at least 10 minutes at a time? (Include walking that you do at work, travelling to and from work, and for sport or leisure).” (“None”: N = 2,528; “One”: N = 7,046; “Two”: 7,046; “Three”: 9,215; “Four”: N = 9,393; “Five”: 18,441; “Six”: N = 11,334; “Seven”: N = 53,125). Error bars represent 95% CI.

Fig 3

Interaction between (A) GS_BMI and (B) rs1558902, with alcohol intake frequency. (A) BMI was plotted against GS_BMI. and stratified by alcohol consumption frequency. The effect of GS_BMI, i.e., the increase in BMI with GS_BMI, was lower in UK Biobank participants who consume alcohol less frequently, compared to participants who consume alcohol more frequently. (B) Mean BMI per genotype of the FTO-linked SNP, rs1558902, is plotted by frequency of alcohol intake. The effect of rs1558902, i.e. the increase in BMI with copies of the A-allele, was lower in high-frequency alcohol consumers, and higher in low-frequency consumers.