Stratification by Smoking Status Reveals an Association of CHRNA5-A3-B4 Genotype with Body Mass Index in Never Smokers

dataset

posted on 2014-12-04, 03:33 authored by Amy E. Taylor, Richard W. Morris, Meg E. Fluharty, Johan H. Bjorngaard, Bjørn Olav Åsvold, Maiken E. Gabrielsen, Archie Campbell, Riccardo Marioni, Meena Kumari, Jenni Hällfors, Satu Männistö, Pedro Marques-Vidal, Marika Kaakinen, Alana Cavadino, Iris Postmus, Lise Lotte N. Husemoen, Tea Skaaby, Tarunveer S. Ahluwalia, Jorien L. Treur, Gonneke Willemsen, Caroline Dale, S. Goya Wannamethee, Jari Lahti, Aarno Palotie, Katri Räikkönen, Aliaksei Kisialiou, Alex McConnachie, Sandosh Padmanabhan, Andrew Wong, Christine Dalgård, Lavinia Paternoster, Yoav Ben-Shlomo, Jessica Tyrrell, John Horwood, David M. Fergusson, Martin A. Kennedy, Tim Frayling, Ellen A. Nohr, Lene Christiansen, Kirsten Ohm Kyvik, Diana Kuh, Graham Watt, Johan Eriksson, Peter H. Whincup, Jacqueline M. Vink, Dorret I. Boomsma, George Davey Smith, Debbie Lawlor, Allan Linneberg, Ian Ford, J. Wouter Jukema, Christine Power, Elina Hyppönen, Marjo-Riitta Jarvelin, Martin Preisig, Katja Borodulin, Jaakko Kaprio, Mika Kivimaki, Blair H. Smith, Caroline Hayward, Pål R. Romundstad, Thorkild I. A. Sørensen, Marcus R. Munafò, Naveed Sattar

We previously used a single nucleotide polymorphism (SNP) in the CHRNA5-A3-B4 gene cluster associated with heaviness of smoking within smokers to confirm the causal effect of smoking in reducing body mass index (BMI) in a Mendelian randomisation analysis. While seeking to extend these findings in a larger sample we found that this SNP is associated with 0.74% lower body mass index (BMI) per minor allele in current smokers (95% CI -0.97 to -0.51, P = 2.00×10⁻¹⁰), but also unexpectedly found that it was associated with 0.35% higher BMI in never smokers (95% CI +0.18 to +0.52, P = 6.38×10⁻⁵). An interaction test confirmed that these estimates differed from each other (P = 4.95×10⁻¹³). This difference in effects suggests the variant influences BMI both via pathways unrelated to smoking, and via the weight-reducing effects of smoking. It would therefore be essentially undetectable in an unstratified genome-wide association study of BMI, given the opposite association with BMI in never and current smokers. This demonstrates that novel associations may be obscured by hidden population sub-structure. Stratification on well-characterized environmental factors known to impact on health outcomes may therefore reveal novel genetic associations.