Gene Variant Associated With Obesity Risk Found With New Statistical Technique

The study involved an international team led by researchers at the Boston University School of Medicine (BUSM) in collaboration with researchers at HSPH, Children's Hospital Boston, GSF National Research Center, Technical University Munich, Ludwig Maximilians University, Channing Laboratory, University of Duisburg-Essen, Loyola University Medical Center, and SeraCare Life Sciences Inc. The paper will appear in the April 14 issue of Science.

The scientists identified the variant by testing 86,604 single nucleotide polymorphisms (SNPs), or DNA sequence variations, for an association to body mass index (BMI), a surrogate measurement for obesity. The SNPs came from blood samples from participants in the Framingham Heart Study (FHS). BMI measurements were gathered by FHS investigators under a contract from the National Heart Lung and Blood Institute administered by Boston University.

Researchers at the BUSM Genetics and Genomics Department, led by lead author and Assistant Professor Alan Herbert and senior author and Department Chair Michael Christman, genotyped the samples, identifying genetic variations within them for comparison. The large-scale genotyping was enabled by 100K DNA microarray technology developed by Affymetrix, a company based in Santa Clara, Calif. Data management and analysis was done by Drs. Norman Gerry and Marc Lenburg at BUSM and HSPH research fellow Matthew McQueen.

The identification of the SNP was made possible by a new statistical technique developed by HSPH Assistant Professor Christoph Lange, who is a senior author on the Science paper, in collaboration with Professor and co-author Nan Laird, and Kristel Van Steen, former postdoctoral fellow.

The technique averts a major statistical difficulty called the multiple comparison problem, which is exacerbated when crunching huge data sets. The more SNPs in a study, the more likelihood of false-positive signals, the increased risk that some SNPs are not tested -- even though they may be viable candidates.

Typically, a study would demand a two-step process of 1) initial testing of all SNPs for associations to specific traits, such as body mass index, and 2) testing only the most promising SNPs in a second data set.

The HSPH analysis method handles the multiple comparison problem with just one data set by estimating the extent to which genetics can explain variations in a specific trait, and then identifying the most promising candidate SNPs for that genetic effect. For more information on the technique, see http://www.hsph.harvard.edu/press/releases/press06052005.html.

In the Science paper, the HSPH method was applied to identify exactly one SNP that emerged from the tens of thousands tested. It lies near the INSIG2 gene (insulin-induced-gene). The variant was present in ten percent of the populations studied and required two copies to be active.

Importantly, the scientists then collaborated with other institutions and replicated the association between the SNP and obesity risk in four other populations. The populations included both families and non-related subjects of different ages and of varying ethnic backgrounds:

• Children's Hospital Boston (Joel Hirschhorn and Helen Lyon) and Genomics Collaborative, SeraCare Life Sciences, Inc., Cambridge, Mass. (Kristin Ardlie)
  --Sample involved self-identified white or Caucasian subjects from the U.S. and from Poland numbering 1,775 cases and 926 controls. The odds of being obese were 1.37 greater for subjects with both copies of the SNP than in controls.
• Loyola University Medical Center (Richard Cooper and Xiaofeng Zhu) and Children's Hospital Boston (Joel Hirschhorn and Helen Lyon)
  --Two samples -- one of which consisted of families and the other consisted of unrelated individuals -- involved self-identified African-Americans in Illinois with a total of 1,268 subjects. Both samples indicated that subjects who had both copies of the SNP were more likely to be obese.
• KORA Group, GSF National Research Center, Neuherberg, Germany, (H.-Erich Wichmann, Iris Heid, Arne Pfeufer, Thomas Illig, and Thomas Meitinger)
  --Sample involved nearly 4,000 subjects in a town near Munich, Germany. Subjects with two copies of the SNP had BMIs that averaged .60 kg/m2 higher than those with other genotypes combined.
• University of Duisburg-Essen, Department of Child and Adolescent Psychiatry, Essen, Germany (Johannes Hebebrand, Anke Hinney, and Kerstin Koberwitz)
  --Sample involved 1,104 subjects (children and their parents) in Western Europe. An analysis confirmed that children who inherited two copies were more likely to be obese than those who did not.

The association was not found in a sample from the Nurses' Health Study. The authors suggest the result may reflect a different distribution of BMI rates among the Nurses' Health Study subjects compared to the other samples, or differences in their environments and lifestyles.

"Several features make the study in Science especially exciting: the large-scale genotyping, the use of a population sample of families, the long-term follow up of the subjects, the replication of the finding in independent samples, and a novel statistical methodology which makes maximal use of the data," said Laird. "Most studies of this type use clinical samples and unrelated subjects that are not measured repeatedly over time, and obtaining replications in independent samples is very difficult."

Approximately 65 percent of Americans are overweight and 30 percent are obese. Obesity is a risk factor for numerous diseases, including hypertension, diabetes, and stroke. Genetics and environmental factors such as poor diet and lack of exercise influence obesity risk. Understanding the genetics underlying obesity is an important piece in the puzzle of how to manage the obesity epidemic.

Components of this work received support from a number of sources: Leadership Award from the Whitaker Foundation, grants from the American Diabetes Association, an ADA Smith Family Pinnacle Program Project, NIH grants, German Ministry of Education and Research through the National Genome Research Network, Bioinformatics for the Analysis of Mammalian Genomes, and National Cancer Institute. DNA samples and phenotypic data were provided by NHLBI-Framingham Heart Study investigators.