|Home | About | Journals | Submit | Contact Us | Français|
The environment is largely responsible for differences in BMI between populations; genetics is largely responsible for differences in BMI within populations. Evidences in support of the environmental effect are many. The prevalence of obesity differs greatly between countries, even among developed countries (1). On a smaller scale, differences in the “built environment” within a city are associated with differences in BMI. In New York City a greater density of bus and subway stops and greater variety in uses of land within a neighborhood are associated with lower BMIs (2). Closely genetically related populations living in different environments differ considerably in average BMI (3). The environment clearly can be more or less obesegenic.
But there is strong evidence that within a population the variance in BMI is largely genetically determined. The weight of adoptees correlates better with the BMI of their biologic parents than with that of their adoptive parents (4). Twin studies indicate that about two-thirds of the variance in BMI is attributable to additive genetic factors (5). The most convincing study of the heritability of BMI in twins was that of Stunkard and colleagues, who studied 93 adult pairs of monozygotic twins who had been separated as young children and reared apart (6). These authors similarly estimated that about two-thirds of the variance in BMI was due to genetic factors.
What are these genetic factors? Two different experimental approaches have been used to identify obesity susceptibility genes in humans: a candidate gene approach and an agnostic method using single-nucleotide polymorphisms (SNPs) in a genome-wide association study (GWAS). Candidate genes have been selected to be screened for variants associated with obesity based on the physiology and molecular mechanisms of energy homeostasis in experimental animals. Very few obese humans have causative variants in genes thought to be the major constituents of this homeostatic system (7). The one exception is the melanocortin 4 receptor (MC4R). About two and a half percent of severely obese individuals have causative variants in this gene (8). These variants are associated with obesity largely due to increased food intake but also due to reduced energy expenditure (9).
GWASs are the most commonly used agnostic approach to identify susceptibility genes for common disease, including obesity. These have been done in populations of European descent and used ~350,000 SNPs, covering more than 75% of the genome (10,11). The two major findings were common noncoding SNPs in an intron of the FTO gene and a common SNP within a presumed regularly site 188 kb downstream of the MC4R gene. The additive genetic effect of these polymorphisms in a population of ~77,000 adults was a difference in BMI of 1.17kg/m2 between the compound homozygotes for the risk alleles (1% of the population) vs. the compound homozygotes for the nonrisk alleles (19% of the population). These two common variants in FTO and MC4R account for less than 2% of the variance in adult BMI. The combined results of the candidate gene and GWAS approach account for very little of the variance in BMI that is heritable.
There could be many more genes with small effects similar to those of the FTO and MC4R polymorphisms. There could also be several rare variants, each with a large effect. Since the GWAS approach depends on the “common phenotype-common genotype” hypothesis to be successful, rare variants would be missed. Copy number variation, rather than SNPs, also may contribute to susceptibility to obesity. But recent work indicates that most of the common (allele freq ≥0.05) diallelic copy number polymorphisms are in strong linkage disequilibrium with SNPs and would have been detected by the GWAS approach (12). Copy number variations that are rare and more variable might contribute to the heritability of BMI and also would have been missed by the GWAS approach.
How many more common variants with small effects or rare variants with large effects have to be found to explain the missing BMI heritability? It has been estimated (13) that for variants with small effects (risk ratio ~1.2), 50 genes with a genotypic prevalence of 10%, or 25 genes with a genotypic prevalence of 20%, are needed to explain fifty percent of the population attributable disease risk. For rare variants (1 per 5,000) with large effects (risk ratios 10–20) as many as 186–556 genes are needed to explain 50% of a common disease.
As an example, all of the 18 recently discovered (and replicated) type 2 diabetes mellitus susceptibility genes combined account for only a small proportion of the heritability of the disease. The reported sibling relative risk is ~3 and the combined sibling relative risk of the 18 variants is ~1.07(14).
What is the potential clinical utility of identifying multiple common variants with small effects on population risk of disease? Knowing an individual’s genotype of the 18 type 2 diabetic susceptibility genes is essentially no better at predicting the disease than relying on the person’s BMI, age, and sex (14). This analysis did not even include knowledge of the individual’s family history. It has been suggested that knowing an individual’s genotype at all these susceptibility loci will be useful for assessing an individual’s risk even though on a population basis these variants account for little risk of disease. But it seems very unlikely that it will be cost-effective any time soon to sequence the entire genome of many individuals to find the very small percentage of people with substantial increased disease risk based on their cumulative genotype at multiple loci. This is especially true considering the marginal, if any, increase in risk for common diseases such as obesity and diabetes over other more easily determined risk factors, including age, sex, and family history.
Thus, there is not likely to be clinical utility of the GWAS results (for BMI and type 2 diabetes at least) in the foreseeable future in populations of European descent.
The utility of the GWAS results for BMI (and type 2 diabetes) is therefore only for identifying new mechanisms of disease. Increased knowledge of disease etiology should lead to better treatments and preventive measures, but this is likely to take many years if not decades.
Finally, whether the genetic susceptibility to obesity in populations not of European descent will also be multigenic, or will be due to a few susceptibility genes with larger effects, remains to be determined. In these populations it remains possible that GWAS will have a different utility.