This guest post was contributed by Karol Estrada, a postdoctoral research fellow in the Analytic and Translational Research Unit at Massachusetts General Hospital and the Broad Institute of MIT and Harvard. It is dedicated to the memory of Laura Riba.
Genome-wide association studies (GWAS) of common variants have successfully implicated more than 70 genomic regions in type 2 diabetes, revealing new biological pathways and potential drug targets. However, most large studies have examined genetic variation only in northwestern European populations, despite the rich genetic diversity in other populations around the world. Most studies have also been limited in their ability to detect variants present in fewer than 5 percent of people. Much remains to be learned.
In this post, we discuss our new paper, published in the Journal of the American Medical Association, on a low-frequency missense variant in the gene HNF1A that raises risk of type 2 diabetes five-fold, and was seen only in Latinos. This variant was the only rare variant to reach genome-wide significance in an exome sequencing study of almost 4,000 people, the largest such study to date. We explain the ramifications for sample sizes of rare-variant studies, note the importance of studying populations outside of northwestern Europe, and caution against simplistic dichotomous interpretations of disease as either complex or monogenic. Finally, we note that a low-frequency or rare variant might guide therapeutic modification.
For the new JAMA paper, we conducted whole-exome sequencing of 1,794 type 2 diabetes cases and 1,962 healthy controls from four studies of Mexicans and other self-identified Latinos. Most large genomic studies of common variants have focused on research participants with northwestern European ancestry. (An exception is the latest paper from the DIAGRAM type 2 diabetes consortium, which identified seven new loci for type 2 diabetes by examining data from people of European, east Asian, south Asian and Mexican and Mexican-American ancestry.) But other ancestry groups have genetic diversity not observed in Europe. Background rates of disease also differ across populations: for instance, type 2 diabetes is the leading cause of death in Mexico, and is twice as common among Mexican Americans as it is among populations with European ancestry. For this reason, the Slim Initiative in Genomic Medicine for the Americas (SIGMA) Type 2 Diabetes Project set out to perform both GWAS and sequencing studies in large, well-characterized samples from Mexico and of Mexican-American ancestry.
The first paper from the collaboration, published on Christmas Day 2013 in Nature, found that a set of four variants in the SLC16A11 gene increases risk of type 2 diabetes. The variants are on the same haplotype, which is common in people of Latin American and East Asian ancestry (it is found in about 50 percent and 10 percent of those populations, respectively) but extremely rare in people of European ancestry and absent in those of African ancestry. Previous GWAS, which focused on people of European ancestry, had failed to find the haplotype because it is so rare in that group.
Like the Nature paper, our JAMA study illustrates the value of studying populations that have not yet participated extensively in genomic research. In our study, a low-frequency variant in HNF1A — present in 2 percent of type 2 diabetes cases and 0.4 percent of healthy controls — quintupled risk of type 2 diabetes, the largest effect ever observed for a type 2 diabetes variant found in more than 0.1 percent of the population. The variant was found only in people who live in Mexico or the southern U.S. and identify as Latino. It was not found in publicly available genetic databases, including 1000 Genomes, Exome Sequencing Project, and dbSNP. Therefore, we would have missed this variant even if we had used the latest genotyping array technology and imputed (i.e., inferred the presence of) variants that were not directly genotyped.
Despite performing exome sequencing of nearly 4,000 people, our study did not find any other low-frequency or rare variants associated with type 2 diabetes above genome-wide statistical significance — the HNF1A variant was the only one. This reflects an important point about studying rare variants. Early efforts to find such variants using exome sequencing involved relatively small numbers of patients, based on the theory that the strength of the variants’ effects might outweigh the need for a large sample size. But in general, identifying rare causal variants will require thousands, and often tens of thousands, of cases and controls. Anything smaller will probably be statistically underpowered.
The HNF1A variant we discovered also has implications beyond study design: it’s a case study in how perceptions of disease are not always subtle enough to match reality. People tend to classify objects into discrete categories: things are big or small, black or white. Classification isn’t a bad thing. Medical doctors figuratively live and die by it, and in some cases their patients literally live and die by it: it’s the basis of medical diagnosis. But it does have its limitations. For instance, the American Diabetes Association suggests that “a fasting glucose of 126 mg/dL or greater, or a 2 hr post 75 gr of glucose load of 200 mg/dl or greater” should be used as a cut-off for diagnosing someone with diabetes. This clearly doesn’t mean that a person with diabetes whose fasting glucose measurement briefly dips below 125 mg/dL should suddenly be considered healthy.
For type 2 diabetes, another such line is between “complex” and “monogenic” the common disease that affects 347 million people worldwide, and the rare “maturity onset diabetes of the young,” or MODY, form of diabetes. Classical type 2 diabetes is typically diagnosed after age 40, associated with obesity, and caused by many genetic and environmental factors acting in concert. MODY, on the other hand, is typically diagnosed before age 25 in patients of average weight, and is considered monogenic: a patient will carry a rare coding mutation in one of 13 known MODY genes, and will have a 50 percent chance of transmitting it to his or her children. MODY has been widely thought of as “fully penetrant” and autosomal dominant — i.e., a person who carries one copy of one of the variants that causes it would be expected to have the disease.
HNF1A — the gene in which we found our type 2 diabetes variant — is one of the 13 known MODY genes. But people who carry the HNF1A variant observed in our study don’t have MODY as doctors would think of it. They look more like other patients with regular type 2 diabetes: overweight to obese, with an onset of disease late in life. And not everyone who carries this variant has a disease — we found 12 carriers who were healthy. We also examined the variant’s effect on the function of the HNF-1A protein, and found that it was only one third as dramatic as one would expect from the known MODY mutations in the gene.
MODY, and in general the way we think about rare diseases, may be more complicated than we’ve assumed. Last year, some of our collaborators showed that 1.5 percent of randomly selected people from the Framingham Heart Study carried so-called MODY mutations, but the vast majority of them had completely normal glucose levels. The upshot is that mutations we think of as fully penetrant for rare diseases may in fact appear surprisingly frequently in healthy people or intermediate phenotypes. It’s unclear because, often, we’ve only studied those mutations in people who have the rare diseases.
Our JAMA paper has one more potential implication for screening and therapeutic modification, which might apply to the 2 percent of people of Latinos with diabetes who carried the HNF1A variant we found. The American Diabetes Association recommends metformin as a first-line treatment for type 2 diabetes, but it has been known for years that patients with MODY respond especially well to a family of drugs called sulfonylureas. If type 2 diabetes patients who carry our HNF1A variant are also especially responsive to sulfonylureas, that might benefit tens of thousands of people. Type 2 diabetes is extremely common in Mexico, and managing the complications of diabetes is expensive, but sulfonylureas are cheap. Much work remains to be done to explore the differential impact of sulfonylureas on HNF1A variant carriers, but this raises the possibility that genotyping a single low-frequency variant could help some 100,000 people suffering from a deadly disease while also reducing healthcare costs.
Geneticists often focus on the importance of statistical “power” in genomics. But the real power of genomics will come from its ability to improve people’s lives.
This work was made possible by a large, multinational consortium supported by the Carlos Slim Health Institute. We thank our scientific collaborators, many of whom are based in Mexico and made critical contributions to the work. The Mexico-based team included investigators from Instituto Nacional de Ciencias Médicas y Nutrición Salvador Zubirán, Instituto de Investigaciones Biomédicas UNAM, led by Teresa Tusié and Carlos Aguilar; the Instituto Nacional de Medicina Genómica (INMEGEN) led by Lorena Orozco; and the Mexico City Diabetes Study led by Clicerio González-Villalpando. The Norway-based team included investigators from the University of Bergen and Haukeland University Hospital, and was led by Pål Njølstad. The Los Angeles-based team included investigators from the Keck School of Medicine of the University of Southern California and was led by Brian Henderson and Christopher Haiman. The Boston-based team included researchers from the Broad Institute, Massachusetts General Hospital, and Harvard Medical School. It was led by Jose Florez, a Broad Institute associate member, an associate professor of medicine at Harvard Medical School, and an assistant physician in the Diabetes Unit and the Center for Human Genetic Research at the Massachusetts General Hospital; David Altshuler, deputy director and chief academic officer at the Broad Institute and a Harvard Medical School professor at Massachusetts General Hospital; and Daniel MacArthur, group leader within the Analytic and Translational Genetics Unit at Massachusetts General Hospital, assistant professor at Harvard Medical School and a researcher at the Broad Institute. The analysis team was led by Karol Estrada, a Research Fellow within the Analytic and Translational Genetics Unit at Massachusetts General Hospital. We thank Mary Carmichael for her substantial contribution to writing this post.