Genetic Dark Matter

Standing over Darwin’s grave in Westminster Abbey, Andrew Feinberg had a realization.

Feinberg, a genetics researcher at Johns Hopkins University in Baltimore, looked to the left and saw Newton’s grave. Just above Newton is a plaque honoring physicist Paul Dirac, a pioneer of quantum theory. Inherent in quantum theory is the idea of uncertainty in the interaction of subatomic particles.

“So I look back at Darwin’s grave and it hits me; there’s nothing like that in biology,” Feinberg says. Nothing that deals with uncertainty.

Yet there is uncertainty in biology. Genes that run in families explain only some of the wide variety of physical appearances among people and their susceptibility to diseases. Much uncertainty in what causes these differences remains.

But biologists don’t just accept this seeming randomness as a fundamental part of reality. Instead, they are seeking an explanation for unknown sources of variation in heritable traits, the way physicists are searching for a mysterious substance dubbed dark matter that could explain puzzling aspects of the cosmos.

And biologists have proposed some solutions. Feinberg’s, scribbled down at a pub in the shadow of the Tower of London, is that chemical modifications to DNA could be the genetic dark matter.

Feinberg is in the minority, though; others have their own favorite theories about what the missing ingredient might be. Some think that researchers just need to hunt harder and longer for common changes in the sequence of genetic letters that make up DNA. But a growing number of researchers are turning to rare genetic changes or absent or duplicated chunks of DNA as important contributors. Others say that interactions among genes deserve more attention.

Whatever the ultimate explanation, understanding the complete heritability picture could allow doctors to design prevention and treatment plans catered to individuals’ needs.

Unclear math

Factors determining how much people have in common with other people fall into two broad categories — nature and nurture. The nature part is a measure of how genetic variation contributes to the range of phenotypes — outward appearances, behaviors or disease risks — among people. Environmental factors, such as exposure to chemicals, nutrition, exercise or smoking habits, make up the nurture part.

Geneticists mainly concern themselves with the stuff of nature, the basis of heredity. Heredity plays a big role in how tall people will grow, their weight or their likelihood of getting the same disease that killed a grandparent, for example. Studies of families, especially identical twins, have helped researchers calculate how much genetic factors contribute to variation in a given characteristic. 

Sometimes the genetic contributions are simple. Inheriting a mutation in a single gene can lead to disorders such as cystic fibrosis, sickle-cell anemia or Huntington’s disease. But usually it’s not just one of the 22,000 or so human genes that is responsible for a trait. Many, many genes play a role in traits like height and weight and in complex diseases such as heart disease and diabetes. Geneticists have been trying for decades to track down how genetic variations in DNA lead to such diseases.

About five years ago, the pieces started coming together. Researchers began scanning the genomes of thousands of people using genetic markers called single nucleotide polymorphisms, or SNPs (pronounced “snips”), to try to pinpoint disease-causing and trait-linked genes. A SNP is a single-letter change in the sequence of A’s, C’s, T’s and G’s that make up the DNA alphabet. Though everyone has thousands of SNPs and some may have no effect, the genome-wide scans use statistical methods to figure out the likelihood that particular spelling changes contribute to a characteristic.

The impetus for these genome-wide association studies was a popular theory that common genetic variants — SNPs found in 5 to 10 percent of the population or more — could add up to cause common diseases. People unlucky enough to be weighed down with a backbreaking number of such variants would get sick.

“Everyone was so excited when these studies first started coming out,” says Teri Manolio, a genetic epidemiologist at the National Human Genome Research Institute in Bethesda, Md. “Everyone was like, ‘Wow, we have a gene for diabetes. We’ve got 10. We’ve got 50.’ But then when we started looking more closely, it was apparent that there was a lot missing.”

How tall a person will grow, for instance, is 80 to 90 percent heritable, meaning that genetic factors account for most of the centimeters and millimeters that separate the tallest from the shortest people. A recent study including more than 180,000 people uncovered 180 common genetic variants associated with height. But those differences account for only about 10 percent of the genetic contribution to height variation (SN: 10/23/10, p. 15). The consortium of researchers conducting the study, published online September 29 in Nature, calculated that more than 1,000 common genetic variants, each of which nudge height up or down a millimeter or two, could be involved in determining how tall or short people will ultimately be. Even then, adding the effects of all 1,000 variants would cover only about 20 percent of heritable variation in the human population, leaving a lot unaccounted for.

For most genome-wide association studies, the story is similar. They identify a large number of genetic variants with small effects, but none add up to fully explain heritability.

Researchers know that the cause of the unexplained genetic variation, dubbed “missing heritability,” must be out there somewhere. Like dark matter, says Olivier Harismendy, a genomicist at the University of California, San Diego, “we know it’s there, but we just can’t grasp it.”

A tagging matter

Unconvinced that common variants will ever explain all the missing heritability, Feinberg is looking under other rocks. His proposal involves a rock that genome-wide association studies have not flipped over.

Chemical tags that don’t change genes but do modulate their activity may account for some of the variation, Feinberg thinks. These changeable tags are known as epigenetic marks, and a genome-wide association study can’t uncover them because they don’t alter the letters of a person’s DNA.

Researchers know about a host of different epigenetic tags that affix to DNA or to proteins closely associated with DNA (SN: 5/24/08, p. 14). The molecular equivalent of the two-sided open/closed signs that hang in business windows, these tags signal that genes are ready to be turned on or shuttered. Certain environmental conditions, including those encountered in the womb, can permanently flip the sign, and that flip can pass from one generation to the next.

Feinberg hypothesized that many epigenetic marks would be largely the same from person to person, but that at a few strategic locations — such as near genes involved in development — people might have widely different tagging. If an epigenetic mark made a gene less active, the tag could affect a person’s outward traits or disease risk, and thus be an uncounted part of heritability. Or an epigenetic mark might change the activity of a mutated gene that would otherwise affect a trait, hiding that mutation’s full contribution.

To see whether these chemical tags make a difference, Feinberg’s team, including Johns Hopkins colleagues Rafael Irizarry and Daniele Fallin, measured a particularly stable kind of epigenetic mark known as DNA methylation. Affixing a methyl group to DNA generally has the effect of shutting down nearby genes.

The team mapped methyl tags in DNA samples taken 11 years apart from 74 Icelandic people (SN: 10/9/10, p. 15). The researchers reported September 15 in Science Translational Medicine finding four places in the Icelanders’ genomes where more heavily methylated DNA correlated with a person’s body weight.

This study is far too small to say how much of the missing heritability the addition or subtraction of chemical marks may account for, Feinberg says. “We’re very careful to say that we don’t have the magic missing answer,” he says. But he thinks his data are strong enough to suggest that other researchers should expand searches to include the chemical marks.

Most haven’t yet gotten on board and continue to look for changes in the DNA letters themselves.

“Right now, we’re hoping for a simpler explanation,” says Rasmus Nielsen, a population geneticist at the University of California, Berkeley, “but if it’s something more complex like epigenetics, it may take a lot longer before we get to the answer.”

Spelling-centric view

Many researchers have pinned their hopes on rare variants, single-letter DNA changes found in only a small portion of the population. Studies of some diseases, including autism, schizophrenia and bipolar disorder, suggest that rare mutations — even some so rare that they appear only in a single person or family — can contribute in a big way to whether a person gets a disease. Some spelling mistakes linked to mental retardation, for example, are seen for the first time in people affected with that disability, according to a study published online November 14 in Nature Genetics.

Genome-wide association studies can look for rare spelling changes by using SNPs present in only 1 to 5 percent of people or fewer, but variants that show up suddenly in just one person may elude these studies.

Enter the 1,000 Genomes Project, an effort to catalog all the genetic variation in 2,500 people from around the world by determining every letter in their genetic instruction books. The project made its debut in a study appearing October 28 in Nature. A consortium of researchers, including David Altshuler, a geneticist at the Broad Institute in Cambridge, Mass., cataloged 15 million SNPs, 1 million small insertions or deletions of parts of chromo­somes, and about 20,000 larger missing or added chunks in more than 800 people (SN: 11/20/10, p. 14). About half of the variants had not been cataloged before. Conducting genome-wide scans that include these new variants may reveal some missing sources of heritability.

“It’s easy to say, ‘Oh, it must be rare variants,’ but now we have the technology to really go after them,” Harismendy says.

A companion study in the Oct. 29 Science also found that about 1,000 genes vary in numbers of copies in people, most falling into the zero- to five-copy range but some reaching as many as 368 copies. The role that multiple copies play in heritability remains to be determined.

“We’re just trying to unveil genetic variation,” says Evan Eichler of the University of Washington in Seattle, who led the study. “But it’s not a panacea. It’s not like everything is solved now.” Researchers should think about genomes as a whole, with all variations, to really understand contributions to heritability, he says.

A fruit fly study suggests the whole-genome approach may be the way to go. The Drosophila Genetic Reference Panel was launched two years ago to try to account for all the genetic factors that go into the fly’s highly heritable traits such as aggressiveness, longevity, stress tolerance, sleep duration and recovery times after cold-induced comas, says Trudy Mackay, a geneticist at North Carolina State University in Raleigh and one of the project’s leaders.

Mackay and her colleagues captured pregnant female Drosophila melanogaster fruit flies from the wilds of a farmers market in Raleigh and then inbred the offspring to create 200 different strains, each with a characteristic response to each test. Instead of doing genome-wide association studies to look for variants linked with each trait, Mackay and her colleagues determined the entire genetic blueprints for the strains. So far, the genomes for 162 are known, though the results are not yet published.

“What we find is that we can account for all the heritability, thank you very much,” Mackay says. Most of the heritability comes from single-letter changes present in about 3 to 5 percent of the fruit fly population. Researchers probably don’t have much further to go to understand heritability in humans either, Mackay says. “I think we’re there, in theory, but we need more data.”

Peter Visscher, a quantitative geneticist at the Queensland Institute of Medical Research in Brisbane, Australia, agrees that missing heritability is almost within researchers’ grasp. He argues that the problem of missing heritability is a mathematical one. Using a statistical model different from the one used in the recent height paper in Nature, Visscher and colleagues calculate that common variants account for about 45 percent of height’s heritability, the team reports in the July Nature Genetics. And if most diseases have underlying genetic contributions similar to those for height, then researchers need only expand the search for common variants with tiny effects in ever bigger studies, or locate rare variants with larger effects, to find all the pieces.

Yeast weigh in

But Feinberg isn’t alone in thinking that some of the work requires looking beyond DNA letters. Leonid Kruglyak, a geneticist and Howard Hughes Medical Institute investigator at Princeton University, finds it unlikely that having complete genetic blueprints for many individual fruit flies and people will tell the whole heritability story for every trait. He has been studying patterns of heritability in baker’s yeast. Yeast have only about 6,000 genes and a much more streamlined genome than humans do, so in theory, it should be far easier to account for heritability.

Kruglyak and his colleagues tested yeast for their ability to withstand 17 different types of chemicals. Resistance to some chemicals is mainly controlled by one gene, while surviving other chemicals may require 20 or more genes, he and his colleagues reported April 15 in Nature. And while the researchers were able to find 14 regions of the genome that account for about 70 percent of the heritability of resistance to a chemical called 4-nitroquinoline, they could explain only about 10 percent of the heritability of resistance to some other types of chemicals, he says.

He is interested in exploring how interactions between genes may affect certain traits. Alone, a genetic variant may have a clear effect, but another genetic variant may cancel out or intensify the effect when it occurs within the same person.

A study of the skin condition psoriasis recently illustrated that point. A consortium of researchers reported in the November Nature Genetics that people who have a disease-associated variant of a gene called ERAP1 have a risk of developing psoriasis that is no greater than average as long as they also carry normal copies of an immune system gene called HLA-C. But a gene variant of HLA-C can work with the ERAP1 variantto dramatically increase psoriasis risk.

As scientists pursue possible hiding places for the missing heritability, many agree that the answer won’t lie under just one rock.

“The missing heritability implies that it is sitting in a box somewhere and it will be obvious when we find the answer,” Altshuler says. “There’s no simple answer because the genetics of these complex diseases is complicated.”

Each of the factors — additional common variants, rare variants, epigenetics, gene-gene interactions — may turn out to explain some of the missing heritability.

“I have no idea how it’s going to shake out when we’re sitting around in 100 years talking about heritability,” says Kruglyak. “Anybody who will give you a precise breakdown is guessing.”