She Has Her Mother's Laugh: The Powers, Perversions, and Potential of Heredity (38 page)

When Hirschhorn and his colleagues published their study in 2001, it was one of the first times that anyone had found a clue about common variants that influence height. But it was a very modest start. Hirschhorn had been able to identify only long stretches of DNA where a genetic variant seemed to be lurking. The variants might reside in one of hundreds of genes in those regions. It was even possible that Hirschhorn's results were a fluke that had nothing to do with height. A number of tall people might have a version of a particular marker thanks simply to chance.

Hirschhorn was not alone in his frustration. Many other scientists were trying to trace traits—especially hereditary risks for certain diseases—to specific genes. At first they enjoyed some high-profile successes, finding links to conditions like diabetes and bipolar disorder.
But very often, the links would melt away when other scientists looked at larger groups of people. Soon scientists worried that they were stuck in a dead end. “
Has the genetic study of complex disorders reached its limits?” two scientists asked in a 1996 article in the journal
Science.

Those two scientists, Neil Risch of Stanford University and Kathleen Merikangas of Yale, argued that the answer was no. But to uncover the variants that raise the risk of common diseases, scientists would have to build new tools. Risch and Merikangas predicted that most variants would not be
powerful, as in the case of Laron syndrome and acromegaly. Instead, the variants behind many diseases would be weak and numerous.

Risch and Merikangas sketched out a new way to carry out this search. Geneticists needed to step away from their beloved pedigrees. Instead, they needed to look at the DNA of hundreds of people, with no regard to their families. They could search for variants that were unusually common in people with a disease, compared to those who did not suffer from it. Risch and Merikangas dubbed their hypothetical method
a genome-wide association study.

It took until 2005 for genome-wide association studies to get their first hit. Josephine Hoh, a geneticist at Yale University, wanted to find genes involved in the leading cause of blindness, a disease called
age-related macular degeneration (AMD for short) that ravages the center of the retina. Hoh knew that having a relative with AMD raised the odds that people would develop it in their own life. But studies on families with AMD had failed to reveal a gene associated with the disease.

Hoh and her colleagues gathered DNA from ninety-six people who had AMD, as well as from fifty people who didn't. They scanned the genetic markers and noticed an unusually common one among people with AMD located on chromosome 1. Closely examining that region, they came across a variant in a gene for a protein made by immune cells, called complement factor H. They found that having two copies of the variant drastically raised a person's odds of developing AMD.

The job of complement factor H is to stick to pathogens, triggering inflammation to fight them. Hoh's research indicated that mutant forms of the protein stick instead to retinal cells, causing the immune system to attack the eye.
Hoh's findings were later confirmed in other studies. But with such a small group of people in her study, she might very well have missed complement factor H if its effects had been any weaker. She was right, and she was also lucky.

To use genome-wide association studies to find subtler variants, scientists recognized they would have to study thousands or even millions of people. In 2007, a consortium of laboratories working through the
Wellcome Trust in England published the first such large-scale study. Examining fourteen thousand people, they identified twenty-four genes with variants that raised the risk of
diseases such as diabetes and arthritis.

After his own frustrating experience studying the height of families, Hirschhorn also turned to genome-wide association studies. He and his colleagues used some of the data from the Wellcome Trust study, adding to it people who had been part of a diabetes study in Sweden. All told, nearly five thousand people became part of the study. The technology for sequencing genetic markers had improved drastically since Hirschhorn had started investigating height. Now, instead of looking at a few hundred markers, he could look at a few hundred thousand of them. The denser spread of genetic markers made it possible to zero in on smaller regions containing fewer genes.

This time, Hirschhorn got a solid hit. One variant, located in
a gene called HMGA2, was significantly more common in tall people than in short ones—so common, in fact, that it couldn't be dismissed as a fluke. Hirschhorn and his colleagues tested the association by looking at HMGA2 in more than twenty-nine thousand other people. In the bigger group, taller people once again were much more likely to carry the same variant of HMGA2.

Yet Hirschhorn couldn't say how precisely HMGA2 influenced people's height. A few experiments carried out over the years offered a handful of clues. In experiments with mice, some mutations to HMGA2 could turn the animals into dwarves. Others turned them into giants (by mouse standards).

The evidence about HMGA2's function in humans was even scarcer. In 2005, geneticists at Harvard Medical School published a case report on
an eight-year-old boy who had a mutation clipping his HMGA2 gene short. He seemed normal at birth, but at three months he sprouted his first tooth. By the time the boy was eight years old, he was over five foot five, the average height for a fifteen-year-old. His legs and fingers grew crookedly, and he developed lumps of fat and blood vessels under some parts of his skin.

These studies suggest that HMGA2 normally acts like a brake, slowing
down our growth-spurring genes. A mutation that shuts down HMGA2 entirely may cause runaway growth. The common variant in HMGA2 that increases height may lift the genetic foot off the brake just enough to make people grow a bit taller—but not enough to lead to deformities or tumors.

The discovery of HMGA2 was like a quarter-carat sapphire: solid, glittering, and tiny. It marked the first time that scientists found a common variant strongly associated with height. Later, when other scientists studied even larger groups of people, they confirmed the link. But the HMGA2 variant accounts for a vanishingly small amount of the variation in the human population. When I got my genome sequenced, I found that I carry one copy of the height-raising form. On average, people with one copy are about an eighth of an inch taller than if they didn't have one. That's the equivalent of putting on a warm pair of wool socks. If I had two copies, it would be like putting on a second pair. And when scientists look at the full range of variation in height, they find that this variant in HMGA2 explains very little—only about 0.2 percent.

Hirschhorn's 2007 study also uncovered some tantalizing clues about many other genes. They contained variants that were more common in tall people than in short ones, or vice versa. But the differences weren't as stark as HMGA2, leaving open the possibility they were the result of chance. To rule out randomness, Hirschhorn would need to measure more people's heights.

Hirschhorn and his colleagues created a new network of hundreds of research groups around the world. They called their consortium the Genetic Investigation of ANthropometric Traits—GIANT for short. The GIANT team
examined the height of tens of thousands of people, then hundreds of thousands, and the bigger numbers allowed them to pick out more genetic variants, first dozens, then hundreds. Most of the genes they discovered had a smaller influence than HMGA2. But they also found a number of genes that had a far bigger one. If people carry two variants of a gene called STC2, for example, those alleles will lift them up about an inch and a half. These powerful genes had gone overlooked in earlier studies of height because they were too rare, found in less than 5 percent
of the population. In 2017, a decade after the first genome-wide association study of height, GIANT published
a study on more than 700,000 people, bringing the total number of genes influencing height to almost eight hundred.

To some observers, however, such results seemed like a colossal disappointment. The combined effect of GIANT's eight hundred–odd genes accounted for just over 27 percent of the heritability of height. The rest remained missing.

Height was not unusual in this regard.
Missing heritability dogged many studies of other traits and diseases, too, even after scientists could study thousands of people. The shortfall was all the more glaring because of all the money that had gone into making genome-wide association studies possible. “The reason for spending so much money was that the bulk of the heritability would be discovered,”
the geneticist Joseph Nadeau told a journalist.

Some critics saw missing heritability as much more than an annoyance. To them, it was a symptom of a scientific disease. In 2015, two French researchers, Emmanuelle Génin and Françoise Clerget-Darpoux, argued that missing heritability revealed the futility of genome-wide association studies. Génin and Clerget-Darpoux describe the research as “
Garbage-In Garbage-Out Syndrome.” The scientists running the studies were trying to use brute force to discover the deepest secrets of biology. Yet their repeated failures simply led them to redouble their efforts, and journal editors to publish more of their papers. To Génin and Clerget-Darpoux, it seemed as if geneticists had become trapped in a game they couldn't stop playing. “Unfortunately, genetics is a clear loser,” they concluded.

Other critics say that missing heritability reveals our profound ignorance about heritability itself. Some attacked twin studies, claiming they lead to estimates of heritability that are much too high. Others argued that heritability studies miss the way some mutations make the effects of other mutations stronger. One plus one, in the world of heredity,
may be far more
than two. Some critics went even further, arguing that
missing heritability is hiding beyond genes, in some other form of heredity scientists have yet to grasp.

When I asked Hirschhorn if missing heritability was giving him existential doubts, he shrugged the problem off. “I think a lot of it is just hidden,” he told me. “If we had all six billion people on Earth in a genetics study, we would actually get to most of the heritability.”

Part of Hirschhorn's confidence came from his own experience over the previous twenty years. The more people he and his colleagues measured, the more heritability they could explain. Some of the genes they found were common but weak, while others were strong but rare. If he could study more people in the future, he expected to find more of both kinds.

Hirschhorn also drew confidence from the work of Peter Visscher, who has given geneticists a new way to study human heritability. Visscher came to research on humans after years of work on livestock.
Animal breeders study the heritability of cows to figure out how to get them to make more milk, of pigs to put on more pork. In the 1900s, they used elaborate pedigrees to track the influence of genes on these traits. But at the end of the century, breeders got their hands on technology for reading genetic markers in their animals.

At first, they searched for candidate genes that might have a big effect on their own. Soon it became clear that a trait like milk output was controlled by many genes, each with a tiny effect. Animal breeders found that they could improve their livestock by comparing all their genetic markers in different animals. Animals that were genetically similar overall tended to have similar traits. Breeders could choose which animals to breed based on these so-called genomic predictions.

When Visscher switched from animals to humans in the early 2000s, he realized that he could use genomic predictions on people, too. Visscher and his colleagues took the method out of the barnyard and adapted it to human genetics, dubbing their method Genome-wide Complex Trait
Analysis. To see how well it worked, they unleashed it on the best-studied complex trait of all, human height.

The researchers delved into the data from earlier genome-wide association studies and looked at the genetic markers from thousands of people. They came up with genetic-similarity scores between each pair of people. Heredity turned out to work a lot in humans as it does in chickens. Pairs of people with high scores tended to have similar heights. That tendency reflects the heritability of a trait. The stronger the tendency, the greater the heritability.

When Visscher and his colleagues estimated the heritability of human height from genetic similarity, they ended up with a number close to what had been estimated in earlier studies on families and twins. In 2015, when they published these results in the journal
Nature Genetics
, they declared the missing heritability of height to be “
negligible.”

Toward the end of my visit with Hirschhorn, I noticed his eyes drifting to the clock on his desk phone. He had a conference call coming up soon with a lot of his collaborators. They were about to take another leap, from 800,000 people to perhaps two million. But before I left, Hirschhorn explained that the years of work he had put into the inheritance of height were not simply to create a catalog of genes. He wanted to use the catalog to understand the mysteries of height. If you stop and think through what it means to grow, the process is astonishing. Each part of the body has to change its shape and size to match every other part. There's no central blueprint for the construction of an adult human. Each cell has to decide for itself, using nothing more than chemical signals and its own network of genes, RNA molecules, and proteins.

As Hirschhorn's list of genes has grown, he and his colleagues have searched them for patterns. They turn out not to be a random assortment. “Most of the action is at the growth plate,” Hirschhorn said.

Growth plates are thin layers of cells located near the ends of limb bones. In children, some of the cells in the plates produce signals, which then trigger neighboring cartilage cells to multiply. As the cells divide, the bones get longer. Eventually the cartilage cells change, producing bone
instead. They finally commit suicide, tearing themselves open to dump out chemicals that make the surrounding bone even harder.