Discover, May 2007[Link]
The town of Bar Harbor, just off the rockbound coast of Maine, is home to spectacular granite cliffs, herds of barking seals, and about 5,000 human residents who live there year-round. Like the rest of us, some of these humble citizens will enjoy long, happy lives, and some will die all too young. According to national statistics, about 1,400 of them will die of heart disease, and 1,100 will die of assorted cancers. Others will struggle with chronic, debilitating diseases. About 1,600 of them are obese. Some 500 suffer from diabetes, and another 150 have osteoporosis. Environment and behavior have their roles, of course, but the different fates of people in Bar Harbor have a lot to do with the different kinds of genes they carry.
In this era of the human genome map, it would seem a simple matter to pinpoint the bit of DNA responsible for each disease and use that knowledge to find a cure. That is indeed what most genetic researchers are doing. After all, Gregor Mendel discovered the laws of genetics simply by crossing pea plants in his garden. But there is a catch. The difference between a wrinkled pea and a smooth one depends on a single gene. Underlying the most common human diseases, however, are entire networks of genes. Some of those networks contain dozens of genes; others contain hundreds. Trying to understand a disease from a single gene is usually like trying to understand an entire magazine article from a single word.
Uncovering the many genes in a network has proved to be maddeningly difficult, and humans are a terrible place to look for them. We each have our own unique combination of genes, and when we have children, we combine two sets of genes into still more combinations. The ideal way to identify a gene network in humans would require an impossible experiment: Take two families, each with dozens of identical twins, and have the families interbreed, combining the same sets of genes together over and over again. Once the children grow up, see who gets sick with a particular disease and who does not. Look at which genes they inherited from their parents. Then find more families with dozens of identical twins, and run the experiment again and again.
No one can actually do this experiment, of course. But something remarkably close to it is taking place in Bar Harbor—on the outskirts of town, to be precise, near the foot of Dorr Mountain, within a sprawling brick compound known as the Jackson Laboratory. When I enter the lab’s front door, the receptionist halts me with a single question: “Have you handled a mouse or any other rodent in the past five days?” I haven’t, so I am allowed into the lab’s remarkable inner sanctum—the Phenome Room, a small windowless space the size of a studio apartment. To enter, I must don paper boots and a lab coat, a surgical mask, and a cap. I step in and I am surrounded, floor to ceiling, with racks of clear plastic boxes containing, collectively, 5,000 mice.
Like the similar-size human population in the surrounding town, each mouse has its own lifestyle and behavior, whether learned or inherited. In one box, the mice are asleep in tight balls. In another, they are grooming one another. In another, they are wrestling. Likewise, each has a distinct genetic profile that results in physical characteristics. Some mice are chunky and some are slim; some are chocolate-colored; others butterscotch or cream. Some have weak bones and others have strong ones. Some live to a ripe old age of three years, and some die within months. They seem like a pretty ordinary collection of rodents, but their DNA is quite special. The animals belong to 40 strains, each of which has been inbred for so long that its mice have identical genes.
“The diversity across these 40 mouse strains is as great as or greater than that of the entire human population," Jackson geneticist Kenneth Paigen tells me, but within each individual strain the DNA is incredibly uniform. This little community of genetically distilled mice is exactly what is needed to crack the mystery of hereditary disease.
Phenome is a twist on genome. The genome is the entire collection of genes in an organism. The phenome is the entire collection of traits those genes give rise to. Scientists who work in the Phenome Room chronicle each mouse’s life in excruciating detail. They X-ray the mice to track their growth of muscle and fat. They slip the tails of the mice into miniature cuffs to measure their blood pressure. They put the mice on treadmills to film their gait. They record how the mice sleep and how much they eat. They draw blood to measure the levels of dozens of compounds.
The Jackson Laboratory mouse population is the largest collection of its kind in the world. It takes time to create so many inbred lines, and workers at Jackson have been doing it longer than just about anyone else, since the lab’s founding in 1929. What is true for the mouse, they are finding, is often true for people too. Humans and mice share 99 percent of their genes, and the two species form genetic networks that are remarkably similar. By going gene hunting in the Phenome Room, Jackson Laboratory scientists hope to steer their colleagues away from the search for single genes in favor of hunting for gene networks. If it works, that could spark a true revolution in the understanding and treatment of heart disease, cancer, and other major DNA-mediated killers.
“What we’re trying to do is change the way that people think about diseases,” says Gary Churchill, a statistical geneticist at Jackson Laboratory. “They’re out there looking for broken genes, but the idea of broken genes just doesn’t make sense. We’re saying, ‘Hey, reality isn’t like that.’”
The search for single disease-related genes has certainly had some major successes. One of the greatest came in 1989, with the discovery of a gene linked to cystic fibrosis. Cystic fibrosis causes inflammation in the lungs, almost like a bad case of flu, and is usually fatal by age 35. Unlike the flu or meningitis, however, cystic fibrosis is not caused by a virus or other pathogen. It runs in families.?
Studies on families with cystic fibrosis showed that the disease gene is passed along in a fairly simple way, much as in Mendel’s smooth and wrinkly peas. Only if a child inherits a defective gene from both parents does he or she get sick. Even so, it took years to find the actual gene. First scientists scanned distinctive stretches of DNA that are spread out across the genome like mile markers on a highway. They then found one particular marker more often in family members with cystic fibrosis than in healthy ones. When they began to search the DNA around the genetic marker for candidate genes, they uncovered a link between cystic fibrosis and a gene for a channel that pumps chloride (electrically charged chlorine atoms) through cell membranes. Mutations to that gene prevent cells from making the channel, disrupting the balance of fluid in the lungs. If scientists can find a way to insert working versions of the gene into lung cells, it may be possible to reverse the disease.
Even in the instance of cystic fibrosis, though, the story is not so simple. The single chloride mutation does not always doom a person to the illness. Many people possess this mutation, “but the clinical outcome is vastly different,” says Rick Woychik, director of the Jackson Laboratory. “Some people have a poor outlook, and others actually live till a ripe old age. So why? One reason is genetic networks.”
Genes do not work in isolation. Many of them make proteins that switch other genes on or shut them off, or that come together to create a giant molecular machine. Other proteins relay signals through the body, like a line of people playing the telephone game. Gene networks tend to be remarkably robust. Even if one gene is hobbled by a mutation, the network still generally manages to do its job. “There are other genes that participate in the network, and they can compensate for the absence of this gene,” Woychik says.?
Because of this, single-gene diseases are rare and affect relatively few people; common diseases tend to be network driven. “If you think about all the major killers—cancer, heart disease, diabetes, stroke—they’re not single-gene diseases,” says Paigen. For instance, if researchers find one particular gene associated with stroke—the kind of discovery that used to make headlines—it will typically have only a small effect on a person’s chances of getting the disease.
In a sense, today’s disease hunters are finally making sense of clues that emerged a century ago in the work of Jackson Laboratory’s founder, Clarence Cook Little. Before there was a high-rise community of inbred mice in Bar Harbor, there was the bathtub full of pet-store rodents that started it all.
As an undergraduate at Harvard in the early 1900s, Little joined in the effort to push genetics beyond Mendel’s peas. One of his professors urged him to study mammals, specifically mice, since they shared many traits with humans but were fast breeding and easy to rear in a lab. So in 1907 Little paid a visit to a local pet store and started raising mice in his tub. He bred the mice incestuously, pairing brothers and sisters in each generation. Over time, each lineage shared more and more gene variants, or alleles. After about 20 generations, they were as similar to one another as are identical twins.
Once Little had created several inbred strains, he could cross yellow mice with black mice, for example, and observe how many pups grew coats of various colors. These studies helped show how Mendel’s simple rules could account for complicated patterns of heredity. Little realized that it takes several genes to determine coat color and learned how to estimate how many genes were involved in a trait based simply on comparing the descendants of his hybrids. In one of his most important experiments, he bred two strains of mice to see how their offspring accepted or rejected tissue grafts. In 1916 he calculated that at least 14 genes were involved. This was a radical idea at a time when nobody even knew what a gene was. “It was the ultimate case of chutzpah,” says Paigen. “Amazingly enough, he turned out to be right.”
After getting his Ph.D., Little served as president of the University of Maine and then the University of Michigan. But his controversial tenure at Michigan (he wanted to ban drinking at fraternities, for one thing) got him fired in 1929. Little promptly persuaded several automobile tycoons to bankroll an independent lab in Bar Harbor where he could go on studying his inbred mice. Jackson Laboratory—named after Roscoe Jackson, a cofounder of the Hudson Motor Car Company—was born.
Little had the misfortune of launching his grand project on the eve of the Great Depression. The funds from his patrons soon dried up. When his scientists went from one room to another, they had to unscrew the lightbulb and take it with them. But Little came up with a way to stave off bankruptcy: He would sell his mice. Inbred mice were rapidly becoming the favorite lab animal for medical studies, and the Jackson Laboratory mice were recognized as the world’s best. Soon Little was doing a booming business in the mouse trade. The tradition continues today. Jackson Laboratory ships 2.5 million mice to labs each year, making it one of the world’s leading suppliers.
Over the years, Little and later scientists expanded the collection of inbred mouse lines. The lab now houses three-fourths of the world’s known mouse varieties. In 1999 Ken Paigen, then the director of Jackson Laboratory, had his team examine a thousand mice from 40 inbred strains. They measured 500 different traits in the mice, from their white blood cell count to the shape of their brains to how salty they liked their water. The mice had a huge range of variations for many of the traits, and the variations stemmed from their genes.
With this newly documented diversity, the scientists were able to explore more thoroughly the genetic networks that underlie these complex traits—almost as if they went from hunting in the dark with a flashlight to hunting in a room full of overhead lamps. For instance, Beverly Paigen, Ken’s wife, has spent the past few years studying genes that influence levels of “good” cholesterol (high density lipoprotein cholesterol, or HDL). One strain of mice has high levels, and another has very low levels. Paigen bred the two strains together and then raised their pups. Once she found genetic markers strongly linked to high or low levels of HDL, she searched around the markers for the genes that were at work. (The hunt has become much faster now that other researchers have sequenced the entire mouse genome.)
Beverly Paigen expected that she would simply build on the enormous amount of work already done in this area, finding a mix of new genes and genes that had already been identified in other experiments. “I thought we would find maybe half and half,” she says. But she was wrong. There were far more novel genes than anyone had expected: “We’re finding mostly new stuff.”
Coming up with a list of genes is a good starting point for mapping a gene network, but it is only a starting point. Imagine if you were an alien trying to understand people on Earth solely by studying satellite pictures. You notice that they put up umbrellas whenever it is cloudy, yet the clouds themselves do not directly cause the umbrellas to go up. Unless you understand that there is rain sprinkling down, the umbrellas will remain a mystery. This is the quandary that geneticists like Beverly Paigen face when they discover genes belonging to the same network. A gene may influence a trait directly, or it may influence another trait that influences the trait in question. Or—to make the puzzle truly migraine inducing—the gene may have different effects on both traits. Unless you understand how those genes work together, the list of genetic markers will remain a mystery.
Gary Churchill is leading Jackson Laboratory’s efforts to cut through that mystery. He is a statistician by training, and he is an expert at recognizing patterns in what may seem like random data. He looks for many markers that tend to turn up together in mice with a single trait, and he looks at whether each gene belongs to other groups that are associated with other traits, using his own statistical tricks to tease all of this apart. Once Churchill has identified potential links among the genes, he creates a mathematical model of the entire network and uses it to predict what sort of mice will be produced from various combinations of the genes. If his predictions hold, he knows he has made a successful model. If not, he can go back and fine-tune it.
Recently Churchill and his Jackson Laboratory colleagues decided to go after some big genetic game, the gene network that controls body weight. With 300 million people now suffering from obesity worldwide, fat has become a global epidemic. For years geneticists have searched for the genes that determine whether people gain weight easily or not, but it has been a frustrating experience. In the 1990s studies on a strain of obese mice developed at Jackson Laboratory guided Rockefeller University scientists to a protein they dubbed leptin. When they injected leptin into the obese mice, the mice lost weight. A major biotechnology corporation, Amgen, seized on the discovery, hoping to create a weight-loss drug. But they could not replicate the effect in humans.
Rather than focus on a single gene, Churchill and his colleagues decided to explore the entire weight-control network. They selected a big, lean strain of mice and mated them with small, fat ones. The offspring of this union grew to many different sizes and weights. Churchill and his team then measured how large the animals grew and how much of their body weight was fat versus muscle. They also measured how the fat was spread out on each mouse. Like us, mice tend to accumulate fat in certain places, like their haunches and their bellies. Finally, the scientists scanned the genome of each mouse for hundreds of markers to see which ones were linked tightly to each trait.
The map they came up with looks like a flowchart from hell. Churchill’s group identified a dozen sites in the mouse genome where genes are influencing the body weight of mice. But the genes have different effects. Some make mice large-bodied, and being big makes mice more likely to get fat. But they also found genes that had separate effects on both body size and fat levels. In some cases, the same gene could make a mouse both big and lean. Other genes influenced only how fat the mice were, with no effect on their body size. Still other genes determined the size of different fat pads. One region of mouse DNA appears to make mice fat overall while actually making the fat pads on their haunches smaller.
While networks like the one that controls body weight may be complicated, Churchill takes some comfort in the fact that they are not so complicated as to be incomprehensible. “The good news is that it doesn’t seem that everything interacts with everything else,” he says. The networks are small enough that it may be possible to understand all their parts, a depth of knowledge that should point to new treatments for disease.
In fact, Beverly Paigen argues that in some cases treating gene networks may be easier than trying to treat single genes. “Suppose you find only one gene, and that’s the cause of the disease,” she says. “Suppose you can’t get a drug to it. Suppose it’s just in a place in the body where it’s intractable. Now suppose the gene is in a network and there are other things in the network that interact with it. You might be able to use a drug target at another piece of the network.”
Despite the recent advances, the mice from the Phenome Room are just barely beginning to divulge their secrets. Jackson Laboratory scientists are still struggling to distinguish the effects of closely spaced genes that are linked to the same genetic markers, for example. The problem, Churchill explains to me, is that 40 strains of mice are just not enough: “If you want to go forward, you’ve got to go big.”
Instead of a Bar Harbor’s worth of mice, he needs the equivalent of a large city. He and other mouse experts around the country have therefore launched a project, the Collaborative Cross, to produce new inbred strains from lab-grown mice, each with its own unique combinations of alleles. So far, the collaboration has started 500 inbred strains. Churchill hopes to have 1,000 by the time they are finished in 2010. For the first time in history, he predicts, we will finally be able to see in crisp, sharp detail the genetic networks that allow us to live and cause us to die.
“We spent a hundred years trying to figure out what the parts are,” Churchill says. “We now have the part list. We can start to ask how the parts are assembled.”
Copyright 2007 Carl Zimmer