A Garden of Mutants

Flowers sprang up suddenly 150 million years ago, and no one knows how. But Elliot Meyerowitz hopes to find out, with a private collection of monster blooms.

By Carol Fletcher|Tuesday, August 01, 1995
A man who has raised a million of the same type of flower might be excused for being a bit indifferent about his latest bloom. But Caltech geneticist Elliot Meyerowitz shows off each blossom with all the affection of a new father passing around baby pictures. And what peculiar babies they are.

One flower has leaves where the petals should be. Another has stamens protruding from strange places. Now, here’s a pretty one, says Meyerowitz, displaying the tiny blossom against his thumbnail. Look at all the extra petals. Tell me if you’re getting tired, he adds. I could stay in here all day.

Here is an old-fashioned meat locker, its meat hooks replaced by ceiling fans and its walls lined with trays, stacked cafeteria-style, each teeming with wispy plants. By studying the blossoms of these unremarkable little weeds--a commercially useless variety of mustard known as Arabidopsis thaliana--Meyerowitz and his colleagues are unlocking the secrets behind the development of flowers.

Meyerowitz and his colleagues have identified the master genes that tell certain of the undifferentiated cells on the stem tip to become petals, others to become stamens and carpels (the sex organs), and others to become sepals (the hardy green leaves on the outside of a flower). By altering these genes, he can design new flowers--blossoms made entirely of green sepals, for instance, or flowers with their parts in all the wrong places. The goal is not merely to dazzle the garden club. Eighty percent of all food comes from flowers, says Meyerowitz. If we really understood how they develop, we might improve crop yields.

Meyerowitz is soft-spoken but articulate, with a disarming way of discounting his work. An eight-year-old could do this experiment on the kitchen table, he says as he charts on a blackboard work published in prestigious journals, careful not to interfere with a picture of a lizard his five-year-old son has drawn there. His conversation slips easily from the ancient origins of maize to an esoteric bit of research on X-ray mutations in Arabidopsis. (The manuscript was seized in Germany at the end of World War II, but the Smithsonian was cleaning a warehouse once and sent it to me. You can look at it if you like, he adds hopefully.) His modesty doesn’t detract from the seriousness of his ambition: to shed light on a great evolutionary question, what Darwin called the abominable mystery.

According to the fossil record, 150 million years ago there were no flowers. Then, like that, Meyerowitz says, snapping his fingers, flowers appeared. And no one knows how. It didn’t happen that long ago. True, there were a lot of dinosaurs milling around and no people, but not a lot has changed since then. Insects are pretty much the same. But land plants have shifted almost completely from nonflowering to flowering.

How the flower suddenly arose--or how, for that matter, the brain, eye, or any morphological novelty arose--is one of the central questions of evolution. At the heart of evolutionary change is change in the genes that regulate development. Thus researchers like Meyerowitz are attempting to crack the mystery by figuring out which genes were added or modified in ancient plants to bring about the development of the flower.

Developmental genes have long been better understood in animals than in plants. During the 1950s, in a lab two floors above Meyerowitz’s meat locker, geneticist Edward Lewis became curious about embryonic development. He wondered how a mass of undifferentiated cells knows which should become a leg, say, or an abdomen. By looking at fruit flies in which organs were growing in the wrong places, he began to identify the genes that dictate organ identity. Mutate one of these genes, and a leg will grow where an antenna should be. Biologists cloning such genes in the 1980s discovered that many shared a similar stretch of DNA, which was dubbed a homeobox. The discovery of homeobox genes in flies soon led to the discovery of homeobox genes that orchestrate development in mice and even humans.

But no one found genes that play a similar role in plants until researchers hit upon Arabidopsis, botany’s answer to the fruit fly. Half a century ago German botanist Friedrich Laibach insisted that Arabidopsis was the perfect plant for studying genetics because it had a short generation time and few chromosomes and produced thousands of seeds per plant. But few heeded his advice until the late 1970s, when plant biologists became interested in gene technology and poked around for a plant conducive to genetic manipulation.

Arabidopsis seemed the perfect candidate. In the mid-1980s Meyerowitz discovered that it has the smallest known genome of any flowering plant. (The corn genome is 30 to 40 times larger.) Moreover, Arabidopsis had almost no junk DNA--the repetitive DNA that researchers get tangled up in when they try to find their way around human or corn chromosomes. It also requires little space or expertise to grow. It’s the people’s plant, says Meyerowitz. In one meat locker, we are growing tens of thousands of plants. To grow that much maize, you’d need a field, a tractor, an irrigation system, and guys in overalls. You’d have to be a land grant university.

By the late 1980s an explosion of Arabidopsis research began. One team discovered a gene that turns on the plant’s photosynthetic machinery. Others found genes responsible for disease resistance. Chris Somerville, at the Carnegie Institution of Washington in Stanford, California, introduced three bacteria genes into Arabidopsis and produced a plant that makes biodegradable plastics. Virtually any question plant biologists are working on, they are working on in Arabidopsis, says Somerville. Among the most striking discoveries have been the genes involved in creating flowers.

The flowers of Arabidopsis, like thousands of other wildflowers, are each made up of four concentric circles, or whorls. The outer whorl is composed of leaflike sepals. The sepals surround the petals, which make up the second whorl. In the third whorl are stamens (which make pollen). And in the center of the flower, or fourth whorl, are carpels (which combine to form an ovary). All these organs arise from a group of undifferentiated cells called the meristem, near the tip of the shoot. But how do cells in the meristem know which organ to become?

Meyerowitz thought he might find an answer the same way Edward Lewis had in flies: by looking at mutants with organs in the wrong places. He wrote to researchers working with Arabidopsis and asked them to pass along plants with bizarre flowers. Meanwhile, he and his colleagues created their own crop of mutants. They soaked seeds in a chemical mutagen, planted them, and crossed the new plants with themselves to ensure that both copies of any mutated gene would be defective. Then they combed through each of the 20,000 progeny looking for mutant flowers.

After several years they identified three common classes of mutations in which flower organs grew in the wrong places. In one the flower sported sepals where petals belonged, and carpels instead of stamens; thus, from the outside in, the organs went sepals-sepals-carpels- carpels. Each of the three common mutations seemed to be caused by the loss of a single class of genes.

To explain how these mutations could arise, Meyerowitz designed a model of gene regulation that was striking in its simplicity. It assumed that just three classes of genes determine organ identity in flowers. He called the genes A, B, and C. Suppose I’m a cell in a developing flower and I’m about to make an organ, says Meyerowitz. If I look within my nucleus and I see gene class A is on, it means I should become part of a sepal. If I see A and B are on, that means be part of a petal. Genes in class B and C mean stamen. And C alone means carpel.

Thus a normal flower is born--as long as one other criterion is met. Each gene class must act only in specific whorls. A should act only in the outer, first and second, whorls; B in the middle, second and third, whorls; and C in the inner, third and fourth, whorls. The choreography is directed by another set of genes--the cadastral, or boundary-setting, genes--that tell A, B, and C where to go on.

Meyerowitz’s model accounted for the three types of mutant flowers the researchers had found; each is caused by the inactivation of one gene class. For instance, suppose B is off. A cell in the outer whorl looks in its nucleus and sees A, as usual; it becomes part of a sepal. So far, everything is okay. But a cell in the second whorl, which would usually see A and B, now sees only A, so it too becomes sepal. In the third whorl, where a cell should see B and C, only C appears, so the cell becomes carpel. In the fourth whorl, only C is on, as usual, so the cell becomes carpel. The flower looks like this: sepal-sepal-carpel-carpel, which is the classic B mutation.

The model also accounted for A mutations, but only after Meyerowitz added a rule. If A is knocked out, C goes on where A should have been. A lot of the ideas in our model had been tossed around earlier, he says. This is what we added that really made it work.

In an A mutation, then, C acts not only in its usual spots (the third and fourth whorls), but also in A’s hangouts (the first and second whorls). So if you are a cell in the first whorl, instead of seeing A, you see C and become part of a carpel. Cells in the second whorl see B and C and become stamens. So do cells in the third whorl. In the last whorl, cells see C and become carpels. The flower looks like this: carpel-stamen- stamen-carpel.

Similarly, when C is knocked out, A reciprocates by leaping in and taking over. Thus, in the C mutation, cells see A in the first whorl and become sepals. In the second whorl, they see A and B and become petals. In the third whorl, instead of B and C, they see B and A, so they become petals here too. (The last whorl, for some reason, seems to vanish.)

There’s a further consequence of the C mutation: evidently, one of the C gene’s responsibilities is to tell the plant when to stop making whorls. When C is knocked out, the plant keeps churning out extra whorls, in an indefinite sequence of sepal-petal-petal, sepal-petal-petal, and so on.

Meyerowitz’s model was based on single gene mutations, but it allowed him to predict how plants with double and triple mutations would look. It predicted, for instance, that if both A and B were knocked out, C would be expressed in every whorl, and the entire flower would become carpels. And in fact, when Meyerowitz crossed single-mutated plants to produce the double mutant, that is exactly what happened.

When A and C were both knocked out, they couldn’t cover for each other, so the flower grew leaves in the first and last whorls. In the two middle whorls, cells saw only B, so they became organs intermediate between petals and stamens. When A, B, and C were all knocked out, the flower had no normal organs but all leaves instead. Because they lack C’s activity, these two mutant flowers also have extra whorls.

It came as an incredible surprise, says molecular geneticist Enrico Coen, that such a small number of simple genetic switches were involved in all this. People had expected it to be very complex. About the time Meyerowitz came up with his model, Coen and his colleague Rosemary Carpenter, at the John Innes Institute in Norwich, England, were piecing together a model of flower development in a distant relative of Arabidopsis, the snapdragon. The models turned out to be identical. Indeed, in every species studied since, including tomatoes, petunias, and tobacco, flower development seems to be regulated in pretty much the same way.

In retrospect, this makes sense. Despite their varied shapes and colors, flowers all follow the same basic blueprint: sepals on the outside, followed by petals, stamens, and carpels. (Some flowers are missing certain organs--tulips have no sepals--but the parts they do have are always in the same order.) The only known exception is a parasitic plant discovered recently in Mexico, Lacandonia schismatica, which has stamens in the center.

The similarity among plants is good news for those who hope to apply what is known about flower development in lab plants to engineering better crops. If you want to change the radial pattern in corn, says Coen, you have a pretty good idea where to start. It has also been a boon to basic researchers. When researchers at the Max Planck Institute in Cologne, Germany, first isolated a B gene in snapdragons, they sent it to Meyerowitz, who used it as a probe to find the B gene in Arabidopsis. Meyerowitz later found a second gene involved in producing B activity.

We have now cloned all the B and C genes, says Meyerowitz, and when we express them in the wrong places, they do exactly what the model predicts. Other labs have cloned several A genes. Scientists can manipulate these genes to design new flowers.

Meyerowitz and others have also begun identifying the cadastral genes, which tell the A, B, and C genes in which whorl they should act. And he has uncovered genes that specify organ number; inactivate one of these genes, and the flower makes five petals instead of four.

Although recent progress has been rapid, many mysteries remain. No one knows how A, B, and C genes allow cells to form organs. It’s easy to say A plus B equals petal, says Meyerowitz, but when you do, you are sweeping a lot of things under the rug. A petal is a complex organ with many cell types, each derived from a different pattern of cell division and elongation. Somehow all those cells have to be organized by a combination of A and B--that’s not many genes, says Meyerowitz. There must be a large cascade of events downstream that allow the cells to form an organ.

Upstream lies another mystery: What sets the whole process of flower making rolling? In 1990, Coen and Carpenter found a gene that acts early on in flower development, telling the meristem to become a flower rather than a shoot, and turning on the A, B, and C genes. Knock out this meristem-identity gene, and a stem grows instead. But how does this gene know when to turn on?

That’s the age-old question of floral induction, says Meyerowitz. How do plants know when to flower?

Environmental conditions, such as temperature or day length, probably trigger meristem-identity genes to begin making flowers. By studying the regulation of meristem-identity genes, says Meyerowitz, we now have a way of getting to the heart of the interaction of plants with the environment.

Meyerowitz is fascinated by such questions; clearly this is a man who loves flowers. In his yard, amid his wife’s hybrid tea roses and camellias, he has planted teosinte, the Mexican grass that gave rise to corn, and several unsightly Madagascar palms--spiny, twisted plants that he finds developmentally curious.

From an evolutionary point of view, his wife’s flowers are the real curiosities. Like almost all ornamental flowers, they are genetic freaks, the product of centuries of crossbreeding.

So are most crops. The ancestor of corn looks very different from corn, says Meyerowitz. But somehow Native Americans recognized the plant’s potential and painstakingly bred it for thousands of years until it became maize, one of the great staffs of life on this planet. Southwest Asians, in the same way, created wheat; native South Americans gave us tomatoes and potatoes.

Such crossbreeding is slow because it randomly mixes all the plant’s thousands of genes. Then breeders must sift through the progeny, selecting those with the best genes for further crossing. If we really understood how plants developed, says Meyerowitz, we could write a list of 200 genes we wanted to change, and do in the lab in a year or so what took our ancestors thousands of years.

First researchers will need to understand better the many genes involved in development, root to shoot. It’s unlikely that all the genetic programs will be cracked as readily as those in flowers. Yet in the world of Arabidopsis research, there is a growing feeling of exhilaration--a sense, as Chris Somerville puts it, that any problem in developmental plant biology is manageable. The key may be identifying all the plant’s 20,000 genes. Last year alone, Somerville and his colleagues sequenced fragments from about half these genes. Each one was entered on an international computer network accessed by some 2,000 Arabidopsis researchers. Work is going on in so many areas of Arabidopsis, says Somerville, that any gene you find, chances are someone will find it useful.

Sequencing the entire Arabidopsis genome would cost less than sequencing a small human chromosome, says Meyerowitz, who has been promoting the project for years. Once done, it would provide the tools to sequence quickly more complex plants, such as corn, which has a genome even larger than that of humans. This in turn could speed progress in engineering new crops.

Still, Meyerowitz’s real obsession is not tinkering with evolution but understanding it. He would especially like to know how flowers popped up so suddenly in the fossil record. The miracle plant, a primitive Angolan plant with two long, straplike leaves, may provide a clue.

A favorite of botanical gardens because of its bizarre appearance, the miracle plant has some flowerlike parts but no real flowers. Meyerowitz and his colleagues are trying to find out whether the genes that control flower development in Arabidopsis are present in miracle plants, and, if so, what they do. By looking at what these genes do, he says, we might learn how their function changed to give rise to flowers.

Already Arabidopsis has revealed a few evolutionary secrets. The genes that control flower development work by producing a protein that binds to DNA on another gene, turning it on. Many of the flower development genes share a similar stretch of DNA--the stretch that produces the DNA- binding part of the protein. This stretch has been dubbed a MADS-box. Such similarity among flower genes suggests that they arose from a common source. One way development evolves, says Meyerowitz, is to start out with a gene that does something you like, then duplicate it and slightly alter its function, until you have a whole set of genes.

MADS-box genes are similar in function, but not identical, to the homeobox genes that regulate animal development. The differences suggest that animals and plants, although they evolved regulation of development in like ways, used a different set of proteins to do it. Plants and animals had the same toolbox, says Meyerowitz, but used different nuts and bolts. He hesitates. At least that’s the way it looks so far.

Meyerowitz hopes that by comparing development in plants and animals, we can begin to answer some of the questions we tried to answer by searching for life on Mars. If we had found life on Mars, it would have been earthshaking, he says. We would have learned which aspects of life are necessary and which evolved by chance. We can’t compare ourselves with aliens, but we can compare animals with plants, which have evolved separately for 2 billion years.

If you really wanted to learn how multicellular development came about, says Meyerowitz, you would take a single-celled organism, put it in a lab under conditions advantageous to evolving a developmental system, then wait 2 billion years to see what happened. That experiment was done. Now we’re coming back to have a look.
Comment on this article