IT ALL STARTED WITH PETUNIAS. In the mid-1980s, Richard Jorgensen and Carolyn Napoli were working as plant geneticists for an Oakland biotechnology startup that specialized in boosting agricultural yields—increasing frost tolerance with a sort of antifreeze bacteria called Frostban and quickening the ripening of fruits, among other advances. Yet if such improvements were apt to delight farmers, they didn’t always impress potential investors. So the researchers decided to try something more obviously spectacular: creating an extraordinary flower. They chose petunias (Petunia hybrida) because of the plant’s large, colorful blooms and because even then, early in the history of genetic research, scientists had developed sound methods for introducing genes into petunia cells.

In the laboratory, petunias can be grown from single cells, so Jorgensen and Napoli inserted into leaf cells a gene known to produce large amounts of the protein responsible for the flower’s purple pigment. They nurtured the cells into full-grown plants, then transplanted them into soil in a greenhouse. But when the blooms appeared, they were white, not vividly violet. Adding a purple pigment gene had somehow caused the plants to make less of the hue. After ruling out an experimental mishap, they realized an unknown process must be at work.

During the next several years, Jorgensen, Napoli and other plant researchers began to unravel the mysteries of a phenomenon they dubbed co-suppression, a form of gene silencing. But remarkable as it may now seem, the discovery had little impact outside the world of plant research until 1998, when a small team of scientists published a paper detailing a similar type of co-suppression they had discovered in a tiny worm. Interest in this type of gene silencing grew exponentially, and today, 20 years later, the same mechanism that drained the color from petunias is being tested in numerous human clinical trials. It appears capable of remarkable things.

Now known as RNA interference, or RNAi, the mechanism has already transformed the way geneticists figure out the function of genes, sparking “a revolution in our understanding of basic biology,” says Judy Lieberman, a biomedical researcher at Harvard Medical School. But the real excitement involves what RNAi could do outside the laboratory, potentially spawning a vast pharmacopoeia that could selectively eliminate harmful proteins produced by wayward genes in difficult-to-treat diseases.

Unlike gene therapy, which attempts—with limited results—to cure disease by replacing defective genes with properly functioning ones, RNAi allows researchers to tap a pathway that primitive organisms use to turn off invading viruses. Because the workings of the mechanism are natural to the cell, RNAi is theoretically much easier to implement than gene therapy, less invasive (because you’re not actually altering a person’s DNA) and has fewer potential side effects. What’s more, if there are problems, it can be washed from a person’s system.

If RNAi works as researchers hope, it might curb cancer genes; inflammatory genes associated with Crohn’s disease and inflammatory bowel disease, among others; and even genes that cause high cholesterol. Already, there are clinical trials of treatments for AIDS, acute renal failure, respiratory syncytial virus and the wet form of age-related macular degeneration.

Not that there aren’t some very real obstacles, such as simply being able to get a drug carrying an RNAi molecule to the right place in the body and avoiding a massive immune system attack against foreign genetic material. Yet while the hype is huge, the research so far is convincing.

ALTHOUGH KNOWN TO RESEARCHERS FOR DECADES, RNA had always been considered a mere servant to the more fundamental DNA. Though both kinds of nucleic acid are made of strings of nucleotides, the building blocks of the genetic code that determines every individual’s unique makeup, RNA generally has just one strand of code, while DNA has two. Encapsulated in the cell’s nucleus, DNA holds an organism’s entire archive of genes. To tap that archive, the organism creates RNA, a complementary string of nucleotides that is a copy of a section of DNA code. Exiting the nucleus, the RNA—in this capacity, called messenger RNA, or mRNA—enters the cytoplasm, where the code is translated into proteins.

By the 1990s, scientists had begun to suspect that RNA might play another important role. Craig Mello, at the University of Massachusetts, and Andrew Fire, then at the Carnegie Institution of Washington and now at Stanford University, were intrigued by studies of worms showing that injected RNA could sometimes interfere with the normal protein production coded by a particular gene. So they decided to inject two forms of RNA into Caenorhabditis elegans, a millimeter-long worm often used as a simple model of human disease. The first form was the better-known single-stranded RNA, while the second was a double-stranded cousin (dsRNA) found naturally only in viruses, in which the second strand contains the complementary code sequence of the first (both strands differ somewhat in structure from DNA).

Mello and Fire used this method to introduce extra copies of certain genes into the worm and then tested whether its behavior and appearance had changed. They hypothesized that the genes they had injected would be turned off. In fact, protein production associated with the genes carried by the double-stranded RNA was almost nil. The shutdown was powerfully specific, much more so than that elicited by the single-stranded RNA. It affected only those genes targeted, and it was easy to elicit. They called the effect RNA interference.

Mello and Fire described their worm experiments in the journal Nature, detailing research for which they were awarded the 2006 Nobel Prize in Physiology or Medicine. But it was only later that they and other researchers discovered how RNAi shuts down protein production. It turns out that double-stranded RNA attaches to a cell enzyme called Dicer, which chops the dsRNA (which the cell thinks came from an invading virus) into little pieces. A complex that contains the enzyme Argonaute 2 attaches to the dsRNA. Argonaute 2 splits those pieces into two single strands; one strand remains bound to the complex and eventually finds its corresponding messenger RNA. That mRNA, without this interference, would deliver the genetic code for the gene in question to the cytoplasm’s protein-making machinery, and the protein coded by the gene would be produced. Instead, Argonaute 2 cleaves the mRNA, rendering it useless. Even tiny amounts of dsRNA are enough to slam the door almost completely on protein production.

RNAi also works in fruit flies, plants, zebrafish and other lower organisms, but for several years that seemed to be as far as it went—no one could get RNAi to work in higher organisms. Double-stranded RNA injected into mammals appeared to turn off all genes. But everything finally changed in 2001 with the publication of a paper in Nature by Thomas Tuschl, a co-founder of Alnylam Pharmaceuticals in Cambridge, Mass. He knew that most RNAi experiments used long strands of dsRNA that strung together hundreds of nucleotides. But Tuschl and others had had success with shorter strands, especially in the fruit fly, so he decided to try that approach in mammalian cells.

Eventually it worked. Tuschl found that to trigger RNAi in a mammal’s cell, the physical structure of the double-stranded RNA molecule—known as small interfering RNA, or siRNA—must be precisely constructed. It had to be short, just 21 nucleotides in length, with an overhang of two nucleotides on one or both ends. Using such a molecule in mammals, Tuschl was able to switch off specific genes. “This made what had been an interesting biologic phenomenon in worms relevant to all of us in the medical profession,” says Johannes Fruehauf, vice president of research at Cequent Pharmaceuticals, another RNAi company in Cambridge. “Suddenly there was the prospect of using this process to make a drug.”

MANY DRUGS TRY TO DEACTIVATE disease-causing proteins. For example, scientists have engineered small molecules that bind to the active part of a cancer-causing protein and disable it. But only a relatively small number of proteins, probably no more than a few thousand, are treatable by these “small molecule” drugs. Other proteins tend to be inaccessible, with chemical structures not easily targeted. In contrast, with RNAi it’s theoretically possible to design a drug that could turn off any of the 30,000 or so human genes—each of which normally codes for a different protein. “RNAi opens up the possibility that the whole universe of genes becomes ‘druggable,’” says Harvard Medical School’s Lieberman.

There’s another potential advantage. Because many drugs are designed to knock out or alter a particular protein, researchers have to consider the target’s physical structure and model a drug that fits it perfectly. Even then, there’s a chance the drug could react with other, similar proteins. With RNAi, protein targeting becomes both simpler and more precise. Suppose a researcher wants to eliminate production of a protein associated with a particular gene. He could systematically test 21-nucleotide sections of that gene with corresponding dsRNA until he finds one that effectively silences the gene. “This gives you a ready-made drug,” says John J. Rossi of the Beckman Research Institute in Duarte, Calif. “It’s very easy to design siRNA for virtually any gene of interest. And with the whole genome now sequenced, we can identify a target instantaneously.”

A final advantage is that rather than attacking a problem protein, RNAi addresses disease at a fundamental level, turning off the gene that codes for production of that protein. “With RNAi, you’re preventing the protein from even being made, versus trying to mop up the protein’s activity,” says Akshay Vaishnaw, vice president of clinical research at Alnylam.

In 2002 Lieberman began a study attempting to cure HIV in a petri dish of human cells. First, she targeted a protein called CD4, a receptor on the outer membrane of human immune cells, to which the HIV virus attaches itself and into which it inserts its genetic material. Lieberman used siRNA to silence the gene that coded for CD4 and found that without CD4 receptors to bind to, the HIV virus was four times less able to enter a cell. This could halt the spread of the virus.

Next, Lieberman tried a different tack, targeting the virus itself. Using RNAi, she turned off a pivotal HIV gene, called gag, that codes for proteins essential to the virus’s structure. She found that this sharply reduced the amount of HIV in the cells, apparently because new copies of the virus could not be made without the gag gene. Finally, to see whether siRNA could treat infection as well as prevent it, she infected the cells with HIV and then dosed them with siRNA. That worked too.

But treating cells in a petri dish is a far cry from achieving the effect in humans. To see what was possible in a living creature, Lieberman moved on to mice. Because overabundance of a protein called Fas is often involved in liver disease, she designed siRNA for the gene that makes that protein and was able to protect mice with hepatitis from liver failure—the first time siRNA had alleviated disease in an animal.

However, the extraordinary measures she had to employ—injecting a dose of siRNA equal to one-fifth of the mouse’s blood volume at high pressure—would never work in humans. But the following year, a group from Alnylam managed to inject siRNA into a mouse at a normal pressure and volume to silence a gene called apoB, which causes high LDL cholesterol. The researchers altered the siRNA slightly, chemically stabilizing the strands and attaching them to a molecule of cholesterol to make it easier for them to pass into cells. This approach also kept the siRNA from being quickly degraded by enzymes—normally a problem—and the treatment had the desired effect, entering the liver and slowing production of LDL. Alnylam followed up with a 2006 study in primates that also reduced cholesterol.

Yet this success hasn’t really solved the biggest problem of using RNAi to treat human disease. Delivery is the elephant in the room, and progress has been slow. In addition to Alnylam’s cholesterol work, Lieberman is developing an approach using antibodies, while Rossi at the Beckman Research Institute is trying to attach siRNA to a molecule called an aptamer that can bind to various parts of a cell. But none of these has been tested in people yet.

Until the delivery issue is sorted out, researchers say, they’re left to pluck the low-hanging fruit, targeting tissues to which siRNA can be applied directly rather than depending on systemic delivery through the bloodstream. For example, three years ago, Acuity (now called OPKO Health) began the first clinical trial of an siRNA for the wet form of age-related macular degeneration (AMD). Researchers injected siRNA for the gene that codes for the protein VEGF directly into the back of the eye. VEGF causes leaky blood vessels to grow in the eye, damaging the macula—the part of the retina with the most vision cells—and harming the ability to see fine detail. In these small trials, the siRNA proved effective in reducing expression of VEGF. Now, in a larger Phase III trial (the first for an RNAi therapeutic), the siRNA drug is being compared with another AMD drug already on the market.

The lung, reached via inhaled drugs, is also an easy target. Alnylam is conducting a clinical trial that attempts to silence a gene important for the replication of the respiratory syncytial virus (RSV). The company’s main drug, ALN-RSV01, was found to be safe and well tolerated in Phase I trials, and it’s now in Phase II trials to test how well it knocks down the virus in the upper respiratory tract.

For RNAi to live up to its hype, however, researchers will have to find a way to go beyond direct delivery. There are other issues too, not least the worry that siRNAs could trigger a damaging immune response in humans. And what would happen if someone took RNAi drugs for a lifetime, a requirement for many diseases such as HIV or hepatitis? Still, hopes are high. “All of us developing RNAi-based drugs think that this will add a whole new class to our arsenal,” says Fruehauf of Cequent Pharmaceuticals.

Rossi hopes his HIV treatment will be one of those. He has just recruited the first of five patients for a small trial. The patient has lymphoma and AIDS, and as part of the treatment for lymphoma, he’ll receive a blood stem-cell transplant. But before the transplanted cells go into the patient, Rossi will add an anti-HIV siRNA, which “we hope will make the other drugs the patient is on more potent. That could let us lower the dosage of those drugs, or enable patients to go on drug holidays.”

Rossi will follow the patients in his trial indefinitely to monitor whether siRNA continues to combat the virus. But he also thinks RNAi will benefit AIDS patients in another major way. Because HIV mutates rapidly, drugs that were once effective eventually lose potency. RNAi could offer an important answer to this persistent problem. “You could just make a new RNA that would counter the resistance mutation,” he says. “It would be so easy to change the drug. We might even be able to develop an injectable once-a-month treatment using siRNA that would take the place of conventional drugs.”