By the late 1960s, molecular biologists had erected an overarching explanation of how genes work–their substance, their structure, their replication, their expression, their regulation or control. Or at least they had done so in outline, for prokaryotes, the simplest single-celled organisms (which include bacteria), and for the viruses, called bacteriophages, that prey upon them. The leaders of the field were now looking to a far more difficult problem: doing it all over again for higher organisms.

What this new generation of molecular biology demanded, and what was developed in just a few years, was a set of methods for investigating and precisely manipulating the genetics of eukaryotes, including animals and plants. With reverse transcriptase, which was discovered independently by ­Howard Temin and David Baltimore in 1970, genes encoded in RNA could be read back into DNA. With Daniel Nathans’s and Hamilton Smith’s work on restriction enzymes, segments of DNA could be snipped out at chosen sites. In a rush, from laboratories chiefly at Stanford University, came ways to link together genetic material from disparate sources. “We will be able to combine anything with anything,” one senior scientist told me at the time. “We can combine duck with orange.” The initial purpose was to get at the most basic questions of cellular biology, to find out exactly what individual genes do and how they do it. Immediately, though, a shining hope dawned: that this toolbox could be carried from the laboratory to the clinic, to cure hereditary diseases caused by genetic defects. Already, some scientists were dreaming of gene therapy.

By 1970, some 1,500 genetically determined diseases had been identified in humans. Some show up in babies; others surface at puberty; a few emerge only toward the end of the victim’s reproductive life. Some can be held in check by dietary restrictions, a few by drugs. But most cannot be cured or even palliated by conventional medicine. Though almost all are rare, some extremely rare, collectively they were coming to be recognized as a burdensome and costly medical problem. Many are marked by gross mental impairment. Victims of Lesch-Nyhan disease, for example, suffer severe mental retardation. They must have their arms splinted, because otherwise they bite their hands and arms. They die in childhood or early adulthood. Though scientists had traced fewer than a hundred of these human diseases to specific genetic deficiencies, they began searching for ways to cure them by safely inserting correcting genes into people suffering from them.

They were still trying nearly two decades later, when on September 29, 1999, the front page ofthe Washington Post carried the headline “Teen Dies Undergoing Experimental Gene Therapy.” Jesse Gelsinger was 18, a recent high-school graduate from Arizona who had a potentially fatal genetic disease. He was one of 18 patients taking part in a trial at the University of Pennsylvania. Viruses carrying a new gene had been injected into one of the arteries supplying blood to his liver. In gene therapy, an engineered virus is often used as a “vector,” delivering the desired gene to the patient’s cells; in this case, however, the virus apparently triggered a series of deadly events.

The New York Times picked up the story the day after it ran in the Post. The National Institutes of Health and the U.S. Food and Drug Administration started investigations, which moved with commendable speed; more details came out. Later, the U.S. attorney general got involved. But with those first newspaper reports, gene therapy seemed dead.

The trial that Gelsinger had been participating in was tainted by accusations of overconfidence, haste, negligent administration, and conflict of interest. Yet all this diverted attention from acute and fundamental problems with gene therapy itself–problems in the science and technology, problems in clinical exploitation of the technology, problems that were by no means new but that Gelsinger’s death made glaringly evident.

I had been following developments in gene therapy for a third of a century, watching as hundreds of millions of dollars were lavished on it, as new hopes cyclically turned to ashes, dramatic claims to sad farce. By 2000, more than 300 gene-therapy trials had been registered with NIH, involving more than 4,000 patients, according to an article printed that year in the Council for Responsible Genetics’ magazine GeneWatch. The Gelsinger affair was the most highly publicized failure. There had been plenty of others.

There were two chief reasons for pessimism about gene therapy. As had been plain from the start, although the total societal load of illness and debility caused by genetic defects is considerable, most individual diseases caused by single-gene defects–the kind that seem most likely to be cured by gene therapy–are rare. (Sickle-cell anemia and some other hemoglobin disorders are among the few exceptions.) Everybody in the field acknowledged this. Nobody seemed to face up to the implications. Because these diseases have different genetic mechanisms and affect different types of tissue, each presents a new set of research problems to be solved almost from scratch. As the millions burned away, it became clear that even with success, the cost per patient cured would continue to be enormous. And success had shown itself to be always glimmering and shifting just beyond reach, an ignis fatuus: from the start, step by step, everybody had underestimated the real difficulties the science presents.

The history of gene therapy can be told as the repeatedly frustrated search for viruses that work well as envelopes for gene delivery, paralleled by the increasingly baffling realization that far more than a few simple genes are needed to produce the desired proteins successfully. For the gene-therapy community, the years had been a calendar of failures. “We totally underestimated the fact that the viruses could present so many difficulties,” Inder Verma–a molecular biologist at the Salk Institute, in La Jolla, CA–told me in August 2006. “We underestimated the fact that it took billions of years for the viruses to learn to live in us–and we were hoping to do it in a five-year grant cycle!” He went on, “You know, the body is designed to fight viral infections. One hundred percent. Luckily for us! And here we are putting billions of viruses back into people and hoping that if we have a good virus, the body will say, ‘It’s okay, because we’re bringing the good stuff.’”

The first attempt at gene therapy in human patients began with a fortuitous observation. In 1959, the physician ­Stanfield Rogers, at the University of Tennessee, was working with the Shope papilloma virus, which causes warts on the skin of rabbits. He reported in Nature that the skin of these warts contained abnormally high levels of arginase, an enzyme that breaks down the amino acid arginine. He then found that some scientists who had worked with Shope virus, even 20 years in the past, had decreased blood levels of arginine.

The possibility that the virus had introduced its gene for arginase into the scientists was a curiosity, nothing more–until 1969, when the Lancet published a paper by ­Heinz-Georg Terheggen, a pediatrician in Cologne, Germany, and colleagues. Two little girls had been brought to Terheggen, deeply mentally retarded and suffering from a form of cerebral palsy, the British journal reported. Tests showed they had high levels of arginine, while very little of the enzyme arginase was detectable. This was a new genetic disease.

Rogers went to Terheggen to urge that he and his colleagues be permitted to inject the girls with Shope virus, hoping to give them a functioning gene for arginase. As an essential precaution, they did try inoculating the virus in a tissue culture of cells from one of the girls. They reported in the Journal of Experimental Medicine that they found arginase activity, apparently from the virus-introduced gene. But in the trial, there was no response, no reduction of arginine, no evidence of arginase activity. After an interval, they gave one child a larger dose. Still no response. The general consensus was that Rogers had made a premature attempt, with inadequate scientific understanding. That judgment was not wrong.

In the spring of 1972, Theodore Friedmann and Richard Roblin published the first extended study of the possibility of treating genetic diseases through gene transfer. “Gene Therapy for Human Genetic Disease?” appeared in Science. Disease by disease and therapy by therapy, the researchers warned of formidable technical problems; much that they laid out was prescient. They were the first to analyze the potential risks that gene therapy posed to patients and the grave ethical concerns it raised.

Nonetheless, the paper was a work of advocacy. With a medical degree from the University of Pennsylvania, ­Friedmann had spent three years in the 1960s at NIH, where, in the laboratory of Jay Seegmiller, he had begun to work on Lesch-Nyhan disease. Seegmiller had discovered that the disease is caused by the absence of the enzyme hypoxanthine phospho­ribosyltransferase, or HPRT, owing to a defect in its gene. Friedmann hoped to find a way to put the correct gene into Lesch-Nyhan cells in culture, perhaps using a virus. His imagination had been caught by the prospect of gene transfer. Indeed, as an assistant professor of pediatrics at the University of California, San Diego, in the early 1970s, he introduced the term “gene therapy.”

In January 1983, Friedmann and colleagues announced that they had isolated the normal gene for HPRT. Inder Verma, with whom Friedmann had struck up a collaboration in the early 1980s, had a potential viral vector: in this case, a type of retrovirus–one for a mouse leukemia. In August 1983, the two researchers reported that they had built the vector and used it successfully to introduce a functioning gene for human HPRT into rodent cells in vitro.

After that initial glimpse of success, Verma says, “very quickly we asked, ‘Can we do it in vivo?’” They began experiments on hemophilia in live mice. The gene defects causing hemophilia were known: the lack of a single protein could prevent blood from clotting. Working in vitro, adding the correct gene to cells in culture, “we could produce the protein forever,” Verma says. “And this is where the first surprise came.” The moment the cells were put back into the mice, “they instantly stopped making the protein. And this is the first limitation we recognized: retro­viruses can only introduce genes when the cells are dividing.” Verma adds, “We could take [the cells] out, grow them in vitro, transfuse them with the virus, put them back–but when we put them back, they shut off.” Why? “We still really have no idea,” he says.

Then, in 1990, an NIH research physician named ­William French Anderson announced to heated publicity that he was launching a gene-therapy trial, treating two young girls for a form of severe combined immune deficiency, or SCID. People with this disease completely lack a normal immune system. The precursor cells in their bone marrow that should make white blood cells are defective, so patients catch all the infectious diseases that white blood cells should fight off. Mild infections become grave; serious ones kill them. They die in early childhood. Anderson said the two girls were suffering from a form of SCID caused by a lack of the enzyme adenosine deaminase (ADA). He was injecting them with correcting genes carried in murine-leukemia virus.

Anderson was a flamboyantly effective publicist of gene therapy and of himself. He announced that the two little girls had been cured. In September 1994, he brought one of them to testify before the Science Committee of the U.S. House of Representatives. She was eight years old by then, lively and apparently well. The chairman of the committee reportedly called her “living proof that a miracle has occurred.” Anderson made sure he was known to the public as “the father of gene therapy,” even displaying the title on his website.

Yet his scientific colleagues and competitors became exasperated, even contemptuous. In point of fact, the trial with the two girls had failed. All along, the girls had also been treated with injections of a synthetic ADA. And Verma and Friedmann had already shown the failure of mouse leukemia virus to introduce genes in vivo. “There was never production of the ADA protein–there never was,” according to Verma. Even before the girl appeared in front of the House committee, the failure was known throughout the medical community.

Since retroviruses presented difficulties in vivo, attention turned to the adenoviruses–which include the viruses that cause certain types of severe upper-respiratory infections in humans. They worked. “They were wonderful,” Verma says. “First of all, you could make billions of virus particles.” Secondly, wherever the particles were introduced, the imported genes would be expressed. Many researchers switched to adenoviruses. But they turned out to be highly immunogenic: they are difficult to use safely because they can provoke strong immune reactions. Next came adeno-associated viruses, AAVs. Because they have only two proteins, AAVs provoke the immune system less than adenoviruses do.

In the fall of 1994, Harold Varmus, the director of NIH, became increasingly skeptical about the quality of gene-therapy research. The agency’s Recombinant DNA Advisory Committee (RAC) was reviewing all protocols for human trials of gene therapy funded by NIH. The committee’s first concern was safety. But as its recommendations passed across his desk for final approval, which was normally routine, Varmus realized that the committee was not systematically evaluating the trials’ scientific merits.

It turned out that Anderson’s were only the most egregious of many extravagant and unsupported claims surrounding gene therapy. Although NIH was giving out $200 million a year for gene-therapy research, and big pharmaceutical firms and swarms of biotechnology startups were thought to be spending as much again, not a single success with humans had been reported in any peer-reviewed journal. In May 1995, Varmus convened a panel headed by Stuart Orkin, a professor at Harvard Medical School, and Arno Motulsky, a geneticist at the University of Washington, Seattle, to review the state of gene-therapy research and assess how funds should be apportioned among gene-therapy research areas.

Orkin and Motulsky reported in December, at length and scathingly. The promise of gene therapy appeared great, but its failures had persisted despite the RAC’s approval of more than a hundred protocols. Most clinical trials were too small and exploratory in nature to evaluate the medical merits of the treatment; they lacked adequate controls and rigorously stated goals. Gene therapy, the panelists concluded, had been widely and harmfully oversold.

The balloon was pricked. The RAC had been considering approximately 15 protocols at each of its regular sessions; but the next meeting, scheduled for March 1996, was canceled. No proposals requiring public review had been submitted.

Three years later came Jesse Gelsinger’s death.

Gelsinger and the 17 other patients in the trial at the University of Pennsylvania were being treated for a deficiency of the enzyme ornithine transcarbamylase, which the liver uses to break down ammonia, a by-product of protein digestion, into harmless waste products. In its most severe form, the deficiency kills babies in their first year. Gelsinger had been kept alive on a strict diet and a regime of pills. When he learned of the gene-therapy trial, he volunteered.

The trial was carried out at the university’s Institute for Human Gene Therapy, which was headed by James Wilson. It was one of the top such centers in the country. The corrective gene was loaded into an adenovirus. The 18 patients were divided into groups that got increasingly large doses. Gelsinger got the biggest–a culture of 38 trillion virus particles. He received the dose on September 13, 1999. By September 15, his vital signs were falling precipitously. With his father’s assent, he was taken off life support, and he died on September 17.

Jesse Gelsinger’s death was the first directly attributed to gene therapy. An alert went out to the hundred or so experimenters using adenovirus vectors. In the press and in scientific journals, the case was reported as a disaster for the field.

NIH investigated and called a special public meeting for December 8, 9, and 10. The problem became clearer. The protocol for the trials, as approved four years earlier by the RAC and the FDA, had called for the adenovirus vector to be injected intravenously. The FDA had subsequently authorized direct injection of the vector into the hepatic artery, which was the method actually used. Nonetheless, Gelsinger’s autopsy found that the vector was widespread in his spleen, lymph nodes, and bone marrow.

Meanwhile, the FDA was conducting its own inquiry. Investigators were harshly condemnatory. Selection of trial participants had been sloppy at best: Wilson and his colleagues were unable to produce proof that any of the volunteers had met the criteria for the trials. Informed-consent procedures had been grossly inadequate. Federal rules require that benefits and risks be explained fully and clearly; Paul Gelsinger, Jesse’s father, told the New York Times that the family had been led to think the treatment might help Jesse, though the trial had been designed only to test the safety of a treatment being developed for infants. Further, the consent form had failed to mention that monkeys had died after a similar though stronger treatment. In 1992 Wilson had founded a private research company, Genovo, in which he held stock. The company had not put money into this particular study, but it did contribute a healthy portion of the Institute for Human Gene Therapy’s overall budget.

On January 21, 2000, the agency ordered a temporary stop to all gene-therapy trials at Wilson’s institute. In 2005, Wilson settled with the U.S. Department of Justice: he was not to lead any clinical trials regulated by the FDA for five years.

Hope for cures based on gene therapy, it appeared, had all but died with Jesse Gelsinger. But in February 2000, Friedmann gave the opening talk at a Monday-morning session of an annual meeting of the American Association for the Advancement of Science, in Washington, DC. He reviewed the fundamental difficulties of gene therapy, spoke of the many hundreds of protocols approved but so far not productive. He reminded his audience of Varmus’s impatient charge in 1995 that the field had been wildly oversold. Then–with a marked change in tone–he said, “We are on the verge of therapeutic efficacy.”

Two lines of work seemed to him to “have the feel of being correct.” A pair of American laboratories were beginning clinical trials of gene therapy for hemophilia. Proper blood clotting requires a cascade of responses, controlled by a series of proteins. Hemophilia A, the most common form of the disease, is caused by a defect in the gene for one of those proteins, factor 8; hemophilia B is caused by a defect in the gene for another, factor 9. The study ­Friedmann thought had that “sense of correctness” came from work with hemophilia B by Katherine High, a hematologist at the Children’s Hospital of Philadelphia. At Stanford, the gene therapist and virologist Mark Kay was also working with hemophilia B. Kay and High had combined their efforts. Their methods worked with animal models of the disease. They were ready to start human trials.

But the most convincing results, Friedmann said, were just then coming from a group of pediatricians in Paris. Their leader was a man named Alain Fischer, a clinician working with small boys who had a form of SCID. Like the girls whom NIH’s Anderson had treated for ADA deficiency, these children produced no T lymphocytes, the white blood cells that fight infection. But their disorder was caused by a different gene. The children had been sick; they were not thriving. Then Fischer and his colleagues tried gene therapy. “These kids are now to all appearances immunologically reconstituted entirely,” ­Friedmann said. “All their immune properties seem to be optimized.” He went on, “And the thing that’s so impressive about it is, first of all, that it came from nowhere. It came from left field.” Experts on immune-system disorders “certainly must have known of Alain Fischer and his group,” Friedmann said, but the gene-therapy community was not as familiar with his work. “And it also is presented in meetings in a very low-key, very modest sort of way,” Friedmann said. “They say straight out there’s nothing new in method–they’ve done just a combination of a fortuitously good disease model [with] a lot of standard retrovirology that’s been developed over many years.”

Fischer and a dozen colleagues reported their method, and their success with their first two patients, in Science on April 28, 2000. They followed up with a report in the New England Journal of Medicine for April 18, 2002.

Meanwhile, Mark Kay and Katherine High reported that when they injected their vector into dogs with hemophilia B, the dogs had a therapeutic response. ­Avigen, a biotech company headquartered in ­Alameda, CA, collaborated with High and Kay to plan clinical tests of the treatment’s safety in people.

In November 2002, the French scientists halted their trials. The number of patients was up to 10, but now one of those patients who’d gained a fully normal immune system had come down with a disease similar to leukemia, out-of-control proliferation of the very white blood cells that had been restored.

Then the June 4, 2004, issue of Science reported that Avigen had backed out of the trials of the hemophilia treatment. Two of seven patients had developed slightly elevated levels of liver enzymes.

On September 28, 2005, I went to see Alain Fischer at the Hôpital Necker, a children’s hospital in Paris. He was direct and clear. “I’m not a specialist in gene therapy,” he said at once. “My real field is immunology and, within immunology, genetic diseases of the immune system.” He had been working with these diseases for 25 years. “I am a physician. And here there is a clinical unit where children with immunological diseases are taken care of. So that’s where I’m starting from.” What kinds of diseases? “All kinds,” he said. “From deficiencies in T lymphocytes, B lymphocytes, innate immunity, there are … ” He drew breath. “We don’t know yet exactly. There are at least 140 different immunological diseases.” He added, “They are all very different.”

Fischer went on, “We are not going to become specialists in gene therapy–that is, to try to adapt gene therapy to different diseases. This is not our goal. We are specialists in these immunological diseases, and gene therapy is one strategy to try to treat these patients.” He was drawn to gene therapy in the early 1990s, when a new gene was identified that, mutated, causes a form of SCID. He had encountered patients with the mutation. “We understood very quickly, within one to two years, the pathophysiology of the disease,” Fischer recalled. “And we realized at that time that this disease could be the best candidate to test gene therapy.” The need for some type of effective treatment was certainly dire. Like all forms of SCID, he said, without treatment this one kills within the first year of life. The only treatment was bone-marrow transplants; but their success rate plummets unless close to identical immune-­system matches can be found, and that’s possible only about 20 percent of the time.

The types of cells affected by the disease also made it a good candidate for treatment with gene therapy, Fischer said. First, when the gene in which the mutation occurs is functioning properly, it encodes a protein that is vital if the precursors of T lymphocytes are to survive and proliferate. Second, unlike other types of immune system cells, T ­lymphocytes can survive for decades–sometimes even for an entire lifetime.

These two facts meant that even if the researchers could genetically alter only a few precursor cells, these cells might develop–or, as the scientists say, “differentiate”–into a large number of mature T cells that had a lasting benefit for the patient. “So we had the hope,” Fischer said, “that a very poor technology could–in that context, with that disease–work.”

Then came the drudgery. “We made vectors, retroviral vectors, the best technology of the time, blah blah blah,” remembered Fischer. But the tests went well. By 1998, Fischer and his colleagues were ready to seek approval to start human trials.

The first trial began on March 13, 1999. “And between ‘99 and 2002, we had treated 10 patients,” Fischer said. The researchers took bone marrow containing the lymphocyte­-precursor cells from the patients. In cell culture, they introduced the vector, a disabled retrovirus with the correcting gene. After several days, they injected the cells back into the patients. “And in nine out of ten, we were pleased to see that it worked,” he said.

As Fischer and his team had expected, the number of treated precursor cells able to generate T cells was very low. However, he said, it was sufficient to produce a normal number of T cells. “After a few months, these children could leave the hospital and start to live normally with their parents. And except for those who had the complication I’m going to describe in a moment, they are living normally still today.”

After the first three years, three of the ten children treated developed a severe complication, an uncontrolled proliferation of T lymphocytes. “I would call it a leukemia-like disease,” said Fischer. Childhood leukemia can usually be cured with massive doses of chemotherapy, and that’s how Fischer and his colleagues treated the three patients. One died. “The other two kids today are doing well, as well as the other seven,” Fischer said.

How much did all this cost? “A lot!” Fischer laughed abruptly. “A lot; but the treatment of a child with such a disease, without gene therapy, costs a lot, too.” Yes, he said, per patient, the cost of the research is huge. But “the cost of the therapy itself is not that big. Let’s assume it’s commercialized. I would assume the cost of the therapy itself, with the cost of the vector–the cell treatment ex vivo–shouldn’t cost more than maybe somewhere between $30,000 and $50,000, something like that. Per patient.” About the same as a heart transplant? “Exactly!” he said. “As it moves toward being a kind of, quote, ‘routine therapy,’ this is not much higher than many other therapies.”

And those complications? “We’ll see when we have enough follow-up to be sure,” he said, adding that if the chances of such a complication were reduced by a factor of 10, he’d consider the risk-benefit ratio “perfectly acceptable.” Fischer said he does not yet know whether his methods can be generalized to other types of genetic defects; he is not making any sweeping claims. His group is moving first to two other immune-deficiency diseases, involving other genes. “So we want to go step by step from the ones that are easiest to the most complex.”

From the first glimmer of possibility to the present day, Theodore Friedmann has written and spoken as gene therapy’s most ardent advocate. He has seen medicine enter a new era, which offers “new and definitive approaches to therapy that were previously only the stuff of dreams and scientific fantasy.” His has also been a voice of caution, of reason. He has had to warn his colleagues that they must openly address their discipline’s difficulties, its limitations, its failures. Yet he continues to marvel at the unprecedented possibilities raised by gene transfer. For the first time, he says, and one can sense his quiet exultation, medicine can do more than treat the signs and symptoms. It can reach the underlying causes. It can cure. “It’s going to be difficult,” he says. “Yet medicine has always had to work with imperfect knowledge and technology.”

Horace Freeland Judson is the author of five books, including The Eighth Day of Creation, a history of molecular biology that was published in 1979 and is still in print.