The Dawn of the Age of Programmable Medicine

I was lucky to grow up in a South Shore Long Island town whose public school system was, by the standards of the time, excellent — and unusually so in two respects: it drew from an ethnically diverse student body, and it took scientific inquiry seriously. Living close enough to New York City, our community was home to a number of working researchers (including both parents of my closest friend), and their influence reached even into the elementary grades. I still remember, in fourth grade, watching a cancer researcher dissect a laboratory rat to remove its tumors — a formative encounter with science as something people actually did for a living. Scientific curiosity, for me, was piqued early and never really subsided. By high school, Science Research Skills was a formal course on the curriculum, and a good number of students went on to NIH-sponsored summer programs at universities across the country. I spent the summer after tenth grade at one such program at the University of Iowa — a transformative experience. I came home wanting to become a biologist.

It wasn’t until I reached college and encountered classes in History of Science that I understood how far biology lagged behind its two main rivals — physics and chemistry — in the great triumvirate of the natural sciences. Of the three, biology was the last to find its mechanistic foundations. Chemistry had its periodic table by 1869, and by 1876 — with Gibbs’s Free Energy equation — could predict which reactions would be thermodynamically favorable. Physics already had Newton’s laws of motion two centuries before that, and by the early twentieth century had upended them with quantum mechanics and the theory of relativity. Biology, by contrast, remained for most of its history a science of observation and classification. It had achieved an impressive body of descriptive work, but without the unifying principles that had given chemistry and physics their explanatory power. The cell had been visible under the microscope since the seventeenth century. Chromosomes — those dense, thread-like structures that appear with dramatic clarity in dividing cells — had been observed and named by the 1880s. Plant and animal breeders had known for generations that traits passed reliably from parent to offspring. And yet the connecting thread between these facts remained unseen for decades. It was not until 1902 that the cytologist Walter Sutton and the biologist Theodor Boveri, working independently, recognized that chromosomes behaved during cell division exactly as Gregor Mendel’s hereditary units — his “factors,” rediscovered just two years earlier — would predict. The physical basis of heredity had been sitting in plain sight, on the slides of a thousand microscopes, waiting for someone to see what it meant.

It was not until 1953 that Watson and Crick described the double helix structure of DNA, revealing the molecular mechanism by which genetic information is stored and copied. And it wasn’t until 1961 — the same year humanity put its first man into orbit — that Nirenberg and Matthaei demonstrated how the cell transcribes DNA into mRNA and then translates it into functional proteins. (I won’t review here the mechanics of transcription and translation, but if you want a quick refresher course in simple language that takes less than 9 minutes, I highly recommend viewing the following Youtube animation video from the Amoeba Sisters.) 

What followed, once biology’s foundations were finally in place, were new technologies advancing with gathering speed. The decades that followed produced an almost unbroken sequence of enabling discoveries: restriction enzymes in the late 1960s and early 1970s gave researchers molecular scissors precise enough to cut DNA at defined sequences; recombinant techniques allowed genes from one organism to be spliced into another; bacteria and then yeasts were engineered to produce human proteins — insulin, growth hormone, clotting factors — at industrial scale. By the 1980s, functional genes could be introduced into mammalian cells, and the question that had once seemed purely theoretical began to acquire the outlines of a clinical program. If a defective gene was responsible for a disease, could a correct copy be delivered to the cells that needed it? The answer, researchers were beginning to believe, was yes. What remained was the harder problem: how.

The tools that were emerging from molecular biology in the late 1980s were genuinely powerful, but they were also new, and their behavior inside living human bodies was far less predictable than their behavior in culture dishes and animal models had suggested. The scientists who moved gene therapy into the clinic in those years were serious researchers who believed, on reasonable grounds, that the field had matured enough to justify the attempt. What they could not yet fully know was how much they did not know — about immune responses to viral vectors, about the unpredictable consequences of inserting genetic material into the genome, about the profound distance between a promising result in a mouse and a safe and durable outcome in a person. The scientists who pursued gene therapy in the late 1980s and 1990s were working, in other words, at the frontier of something genuinely new — in a field where the tools were outpacing the understanding of the systems those tools would enter. What they built, and what failed, and what they learned from the failing — all of it would prove indispensable to what happened in that Philadelphia hospital room thirty years later. The story of KJ Muldoon does not begin in 2024. In important ways, it began here.

***

The concerns over the emerging technologies were not new. By the 1970s, the tools for manipulating genetic material were beginning to arrive. Restriction enzymes — molecular scissors that could cut DNA at specific sequences — made it possible to isolate and examine individual genes. Recombinant DNA technology allowed researchers to splice genes from one organism into another, a development that alarmed the public and electrified the scientific community in roughly equal measure. In 1975, leading scientists gathered at the Asilomar Conference Center in California to debate, voluntarily and with unusual candor, whether the technology they were building was moving faster than their ability to assess its risks. The moratorium they proposed — a temporary, self-imposed pause on certain experiments — was brief, but it established a precedent: that the scientists most capable of advancing this work were also responsible for questioning it.

By the early 1980s, the moratorium had given way to momentum. Recombinant techniques were being used to produce human insulin in bacterial cultures — the first commercially successful product of genetic engineering. The polymerase chain reaction (PCR), conceived by Kary Mullis in 1983 and published two years later, gave researchers the ability to amplify vanishingly small quantities of DNA into amounts large enough to study. Genes responsible for hereditary diseases were being identified one after another: sickle cell anemia, Huntington’s disease, cystic fibrosis. The map was filling in. And in the minds of the field’s most ambitious practitioners, a question was beginning to solidify from speculation into program: if we can read these mutations, can we correct them?

The first formal proposal for human gene therapy appeared in 1972, when the biochemist Theodore Friedmann and his colleague Richard Roblin published a paper in Science laying out the ethical and scientific conditions under which gene therapy might one day be considered. It was a remarkably prescient document: cautious in its recommendations, clear about the magnitude of what would be required, and aware that the question was not merely technical but moral. The idea of treating genetic disease by supplying a patient with a functional copy of a defective gene was worth pursuing — but only after rigorous preclinical validation that the field was not yet in a position to provide.

That caution eroded, unevenly, over the decade that followed. By the late 1980s, Dr. W. French Anderson — working at the National Institutes of Health — had become the foremost advocate for moving gene therapy from the laboratory into the clinic. He was a compelling and tenacious figure, convinced that science had progressed far enough, and that patients were dying while researchers debated. His target disease was adenosine deaminase deficiency — ADA-SCID — a rare and devastating genetic disorder that left children without a functioning immune system. The condition had acquired a name in the press that conveyed its human cost with terrible efficiency: bubble boy disease, after David Vetter, a boy from Texas who had spent his entire short life inside a sealed sterile enclosure, his story covered so widely that the disease’s cruelty had become impossible for the public to ignore. He died in 1984. ADA-SCID was caused by a single gene defect; it was well characterized; and it seemed, in principle, like an ideal candidate for genetic correction.

The approach Anderson and his colleagues proposed was what would come to be called ex vivo gene therapy — cells would be extracted from the patient, corrected in the laboratory, and returned to the body. The corrective gene would be delivered using a retrovirus — a virus modified to be harmless, but retaining its natural ability to insert genetic material into the DNA of the cells it infected. On September 14, 1990, a four-year-old girl named Ashanthi DeSilva became the first human being to receive gene therapy. The infusion was administered at the NIH Clinical Center in Bethesda, Maryland. Ashanthi’s condition improved — though she continued her other medications, and the precise contribution of the gene therapy to her recovery remained, then as now, difficult to isolate. The field exhaled nonetheless.

What followed was a decade of rapid expansion and growing unease. Clinical trials multiplied. Hundreds of patients with dozens of different diseases entered studies for gene therapy. The enthusiasm was genuine, and in some cases so was the early promise. But replication was difficult. The immune systems of patients responded to viral vectors in ways that had not been predicted. The clinical landscape was more complex than the laboratory had suggested, and the gap between promising results in mice and durable benefit in humans was wider than many had allowed themselves to believe.

The field was poised, though it did not fully know it yet, for a reckoning. That reckoning did not arrive as a single event but as a series of disasters separated by time, geography, and mechanism — each one revealing a different dimension of how much the field still did not understand. The first arrived in Philadelphia, in the form of a young man who had volunteered not out of desperation, but out of something rarer: altruism.

***

Jesse Gelsinger was a recent high school graduate when he entered the University of Pennsylvania’s Institute for Human Gene Therapy in the summer of 1999, applying to take part in a Phase I trial of a new genetic therapy. He was not in danger of dying. That fact matters enormously in any honest accounting of what happened next — and it sets him apart, in the most poignant way, from KJ Muldoon.

Like KJ, Jesse had been born with a metabolic disorder rooted in a single nucleotide genetic mutation. His disease, ornithine transcarbamylase deficiency — OTC deficiency — was, like KJ’s CPS1 deficiency, a disorder of the urea cycle, the same metabolic pathway responsible for converting ammonia into urea, a form the body can safely excrete. In patients with a complete absence of OTC enzyme function, the disease is catastrophic and almost universally fatal in the neonatal period. Newborn boys with the most severe form of the disorder do not typically survive infancy.

Jesse did not have the most severe form. He had a mosaic version of the disease — a consequence of the timing of his mutation, which had occurred not at conception but early in embryonic development, leaving some of his cells carrying the defective gene and others functioning normally. The result was a partial enzyme deficit that could be managed, though with considerable discipline. He had been diagnosed at age two, after a dietary mishap sent his ammonia levels spiking and nearly killed him. Through a strict low-protein diet and a daily regimen of between thirty-two and fifty pills, he was able to live a mostly normal life — going to school and graduating from high school with his class. He worked with his father, a contractor in Tucson, and held a part-time job as a supermarket clerk. He loved riding his off-road motorcycle. He was, by multiple accounts, an easy and generous presence — the kind of young man whose uncomplicated warmth made people like him immediately and remember him clearly.

Jesse knew that his form of OTC deficiency was not what the clinical trial he was joining was designed to treat. He knew the trial was designed to test the safety of a new gene therapy vector — an adenovirus, in this case, rather than the retrovirus used in earlier trials — and that it was aimed, ultimately, at developing a therapy that could help the neonatal patients for whom the disease was a death sentence. He volunteered because he understood this. Because he had grown up knowing what OTC deficiency could do in its most merciless form, and because someone had to be among the first. “What’s the worst that could happen to me?” he reportedly told a friend before the trial. “I could die, and it would be for the babies.”

His disorder was not without consequences in the meantime. Severe fatigue and ammonia spikes that followed even routine viral infections had made long-term planning difficult, and Jesse had deferred decisions about his future. While many of his friends were settling into their first weeks of college, Jesse was preparing to enter the Phase I trial at the University of Pennsylvania. The treatment was administered on September 13, 1999 — nine years, almost to the day, after Ashanthi DeSilva’s landmark infusion. Within sixteen hours, Jesse began to show signs of an immune response that quickly spiraled into a medical crisis. His immune system was producing massive quantities of inflammatory cytokines — what clinicians call a cytokine storm — triggered not by anything the vector had done after delivery, but by the immune system’s recognition of the adenoviral capsid itself: the protein shell of the virus, which the body identified as a pathogen and attacked with overwhelming force. Within two days the cytokine storm had cascaded into multiorgan failure. On September 17, 1999, four days after the infusion, Jesse Gelsinger died. He was eighteen years old.

The Institute for Human Gene Therapy had been built by Dr. James M. Wilson, one of the field’s most prominent and ambitious scientists. Wilson was not a reckless man by temperament or by training. He was, if anything, the opposite: methodical, driven, and possessed of the kind of focused determination that had built one of the most well-resourced gene therapy programs in the world. But tenacity has its own blind spots. A mind organized around forward momentum — around the next experiment, the next result, the next justification for pressing ahead — can become subtly resistant to the signals that argue for pausing and reevaluating. Wilson had not abandoned scientific rigor. He had, perhaps, allowed his confidence in the direction of travel to outrun his caution about the road.

The investigation that followed Jesse’s death revealed something more complicated than simple negligence — and, in some ways, more troubling. Researchers at the Penn institute had not fully disclosed prior adverse events to regulators. Animal data suggesting the possibility of serious immune responses had not been given adequate weight. Commercial interests had created conflicts that had not been properly disclosed to patients: Wilson and his colleagues held equity stakes in Genovo, a company with direct financial ties to the therapy being tested. Jesse himself had had elevated ammonia levels on the morning of his treatment, a finding that might, under the trial’s own eligibility criteria, have been grounds for postponing the infusion. The treatment proceeded as scheduled.

What the investigation ultimately revealed was not a story of recklessness per se, but of a failure of culture: a laboratory environment in which commercial entanglements had subtly distorted scientific judgment, and in which the distance between animal data and human safety had been crossed with insufficient caution. Wilson was eventually barred from conducting human clinical trials. The full accounting of what went wrong at Penn involves a braided failure of institutional oversight, regulatory compliance, and individual judgment that does not reduce easily to any single cause. What matters here is not the assignment of blame — though blame was warranted — but what Jesse’s death revealed about the state of the field at the moment it claimed him. Gene therapy, in 1999, was attempting to do in patients what it had not yet learned to do reliably in the laboratory. The viral vectors being used to deliver genetic material were powerful, imprecise, and immunogenic in ways that had been underestimated. The transition from animal models to human patients was being made with insufficient caution. The science was real, and its eventual promise was real, but it was not yet ready for what was being asked of it.

Jesse Gelsinger’s death did not merely end a trial. It ended an era.  The field essentially collapsed. Clinical programs were suspended. The FDA imposed clinical holds. The NIH convened a special panel. Enrollment in gene therapy trials plummeted from hundreds to handfuls. The scientists who had built the field largely retreated — some into other areas of research, others into a period of methodical reassessment that would eventually prove more productive than the preceding decade of expansion had been. For several years, it was genuinely unclear whether gene therapy as a clinical enterprise would survive.

***

Jesse’s death was not the field’s only tragedy that decade. The second disaster unfolded not in Philadelphia but in Paris, and it arrived through an entirely different mechanism — one that would prove, in some ways, even harder to anticipate, because it announced itself not in hours but in years.

In the late 1990s, the French immunologist Alain Fischer and his colleague Marina Cavazzana-Calvo were treating children born with X-linked Severe Combined Immunodeficiency (X-SCID), a disease caused by a defect in a gene encoding a protein called the common gamma chain, essential for the development of immature immune cells into functional ones. The vector Fischer used was a gamma-retrovirus — a retrovirus engineered to deliver a corrective gene to the patients’ bone marrow cells. Unlike the adenovirus that had killed Jesse Gelsinger, gamma-retroviruses integrate their genetic payload directly and permanently into the host cell’s chromosomes. This integration was the point: it meant that every time a corrected cell divided, its daughters would carry the therapeutic gene as well, producing a self-renewing population of immune cells. For children born with no immune system at all, the results were remarkable. Treated children developed T-cells, went home, and began living something approaching normal lives. It was the clearest proof of concept the field had seen. Fischer and Cavazzana-Calvo had, it seemed, done what Anderson had only partially achieved — they had actually cured a genetic disease.

Then, two to three years after treatment, children began developing leukemia.

To understand why, we need to understand what retroviruses do — and what happens when that process goes wrong.

A retrovirus carries its genetic material not as DNA but as RNA. When it infects a cell, it deploys an enzyme called reverse transcriptase to convert that RNA into DNA, and then a second enzyme called integrase to splice that DNA directly into the host cell’s chromosomes. This is not an accident of infection; it is the virus’s reproductive strategy. Once integrated, the viral DNA — now called a provirus — is copied every time the cell divides, passed to every daughter cell, and read by the cell’s own machinery as if it were part of the host’s original genome. The virus has, in effect, made itself part of the permanent machinery of the cell.

Gene therapists saw this viral mechanism as opportunity. If you could strip out the viral genes responsible for making new virus particles and replace them with a therapeutic gene, you’d have a delivery vehicle — a vector — that could efficiently insert a corrective sequence into a patient’s cells. The virus’s pathogenic capacity could be neutered; its integrative capacity could be utilized for good.

What they did not properly account for, in those early years, was a critical feature of how retroviral integration works: it is not random, but it is also not precise. Retroviruses don’t insert themselves at a single defined location in the genome. They have preferences — they tend to favor regions that are actively being transcribed, areas where the DNA is, so to speak, open for business. But within those broad preferences, the exact landing site varies from cell to cell, and cannot be controlled by the researcher engineering the vector.

This matters enormously because the 3.1 billion base pairs of the human DNA molecule are not simply genes dotted by random space in between them.  The genome is densely annotated — packed with regulatory sequences, promoters, enhancers, silencers, and other control elements that govern when genes are switched on, how strongly they are expressed, and in response to what signals. A retroviral insertion doesn’t just add a sequence to the genome; it lands in a neighborhood, and it can disrupt that neighborhood in several ways.

The most straightforward is insertional disruption: if the viral vector integrates within an existing host gene — into one of its exons or introns — it can physically interrupt that gene’s coding sequence, rendering the corresponding protein nonfunctional. Cell biologists call this a knockout — essentially a man-made mutation. Researchers were aware of this risk but were not, at the time, overly alarmed by it. Of the human genome’s three billion base pairs, only one to two percent are coding sequences; the odds of a random insertion landing squarely in a critical gene seemed small. And if an insertion did disable a gene essential to cell survival, the cell would likely die — in the laboratory or shortly after being returned to the patient. Dead cells that cannot replicate are, generally speaking, not dangerous.

More insidious, and ultimately more consequential, is insertional activation. Retroviral vectors carry their own regulatory sequences — strong promoters designed to drive high-level expression of the therapeutic gene. If the viral vector integrates near a host proto-oncogene, one of the many genes that, when overactivated, can drive uncontrolled cell proliferation, those viral promoters can act at a distance, switching the neighboring gene on at inappropriate levels. The cell doesn’t need a mutation in the oncogene itself; it just needs the wrong neighbor. Unlike the cell that dies from a disruptive insertion, this cell — with its newly activated growth switch — does not die. It divides. One cell becomes two, two become four, and within a short number of generations, what began as a single misplaced integration event can produce a population of abnormal cells numbering in the millions.

This is precisely what happened in the French X-SCID trials. In multiple patients, the gamma-retroviral vector had integrated near a gene called LMO2 — a proto-oncogene involved in blood cell development. The vector’s promoter sequences drove LMO2 into abnormal overexpression, triggering uncontrolled proliferation of T-cells. Of the ten children Fischer treated in Paris, and ten more treated in a parallel British trial led by Adrian Thrasher at London’s Great Ormond Street hospital, five of the twenty children eventually developed T-cell leukemia. One died.

The tragedy illuminated something researchers had known in principle but had not fully reckoned with in practice: that the genome is a system, not a collection of independent parts, and that inserting a sequence into it — however precise the delivery mechanism — is an intervention with consequences that extend beyond the insertion site. A vector that lands in one location does one thing; the same vector landing three hundred base pairs to the left can do something entirely different. And in a population of treated cells, the vector lands in thousands of different locations simultaneously. Most of those landings are inconsequential. Some are not. And there is, with retroviral vectors of that era, no way to know in advance which is which.

What is striking, viewed together, is how different the two disasters were from each other. Jesse Gelsinger died because his immune system recognized the adenoviral vector as a foreign pathogen and destroyed it — and him — before it could do anything at all. The French children developed leukemia because their cells accepted the gamma-retroviral vector completely, incorporating it so seamlessly into their genomes that it began to behave like a native regulatory element. One failure was too much immune response. The other was too little. The field was not suffering from a single problem it had failed to solve. It was navigating a landscape of multiple independent failure modes, each invisible until someone fell into it.

***

Wilson spent the years following Jesse’s death doing something that might, in a different kind of story, count as straightforward redemption — though the reality is more ambiguous than that. Working largely outside the clinical arena to which he had been barred, he threw himself into the basic science of viral vectors, and in particular into the refinement of adeno-associated viruses — AAVs — as delivery vehicles for genetic material. AAV vectors do not integrate into the genome the way retroviruses do; they generally persist as separate circular DNA inside the cell, expressing their therapeutic payload without the insertional risks that had undone the French trials. The work was painstaking, iterative, and largely invisible to the public. It was also consequential. The AAV platforms Wilson’s lab helped to develop and characterize became the foundation for some of the most successful gene therapies ever brought to patients — among them Zolgensma, approved by the FDA in 2019 for children with spinal muscular atrophy, which uses an AAV9 vector to deliver a working copy of the missing SMN1 gene in a single infusion; and Hemgenix, approved in 2022 for hemophilia B, the first liver-directed AAV gene therapy to reach patients. Those successes rebuilt the field’s credibility and sustained the institutional and commercial infrastructure within which the next generation of approaches could be developed — including the base editing technology, delivered not by any virus but by lipid nanoparticles, that would eventually reach KJ Muldoon’s liver cells. The field Wilson had almost fatally wounded was the field he also helped to rebuild. Whether that constitutes redemption is a question each reader will answer differently. What is not in question is that the science mattered greatly.

Jesse Gelsinger’s story should not amount to a purely cautionary tale, though he is sometimes reduced to one. He was a person — generous, curious, apparently fearless in the way that certain eighteen-year-olds are fearless, before the world has given them sufficient cause for caution. The baby who went home from CHOP in June 2024 carries, in some sense, a debt to him — and to the children in Paris whose leukemias forced a reckoning the field had been avoiding. Not because those deaths and near-deaths made KJ’s treatment possible in any simple, linear way — the path between them is far more complicated than that. But because the rigor that followed, hard-won and overdue, is part of what made the precision of KJ’s therapy possible at all. Both Jesse and KJ were delivered into a moment they could not have anticipated. KJ by birth, Jesse by choice. One arrived at the beginning of the story. The other, in some ways, was its turning point.