Gene therapy is given once, and almost every hard problem in the field comes from that

A gene therapy sounds like the simplest idea in medicine. A gene is broken, so supply a working copy. The complications are not in the genetics — a functional gene sequence is trivially available — but entirely in delivery, and one property of that delivery shapes everything else: you generally get a single attempt, for life.

The reason is straightforward. The vehicle used to carry the gene into cells is a virus shell, and the immune system responds to it the way it responds to any virus. Within weeks of the infusion the patient has strong neutralising antibodies against that shell. A second dose meets an immune system already prepared for it and is destroyed before it reaches anything. Everything difficult about this field — the dose sizes, the exclusions, the price, the serious adverse events — follows from having exactly one shot.

The delivery vehicle and what it can carry

The workhorse is adeno-associated virus, AAV: small, not known to cause disease, and stripped of its own genes so that only the therapeutic cargo remains inside the protein shell, the capsid.

That shell imposes a hard physical limit. An AAV capsid holds roughly 4.7 kilobases of genetic material, and that has to include the gene, a promoter to switch it on, and regulatory sequence. Plenty of genes fit comfortably. Plenty do not — the dystrophin gene behind Duchenne muscular dystrophy runs to about eleven kilobases of coding sequence alone, which is why work in that disease uses deliberately shortened micro-dystrophin constructs that keep the essential domains and discard the rest. A whole sub-field exists to work around the size of a protein shell.

Different serotypes — capsid variants — favour different tissues. One crosses into the nervous system reasonably well, another concentrates in liver, another is used in the eye. Choosing the serotype is choosing which organ receives most of the dose, and it is one of the few steering mechanisms available.

The gene arrives, but it does not settle in

Once inside the nucleus, the delivered gene mostly does not integrate into the patient's chromosomes. It persists as a separate circular episome, transcribed normally but not copied when the cell divides.

For tissue that does not divide, this is close to ideal: retina, neurons, mature muscle. Treat those and the effect can last years, with no insertional risk to the genome. For tissue that does divide, the episomes dilute with every cell division and expression fades. This is the reason treating a young child raises a question a textbook rarely asks: the organ being treated is still growing, so a therapy that works beautifully at age one may weaken as that liver doubles in size, and the obvious remedy — treat again — is the one thing the immune response forbids.

The therapy is not permanent because the gene is durable. It is durable only where the cells that received it stop dividing. Everywhere else, one dose is a fading asset.

Who is excluded before anyone examines them

AAV serotypes are derived from viruses that circulate widely in the human population, and a substantial share of adults have already met them. Depending on the serotype and the region, anywhere from a large minority to a majority of people carry pre-existing neutralising antibodies — and if they do, an infusion is neutralised in the bloodstream before it does anything.

So screening for those antibodies is a standard eligibility test, and failing it excludes a patient not because their disease is unsuitable but because of an infection they may have had decades earlier and never noticed. Seroprevalence also varies geographically, which means eligibility rates in India or sub-Saharan Africa can differ from those in the trial populations where these products were developed — a genuinely under-studied problem for equitable access.

Why single-dosing makes the dose dangerous

Because there is no second attempt, the first has to work, which pushes doses to the physical limit of what can be manufactured and tolerated. Systemic AAV therapy is administered in vector genomes per kilogram of body weight, at magnitudes that mean an infusion for an adult contains a staggering number of particles.

At that scale the immune system stops being a background consideration. A T-cell response directed against the capsid can attack the very liver cells that took up the therapy, producing the liver enzyme rises seen in many trials and typically managed with steroids. Complement activation can trigger clotting and kidney complications. Serious adverse events, including deaths, have occurred in high-dose systemic trials, particularly in patients with pre-existing liver disease. None of this reflects the gene doing something wrong. It is the response to the delivery vehicle, at the dose that single-shot delivery demands.

Manufacturing carries the same signature. Producing AAV at clinical scale is genuinely difficult, and a persistent problem is that a large proportion of assembled capsids come out empty — correct shells with no genetic cargo. Empty capsids deliver nothing while contributing fully to the immune burden, so separating them out is a major purification challenge and a substantial part of why a dose costs what it does. Several approved therapies are priced above two million dollars, and while that figure reflects rare-disease economics, it also reflects a manufacturing process making an extraordinary number of particles for one person.

What would change the picture

Three directions matter, and they attack different parts of the constraint.

Enabling a second dose. If capsid antibodies could be cleared or suppressed around the time of infusion, redosing becomes possible — approaches under investigation include enzymes that cleave circulating antibodies and targeted immune suppression. Success here would relax nearly every other constraint at once, including the need to give a maximal dose the first time.

Engineered capsids. Rather than using naturally occurring serotypes, capsids can be evolved in the laboratory for stronger targeting of a chosen tissue and reduced recognition by existing antibodies. Better targeting means a smaller dose, and a smaller dose means less of everything that goes wrong at high doses.

Leaving viruses behind. Lipid nanoparticles already deliver genetic material without a capsid, they provoke nothing that prevents redosing, and they have been used to deliver gene-editing machinery to the liver in human trials. They do not yet reach most tissues well, which is the whole problem, but they sidestep the immunological trap entirely.

There is also a strategic point about vector choice that Indian work illustrates. In 2024 a team at CMC Vellore reported the country's first gene therapy trial for haemophilia A, using a lentiviral approach applied to the patient's own blood stem cells rather than a systemic AAV infusion — a route that changes the immune and manufacturing problem completely. Which vector you choose is not a technical footnote; it determines which patients are eligible, what it costs and whether it can ever be repeated.

Why it matters for students and researchers

Gene therapy is the clearest example in modern medicine of a field where the biology was the easy part. Identifying the causative gene in a monogenic disease is now routine. Synthesising a correct copy is trivial. Everything hard is delivery: getting a molecule into the right cells, in enough of them, without the immune system treating the attempt as an infection — which, structurally, it is.

That makes it a bioengineering discipline rather than a genetics one, and the open problems are correspondingly practical: capsid engineering, immune modulation, purification and full-to-empty ratios, potency assays, and scale-up. For India the questions are sharper still, because seroprevalence differs, because the disease burden in haemophilia and thalassaemia is large, and because a two-million-dollar therapy is not a therapy for this population — it is a proof that the biology works, waiting for someone to solve the manufacturing.

Frequently asked questions

Why can gene therapy usually be given only once?

Because the viral capsid that delivers the gene provokes a strong antibody response. Within weeks the patient has neutralising antibodies against it, so a second dose of the same vector is destroyed in the bloodstream before reaching its target cells.

What is AAV and why is it used?

Adeno-associated virus is a small virus that does not cause known disease. Its own genes are removed and replaced with a therapeutic gene, so the shell acts purely as a delivery vehicle into human cells.

Why can't gene therapy treat every genetic disease?

An AAV capsid holds only about 4.7 kilobases, so large genes such as dystrophin do not fit and must be shortened. Delivery is also limited to tissues a given capsid can reach, and effects fade in tissues whose cells keep dividing.

Why are some patients ineligible for gene therapy?

Because they already carry neutralising antibodies from an earlier natural exposure to a related virus. Depending on serotype and region this can exclude a large share of adults, regardless of how suitable their disease otherwise is.

Why is gene therapy so expensive?

Manufacturing clinical-grade viral vector at the doses a single administration requires is difficult and low-yield, with many assembled capsids emerging empty and needing to be separated out. That cost, combined with very small patient populations, produces prices above two million dollars for some approved products.