In Vivo Base Editing: The Short Read

The Cut Is the Problem

In the summer of 2024, a baby boy in Pennsylvania started to die from the food he was fed. His parents called him KJ. Within days of birth, ordinary protein, the protein in milk, was poisoning him. His body couldn’t get rid of ammonia, the toxic waste left over when the body breaks protein down. The ammonia kept climbing toward the levels that swell the brain and kill. His disease is so rare that maybe one baby in a million is born with its severe form, and about half of those infants don’t survive infancy. There was no drug for it. So a team of doctors and scientists built one. A medicine for exactly one patient on Earth, made from scratch in roughly six months, and dripped into KJ’s vein in early 2025. The most surprising thing about that medicine is what it didn’t do. It never cut his DNA.

That cuts against almost everything the public has been told about gene editing. The technology that made editing famous is CRISPR. The popular image of CRISPR is a pair of molecular scissors. The image is earned. In 2012, Jennifer Doudna and Emmanuelle Charpentier showed that a system bacteria use to chop up invading viruses could be reprogrammed to cut DNA at any address a scientist chose. A short guide molecule acts like a postal code, steering a cutting protein to one exact spot among the three billion letters of the human genome. Doudna and Charpentier shared the 2020 Nobel Prize in Chemistry for it.

But scissors cut, and a cut in DNA is more violent than it sounds. DNA is a double strand, the famous twisted ladder, and classic CRISPR slices clean through both rails at once. To the cell, a clean break through both strands is an emergency. It drops everything to glue the ends back, fast and without a blueprint. In the rush it often loses a few letters or jams in extra ones at the splice. Sometimes it deletes long stretches to either side, or stitches the wrong ends together and scrambles a whole region. The break gets sealed, but the neighborhood around it can come back damaged. If all you want is to switch a gene off, that’s tolerable. Broken is broken. But if you’re trying to repair one, where the whole point is landing on an exact sequence, it’s a serious liability.

So a different idea took hold. Keep CRISPR’s genius. Throw away its blade. Keep the postal-code targeting that finds one spot in three billion letters. But on arrival, instead of cutting, quietly rewrite a single letter of the code in place. In 2016, a Harvard chemist named David Liu and his lab built the first tool that could. They took CRISPR’s targeting machinery, dulled the scissors so they could grip the DNA without slicing through it, and bolted on a second part: a chemical tool that converts one DNA letter directly into another. DNA is written in just four letters (A, C, G, and T). A huge number of genetic diseases come down to one of those letters being wrong. Liu’s tool turns the wrong letter into the right one without ever breaking the strand. He puts the contrast plainly. The original CRISPR was a pair of scissors. This is a pencil.

The first CRISPR medicine ever approved treats sickle-cell disease the cautious way. Doctors remove a patient’s blood-making cells, edit them in a dish, check the result, and infuse the corrected cells back. The harder and more ambitious route is to edit inside the living patient (in vivo, Latin for “in the living body”) by injecting the machinery and letting it hunt down its target among the body’s own organs. There’s no dish, no inspection, no taking it back. Once the injection is in, the editor changes whatever it reaches, in a body you can’t pause. That’s the danger. It’s also the only way to reach the organs you can’t remove and return: the liver, the brain, the eye, the muscle. In vivo is where the stakes are highest, and where the real frontier is.

One Letter at a Time

KJ’s disease lives in the liver. His condition, CPS1 deficiency, knocks out the first step of the urea cycle, the chemical assembly line the liver uses to pack ammonia into urea and flush it out in urine. Both of KJ’s copies of the gene were misspelled, one error inherited from each parent. The team that took his case (at Children’s Hospital of Philadelphia and the University of Pennsylvania, led by the physician Rebecca Ahrens-Nicklas and the gene-editing researcher Kiran Musunuru) read out KJ’s exact misspelling and designed a base editor to correct one of his two faulty copies. Rewrite the single wrong letter, switch the dead enzyme back on. The whole therapy came together in about six months. The FDA cleared it about a week after the application was filed. KJ got his first infusion in late February 2025. Within weeks he could handle more protein. His scavenger drugs were cut by roughly half. He went home.

A base editor is built from three parts fused into a single molecule you can inject. It has to find one address among billions of letters. It has to stop there and hold absolutely still. And it has to perform the chemical rewrite. The first part is the address-finder. It works like GPS coordinates written in the same chemical language as the target. A short single strand of RNA, carrying a run of about twenty letters, threads along the genome until its letters meet their matching partners in the DNA and lock on. Load the machine with a different guide and you send it to a different address. The aim is reprogrammable by rewriting twenty letters, nothing more.

The second part is the very protein that, in ordinary CRISPR, does the cutting: Cas9. In a base editor, though, its blades are deliberately broken. A small change to its structure leaves Cas9 able to clamp onto the double helix and grip it firmly, but no longer able to slice it. The third part is the rewrite tool itself, a small enzyme bolted onto the parked Cas9. When the guide RNA pairs with its matching DNA strand, it shoulders in between the two rails. Locally, over a short stretch, the double helix unzips. One strand is now held by the guide. The other strand is left dangling loose and single. This open pocket is the bubble. The loose strand inside it is exactly what the chemical tool has been waiting for. A short run of exposed letters, finally reachable.

What happens inside that bubble is gentler than the word rewrite suggests. The tool performs a tiny piece of chemical surgery on the letter already sitting there. It shaves a small chemical tag off the letter, and the cell’s reading machinery now mistakes it for a different letter entirely. A C gets shaved down until the cell calls it a T. An A gets shaved down until the cell calls it a G. The strand stays whole the entire time. Once the cell copies the new reading and locks it in, the change is permanent. Indistinguishable from a letter that had been correct all along.

The molecular pencil works. The chemistry is no longer what’s keeping these therapies out of patients. What’s keeping them out is humbler. Plumbing. A human body is not a dish. It’s a sealed, churning system of some thirty trillion cells, sorted into organs, most of them walled off behind membranes and barriers and nowhere near a needle. The vehicle that actually carried base editing into living people came from a completely different direction. And most of the world already has it coursing through their arms. It’s the technology behind the mRNA COVID-19 vaccines: the lipid nanoparticle, a microscopic bubble of fat. You wrap the instructions for the editor inside a droplet of specially designed fat, small enough to slip into a cell. The cell takes in the bubble, reads the blueprint, and builds the editing machine on its own factory floor.

Once the bubble is in the bloodstream, where does it actually go? Left to itself, it goes to the liver. The reason is a happy accident of the body’s own housekeeping. Part of the liver’s day job is filtering fats out of the blood, so it’s built to grab fatty particles. Injected into a vein, the bubbles drift downstream and the liver inhales them. The organ isn’t so much targeted as it is the path of least resistance. Which is why the opening wave of in vivo base-editing therapies reads like a catalog of liver diseases. The protein that keeps cholesterol high in the blood is made in the liver. So is the enzyme whose absence poisoned the infant KJ. So is the rogue protein behind transthyretin amyloidosis. The liver’s lucky address is also a cage. Everything the technology most wants to reach next sits outside the liver: the muscles, the neurons, the blood-forming stem cells in the marrow, the retina. Re-addressing the bubble is the frontier on which the next decade of this technology will be won or lost.

A Vaccine for Heart Attacks

The bolder bet belonged to a second company. In 2018 a Boston cardiologist named Sekar Kathiresan left his hospital post to chase it, founding Verve Therapeutics with the gene-editing researcher Kiran Musunuru. Kathiresan’s younger brother had died suddenly of a heart attack at 41, a man who felt perfectly fine right up until he didn’t. Kathiresan had spent his career studying the genetics of heart attacks at Harvard. His brother’s death turned a question into an obsession. Why do hearts fail like this, without warning, in people who feel perfectly fine?

Heart disease is the single largest cause of death on the planet. Some 18 million people a year, nearly 1 in 3. Verve set out to stop it in the most audacious way imaginable. Not a daily pill taken for life. One injection, given once. The target was handed to Verve by a quirk of human biology. Around the turn of the millennium, Helen Hobbs and Jonathan Cohen at UT Southwestern combed the genes of thousands of ordinary Dallas residents. One gene kept surfacing: PCSK9. About 1 in 50 of the Black participants carried a version that was simply broken. These people walked around with strikingly low LDL (the so-called bad cholesterol) their entire lives. Carriers of a dead PCSK9 gene suffered 88% fewer cases of coronary heart disease than everyone else. Nature had already run the experiment. Switch off this one gene and you appear to be protected from the world’s leading killer for as long as you live, at no apparent cost.

Verve’s bet is to reach into an ordinary adult’s liver and switch off PCSK9 with a single edit. Verve dosed its first patients in 2022. The first attempt stumbled. The original therapy ran into a safety signal in early patients. But the trouble didn’t trace to the edit itself. It traced to the delivery. So Verve re-engineered the vehicle. The reworked version, VERVE-102, proved far better tolerated. It’s the candidate the whole enterprise now rides on. In June 2025, Eli Lilly agreed to buy Verve outright. Roughly $1 billion in cash up front, plus a contingent promise of up to $300 million more, all in up to about $1.3 billion. It closed that July.

The radical idea hidden inside what looks like an ordinary cholesterol drug is the one the company is named for. Medicine, as practiced, waits for disease. The chest pain, the narrowed artery, the first heart attack. Only then does it intervene. Verve’s logic runs the other way. The Dallas carriers weren’t spared because they got a good drug late in life. They were spared because their cholesterol was low from birth. That’s not really treating heart disease. It’s vaccinating against it. One shot, delivered to someone who feels perfectly well, to head off a threat that hasn’t yet arrived.

Beam Therapeutics, founded in 2017 by Liu and two other gene-editing pioneers, is the other half of the field. It’s what the industry calls a pure-play. A company whose fortune rides on one idea, the molecular pencil, working inside human beings. In January 2022, Pfizer agreed to pay Beam $300 million in cash up front, plus up to $1.35 billion if three liver, muscle, and brain programs panned out. Pfizer was buying a promise: that a tool which had barely touched a patient could be aimed, organ by organ, at one disease after another. When the four-year collaboration ran its course at the end of 2025, Pfizer made good. It took a worldwide license to one of the liver programs and committed to carrying it the rest of the way. By early 2026 Beam was sitting on roughly $1.2 billion in cash.

So far, importantly, the trouble in human trials has come not from off-target edits but from the rest of the procedure. Verve’s second-generation version has so far stayed clean: no treatment-related serious events in its first fourteen patients, and cholesterol cut by as much as sixty-two percent. The measurable edits look reassuring. It’s the ones that might be slipping past the measurements (sitting silent and permanent in an otherwise healthy genome, waiting thirty years to matter) that keep the safety bar for vaccinating the well so far above the bar for rescuing the sick.

Medicine Becomes Programmable

For all that ingenuity, base editing can only do so much. The ceiling is built into the chemistry. The four DNA letters fall into two shape-families. The nip the editor performs can only convert a letter into its same-family partner: C into T, or A into G, and the reverse. The other kind of change is the problem. Swapping a letter for one from the opposite family would mean tearing out the whole letter and dropping in a structurally different one. The shaving trick can’t do that. Nor can a base editor add a missing letter or delete an extra one. The gap was glaring enough that Liu’s lab built a second, more versatile tool to close it. Prime editing, reported in 2019, works less like a chemical eraser and more like find-and-replace in a word processor. It can write a short new stretch of sequence into a chosen spot. In principle it reaches nearly nine in ten of the disease variants catalogued so far.

But even hemmed in, base editing’s reach is enormous. Among the tens of thousands of single-letter typos catalogued as causes of human disease, the within-family transitions are the clear majority. Roughly three in five. The adenine editor alone, in principle, could reverse close to half of every disease-causing point mutation on record.

The deeper shift in this technology is that medicine is becoming programmable. KJ’s case showed it most clearly. The whole therapy was designed, tested in animals, manufactured to clinical standards, and cleared by regulators in about six months. Swap a different guide RNA, twenty letters’ worth of sequence, and the editor goes to a different address. The platform is the same. The drug is different. A pipeline that used to take a decade per disease can, for a base editor aimed at a liver protein, run in months. The molecular pencil that fixed KJ’s enzyme could be retargeted at an entirely different patient’s mutation by swapping a single component.

And there’s a bigger asymmetric story now coming into focus. The same playbook (precise rewrites, lipid-nanoparticle delivery, organ-by-organ rollout) extends beyond rare diseases. Verve’s PCSK9 program is the leading example: an edit aimed at common cardiovascular disease, with potentially hundreds of millions of patients eventually in scope. Behind PCSK9 sit other once-and-done liver edits for diabetes, obesity, and metabolic disease. Beyond the liver, the technology is being aimed at the eye, the muscle, the immune system, the bone marrow. Targeted lipid nanoparticles are being engineered to deliver editors to specific cell types, edit a marker on the surface of blood-forming stem cells, and rebuild an immune system from inside a patient who never left the doctor’s office.

The unease grows in step with the power. A base edit can’t be undone. It’s written into the genome and copied into every descendant of every cell it touches, for the rest of a life. That permanence is a gift when the edit is right and a liability when it’s wrong. And the more ambitious the use (editing people who aren’t sick yet, to head off a disease that might never have come) the higher the bar climbs. Every harm is now a harm done to someone who was perfectly fine. The cautious path Verve has walked is to start with the sickest, where the math of risk and benefit tilts sharply in the edit’s favor. Proving the same trade makes sense for a basically healthy person in their 40s is a different and far longer road.

A child born today will likely live in a world where a single shot at fifty switches off a gene whose normal job was to raise their cholesterol, and the heart attack they would otherwise have had at sixty-five never comes. Where the rare disease their parents would have been told was untreatable is treated with a custom medicine designed and manufactured in months. Where a single permanent edit, delivered into the bloodstream, takes the place of a daily pill they would have remembered to take maybe half the time.

The cut is the blunt instrument. It breaks, it disables, it scrambles. The edit that refuses to cut is the subtle one. It settles onto one wrong letter and quietly changes it. And it may reshape medicine far more profoundly than the famous cut ever will.