Two Labs Just Edited Human Embryos Without Breaking Their Chromosomes. The Last Team That Tried Couldn't Say That.
Base editing achieved zero aneuploidy and zero large deletions in embryos at two disease-relevant gene targets, yet persistent mosaicism and a widening ethics rift mean the clinic is still a decade away.
Zero. That is the number of chromosomal abnormalities found in human embryos edited with base editing technology across two independent studies published in June 2026: one in Nature by a Cambridge-led team, the other in a Columbia University preprint on bioRxiv. Compare that to the CRISPR/Cas9 results from six years earlier, when the same Columbia lab showed that standard gene editing caused frequent aneuploidy and large deletions in embryos, findings published in Cell in 2020 that effectively shut down serious talk of using CRISPR for heritable genome editing.
Safety math just changed. Not enough to start editing babies, but enough that the conversation has shifted from "can we avoid destroying embryonic chromosomes?" to "can we solve the remaining problems fast enough to matter for couples who need it?"
What the two studies actually did
Professor Kathy Niakan's Cambridge team at the Loke Centre for Trophoblast Research used base editing to knock out a gene called NANOG in early-stage embryos donated by IVF patients. Their goal was not therapeutic but developmental: understanding how human embryos assemble themselves in their first days of existence. They discovered that NANOG is essential for forming the epiblast, the cell layer that becomes the body, but dispensable for the placenta or yolk sac. In mice, losing NANOG disrupts both structures. Because base editing made it safe to run the experiment, they could check an assumption that nobody had been able to test directly, and it turned out the mouse blueprint does not transfer cleanly to human development.
At Columbia, Dieter Egli's team took a more provocative path, editing two disease-relevant genes in human zygotes at the single-cell stage: PCSK9, which regulates cholesterol, and HBG, which controls fetal hemoglobin. Editing worked with high efficiency when delivered as a protein, producing no chromosomal abnormalities and no large deletions while keeping small insertions or deletions rare. Embryos developed normally to the blastocyst stage, and the team even derived edited stem cell lines from them.
Then came the stark surprise: when Egli's team delivered the base editor as RNA instead of protein, the embryos arrested and simply stopped developing. Authors do not fully explain why, but the implication is concrete and unsettling because the delivery vehicle matters as much as the edit itself, and the wrong choice kills the embryo outright.
A safety ledger: what base editing fixes and what it does not
| Safety dimension | CRISPR/Cas9 (2020 data) | Base editing (2026 data) |
|---|---|---|
| Aneuploidy (abnormal chromosome count) | Frequent | Zero cases across both studies |
| Large deletions | Frequent | Zero cases |
| Small insertions/deletions | Common | Rare |
| Off-target edits | Significant | Present but guide-RNA dependent |
| Mosaicism | Common | Still present (needs 5 to 12 hr delivery window) |
| Delivery method | Multiple approaches tested | Protein works; RNA causes embryo arrest |
Those top three rows represent the catastrophic failures that made CRISPR/Cas9 unusable in embryos, and base editing eliminates all three. That is not a marginal improvement but the difference between a tool that randomly shreds chromosomes and one that does not.
Mosaicism, where not every cell in the embryo carries the intended edit, remains the stubborn holdout. According to the Columbia preprint, base editing would need to occur within five to twelve hours of fertilization to prevent mosaic outcomes, a timing window that adds real technical complexity to any clinical protocol. Off-target edits, while reduced, have not vanished either; their severity depends on which guide RNA sequence is used, meaning each target gene requires its own safety validation.
Where editing could actually matter: the transferable embryo problem
To understand why anyone would consider editing an embryo instead of just selecting an unaffected one, you need to see the numbers that IVF clinics deal with every day.
In preimplantation genetic testing for monogenic disorders (PGT-M), clinics biopsy embryos and test them for both the disease-causing mutation and chromosomal normality. A retrospective study of 2,344 embryos tested for monogenic conditions found that only 36.2% were both unaffected by the mutation and chromosomally normal, meeting the dual requirement for transfer.
For autosomal dominant conditions, where a single copy of the mutant gene causes disease, the math gets brutal: only 30.0% of tested embryos were suitable for transfer, and sixty percent were ruled out entirely. Among families carrying neurological disorders like Duchenne muscular dystrophy, spinal muscular atrophy, and Huntington's-related conditions, one study of 43 families found that just 12.45% of embryos met the bar for transfer.
Here is where base editing changes the arithmetic, at least on paper. In autosomal dominant conditions, Mendelian genetics predicts roughly half of embryos will carry the pathogenic variant. Of the 1,259 embryos tested in the autosomal dominant cohort, 378 (30.0%) were suitable, meaning they were both unaffected and euploid. Roughly 630 mutation-carrying embryos split into two groups: those that are also aneuploid (editing will not help) and those that carry the mutation but have normal chromosomes (editing could theoretically rescue them). If we apply the same ~60% euploidy rate observed in the unaffected group, roughly 380 mutation-carrying embryos had normal chromosomes but were discarded solely because they carried one bad gene.
If base editing could correct that single pathogenic variant without introducing new chromosomal damage, which is precisely what the 2026 data suggests is possible, the pool of transferable embryos would nearly double from 378 to approximately 758. For a couple who produced only three or four embryos in a cycle and got zero transferable results from PGT-M, that mathematical shift separates "try again" from "we have an embryo."
Nobody has run this calculation in the context of the new base editing safety data, and it deserves scrutiny alongside the caveats that follow.
Limitations
This analysis assumes base editing efficiency comparable to what Egli's team achieved at their two test targets, yet different disease genes will behave differently because off-target profiles vary by guide RNA and some loci may prove more resistant to clean editing. Mosaicism is not a theoretical concern: if only 70% of cells in an embryo carry the corrected gene, the child could still develop disease in the remaining 30%. We do not know the clinical threshold for "enough cells edited," and nobody will until long-term human follow-up data exists, which requires implanting edited embryos that nobody is currently permitted to transfer. Our transferable embryo pool calculation also does not account for the possibility that editing itself could introduce subtle damage not captured by current sequencing methods. Whole-genome sequencing at single-cell resolution would be needed to truly clear each embryo, a cost and complexity barrier that does not yet exist in routine IVF practice.
A strong case against moving forward
Krishanu Saha, a biomedical engineer at the University of Wisconsin-Madison, does not see medicine here. "I would not call it a breakthrough, and it does not establish genome-wide safety or clinical readiness," he told Scientific American. "I find it hard to think about a scenario where this is medicine." His point cuts deeper than technical safety: PGT-M already exists, embryo selection works for most couples, and the marginal benefit of editing goes to the hardest cases where couples have few embryos or unlucky Mendelian draws. But that marginal benefit requires accepting a technology whose long-term safety profile in humans is, by definition, unknown, because every edited child would be a decades-long experiment in a sample size of one. As Saha puts it, the "real safety involves the birth and long-term follow-up of the child."
Three major professional societies in gene therapy jointly proposed a 10-year moratorium on heritable human genome editing in May 2025. David Barrett, CEO of the American Society of Gene and Cell Therapy, called the Columbia preprint "unfortunate" and said it "flies in the face of the moratorium." Congress has, for roughly 30 years, barred federal funding for research in which human embryos are created or destroyed, so Egli's work was privately funded by the New York Stem Cell Foundation, Genomic Prediction, Inc., and international partners, sidestepping the federal ban without violating it. Alexis Komor of UC San Diego was blunter: "The cat's out of the bag... It kind of opens the floodgates."
R. Alta Charo, an emerita professor of bioethics at Wisconsin-Madison, offered the counterweight: the technology is "morally defensible" if proven safe and effective, especially given that IVF itself is "an expensive and difficult procedure" whose risks are "almost entirely absorbed by the woman." When embryos are limited, she argues, using genetic editing to avoid further rounds of IVF "really needs to be discussed."
A regulatory no-man's-land
Niakan's Cambridge study ran under a licence from the UK's Human Fertilisation and Embryology Authority, with ethics review from an NHS committee, and the system worked as designed: clear oversight, published approval, embryos destroyed within 6.5 days.
Egli's Columbia study followed a less transparent oversight path. He says an independent ethicist assessed the process and that IRB and embryonic stem cell research oversight (ESCRO) reviews were conducted, but Megan Allyse, a visiting bioethicist at Case Western Reserve University, noted that "there are lots of guardrails on here, and it's not clear how this particular research study threaded its way through those guardrails." Universities generally route privately funded embryo research through IRB and SCRO review voluntarily, and there is no legal requirement to do so. Until peer review of the submitted manuscript concludes, the oversight question lingers.
A gap between the UK's regulated framework and America's patchwork of voluntary committees and congressional funding restrictions is, itself, a finding worth examining. If base editing in embryos is going to advance toward clinical use, it will need a regulatory home, and the US does not currently have one.
What You Can Do
If you're a genetic counselor or reproductive endocrinologist: base editing data does not change clinical practice today, but it changes the conversation with patients who ask why editing is not an option. Where the honest answer used to be "it destroys chromosomes," it has become "it does not destroy chromosomes, but it creates mosaic embryos and we lack a regulatory path." Patients will learn that distinction whether clinicians teach it or not.
If you're an IVF patient carrying a monogenic disease: PGT-M remains the standard of care, and base editing in embryos is research, not treatment, with no clinic authorized to offer it. If you are in the 60 to 88% whose embryos do not pass PGT-M screening, the pipeline you are watching runs through three stages: mosaicism must be solved (technical, 3 to 5 years optimistically), then a regulatory framework must be built (policy, 5 to 10 years), then clinical trials must run (safety, 5+ years). Realistic timeline to clinical availability: mid-2030s at earliest.
If you're a policymaker: a 10-year moratorium was proposed before the safety data arrived, and the data that just arrived eliminates the worst technical objections while leaving the deepest ethical questions intact. A moratorium that forbids research is fundamentally different from one that forbids clinical use, and right now Congress conflates the two through funding restrictions rather than purpose-built regulation. Britain's HFEA model demonstrates that embryo research can be tightly overseen without a blanket ban, and whether you support or oppose heritable editing, the current US framework of voluntary university committees plus a congressional funding rider is not where you want the world's most consequential genetic technology to be governed from.
Bottom Line
Base editing has eliminated the chromosomal carnage that made CRISPR/Cas9 unusable in human embryos, and for the first time scientists can edit disease-relevant genes in early embryos without triggering the DNA repair failures that produce aneuploidy and large deletions. PGT-M data shows why this matters: for autosomal dominant conditions, only 30% of embryos pass screening, and editing could theoretically rescue a comparable number of mutation-carrying but chromosomally normal embryos from the discard pile. But "theoretically" is doing heavy lifting in that sentence because mosaicism is unsolved, delivery method is finicky (protein works, RNA kills the embryo), off-target editing has not vanished, a regulatory framework for clinical heritable editing does not exist in the US, and three major professional societies want a decade-long ban. What has changed is the ceiling: the technology no longer destroys what it touches. Whether society is ready to let it build is a different question, and it is the one that will define the next decade of genomics.