The Daily Inference
Science & Space · News

Safety Checks for Gene Editing Can Now Ask What Each Stray Edit Did

Salk Institute researchers have built a way to see unintended gene edits and their effects on gene activity in single cells. It is a laboratory demonstration, not yet a safety test for any therapy.

Researchers at the Salk Institute have built a method that reads where CRISPR-Cas9 cut a cell's DNA and what those cuts did to its genes, one cell at a time.

The tool, called Superb-seq, was published in Nature Methods on November 16, 2026. The team was led by Michael Lorenzini, Brad Balderson and senior author Graham McVicker. [1][2]

It targets a worry that sits behind every plan to edit human genes.

Cas9 is a molecular pair of scissors steered to a spot in the genome by a short guide RNA. Sometimes it cuts elsewhere, leaving what are called off-target edits. They are changes nobody planned, in places nobody was watching.

The gap it fills

An observed guide RNA does not prove the intended edit happened, and it does not reveal what else happened in the same cell. Bulk sequencing, which grinds up many cells and reads them together, can flag candidate off-target sites. But it blends the cells into an average, so it cannot say which cell carries which edit, or how that cell's genes behave afterward. [2]

A popular technique called Perturb-seq has the same limit. It reads guide RNAs as a stand-in for edits rather than measuring the edits each cell actually received. The Superb-seq authors report that their approach gave better estimates of how gene perturbations change gene activity than Perturb-seq did. [2]

How it works

The trick is to make every cut leave a mark. Cas9 cuts the DNA, and the cell's own repair machinery inserts a small DNA tag into the break. The tag contains a promoter from the T7 virus, a start signal that lets an enzyme make RNA copies of the DNA next to it. [2]

Those copies are made inside cells that have been chemically fixed in place. Single-cell RNA sequencing then reads two things at once from each cell: the T7 transcripts that reveal where the edit landed, and the cell's ordinary RNA, which shows which genes are switched on. [2]

The result is a ledger with one line per cell. It lists the edits that cell carries and the activity of its genes.

What it found

The team tested Superb-seq on 35,000 cells from three cell lines. In one 10,000-cell experiment, they used seven guide RNAs aimed at four chromatin-remodeling genes in K562 cells, a human leukemia line. Chromatin-remodeling genes help control how tightly DNA is packed and so which genes can be read. [2]

The method detected 5,761 edited cells and 43 edit sites. Of those, 36 were off-target. [2]

That is 36 of 43 sites, about 84 percent, outside the intended targets. The experiment used a small set of guides on a cancer cell line, so the figure describes that experiment, not CRISPR in general.

Across the study, 19 off-target edits were associated with differences in gene expression. Nine of those associations appeared only in a particular cell type. [2]

That second finding is the one with consequences. The same stray edit can matter in one kind of cell and do nothing visible in another, which is hard to see when cells are averaged together.

Balderson said some off-target edits could disrupt genes that protect against cancer or other unhealthy cell states. Knowing which cells carry edits, and whether their activity changes, helps tell likely harmless edits from potentially important ones, he said. [1]

What it does not yet show

The work was done on cells in a dish. It did not test a therapy in patients, and it did not evaluate agricultural gene drives.

The researchers propose uses in basic research and in early assessment of gene-therapy candidates. They also suggest that differences between people's genomes could affect where off-target edits fall, and that single-cell data could help compare those risks. All of that is proposed, not demonstrated. [1][2]

The method has limits of its own. In validation work, the promoter tag labeled edits less efficiently in Jurkat and U2OS cells than in K562. [2]

No one yet knows how many edits Superb-seq misses, including cuts that never pick up the tag, or how detection varies by guide, cell type and location in the genome. The gene-expression changes it links to off-target edits have not been shown to cause lasting harm to a cell.

The larger tests are still ahead. Therapies are built on primary cells taken from patients, not on long-established laboratory lines. Whether Superb-seq works as well there is open.

So is whether it can follow editors that do not make Cas9's clean double-strand cut. And the study ran no gene-drive experiment, so whether the method would help assess agricultural gene drives is untested. [2]

What the paper does establish is narrower and still useful. For the first time in this setup, a cell's stray edits and its gene activity can be read from the same cell, so a safety check can ask what each edit did instead of how many there were.

Topics: Gene editing · Medical research

Every edition in brief, three times a day, on our Telegram channel, on Bluesky and on Threads.

Sources
  1. New tool tracks unintended CRISPR edits and their effects in individual cells Phys.org
  2. Joint single-cell capture of Cas9 edits and transcriptomes reveals on- and off-target effects on gene expression | Nature Methods nature.com