The standard way to insert a gene into a human cell requires breaking both strands of the DNA double helix at a precise location. The cell, sensing catastrophe, rushes to repair the damage, and if the right donor template is nearby, it sometimes uses that template to patch the break, incorporating the new genetic material. This approach, homology-directed repair after a Cas9-induced double-strand break, has been the backbone of gene knock-in for a decade. It is also fundamentally constrained by the biology it depends on.
The double-strand break is the problem. The cell has multiple ways to repair a clean cut through both strands of DNA, and the one the researcher wants, homology-directed repair (HDR), is the cell’s least preferred option. The cell opts instead for non-homologous end joining (NHEJ), a faster, error-prone pathway that stitches the broken ends back together but frequently introduces small insertions or deletions, so-called indels. Even when HDR succeeds, the presence of indels in cells that also underwent NHEJ complicates the edited population. At the scale of whole organisms, undetected translocations, large-scale rearrangements in which broken chromosomes fuse to the wrong partners, remain a persistent concern.
A team led by Haifeng Wang at Tsinghua University has published a fundamentally different strategy in Nature. Called CRISPR kilobase-scale nickase-targeting (KNIT) editing, the approach does not break the genome at all. It nicks it, cutting only one strand of the DNA double helix, and uses a molecular recruitment system to deliver the donor DNA directly to the nick site. The distinction is not incremental. By avoiding the double-strand break, KNIT sidesteps NHEJ entirely, eliminates nearly all indels, minimizes translocations, and still achieves knock-in efficiencies that rival or exceed conventional methods.
How KNIT works
The system relies on two molecular components. The first is a Cas9 nickase, a variant of the standard CRISPR enzyme in which a single amino acid change (D10A) inactivates one of its two nuclease domains. The nickase cuts only the strand complementary to its guide RNA, producing a single-strand break, or nick, rather than a double-strand break. The second component is a fusion protein: the nickase is linked to monomeric streptavidin (mSA), a modified version of the bacterial protein that binds biotin with extremely high affinity.
The donor DNA, a double-stranded fragment carrying the gene to be inserted flanked by homology arms matching the target site, is chemically modified with biotin. When the nickase-mSA fusion cuts at the target site, the streptavidin domain immediately captures the biotinylated donor molecule and holds it in place. The cell’s endogenous repair machinery, operating on a single-strand nick rather than a clean break, then uses the tethered donor as a template to synthesize the new DNA across the nick. The result is precise integration of the donor sequence without ever cutting both strands of the genome.
This design solves a problem that has limited previous nickase-based approaches. Single nicks alone are poor at stimulating HDR because the cell treats a nick differently from a break. The repair machinery that promotes HDR is activated primarily by double-strand breaks. By recruiting the donor directly to the nick site, KNIT artificially concentrates the template where it can be used, compensating for the weaker repair signal.
Performance across cell types and insert sizes
The authors tested KNIT across multiple human cell lines, including HEK293T (embryonic kidney), K562 (chronic myeloid leukemia), Jurkat (T cell leukemia), and most importantly, primary human T cells isolated from healthy donors. Across these cell types, knock-in efficiencies reached up to 89 percent, depending on the locus and donor design.
The system handled insert sizes from 0.7 kilobases up to more than 10 kilobases, covering the range needed for most therapeutic transgenes, including full-length coding sequences with regulatory elements. Multiple genomic loci were targeted with similar success rates, suggesting the system is not highly sensitive to local chromatin context.
The safety profile represents the most significant advance. In standard CRISPR knock-in experiments, indel rates at the target site typically range from 10 to 50 percent or higher, reflecting the competition between NHEJ and HDR repair pathways. With KNIT, indels fell below 1 percent. Because there is no double-strand break, NHEJ has nothing to repair. High-throughput sequencing of edited cells revealed minimal off-target editing, and translocation assays showed a dramatic reduction compared with conventional Cas9-based methods.
Repeated editing without accumulating damage
One practical consequence of KNIT’s low indel profile is that cells can be edited more than once. Conventional Cas9 editing at a given locus creates a mix of correctly edited, indel-containing, and unedited cells. A second round of editing cannot distinguish the correctly edited cells from the unedited ones, but it will generate new indels in both populations, gradually degrading the quality of the cell pool. KNIT, with its sub-1-percent indel rate, allows sequential rounds of editing without accumulating junk mutations. The authors demonstrated this by successfully performing two rounds of KNIT editing at the same locus and observing no meaningful increase in indels.
The system also supports simultaneous editing at multiple loci, a capability increasingly important for engineering complex cell therapies. Multiplexed knock-in, targeting two or three loci in the same cell, produced efficiencies that remained useful for therapeutic applications, and translocation rates between the targeted sites were minimal.
KNIT Editor 2: one transfection, higher efficiency
The original KNIT system requires delivering both the nickase-mSA fusion protein and the biotinylated donor DNA into cells. The authors developed an enhanced version, KNIT Editor 2 (KE2), that consolidates the components into a single optimized vector, eliminating separate donor transfection and improving the stoichiometry between the nickase and the donor template. KE2 showed further improvements in knock-in efficiency across multiple cell types while maintaining the same low indel profile.
Restoring IL2RG in mutant cells
To test therapeutic utility, the team applied KNIT to cells carrying a pathological mutation in the IL2RG gene. Loss-of-function IL2RG mutations cause X-linked severe combined immunodeficiency (SCID-X1), a condition in which patients lack functional T cells and natural killer cells. KNIT inserted a functional copy of IL2RG into either the native locus or the AAVS1 safe harbor locus. In both cases, edited cells restored normal IL2RG expression, demonstrating correction of a disease-relevant genetic defect without the risks of double-strand breaks.
Building CAR-T cells without viral vectors or double-strand breaks
The most immediately translatable application may be in cell therapy engineering. Chimeric antigen receptor (CAR) T cells, in which a patient’s T cells are reprogrammed to recognize and kill cancer cells, represent one of the most successful classes of cell therapy. Current manufacturing relies on viral vectors (lentivirus or retrovirus) to deliver the CAR transgene, which adds cost, regulatory complexity, and a small but nonzero risk of insertional mutagenesis. Non-viral engineering methods exist, but most rely on Cas9-induced double-strand breaks, which trigger the same NHEJ problems.
KNIT enabled the production of functional CAR-T cells using only electroporated plasmid components, with no viral vectors and no double-strand breaks. The resulting CAR-T cells expressed the CAR at clinically relevant levels and exhibited potent cytotoxic activity against target cancer cells in vitro. In a mouse xenograft model of leukemia, KNIT-engineered CAR-T cells controlled tumor growth as effectively as conventionally engineered cells, without the indel burden.
The bigger picture
KNIT is not the only recent advance in large-DNA insertion without double-strand breaks. Systems based on prime editing, bridge RNA recombinases, and quadruple-pegRNA approaches have all demonstrated programmable knock-in with reduced collateral damage. What distinguishes KNIT is its simplicity: a single nickase fusion protein, a biotinylated donor, and standard electroporation conditions already used in clinical cell manufacturing.
The field is converging on a principle that would have seemed radical a few years ago: the safest way to insert DNA into the human genome is not to break it in the first place. KNIT provides a practical implementation with efficiencies that suggest clinical translation is worth pursuing. For ex vivo cell engineering, where cells are removed from the patient, edited, and returned, the pathway to clinical testing is reasonably clear. The first KNIT-engineered CAR-T cells may enter human trials within the next few years, and if the safety profile observed in the lab translates to patients, the approach could change how cell therapies are manufactured.