The CRISPR gene editing field spent a decade convinced the hard problem was cutting precisely enough. It was not. The hard problem was always delivery: getting the molecular scissors into the right tissue at the right dose without poisoning the patient in the process. A paper published this week in Nature Structural & Molecular Biology suggests the cutting problem is largely solved. What comes next is the harder one.
Researchers at the University of Texas at Austin, working with the Bay Area biotech Metagenomi Therapeutics and funded by the National Institutes of Health, identified a naturally occurring miniature CRISPR enzyme called Al3Cas12f and engineered it into a variant that achieves 90% editing efficiency in human cells. The work appears in a paper by Guan and colleagues, confirmed also by the NIH and by EurekAlert.
The size comparison explains why this matters. The gold-standard gene editing enzyme, Cas9, runs roughly 1,368 amino acids. AAV, the modified virus most commonly used to deliver gene therapies, has a packaging limit of about 1,000 amino acids. The new enzyme is 433 to 488 amino acids. For a decade, the field has been trying to squeeze oversized scissors into a container too small to hold them. This one fits.
The researchers used cryo-electron microscopy to solve the atomic structure of the enzyme and found its efficiency comes from a structural feature no one had predicted: an unusually extensive interface between its two subunits that locks the DNA-cutting complex into a stable, pre-assembled configuration before it needs to act. Other miniature CRISPR enzymes float loose inside cells. This one arrives ready.
They called the engineered version Al3Cas12f RKK. In human leukemia cells, the RKK variant reached 90% editing efficiency at a commonly used genomic target, up from under 10% for the natural enzyme. That number will appear in press releases. The more important experiment has not been done yet.
The next step, as the NIH frames it, is testing RKK when packaged inside AAV vectors and administered to a living animal. That is where the gap between cell line data and tissue data has killed most prior claims of compact CRISPR breakthrough. Prior Cas12f nucleases tested via AAV in mice for muscular dystrophy showed high efficiency in cell models and failed to replicate in animals. Whether RKK escapes that pattern is the question that matters.
AAV has its own delivery constraints. The same size advantage that makes compact nucleases attractive also means less room in the viral particle for the guide RNA and other components needed to target the nuclease. A 2025 review in Molecular Therapy Nucleic Acids noted that AAV delivery to the central nervous system remains inefficient and that pre-existing immunity in patients can neutralize the vector before it reaches its target. The review also flagged liver toxicity and dose-limiting immune responses as constraints at the systemic doses needed for broad tissue coverage.
Metagenomi, which provided the bacterial enzyme discovery and holds rights to the platform, is one of several companies racing to solve the delivery problem. The company has two compact nuclease programs: the UT Austin collaboration covered in the new paper, and a separate internal effort. Its January 16, 2025 pipeline update described ongoing IND-enabling work for its Wave 1 development candidates, with additional nominations planned for 2026.
If in vivo delivery succeeds, it opens tissues beyond the reach of ex vivo approaches: the brain, the heart, the muscle. If the AAV safety problem reasserts itself, non-viral alternatives like lipid nanoparticles regain relevance, and companies that built ex vivo pipelines keep their commercial moat for longer than the press release suggests.
