Engineered mini CRISPR enzyme gets a 60-fold power boost for gene editing

A tiny molecular scissor that once barely worked in human cells has been transformed into a high-performance genome-editing platform, thanks to a clever engineering strategy that swapped a handful of amino acids and trimmed the enzyme’s RNA guide. In a study published in Molecular Systems Biology, researchers report that a compact CRISPR nuclease called CgCas12n, which showed essentially no editing activity in mammalian cells in its natural form, can be rationally modified to achieve editing efficiencies up to sixty times higher than the wild-type enzyme. The work offers a blueprint for rescuing underperforming miniature CRISPR systems and adds a new, delivery-friendly tool to the growing arsenal of compact genome editors.

The appeal of miniature CRISPR nucleases is straightforward. The most widely used editors, such as Streptococcus pyogenes Cas9, are large multidomain proteins whose coding sequences, together with their guide RNAs and regulatory elements, often exceed the packaging capacity of adeno-associated virus (AAV) vectors, the workhorses of in vivo gene therapy delivery. Compact type V effectors sidestep this bottleneck. Cas12n nucleases, classified as type V-U4, are among the smallest RNA-guided DNA endonucleases known, spanning roughly 400 to 700 amino acids, and are considered evolutionary intermediates between ancestral transposon-encoded TnpB proteins and the larger type V CRISPR effectors such as Cas12a.

Cas12n also brings a distinctive targeting preference. Unlike Cas12f nucleases, which must dimerize to cut double-stranded DNA and favor T-rich PAM sequences, Cas12n enzymes are presumed to function as monomers and recognize an A-rich protospacer-adjacent motif. That preference is orthogonal to the G-rich PAM of SpCas9 and the T-rich PAM of AsCas12a, meaning Cas12n can reach genomic sites inaccessible to existing tools. Yet the family’s performance in mammalian cells is strikingly uneven. While some orthologs, such as those from Actinomadura craniellae and Rothia dentocariosa, edit human genomes, CgCas12n from Corynebacterium glutamicum, despite its compact 537-amino-acid length and close structural similarity to active variants, showed negligible editing activity in prior studies.

To find out why, the research team took a structure-guided approach rooted in electrostatics. Inside the Cas12n enzyme’s active complex, the crRNA and target DNA form a highly negatively charged RNA-DNA heteroduplex. The team hypothesized that boosting local positive charge on the protein surface could stabilize this interaction. Comparing CgCas12n sequences with those of AsCas12a and the active AcCas12n, they identified positions where CgCas12n carried neutral or negatively charged residues while its more capable relatives carried positively charged ones. From this analysis they built 52 single-amino-acid substitution variants and screened them using a fluorescence reporter assay in HEK293T cells, in which nuclease cleavage restores the reading frame of an out-of-frame EGFP gene through repair of the break.

Eleven variants showed significantly improved activity, with gains ranging from about 1.2-fold to 5.2-fold. The team then combined beneficial substitutions into double, triple, and quadruple mutants. The standout was a quadruple mutant designated v4.6, carrying the substitutions D157R, Q170R, T184R, and T423R, which exceeded 20 percent editing efficiency in the reporter assay, an approximately 60-fold improvement over the wild-type enzyme. At endogenous genomic loci with a 5′-AAG PAM in HEK293T cells, the variant achieved indel frequencies of 58.7 percent and 71.4 percent at two target sites, and comparable activity in HeLa and K562 cells. AlphaFold2 structural modeling placed all four introduced arginines near the crRNA-target DNA duplex, consistent with the idea that enhanced electrostatic interactions underlie the improved performance.

Protein engineering was only half the story. The researchers also dismantled and rebuilt the single-guide RNA scaffold. RNAfold-based secondary structure prediction divided the scaffold into four stem-loop domains, and complete deletion of any one abolished activity, confirming that all four contribute to function. Partial truncation screening revealed that only cuts within the second stem-loop retained substantial activity. The team further shortened the scaffold by deleting single base pairs from the tracrRNA-crRNA pairing region, and the best variant, called D19, preserved roughly 92 percent of wild-type activity while shaving 38 base pairs off the guide. Combining that truncation with the optimal stem-loop-2 trim produced the T6D19 scaffold, just 131 base pairs long, which actually outperformed the wild-type guide. Pairing the v4.6 protein with T6D19 yielded the fully engineered system, eCgCas12n, which roughly doubled average editing efficiency at endogenous loci compared with protein optimization alone, lifting average indel frequencies from about 20 percent to about 40 percent.

Importantly, the gains did not come at the cost of accuracy. Deep sequencing of 18 predicted off-target sites, identified with Cas-OFFinder allowing up to four mismatches, detected no significant indel formation, and the engineered system showed no increase in off-target editing relative to the wild-type configuration. Mismatch tolerance testing showed that editing dropped sharply when the guide carried single- or double-nucleotide mismatches, especially in the PAM-proximal region, reflecting the stringent target recognition typical of type V nucleases. In head-to-head comparisons at 18 shared target sites, eCgCas12n matched the editing efficiencies of SpCas9 and AsCas12a, cementing its credentials as a compact but competitive alternative.

The team then converted the nuclease into a base editor. They introduced a D297A mutation in the conserved RuvC domain to create a catalytically dead version, but early fusions with the deaminases TadA-8e and rAPOBEC, both optimized for Cas9-based systems, failed to produce detectable base conversion, suggesting that Cas9-compatible deaminases do not transfer readily to the single-RuvC architecture of Cas12-family effectors. Switching to the eCDA1 cytidine deaminase, fused to the N-terminus through an XTEN linker and augmented with tandem uracil glycosylase inhibitor domains, solved the problem. The resulting eCgCas12n-CBE mediated efficient C-to-T conversion across a broad editing window spanning spacer positions 2 through 17, reached efficiencies up to about 60 percent at individual loci with an average near 15 percent, showed little sequence-context bias, and produced desired edited alleles with purity above 95 percent while generating virtually no unintended indels. RNA sequencing and an orthogonal R-loop assay using catalytically inactive SpCas9 indicated only minimal off-target deamination at the transcriptome and genome levels.

As a functional demonstration, the researchers aimed the base editor at the mouse Dmd gene, which encodes dystrophin, the membrane protein whose loss causes Duchenne muscular dystrophy. By converting cytosines within glutamine, arginine, and tryptophan codons, the editor can install premature stop codons that trigger nonsense-mediated mRNA decay. In N2a cells and C2C12 myoblasts, eCgCas12n-CBE achieved editing efficiencies above 26 percent at the best target sites, with stop codon induction averaging about 10 percent. Edited C2C12 cells showed markedly reduced dystrophin and myosin heavy chain protein, decreased Dmd mRNA, impaired myotube formation, and downregulation of key myogenic markers, faithfully recapitulating the differentiation defects seen in DMD models.

The most striking result came in vivo. Because of its compact size, the entire eCgCas12n-CBE cassette fit inside a single AAV9 vector. Intramuscular injection into the tibialis anterior muscles of C57BL/6 mice produced C-to-T conversion at the Dmd target site with an average efficiency of about 11 percent, alongside significantly reduced dystrophin-positive myofibers, lower Dmd mRNA, and diminished dystrophin protein four weeks after treatment. The study also showed that eCgCas12n-CBE outperformed TSmini-CBE, a cytosine base editor built on the engineered ultra-compact enUnCas12f scaffold, at most of twelve disease-relevant human loci drawn from ClinVar. Together, the findings establish a generalizable strategy, pairing arginine enrichment of the protein with rational guide RNA minimization, for converting naturally weak miniature nucleases into practical editing platforms, and position eCgCas12n as a promising candidate for therapeutic applications where AAV delivery constraints make every base pair count.

Subject of Research: Engineering the compact CRISPR-Cas12n nuclease for efficient mammalian genome and base editing

Article Title: Engineering a compact CgCas12n platform for efficient genome and base editing in mammalian cells

Article References: Hou, B., Liu, X., Fan, J., Wang, W., Jin, C., Li, G., Zhang, Y., Yu, W., Huang, L., Li, X., Huang, X., & Li, K. (2026). Engineering a compact CgCas12n platform for efficient genome and base editing in mammalian cells. Molecular Systems Biology. https://doi.org/10.1038/s44320-026-00248-z

Image Credits: AI Generated

DOI: 10.1038/s44320-026-00248-z

Keywords: CRISPR, Cas12n, CgCas12n, genome editing, base editing, cytosine base editor, AAV9 delivery, guide RNA engineering, Dmd gene, Duchenne muscular dystrophy, miniature CRISPR, off-target specificity