Epigenome editing has emerged as a programmable alternative to cutting DNA. Rather than introducing double-strand breaks that can carry risks of unintended mutations and chromosomal rearrangements, epigenetic editors park a catalytically inactive CRISPR protein at a chosen genomic address and ferry enzyme domains that chemically modify histones or DNA, dialing gene expression up or down. The approach holds obvious therapeutic appeal, particularly for diseases in which silencing a gene, an enhancer, or a viral genome could produce lasting benefit. But the most widely used scaffolds, based on the large nuclease-dead Cas9 and Cas12a proteins, are simply too big for the delivery workhorses of gene therapy. Adeno-associated virus vectors, prized for their safety record and tissue tropism, carry only about 4.7 kilobases of genetic payload, and a full epigenetic editor with its guide RNA and effector domains routinely overflows that budget, forcing cumbersome dual-vector strategies that reduce efficiency and complicate manufacturing.

The Peking University team’s solution began with a systematic search of microbial genomes. Through iterative bioinformatic analysis using hidden Markov model profiling and sequence similarity searches, they cataloged miniature Cas12m effectors belonging to CRISPR subtype V-M, with particular attention to homologs that lack the canonical DED catalytic motif required for DNA cleavage. Previous work had established that some Cas12m proteins bind DNA programmably without cutting it, but their utility as epigenome-editing scaffolds remained unproven. Among the candidates, the enzyme from P. methylotrophicum, PmCas12m, stood out. Functional assays showed that it recognizes a flexible 5′-YTN-3′ protospacer-adjacent motif, a permissive target range far broader than many larger editors, and that it grips double-stranded DNA robustly while displaying no detectable DNA cleavage activity in vitro. Those properties, tight binding without cutting, are precisely the profile desired for a protein whose job is to serve as a molecular delivery truck for epigenetic modifiers.

To understand why PmCas12m binds but never slices, the researchers turned to cryo-electron microscopy, solving structures of the protein in complex with its CRISPR RNA and target DNA. The resulting maps, deposited in the Protein Data Bank and the Electron Microscopy Data Bank, reveal the architecture of a compact single-domain effector and trace the network of contacts that PmCas12m makes with the guide RNA, the DNA-RNA hybrid formed upon target recognition, and the displaced non-target strand. The structures explain how the enzyme accommodates its permissive YTN PAM and show that although the RuvC nuclease domain region is present in degraded form, the catalytic machinery needed to position metal ions for phosphodiester cleavage is functionally absent. In other words, PmCas12m is nuclease-dead not by accident of a single mutation but as an innate, evolutionarily entrenched property, a feature that removes the need for artificial inactivation and the performance penalties that often accompany it.

Structural insight then became a design tool. Guided by the cryo-EM maps, the team examined which parts of the protein were dispensable and which residues constrained or enabled function, and they applied deep mutational scanning, a technique that simultaneously tests the effects of thousands of amino acid substitutions, to map the protein’s tolerance landscape. Iterative rounds of rational deletion and mutation, informed by both the experimental scan and computational structure predictions, yielded xCas12m, a hypercompact variant that retains specific DNA recognition while acquiring highly potent epigenome-editing capability in human cells. The engineering strategy mirrors approaches that have successfully miniaturized other compact editors, such as Cas12f and IscB, but here it was applied to a protein that needed no catalytic rescue, only optimization of binding, expression, and compatibility with effector fusions.

With the scaffold in hand, the researchers built xCas12m-CRISPRoff, a silencing platform in which the miniature protein is fused to repressive chromatin-modifying domains of the kind used in canonical CRISPRoff systems, including KRAB repressor domains and methyl-CpG-binding protein modules. Tethered to gene promoters by single guide RNAs, the system laid down repressive epigenetic marks that persisted after the editing machinery itself was no longer detectably active, reproducing the durable, heritable transcriptional memory that makes CRISPRoff-style tools attractive for both research and therapy. The team also demonstrated the converse capability, showing that xCas12m fused to activation domains such as VPR could drive expression of endogenous genes in human cells, confirming that the miniature scaffold is a versatile platform for bidirectional control of transcription rather than a silencing-only tool.

The delivery advantage then became the centerpiece. Because xCas12m is so small, the entire xCas12m-CRISPRoff system, including the protein, its guide RNA expression cassette, and the necessary regulatory elements, fit within the packaging capacity of a single AAV vector. In a mouse model harboring hepatitis B virus covalently closed circular DNA, the persistent form of the viral genome that current antiviral drugs rarely eliminate, a single administration of the AAV-delivered editor produced durable epigenetic silencing of viral sequences. In treated animals, the researchers measured reduced levels of viral RNA, including pregenomic RNA, and diminished expression of viral antigens, indicating effective inhibition of hepatitis B virus infection through an approach that reprograms the chromatin state of the viral genome rather than cutting it. Because silencing leaves the viral DNA physically intact, the strategy could, in principle, be reversible or tunable, a safety profile distinct from nuclease-based antiviral editors.

The hepatitis B results carry particular weight because chronic HBV infection remains one of the world’s most consequential viral diseases, affecting hundreds of millions of people and driving cirrhosis and liver cancer. Existing therapies suppress replication but seldom achieve a functional cure, largely because the cccDNA reservoir and integrated viral DNA persist in hepatocytes. CRISPR nucleases have been explored as a means to excise or disrupt viral DNA, but they raise concerns about off-target cuts in the human genome and about delivery, since Cas9-based antiviral systems typically strain or exceed AAV capacity. An epigenetic editor that arrives in one viral particle, targets the viral genome with programmable guides, and extinguishes its transcription without touching the host sequence offers a conceptually different route toward functional cure, one that the authors position as a step toward clinical translation of epigenetic therapies.

Cautions remain before such promises can be evaluated in patients. Off-target DNA binding by any guide-directed protein must be mapped comprehensively across the genome, the durability and stability of the silenced state must be tracked over long periods, and immune responses to a bacterial protein delivered by AAV in the liver will need careful assessment. The specificity data reported for xCas12m in human cells are encouraging, and the compact size may itself reduce immunogenic surfaces, but these questions will only be settled in larger animal models and, eventually, carefully designed trials. What the study establishes now is a proof of principle that structure-guided discovery plus deep mutational scanning can compress an epigenome editor into a single-vector package without sacrificing potency. In doing so, it hands the field a new miniature chassis, xCas12m, that could carry epigenetic payloads to targets previously out of reach, from repressing disease genes to quieting persistent viral genomes, and it demonstrates how atomic-level structural biology continues to reshape what therapeutic genome engineering can look like.

Subject of Research: Structure-guided discovery and engineering of the miniature nuclease-dead CRISPR–Cas12m protein for AAV-deliverable epigenome editing and hepatitis B virus silencing

Article Title: Structure-guided discovery and engineering of miniature CRISPR–Cas12m for epigenome editing

Article References: Yu, T., Ji, M., Yu, D., Guan, Z., Zhu, R., Jiang, Y., Yang, Z., Qiu, L., Zhang, Z., Mu, J., Mao, F., Xiang, K., Bai, L., & Li, K. (2026). Structure-guided discovery and engineering of miniature CRISPR–Cas12m for epigenome editing. Nature Structural & Molecular Biology. https://doi.org/10.1038/s41594-026-01890-9

Image Credits: AI Generated

DOI: 10.1038/s41594-026-01890-9

Keywords: CRISPR, Cas12m, xCas12m, epigenome editing, CRISPRoff, AAV delivery, gene silencing, hepatitis B virus, cryo-electron microscopy, deep mutational scanning, PmCas12m, gene therapy