Editing RNA with Our Own Proteins: Argonaute 2 Reprogrammed as a Base Editor

Nearly every programmable tool for manipulating RNA in human cells has been borrowed from microbes, and Cas13, the RNA-targeting arm of CRISPR, remains the dominant scaffold. That foreign ancestry carries a cost: bacterial proteins are bulky, they provoke immune responses in patients who may already harbor pre-existing immunity, and they cleave bystander transcripts collaterally. A study published on October 20, 2026, in Nature Biotechnology by Robert Lusi, Steven Banik, and colleagues at Stanford proposes an elegant alternative: reprogram a human protein that already does this kind of work, namely Argonaute 2 (AGO2), the catalytic engine of RNA interference, which a short guide RNA directs to any complementary transcript.

The obstacle is that AGO2 ordinarily destroys what it binds, and the authors circumvented it by redesigning the guide rather than disabling the enzyme: complementary along both flanks but interrupted at its center, the guide forces the target mRNA to bulge into a single-stranded loop of some fourteen nucleotides, an ideal substrate for APOBEC3A, a human cytidine deaminase fused directly to AGO2. The resulting editor, ALTER, converts sequence-specific binding into a C-to-U edit without appreciably degrading the transcript; successive substitutions in both halves of the fusion then raised signal and lowered background. On endogenous transcripts such as PPIB and TYMS it matched or exceeded the Cas13-based editors CURE and RESCUE, averaging a 1.9-fold advantage with roughly four-fold fewer off-target edits across the transcriptome; at a dose matching Cas-level on-target activity, off-targets fell a further twenty-five-fold.

Because the editor is assembled from human proteins, the only foreign sequences are the engineered substitutions, the linker, and the junction between the two domains; screened through a neural network, these predicted a markedly smaller immunogenic footprint than the bacterial comparators. Delivery is a separate matter: ALTER reached target cells both as a ribonucleoprotein preloaded with its guide inside virus-like particles and as a gene in a single AAV vector.

Unfortunately, several caveats temper the promise: the work is confined to HEK293T cells, to a fluorescent reporter and housekeeping transcripts rather than a pathogenic mutation or an animal model; ALTER perturbed more genes than its Cas counterparts, even if no coherent pathway emerged; most residual off-target editing reflects intrinsic deaminase activity rather than misdirected guides; and nuclear transcripts remain poorly accessible without an added localization signal.

What remains open is inviting: whether design rules for loop-inducing guides generalize through high-throughput screening, whether other human effector domains can be grafted onto the same scaffold, and whether the predicted drop in immunogenicity survives a living immune system. The underlying proposition — that the toolkit of RNA medicine might be assembled from human parts — deserves rigorous testing.

Source: Lusi RF, Moncayo von Hase S, Model C, Banik SM. Engineered human Argonaute 2 for targeted C-to-U RNA base editing. Nature Biotechnology 2026. DOI: 10.1038/s41587-026-03359-y | PMID: 42902027 (open access)