Bird-Derived Jumping Gene Rewritten to Insert DNA into Human Cells with RNA Alone

A team of researchers in China has scoured more than a thousand avian genomes for ancient molecular parasites and turned what they found into a new class of gene-writing tools. In a study published in Nature Biotechnology, the group led by Wei Li, Qi Zhou and Haoyi Wang of the Institute of Zoology at the Chinese Academy of Sciences reports the discovery of 159 R2 retrotransposons from 1,139 bird genomes and the engineering of variants that achieve site-specific gene integration in human cells at efficiencies of up to 60 percent in primary cells. The work represents a substantial expansion of a resource that many genome engineers consider one of the most promising alternatives to viral vectors and CRISPR-based integration systems.

R2 retrotransposons are non-long-terminal-repeat retroelements with a remarkable lifestyle. Rather than scattering randomly across the genome like many transposable elements, R2 elements insert with exquisite precision into a conserved site within the 28S ribosomal RNA gene cluster. They accomplish this through a mechanism called target-primed reverse transcription. The R2 protein recognizes and cleaves the target DNA site, and the exposed DNA end then primes reverse transcription of an RNA template carried by the element itself. Because the RNA donor both encodes the protein machinery and serves as the template for integration, the entire system can in principle operate with RNA as the only genetic material delivered into a cell, a property that has earned the approach the label all-RNA-mediated DNA integration.

The appeal of such a system for genome editing is considerable. Viral vectors, including lentiviruses and adeno-associated viruses, can insert therapeutic genes but do so with limited control over where the DNA lands, raising the risk of insertional mutagenesis. CRISPR-associated technologies, by contrast, typically require a double-strand DNA break at the target site, which can trigger chromosomal rearrangements, p53-mediated toxicity and unpredictable repair outcomes. R2-based integration sidesteps both problems: it makes no double-strand break and its natural target specificity confines insertion to the repetitive ribosomal DNA array, a locus that has been proposed as a genomic safe harbor because thousands of copies exist and disruption of individual copies is well tolerated.

The new study builds on earlier work from the same laboratory. In 2024, the team reported in Cell that a rationally engineered R2 retrotransposon could mediate targeted gene integration in mammalian cells using only RNA components. That first-generation system, however, was derived from a limited set of source elements, and its efficiency in clinically relevant primary cells remained modest. Subsequent structural and biochemical studies by other groups, including cryo-electron microscopy analyses of R2 complexes captured during target-primed reverse transcription, revealed how the protein and its RNA cofactors cooperate to nick, unwind and copy DNA, providing a mechanistic foundation for rational engineering.

To widen the search space, the researchers turned to birds. R2 elements are widespread across animal lineages, but avian genomes had been comparatively underexplored. The team downloaded and screened 1,139 avian genomes using a computational pipeline built on sequence homology searches, clustering tools such as CD-HIT, protein domain annotation with InterProScan, and multiple sequence alignment with MAFFT followed by maximum-likelihood phylogenetic reconstruction with IQ-TREE. From this survey they identified 159 full-length avian R2 retrotransposons, distributed across multiple bird orders, and systematically catalogued which protein regions and RNA untranslated-region elements were conserved across the family and which varied.

That comparative map proved essential for engineering. Focusing on an element designated R2Tg, the team dissected two previously under-characterized protein regions. Region 1, located near the amino terminus, turned out to be indispensable: truncating it sharply reduced integration efficiency, and the defect could be rescued by fusing the chromatin-remodeling factor HMGN1 or by substituting domains such as nucleosome-binding or chromatin-unfolding modules. Region 2, comprising spacer segments and N-terminal extensions, modulated activity in a length-dependent manner. The analysis also clarified the roles of nuclear localization signals and showed that Region 1 must act in cis with the rest of the protein, since supplying it in trans failed to restore function.

The RNA donor itself proved equally amenable to optimization. The researchers varied the length of the right homology arm, the poly(A) tail, and the 5′ untranslated region, finding that a short four-nucleotide right homology arm sufficed when paired with an appropriate poly(A) tract and an intact 5′ UTR. Covariation modeling across the 165 avian R2 5′ ribozymes revealed a conserved J1/2 secondary structure whose disruption by heterologous substitutions reduced integration, confirming that the ribozyme fold, not merely its sequence, is functionally important. A conserved buffer region in the 3′ UTR was likewise characterized for length and nucleotide composition. These donor-design rules distinguish the new toolkit from earlier R2 systems and contributed to marked gains in both integration efficiency and the proportion of full-length, intact insertions.

Whole-genome sequencing of edited cells confirmed high on-target specificity, with insertions concentrated in the ribosomal DNA array and minimal off-target activity. Long-term culture experiments showed that integrated transgenes, including a GFP reporter, maintained stable expression over months, and that copy numbers remained constant. The team also examined the fate of the integrated cassette, documenting DNA methylation dynamics at the CMV promoter and testing promoter and orientation configurations that maximize durable expression. Notably, re-delivery of R2 protein mRNA induced only negligible indel formation at ribosomal DNA loci, below 0.1 percent, and did not erode transgene expression, suggesting the system can be re-dosed without progressive damage to the target locus.

The most clinically significant results came in primary human cells. In primary T cells, the researchers screened viral accessory proteins for their ability to enhance integration and identified Vp4, the ORF4a protein from Middle East respiratory syndrome coronavirus, which suppresses antiviral stress responses. Co-delivery of Vp4 with an engineered R2 element designated R2SPs substantially improved integration while preserving cell viability and expansion. The team went on to integrate a full CD19 chimeric antigen receptor cassette into primary T cells, generating CAR-T cells that were positive for the receptor by day two and demonstrated cytotoxic activity against target tumor cells in coculture assays. Parallel experiments in primary natural killer cells and primary human foreskin fibroblasts showed that the same tool achieved robust, stable integration across cell types that are notoriously difficult to engineer.

The authors report efficiencies reaching 60 percent site-specific integration across human primary cells, a figure that, if reproduced across laboratories, would place RNA-only gene writing among the most efficient nonviral integration methods available. The sequencing data have been deposited in the Genome Sequence Archive for Human under accession HRA013312, and the analysis code is publicly available on GitHub, lowering the barrier for other groups to adopt and extend the resource. Several of the authors have filed patent applications related to the technology, signaling commercial interest in R2-based gene writing.

Challenges remain before such systems reach the clinic. The ribosomal DNA target, while abundant and tolerant of insertion, constrains applications to loci where expression from the rDNA array is adequate, and controlling transgene copy number and expression level will require further tuning. Delivery of large RNA molecules into primary cells still relies on electroporation, and the immunogenicity of both the R2 protein and coronavirus-derived accessory factors in vivo is untested. Nevertheless, by mining evolution’s own library of site-specific integrases and applying systematic protein and RNA engineering, the study demonstrates that retroelements long studied as genomic fossils can be refashioned into programmable tools, opening a path toward gene insertion therapies that require no virus, no DNA break and no DNA cargo at all.

Subject of Research: Engineering avian R2 retrotransposons for RNA-mediated targeted DNA integration in human cells

Article Title: Discovery and engineering of avian R2 retrotransposons for all-RNA-mediated targeted DNA integration in human cells

Article References: Chen, Y., Hu, Y., Luo, S., Mao, B., Tang, N., Wang, C., Wang, Y., Bai, H., Wang, X., Peng, C., Chen, Q., Xiao, J., Zou, Y., Wang, P., Niu, R., Zhang, J., Zhao, Y., Liang, C., Wang, Y., … Li, W. (2026). Discovery and engineering of avian R2 retrotransposons for all-RNA-mediated targeted DNA integration in human cells. Nature Biotechnology. https://doi.org/10.1038/s41587-026-03315-w

Image Credits: AI Generated

DOI: 10.1038/s41587-026-03315-w

Keywords: R2 retrotransposon, gene integration, genome editing, retroelement, avian genomics, target-primed reverse transcription, CAR-T cells, ribosomal DNA, nonviral vector, RNA donor, primary T cells, Nature Biotechnology