← Back to Home

AI-Designed Protein Carriers: RNA Delivery Advances to Gene-Editing Experiments in Mice and Pigs

Researchers selected STV-C8 from more than 100 synthetic carriers, improving RNA delivery efficiency in cell culture and demonstrating function in animal experiments. But a critical gap remains between getting into cells and safely reaching the tissues needed in humans.

By SURL BioNews

RNA can enable cells to make needed proteins and carry instructions for producing gene-editing tools; the challenge is delivering it intact into cells. A German research team combined natural protein components with structures designed using generative AI to build “synthetic transfer vehicles” (STVs), seeking a new delivery route for RNA therapies.

The study, published in *Nature* and involving Helmholtz Munich, the Technical University of Munich, and other institutions, used AI to design the carriers’ protein scaffolds. Researchers produced and tested more than 100 combinations of different sizes and shapes, finding that some structures departing from the geometry of natural viral shells performed particularly well, with STV-C8 proving the most efficient. Whether a design worked was ultimately determined by actual delivery experiments.

In cell-culture comparisons, STV-C8 outperformed the tested lipid nanoparticles and virus-like particles in RNA delivery. The paper’s abstract reported efficiency differences of several orders of magnitude in some comparisons; institutional accounts also noted that STV-C8 required less RNA to achieve similar protein output. These advantages are limited to the cells, payloads, and comparison conditions used in the study and cannot be directly taken to mean that it is broadly superior to existing delivery technologies.

Another feature of the carriers is that their components can be adjusted. Researchers added computationally designed peptides capable of binding to specific targets to change which cells the carriers preferentially entered, and tested different RNA payloads. This extends the role of protein design from building containers to adjusting delivery targets; however, recognizing targets in cultured cells and precisely reaching them within complex organs remain challenges at different levels.

Animal experiments provided the next layer of evidence. After intravenous injection in mice, the delivered RNA was expressed mainly in the lungs, and the team analyzed its distribution in the body at near-single-cell resolution. The research institutions said the experiments showed no evidence of immune or toxic side effects; this observation is limited by the scope of the testing and is not yet sufficient to establish safety for human use or repeated dosing.

The study also tested patient-derived cells and pigs using a gene-editing strategy related to Duchenne muscular dystrophy. In experiments involving local injections into pig muscle, STV-C8 delivered relevant components of the CRISPR/Cas9 system and successfully removed a disease-associated sequence from the dystrophin gene. This demonstrates that the carrier can support a specific editing step, but does not mean that muscle function has improved, nor does it answer how to deliver the tools to muscles distributed widely throughout the body.

The team plans to establish a spin-off company to advance the technology, but STV-C8 remains an experimental platform. The central next steps are to clarify how the carrier is distributed in the body, whether it can consistently target the required cells, and whether its safety holds up under more comprehensive evaluation. The study demonstrates that AI protein design can translate into testable biomedical function; becoming a therapeutic tool will still require clearing each of these hurdles.

References

  1. Phys.org
  2. Helmholtz Munich
  3. Technical University of Munich