Background: Assessing delivery efficiency and tissue selectivity together

A lipid nanoparticle (LNP) designed with artificial intelligence (AI) achieved more than 90% transfection efficiency in mouse articular chondrocytes and showed greater knee-to-liver selectivity than a comparator, according to a study published in Nature Biotechnology. The study also reports Mmp13 gene editing in chondrocytes and cartilage protection in a mouse model of osteoarthritis. The findings are limited to experiments in mice; therapeutic effects in humans have not been established.

Delivery of RNA therapeutics requires assessing both performance in the target tissue and effects on non-target tissues. The researchers identify low tissue selectivity and off-target toxicity as constraints on LNP delivery. They explain that existing high-throughput screening has focused mainly on delivery efficacy for a single target, without adequately accounting for uptake by non-target cells. They therefore propose a design strategy that considers target-tissue performance and off-target delivery together.

Key findings: Cartilage delivery and knee-to-liver selectivity of K9 LNP

The researchers developed MOLEA, an AI-based, multi-objective LNP design system. It designs ionizable lipids by integrating high-dimensional representations of lipid properties, cell-type-specific transfection data, and multi-task optimization. Here, transfection means introducing genetic material into cells. MOLEA was designed to learn relationships between lipid structure and function across different cellular environments and identify candidates that improve mRNA delivery to target tissues while minimizing transfection of hepatocytes.

The cartilage-directed K9 LNP achieved more than 90% transfection efficiency in mouse articular chondrocytes. Its knee-to-liver selectivity was 13.5 times that of SM-102, the comparator described in the abstract as a clinical benchmark. The greater-than-90% figure refers to transfection efficiency in chondrocytes, while the 13.5-fold figure refers to the relative improvement in selectivity between the knee and liver. These results do not support interpreting the finding as a 13.5-fold increase in mRNA expression in the knee or as evidence that no delivery components were detected in the liver.

In a mouse model of osteoarthritis, the researchers achieved chondrocyte-specific Mmp13 editing. They report sustained cartilage protection, together with suppression of disease-related immune processes and matrix remodeling. However, the provided abstract does not specify the editing tool or editing rate, administration route or dose, or observation period. It is therefore difficult to quantify the magnitude or duration of cartilage protection.

Implications and outlook: A design strategy that also considers off-target delivery

This study presents an RNA-carrier design approach that considers both delivery efficiency in target cells and transfection of non-target cells. It incorporates tissue selectivity as an evaluation criterion from the candidate-lipid selection stage and links that approach to gene-editing and cartilage-protection results in a mouse model of osteoarthritis.

The abstract alone does not allow assessment of sample size, statistical uncertainty, how selectivity was calculated, or detailed toxicity results. Efficacy and safety in humans, as well as the possibility of low-dose treatment, remain unverified. Application to other tissues is a potential use proposed by the researchers and should be distinguished from the cartilage-delivery results reported here. Further studies will need to determine the administration conditions under which delivery selectivity is maintained and assess long-term safety and effects.

The key findings and methods can be checked in the provided Europe PMC bibliographic record and abstract.