Gene editing tools have matured — delivery systems are unlocking the clinical bottleneck for therapeutics
Background
Gene editing technology has advanced dramatically over the past 20 years. From zinc finger nucleases (ZFNs) and TALENs to the CRISPR-Cas system, and subsequently base editors and prime editors, the toolset has expanded. The ability to cut or modify DNA at desired locations has already reached a sufficient level.
However, editing capabilities demonstrated in the laboratory do not immediately translate into patient treatment. Delivering the editing tool safely and efficiently to target cells remains the biggest challenge. The requirement to selectively reach specific tissues, minimize immune responses, and enable large-scale production further complicates the problem.
Key Findings
This review systematically organizes currently used delivery systems, dividing them into viral and non-viral categories.
Among viral vectors, adeno-associated viruses (AAVs) are the most widely used. They demonstrate sustained editing effects in specific organs, but their limited cargo capacity of approximately 4.7 kb, immunogenicity, and complex manufacturing process pose limitations. It is difficult to load large editing tools, and repeated administration reduces efficacy due to neutralizing antibodies.
Among non-viral platforms, ionizable lipid nanoparticles (LNPs) are leading the way. Clinical trials targeting the liver have reported up to 93% protein knockdown with a single dose. The fact that the vector does not remain in the body for a long time, as it temporarily expresses mRNA-based editing tools, is also an advantage.
Novel platforms, such as virus-mimicking nanosystems, cell-derived extracellular vesicles, cell-penetrating peptides (CPPs), and stimulus-responsive multifunctional scaffolds, are also emerging. These aim to target tissues other than the liver or to simultaneously achieve transient expression of the editing tool and programmable targeting.
The pace of delivery vector optimization is also accelerating. High-throughput barcoded screening combined with machine learning is being used to simultaneously evaluate thousands of candidate vectors and rapidly identify effective combinations. Strategies to introduce chemical modifications to the payload itself to increase in vivo stability and tissue specificity are also being pursued.
Significance and Outlook
The clinical expansion of gene editing therapeutics ultimately depends on the maturity of delivery technology. LNP-based liver-targeted editing has already yielded significant clinical data, and it can be considered that it is within reach of commercialization in the liver disease area.
The remaining challenge is tissues other than the liver. To expand the delivery range to various organs such as the brain, muscles, and lungs, new targeting strategies and immune evasion technologies must be in place. Cell-derived extracellular vesicles or stimulus-responsive scaffolds are attracting attention as candidates to fill this gap, but most are still in the preclinical stage, so it will take time to verify them clinically. How quickly machine learning-based vector design can shorten this process will be a key variable in the future.
Therapeutic genome editing has advanced rapidly with the development of diverse programmable nucleases, from zinc-finger nucleases and transcription activator-like effector nucleases to clustered regularly interspaced short palindromic repeats (CRISPR)-based systems such as base and prime editors. Despite these breakthroughs, clinical translation remains constrained by the challenge of achieving safe, efficient, and tissue-specific delivery. Viral vectors, particularly adeno-associated viruses, have enabled durable editing in selected organs but are limited by their restricted cargo capacity, immunogenicity, and complex manufacturing. Nonviral platforms, most notably ionizable lipid nanoparticles, have demonstrated remarkable efficacy for hepatic targets, with clinical trials reporting up to 93% protein knockdown after a single dose. An expanding set of emerging modalities, including virus-mimicking nanosystems, cell-derived extracellular vesicles, cell-penetrating peptides, and intelligent-responsive multifunctional scaffolds, further enriches the delivery toolbox by supporting transient expression and programmable targeting across diverse editors and tissues. Parallel advances in high-throughput barcoded screening and machine learning are accelerating vector optimization, while rational chemical modification of payloads improves in vivo stability and specificity. This review provides a comprehensive overview of current and emerging delivery systems for genome editing, highlighting key innovations, unresolved challenges, and interdisciplinary strategies poised to unlock broader therapeutic potential.
This review provides practical guidelines for delivery platform selection for gene editing therapeutic developers. If the target is liver disease, LNP is the best option with clinical data, and if the target is tissues other than the liver, it is necessary to accept the limitations of AAV's cargo capacity and immunogenicity or wait for preclinical data accumulation of new non-viral platforms. The practical value lies in clearly showing this decision-making branch. Barcoded screening and machine learning pipelines reduce the cost of exploring vector candidates, opening up a path for small biotech companies to optimize their own delivery vectors.