Gene therapy's central engineering problem is not writing the genetic code—it is delivering it. Every therapy must get nucleic acid into specific cells without triggering dangerous immune reactions, and the field splits into two broad answers: hijack a virus, or build a synthetic carrier. Viral vectors such as lentivirus and adeno-associated virus (AAV) deliver genes with high efficiency because they exploit natural infection machinery, but they carry immunogenicity, cargo-size, and manufacturing burdens. Non-viral systems, especially lipid nanoparticles (LNPs), offer lower immunogenicity, larger payloads, and re-dosability, at the cost of delivery efficiency that still needs improvement. As industry analyses of gene therapy development put it, there is no universally optimal platform: vector selection is a product-specific optimization problem balancing efficacy, safety, durability, manufacturability, and cost.
Viral Vectors: Nature's Delivery System
Viral vectors work because viruses have spent millions of years solving exactly the problem gene therapy faces—getting genetic material inside cells. As the American Society of Gene and Cell Therapy's patient education materials explain, vectors are "essentially vehicles designed to deliver therapeutic genetic material, such as a working gene, directly into a cell," and researchers derive them from viruses "because viruses have proven to be very efficient at finding their way into cells." The four main viral vector families—adeno-associated viral, adenoviral, lentiviral, and retroviral—each carry distinct characteristics, uses, and limitations.
Lentiviruses are the workhorse of ex vivo cell engineering. They integrate their genetic cargo into the host cell's genome, which enables long-term expression and effective transduction of both dividing and non-dividing cells. That integration also supports complex payloads: multicistronic designs incorporating elements such as chimeric antigen receptors, regulatory sequences, and in some cases safety or control mechanisms, according to coverage of delivery platform choices for developers.
Integration, however, cuts both ways. Because lentiviral cargo inserts into the host genome, there is a theoretical risk that integration could alter the expression of endogenous genes, including genes involved in cell growth and proliferation. Modern self-inactivating vector designs have substantially improved the safety profile by reducing the potential for enhancer-mediated activation of nearby genes, but the consideration remains part of any lentiviral risk assessment.
AAV takes a different approach. It is known for safe and efficient in vivo gene therapy—injecting the therapy into the body so new genetic instructions reach cells directly. AAVs are usually non-integrating: the DNA they carry does not typically insert into the genome. That improves safety, but it means the therapeutic gene is not copied when a dividing cell divides, so the effect can be diluted over time in dividing tissues. The ASGCT materials also note that AAV's size capacity may determine which rare diseases it can target—a constraint that shapes the entire pipeline of AAV-based candidates.
Non-Viral Vectors: Synthetic Solutions
Non-viral delivery takes the opposite bet: rather than borrowing viral entry machinery, build a carrier from scratch. As Robin Oliveres writes in an overview of non-viral gene delivery methods for Inside Therapeutics, introducing genes into cells is difficult because negatively charged oligonucleotides struggle to cross the cell membrane, so synthetic systems must be specifically engineered for transfer efficiency, specificity, gene expression, and low toxicity.
Lipid nanoparticles are the leading example. LNPs encapsulate nucleic acids, protecting them from degradation and facilitating cellular uptake through endocytosis. Following the success of mRNA COVID-19 vaccines, LNPs are now being developed for mRNA therapeutics, CRISPR delivery, and siRNA modulation. They are re-dosable and non-integrating—two properties viral vectors struggle to match, since an immune response to a viral vector can block repeat dosing.
The safety advantages are consistent across sources. Single Use Support's Sonja Rieger notes that non-viral vectors are generally regarded as safer than viral vectors, posing a lower risk of triggering an immune response, with lower cytotoxicity and mutagenesis, and can transport a wide range of cargo including plasmid DNA, mRNA, and siRNA. Oliveres reaches the same conclusion: non-viral vectors offer lower toxicity, reduced immunogenicity, and the ability to deliver larger genetic payloads than viral vectors can.
The trade-off is efficiency. A scientific review of viral and non-viral vectors in gene therapy observes that non-viral vectors have low immunogenicity and short expression duration, making them more suitable for transient gene expression or suppression, but their infection efficiency still needs improvement. And LNPs currently target the liver preferentially—a limitation that LNP targeting strategies are actively trying to overcome to expand RNA therapeutics beyond liver-centric applications. Polymer and peptide carriers offer tunable chemistry but have even lower intracellular uptake relative to viral vectors.
Gene Delivery Systems: A Comparative Overview
The table below compares the two most consequential viral platforms and the leading non-viral platform across the parameters that actually drive development decisions: how the vector works, how well it delivers, how the immune system responds, how much cargo it can carry, whether it integrates into the genome, and how easily it can be manufactured at scale.
| Parameter | Lentivirus (LV) | Adeno-associated virus (AAV) | Lipid nanoparticles (LNP) |
|---|---|---|---|
| Delivery mechanism | Integrates genetic cargo into the host genome via natural viral infection processes | Engineered virus delivering genetic instructions directly to cells in the body (in vivo) | Synthetic lipids encapsulate nucleic acids, protecting them and enabling uptake via endocytosis |
| Delivery efficiency | Highly efficient; transduces both dividing and non-dividing cells | Efficient for in vivo delivery; high in vivo infection efficiency typical of viral vectors | Lower intracellular uptake than viral vectors; efficiency still needs improvement |
| Immunogenicity / safety | Can elicit immune response; theoretical risk that integration alters endogenous gene expression, mitigated by self-inactivating designs | Known for safety in in vivo use, but viral vectors can provoke severe immune responses that limit re-dosing | Lower immunogenicity, lower cytotoxicity and mutagenesis; relatively well-tolerated and re-dosable |
| Cargo capacity | Limited; supports multicistronic payloads such as CAR constructs | Limited; size capacity may determine which rare diseases it can target | Can deliver larger genetic payloads, including plasmid DNA, mRNA, and siRNA |
| Gene integration | Integrating—enables long-term expression | Usually non-integrating; effect may dilute in dividing tissues | Non-integrating; suited to transient expression or suppression |
| Manufacturing scalability | Relatively low volumetric productivity; complex downstream processing and scaling challenges contribute substantially to cost | Complex, costly manufacturing and purification; costs often $1–3M per patient, limiting accessibility | Synthetic production avoids viral manufacturing complexity; RNA-LNP manufacturing is advancing but still being scaled |
Read across the rows, the pattern is clear: every strength in one column is paid for in another. Lentivirus buys durable expression and high efficiency with integration risk and difficult manufacturing. AAV buys in vivo safety and efficiency with a small cargo bay and immune barriers to re-dosing. LNPs buy safety, payload flexibility, and re-dosability with efficiency and targeting limitations—currently concentrating delivery in the liver.
Choosing the Right Vector: Trade-offs and Applications
Why does this matter in practice? Because each therapeutic application weighs these parameters differently. A therapy that must permanently modify a patient's cells favors an integrating vector; a therapy that needs only brief expression of a gene-editing component favors a transient, non-integrating one. A small target gene fits AAV; a large or multi-part payload pushes toward non-viral systems. A disease in a tissue the immune system will remember after one viral dose raises the value of re-dosable LNPs.
The developer-focused analysis frames the decision explicitly: the choice depends on the target tissue and cell type, cargo size and composition, required duration of expression, dosing strategy, immunogenicity, manufacturing requirements, and cost of goods. The most effective approach is not necessarily the technology with the highest delivery efficiency in isolation, but the one providing the appropriate balance of efficacy, safety, durability, manufacturability, and scalability for a particular application.
Both vector classes share one unresolved risk: off-target expression. The clinical review notes that many pathogenic genes are tissue- or cell-specific in their expression, and widespread delivery by either viral or non-viral vectors can cause therapeutic genes to be expressed in non-target organs, leading to toxicity. It also flags that using viral vectors to deliver gene-editing tools can increase off-target effects due to prolonged expression—another reason transient non-viral delivery appeals for editing applications.
Rather than a winner-take-all contest, the field is described as entering a convergence era: combining platforms based on biological need, with AAV potentially delivering gene replacement, LNPs delivering CRISPR repair templates, and lentiviruses engineering immune cells that are later modulated in vivo. The practical takeaway for anyone evaluating a gene therapy program is to match the platform to the biology, not to a league table.
Strategic Vector Selection for Gene Therapy
When developing a gene therapy, carefully evaluate the specific therapeutic goals, target cell characteristics, required expression duration, and manufacturing constraints to select the most appropriate delivery vector, acknowledging that no single platform is universally optimal. The decision rests on the trade-offs documented above: lentivirus for durable integration at the cost of insertional risk and hard-to-scale manufacturing; AAV for efficient, safe in vivo delivery within a tight cargo limit and with re-dosing barriers; LNPs for safer, re-dosable, larger-payload delivery where efficiency and liver-preferential targeting are acceptable or engineerable.
The measurable indicator that vector selection is being made well is successful clinical outcomes and regulatory approvals for gene therapies, reflecting appropriate vector selection for specific indications. Watch, in particular, whether emerging LNP targeting strategies succeed in expanding delivery beyond the liver and whether manufacturing costs for viral products fall from their current per-patient range—two developments that would shift the trade-off calculus for the next generation of therapies.
