The question has a real answer — it depends on four parameters
Gene therapy program teams debate AAV versus LNP from first principles at every platform decision point. The debate is sometimes framed as ideology (viral vs. non-viral) but the actual decision logic is clinical: payload size, immunogenicity, redosing requirements, and target tissue determine which vector class can physically accomplish the therapeutic goal. This article provides a structured framework for making that determination.
One note on scope: this comparison focuses on in-vivo delivery programs using systemic or near-systemic routes of administration. Ex-vivo programs (lentiviral transduction of harvested cells) operate under different constraints and are not the subject of this analysis. We also acknowledge that this comparison is not vendor-neutral — Gendelivr optimizes LNP formulations, not AAV capsids. We have tried to present the comparison accurately, including the cases where AAV is the stronger choice.
Decision axis 1: Payload size
The most fundamental difference between AAV and LNP as delivery vehicles is packaging capacity.
AAV has a hard physical limit of approximately 4.7 kb for single-stranded DNA genome packaging, with the effective coding capacity reduced to about 4.5 kb after accounting for the inverted terminal repeat (ITR) sequences required for packaging. The ITRs consume approximately 200 bp on each end.
In practice, this limit is severely constraining for base editors and prime editors. SpCas9 (the most common Cas9 ortholog in clinical use) encodes a protein of 1,368 amino acids, requiring approximately 4.1 kb of coding sequence. The sgRNA expression cassette adds another 250 bp. A promoter (e.g., hepatocyte-specific ApoE/hAAT) adds 700–800 bp. Total: approximately 5.2 kb — which exceeds the AAV packaging limit by more than 500 bp.
Various strategies have been developed to accommodate SpCas9 in AAV: using smaller Cas9 orthologs (SaCas9 at 3.2 kb, CjCas9 at 2.95 kb, ScCas9v1 at approximately 3.3 kb), split-AAV dual-vector delivery (two vectors with split-intein reconstitution of Cas9 activity), and truncated promoter/polyA sequences. Each strategy involves tradeoffs: smaller Cas9 orthologs have different PAM requirements and targeting efficiency profiles; dual-vector delivery requires co-transduction and reduces effective titer; truncated regulatory elements may reduce expression.
For adenine base editors (ABEs) and cytosine base editors (CBEs), the size problem is worse. ABE8e (a widely used ABE variant) fuses SpCas9 to the engineered adenine deaminase TadA-8e, producing a construct of approximately 5.5 kb. CBE3max is similar. Prime editors carrying pegRNA and the reverse transcriptase domain of MMLV RT exceed 6 kb. These editors are practically incompatible with single-vector AAV delivery without extensive engineering compromises.
LNP carries no packaging limit. mRNA encapsulation capacity scales with particle size and RNA encapsulation efficiency, not with a hard capsid volume. SpCas9 mRNA (approximately 5 kb) is routinely encapsulated in LNP formulations used in clinical CRISPR programs. ABE and CBE mRNAs are similarly accommodated. The payload flexibility of LNP is an unambiguous advantage for next-generation CRISPR cargo beyond standard SpCas9.
Decision axis 2: Immunogenicity
Both AAV and LNP activate the immune system. The relevant immunogenicity profiles are different enough that they require separate analysis.
AAV immunogenicity. Adeno-associated virus is a naturally occurring pathogen (or near-pathogen — it is replication-incompetent without a helper virus). Human immune systems have co-evolved with AAV, producing two immunological consequences: (1) substantial fractions of human populations have pre-existing neutralizing antibodies (NAbs) from natural AAV exposure, with seroprevalence varying by serotype (AAV2: 40–70%; AAV8: 20–40%; AAV9: 35–50%); and (2) T-cell responses to AAV capsid peptides presented by transduced cells can trigger cytotoxic killing of transduced cells, which has been observed clinically at high doses.
Pre-existing NAbs are the primary clinical gating factor. NAb-positive patients are typically excluded from AAV gene therapy trials. Depending on serotype, NAb seroprevalence is high enough to significantly narrow the eligible patient population — and to raise questions about the durability of benefit in treated patients as anti-AAV IgG titers rise post-treatment.
LNP immunogenicity. LNPs activate innate immunity primarily through the RNA cargo. Single-stranded RNA activates Toll-like receptors TLR7 and TLR8 in the endosomal compartment, producing type I interferon and pro-inflammatory cytokine release (IL-6, TNF-α). This response is substantially reduced — though not eliminated — by N1-methylpseudouridine (m1Ψ) modification of mRNA, now standard in clinical mRNA products.
LNP lipid components can also activate complement and trigger cytokine release independently of the RNA cargo, at doses sufficient to saturate Kupffer cells or exceed complement threshold concentrations. The dose-limiting toxicity for IV LNPs in clinical trials has generally been hepatotoxicity (elevated liver enzymes) and inflammatory cytokine release — both manageable with dose fractionation and anti-inflammatory premedication.
LNPs do not generate adaptive immune responses to a protein antigen (unlike AAV capsid), which means re-dosing is less constrained by T-cell memory. Anti-PEG IgM and IgG (discussed separately above) are the primary adaptive immune concern for LNP repeat dosing.
Decision axis 3: Redosing requirements
Single-dose, permanent correction: AAV is viable. AAV2 genome integrates episomally (extrachromosomal) in non-dividing cells and can persist for years without dilution by cell division. For slowly dividing or post-mitotic cells (hepatocytes in adults divide approximately once every 200–400 days), a single AAV dose can produce durable correction. This is the clinical experience with approved AAV gene therapies in hemophilia and lysosomal storage disorders.
Single-dose, transient expression with permanent editing: LNP delivers Cas9 mRNA transiently (t½ approximately 8 hours in hepatocytes). The editing event itself is permanent (as long as the genome is edited correctly), even though Cas9 expression is short-lived. For programs targeting hepatocytes with CRISPR mRNA delivery, a single LNP dose can achieve permanent genomic correction without prolonged Cas9 expression — which reduces off-target editing risk (less time for Cas9 to cut non-target sites) and avoids immune recognition of persistently expressed Cas9 protein.
Programs requiring multiple doses: LNP has a meaningful advantage. Anti-AAV IgG generated after first administration limits second-dose efficacy to near zero in most patients. The IgG response is robust and long-lived. Some programs have explored serotype switching (administering AAV8 first, then AAV9 second) to avoid pre-existing NAbs from prior treatment, but cross-reactive antibodies limit this approach. LNP can be re-dosed — with the caveat of anti-PEG antibody buildup on repeat administration. Pre-dose anti-PEG monitoring and potential PEG chemistry modification (shorter PEG chains, cleavable PEG-lipids) can mitigate the repeat-dose limitation.
Decision axis 4: Target tissue
Both vector classes show preferential hepatic delivery via IV administration — AAV through serotype-specific tropism (AAV8, AAV-LK03 are hepatotropic), LNP through ApoE-LDLR endocytosis. For liver targets, both are viable from a biodistribution standpoint.
For extrahepatic targets, the comparison changes substantially. AAV has natural tropism for CNS (AAV9 via IV at high dose, AAV2/9 via intrathecal or intraparenchymal injection), retina (AAV2, AAV5), and muscle (AAV1, AAV6, AAV9). These tropisms arise from natural capsid interactions with specific cell-surface receptors. Intrathecal and subretinal injection allow targeted delivery that largely bypasses the systemic immunogenicity concerns.
LNP hepatic selectivity via IV is difficult to redirect to extrahepatic targets without active targeting modifications (surface ligands, antibody conjugates). Unmodified IV LNPs achieve >90% hepatic biodistribution. CNS delivery via LNP requires either direct injection into CSF/brain parenchyma or surface modification with CNS-targeting ligands — neither approach is as clinically established as AAV for CNS programs.
The decision matrix
| Program type | LNP preferred | AAV preferred |
|---|---|---|
| Base editing or prime editing (large cargo) | ✓ | Constrained |
| Standard SpCas9 hepatic correction, single dose | ✓ | ✓ |
| Hepatic delivery, high NAb-prevalence population | ✓ | Excluded |
| Repeat dosing required (e.g., dose escalation) | ✓ | Severely limited |
| CNS delivery via systemic route | Difficult | ✓ (AAV9 IV) |
| Retinal delivery | Experimental | ✓ (AAV2 subretinal) |
| Small SpCas9 gene (<4.5 kb) hepatic, no redosing | Viable | ✓ (if NAb screen passes) |
Manufacturing and cost
Manufacturing considerations increasingly influence vector selection for programs beyond Phase 1.
LNP manufacturing via microfluidic ethanol injection is straightforward, scalable, and GMP-compatible. Process development is well-understood. Cost of goods is dominated by the ionizable lipid synthesis cost (custom ionizable lipids can be expensive at development scale) and mRNA synthesis. At commercial scale, LNP manufacturing cost per dose is competitive with small-molecule biologics.
AAV manufacturing is considerably more complex. Clinical-grade AAV requires baculovirus-based production in Sf9 insect cells (for large-scale) or transient transfection of HEK293 cells, followed by multi-step purification by affinity chromatography (using AVB Sepharose or similar) and ultracentrifugation to achieve genome titers of 10¹³–10¹⁴ vg/mL with acceptable full:empty capsid ratios. The cost of goods for high-dose IV AAV (doses of 10¹⁴–10¹⁵ vg for systemic programs) is very high, with production batches for a single patient in some CNS programs costing seven figures.
For hepatic programs where LNP is mechanistically viable, manufacturing scalability and cost of goods favor LNP — particularly for large patient populations where the per-patient dose cost of high-titer AAV becomes prohibitive at commercial scale.
The 2025 clinical landscape
As of mid-2025, the clinical pipeline has matured to a point where vector selection patterns are visible:
LNP-based CRISPR delivery has achieved clinical proof of concept in hepatic programs targeting transthyretin (TTR) and other liver-expressed disease targets. Programs from multiple companies have demonstrated in-human CRISPR editing with LNP-delivered Cas9 mRNA and chemically synthesized sgRNA. The durable editing efficiency and safety profile from these programs has validated the LNP + CRISPR mRNA approach as a clinical-grade hepatic delivery strategy.
AAV remains the dominant vector for CNS programs (SMA, spinal muscular atrophy), retinal dystrophies, and inherited metabolic diseases where the payload fits within the packaging limit, re-dosing is not required, and direct delivery to a privileged anatomical compartment (retina, CSF) limits systemic immunogenicity concerns.
The programs where the debate is most live are hepatic enzyme replacement candidates (some fit within AAV limits, some don't), base editing programs (most exceed AAV limits), and metabolic liver diseases requiring periodic correction (favoring LNP re-dosability over AAV single-dose durability).
Gendelivr focuses exclusively on LNP optimization. For programs where the analysis above points to AAV, we will say so — and refer to capsid engineering groups better positioned to serve those programs. The decision framework should follow the biology, not the vendor.