AAV Dose Translation from Mouse to Human

Adeno-associated virus (AAV) has become one of the most important delivery platforms in modern gene therapy. For many rare genetic diseases, especially disorders caused by loss-of-function mutations, AAV-mediated gene replacement offers the possibility of durable therapeutic benefit after a single or infrequent administration. Yet moving an AAV program from animal studies into first-in-human clinical testing raises a central translational question: how should the starting human dose be selected and justified?

A common mistake is to convert a mouse dose directly into a human dose using the same vector genomes per kilogram (vg/kg) value, perhaps with a generic safety factor. This approach can be misleading because AAV dose-response is shaped by target-organ size, species-specific transduction, capsid biology, immune recognition, biodistribution, product quality, potency, and route of administration. The result is that two species receiving the same nominal vg/kg dose may experience very different biological exposure and safety profiles.

A defensible AAV first-in-human dose rationale is therefore not a single mathematical calculation. It is a structured scientific argument that integrates efficacy, biodistribution, toxicology, potency, manufacturing comparability, route-specific constraints, patient risk, and regulatory strategy.

• Mouse data are valuable for proof of concept, but they should rarely be used alone to define a human starting dose.
• For systemic AAV, non-human primate or other relevant large-animal data often provide critical information on biodistribution and safety.
• For local delivery, such as ocular or CNS administration, dose units such as vg/eye, vg/injection site, or total vg in a compartment may be more meaningful than vg/kg.
• Regulators generally expect a product-specific, scientifically justified dose rationale rather than a fixed conversion formula.
• CMC and analytical comparability are central to dose translation because a nominal vg dose is only meaningful when vector potency and quality are well characterized.

Why AAV Dose Conversion Is More Than a vg/kg Calculation

AAV doses are commonly reported as vector genomes per kilogram of body weight (vg/kg). This unit is useful for systemic delivery because it normalizes dose by body weight and allows comparison across animals and patients. In localized applications, doses may instead be reported as total vector genomes per eye, per injection site, per brain region, or per administration compartment.

The problem is not the vg/kg unit itself. The problem is the hidden assumption behind a direct mouse-to-human conversion: that each kilogram of body weight has the same relationship to vector distribution, target-organ exposure, immune activation, and therapeutic response across species. For AAV, this assumption is often false.

AAV vectors are biological delivery systems. They circulate, distribute, bind to cell-surface glycans or receptors, enter cells, traffic intracellularly, deliver vector genomes to the nucleus, and drive transgene expression. They can also trigger innate and adaptive immune responses. Therefore, AAV dose selection must account for both pharmacologic exposure and biological activity.

Current regulatory expectations reflect this complexity. FDA guidance for human gene therapy in rare diseases states that preclinical programs should help identify a biologically active dose range, recommend an initial clinical dose and escalation plan, establish the feasibility and reasonable safety of the clinical route of administration, and identify toxicities and monitoring parameters. EMA guidance similarly emphasizes that dose selection should be linked to product potency and supported by quality and nonclinical data.

Why Simple Mouse-to-Human AAV Scaling Can Be Misleading

Several biological and technical factors can systematically change the effective AAV dose when moving from mouse models to humans.

Target-organ size and cell number are not proportional to body weight. In liver-directed AAV therapy, for example, the liver represents a different fraction of total body weight in mice, non-human primates, infants, and adults. Even when the same vg/kg dose is administered, the number of vector particles available per hepatocyte may differ substantially. The same principle applies to retina, muscle, CNS regions, and other target tissues.

AAV receptor biology and glycan presentation differ across species. AAV serotypes rely on different cellular attachment factors and entry pathways. AAV9, for example, is influenced by terminal galactose-containing glycans rather than a single universal receptor. Differences in glycan abundance, accessibility, and tissue distribution can make rodent transduction efficiency a poor standalone predictor of human transduction.

Human immune biology is not captured by specific-pathogen-free mouse models. Patients may have pre-existing anti-AAV antibodies, memory T cell responses, complement activation risk, baseline inflammation, or disease-related organ vulnerability. Even when patients are seronegative by a neutralizing antibody assay, innate immune activation and anti-capsid T cell responses may alter effective delivery and safety.

Product quality affects the meaning of dose. Two AAV lots with the same genome titer may differ in empty/full capsid ratio, genome integrity, potency, aggregation, residual DNA, residual proteins, infectivity, and overall biological activity. A qPCR or ddPCR genome titer is not equivalent to a functional potency measurement. A dose reported as vg/animal in an early publication may therefore not be directly comparable to a clinical-grade lot.

Allometric scaling may provide supportive context, but it is not a universal answer for AAV. Classical allometric scaling is useful in conventional pharmacology because many physiological processes scale nonlinearly with body weight. AAV, however, behaves as a biological particle that distributes, binds, enters cells, expresses genetic cargo, and triggers immune responses. For that reason, allometric scaling should support, not replace, AAV-specific biodistribution, pharmacology, toxicology, and potency data.

Four Common Approaches to AAV Dose Translation

No single method applies to every AAV program. The most appropriate dose-translation strategy depends on the route of administration, target tissue, disease biology, capsid, vector genome, therapeutic window, patient population, and available preclinical data.

Table 1. Common AAV dose-translation methods and when to use them

Dose translation approach	Core logic	Best-fit use cases	Key limitations
Body-weight scaling	Converts animal dose directly into vg/kg for humans, often with a safety factor.	Early internal estimation; systemic programs without a clearly dominant target organ.	Oversimplifies species differences in biodistribution, target-organ size, transduction efficiency, and immune response.
Body surface area or allometric scaling	Applies nonlinear scaling based on body weight or body surface area.	Supportive calculation for systemic exposure or toxicity assessment.	AAV is not a conventional small molecule; capsid biology, receptor binding, and tissue tropism limit direct applicability.
Target-organ scaling	Adjusts dose based on target-organ mass, cell number, or anatomical compartment.	Liver-directed, ocular, CNS, and other organ-focused AAV programs.	Requires reliable organ-size, cell-number, and distribution data; may not capture immune or off-target effects.
NHP-anchored scaling	Uses non-human primate efficacy, biodistribution, and toxicology data as a translational anchor.	High-dose systemic AAV programs and IND-enabling dose justification.	NHP studies are costly, often small, and still imperfect predictors of human response.

 

In practice, most clinical AAV programs use a combination of these methods. A body-weight calculation may provide a rough starting point, target-organ scaling may refine expected exposure, and NHP data may help define safety margins and monitoring priorities. The final clinical dose rationale should make the assumptions behind each method explicit.

Route of Administration Changes the AAV Dose-Scaling Logic

AAV dose translation should begin with the delivery route. The same total vector dose can carry very different implications depending on whether it is administered intravenously, intrathecally, intracisternally, subretinally, intravitreally, intramuscularly, or directly into a specific tissue compartment.

For systemic intravenous AAV, total vector burden is a major safety consideration. Many AAV serotypes expose the liver substantially after systemic delivery, even when the intended target is muscle, CNS, or another tissue. For local delivery, body weight may be less relevant than local anatomy, injection volume, tissue accessibility, target-cell number, and local inflammatory risk.

Table 2. Route of administration changes the AAV dose-scaling logic

Route or indication type	Preferred dose unit	Key scaling considerations	Practical recommendation
Systemic intravenous AAV	vg/kg and total vg	Total vector burden, liver exposure, immune activation, body weight, and total capsid load.	Use mouse data as an early reference; anchor clinical dose with biodistribution and relevant large-animal toxicology.
Liver-directed AAV	vg/kg, total vg, and estimated vg per hepatocyte	Liver mass, hepatocyte number, transduction efficiency, secreted protein target level, and liver safety biomarkers.	Combine target-organ scaling with NHP or other relevant large-animal data whenever possible.
Ocular AAV	vg/eye or vg/injection	Injection volume, retinal target-cell number, local inflammation, delivery route, and vector spread.	Do not use vg/kg as the primary logic; benchmark against similar ocular AAV programs and delivery routes.
CNS-directed AAV	vg/injection site, vg/CSF compartment, or total vg	CSF volume, local diffusion, target-region coverage, DRG exposure, and route-specific anatomy.	Select dose based on anatomical distribution and safety margins, not body weight alone.
Neuromuscular systemic AAV	vg/kg and total vg	Muscle mass, liver uptake, immune activation, patient age, disease stage, and total vector burden.	Requires conservative escalation, strong safety monitoring, and relevant large-animal data.

 

A Practical Framework for AAV First-in-Human Dose Selection

A strong AAV first-in-human dose rationale can be built as a layered evidence package. Each layer should connect the proposed clinical dose to the biology of the disease, the behavior of the vector, the route of administration, and the quality of the product.

Step 1: Define the therapeutic target and required biological effect

The first question is not what dose worked in mice, but what biological effect is required in humans. For a secreted protein therapy, the goal may be to reach a circulating protein level above a disease-modifying threshold. For a CNS therapy, it may be transduction of a specific neuron or glial population. For a retinal therapy, it may be sufficient expression in photoreceptors or retinal pigment epithelium. For a muscle disease, it may be transgene expression across enough muscle fibers to change function.

• Transgene protein concentration or enzyme activity
• Vector genome copies per target cell or target tissue
• Biomarker normalization in tissue, blood, CSF, or ocular fluid
• Functional rescue in a disease-relevant model
• Histological, imaging, or electrophysiological evidence of target engagement

Step 2: Identify the minimally effective dose range

A minimally effective dose should be supported by relevant in vitro, ex vivo, and in vivo studies. Mouse models can be highly informative when they capture the disease mechanism, but a single effective mouse dose is not enough. Whenever possible, the dose-response relationship should include a no-effect dose, minimally active dose, pharmacodynamically active dose, plateau dose if present, and toxic or poorly tolerated dose.

Step 3: Generate biodistribution and expression data

Biodistribution data are central to AAV dose translation. They help determine where the vector goes, how long it persists, which tissues express the transgene, and which non-target tissues may be at risk. For systemic AAV, biodistribution should include target organs and major off-target tissues such as liver, spleen, heart, gonads, dorsal root ganglia, and relevant immune tissues. For local delivery, biodistribution should address the injected compartment and adjacent tissues.

Step 4: Establish toxicology and safety margins

Toxicology studies should reflect the proposed clinical route, dose range, dosing schedule, and monitoring plan as closely as feasible. For AAV, safety evaluation often pays particular attention to liver injury, complement activation, thrombocytopenia, dorsal root ganglia findings, inflammatory cytokines, transgene-related toxicity, vector shedding, and germline biodistribution risk.

A safety factor may be applied, but there is no universal one-size-fits-all safety factor for every AAV program. The appropriate margin depends on disease severity, treatment alternatives, patient age, route of administration, dose-limiting toxicities, immune risk, therapeutic window, and the strength of the nonclinical package.

Step 5: Integrate product quality and potency

Dose translation is only meaningful if the vector product is well characterized. A clinical dose based on genome copies must be interpreted alongside potency, empty/full ratio, genome integrity, residual host-cell DNA, residual plasmid DNA, protein impurities, aggregation, endotoxin, sterility, replication-competent AAV testing, and comparability between preclinical and clinical lots.

Step 6: Design dose escalation and stopping rules

Clinical safety depends not only on the starting dose, but also on the escalation design. High-risk systemic AAV programs may require sentinel dosing, staggered enrollment, predefined stopping rules, intensive laboratory monitoring, an immunosuppression plan, and independent data safety monitoring. Delayed immune-mediated toxicities may emerge weeks after dosing, so the monitoring period should be matched to the expected biology of the vector and transgene.

Table 3. Key questions for AAV first-in-human planning

Development question	Why it matters	Data needed
What biological effect is required in humans?	Defines the minimum therapeutic exposure needed.	Target protein level, enzyme activity, vector copies per target cell, biomarker response, and functional rescue.
What is the minimally effective dose in relevant models?	Helps avoid starting too high or too low.	Dose-response studies, disease-model efficacy, and pharmacodynamic markers.
Where does the vector go after administration?	Identifies target and off-target tissue exposure.	Biodistribution, persistence, clearance, and transgene expression.
What toxicities appear at clinically relevant doses?	Defines safety margins and monitoring needs.	Toxicology, liver enzymes, complement markers, DRG histology, hematology, and cytokines.
Is the clinical product comparable to the preclinical product?	Dose translation depends on product quality.	Genome titer, potency, empty/full ratio, genome integrity, impurities, and aggregation.
How will risk be controlled clinically?	Safety depends on protocol design, not just starting dose.	Sentinel dosing, staggered enrollment, stopping rules, immunosuppression plan, and long-term follow-up.

 

Case Study: What AAV9-SMN1 Development Taught the Field

Spinal muscular atrophy (SMA) provides one of the most important examples of AAV dose translation. Zolgensma (onasemnogene abeparvovec) is an AAV9-based gene therapy designed to deliver a functional SMN1 transgene to eligible pediatric patients with SMA. The approved product is administered as a single intravenous infusion at 1.1 × 10^14 vg/kg.

The SMA program illustrates why mouse efficacy is only one part of dose selection. In neonatal SMA mouse models, AAV9-SMN delivery can produce strong survival and motor benefits. However, the clinical dose could not be justified by a direct mouse-to-human vg/kg conversion alone. Developers also needed to understand biodistribution, large-animal safety, liver enzyme elevations, dorsal root ganglia findings, patient age, baseline disease status, steroid management, and clinical monitoring.

High-dose systemic AAV studies in large animals have also demonstrated that toxicities may appear in species and settings where mouse models are less predictive. In one published study of high-dose intravenous AAV carrying human SMN, severe toxicity was reported in non-human primates and piglets, including liver and sensory neuron findings. This does not mean that all systemic AAV programs will show the same profile, but it highlights why dose translation must be built from converging evidence rather than simple arithmetic.

Table 4. Lessons from AAV9-SMN1 development in SMA

Development stage	Model or population	Dose information	Key lesson
Mouse efficacy studies	Neonatal SMA mouse models	AAV9-SMN delivery showed strong survival and motor benefit in early postnatal treatment.	Mouse efficacy supports proof of concept but cannot define a human dose alone.
Large-animal evaluation	NHP and other large-animal models	High-dose systemic AAV studies revealed liver and sensory neuron toxicity risks not fully predicted by mice.	Large-animal data are critical for systemic AAV safety assessment.
Early clinical testing	Pediatric SMA patients	Clinical development evaluated systemic AAV9-SMN1 dosing in infants with SMA.	Clinical dose selection required integration of efficacy, biodistribution, toxicology, safety monitoring, and patient context.
Approved clinical use	Eligible pediatric SMA patients	Zolgensma is administered as a single IV infusion at 1.1 × 10^14 vg/kg.	The final dose reflects product-specific translational evidence, not direct mouse-to-human scaling.

 

The broader lesson is clear: a successful AAV program must connect the intended human dose to product-specific potency, target-tissue exposure, safety margins, and clinical risk controls. A clinically justified dose is not simply the mouse dose rewritten in human units.

Common Pitfalls in AAV Dose Translation

Many dose-translation challenges arise because teams do not clearly state the assumptions behind their calculations. The following pitfalls are especially common during pre-IND planning.

Table 5. Common pitfalls in AAV dose translation

Pitfall	Why it creates risk	Better approach
Directly converting mouse vg/kg to human vg/kg	Ignores species differences in organ size, biodistribution, transduction, and immune response.	Use mouse data as proof of concept, then integrate target-organ scaling and large-animal data.
Using total vg without potency context	Genome copies do not always reflect functional activity.	Include potency, infectivity, genome integrity, and empty/full capsid analysis.
Treating NHP data as perfectly predictive	NHPs are informative but not identical to humans.	Use NHP data as a translational anchor while clearly stating uncertainty.
Relying on immunosuppression to manage dose-related toxicity	Immunosuppression may reduce immune injury but cannot make an unsafe dose safe.	Reassess starting dose, escalation plan, and stopping criteria.
Applying vg/kg to ocular or local CNS delivery	Local exposure is driven by anatomy and injection volume, not body weight.	Use compartment-specific units such as vg/eye or vg/injection site.
Assuming scAAV is always more potent than ssAAV by a fixed factor	Potency depends on tissue, promoter, transgene, capsid, and model.	Generate product-specific comparative potency data.

 

How to Think About AAV Dose Translation by Program Type

A practical way to reduce translational uncertainty is to match the dose logic to the program type. Liver, retina, CNS, muscle, and pediatric programs each require different assumptions and different evidence packages.

Table 6. Practical AAV dose-selection framework by program type

Program type	Recommended dose logic	Key data to prioritize
Liver-directed AAV	Combine target-organ scaling with NHP or other relevant large-animal validation.	Liver biodistribution, hepatocyte transduction, secreted protein level, ALT/AST, bilirubin, and coagulation markers.
Ocular AAV	Use vg/eye or vg/injection instead of vg/kg.	Target-cell number, injection volume, local inflammation, retinal anatomy, and vector spread.
CNS AAV	Base dose on route-specific anatomy and distribution.	CSF volume, target-region coverage, DRG exposure, local tolerability, and biodistribution.
Systemic neuromuscular AAV	Evaluate both vg/kg and total vg burden.	Muscle expression, liver uptake, complement activation, body weight, and disease stage.
Pediatric AAV	Consider developmental physiology and long-term risk.	Organ proportion, immune maturation, growth, pediatric safety monitoring, and long-term follow-up.

 

What to Include in a Pre-IND AAV Dose Rationale

A well-prepared pre-IND briefing package should present dose selection as an integrated rationale. The goal is not to prove that the calculation is perfect. The goal is to show that the assumptions are scientifically justified, the uncertainties are understood, and the clinical design manages risk appropriately.

Table 7. What to include in a pre-IND AAV dose rationale

Section	What to include
Proposed clinical dose range	Starting dose, escalation doses, maximum planned dose, and rationale.
Animal efficacy data	Minimally effective dose, dose-response relationship, and disease-model relevance.
Biodistribution	Target and non-target tissue exposure, persistence, clearance, and vector copies per tissue.
Toxicology	NOAEL or relevant toxic dose, dose-limiting findings, reversibility, and safety margins.
Product characterization	Titer method, potency, empty/full ratio, genome integrity, impurities, and comparability.
Route of administration	Rationale for delivery route and route-specific safety considerations.
Patient population	Disease severity, age, weight, organ function, prior treatment, and immune status.
Clinical risk mitigation	Sentinel dosing, staggered enrollment, stopping rules, monitoring plan, and immunosuppression strategy.
Uncertainty and assumptions	Clear explanation of what is known, what is estimated, and what remains uncertain.

 

Frequently Asked Questions

How should vg/animal be converted to vg/kg?

The basic calculation is straightforward: vg/kg equals the total vector genomes administered divided by animal body weight in kilograms. However, the more important question is whether the reported titer accurately reflects active vector. Before using a literature dose, check how the titer was measured, whether empty/full ratio was reported, whether potency was measured, and whether the vector preparation is comparable to the intended clinical product.

Do FDA or EMA provide a fixed formula for AAV first-in-human dose selection?

No. Regulators generally expect a scientifically justified dose rationale rather than a fixed conversion formula. FDA guidance notes that dose selection should be informed by available clinical information, experience with similar products, non-human data, in vitro data, predictive models, and allometric scaling where appropriate. EMA guidance states that dose selection should be based on quality and nonclinical findings, linked with product potency, and justified by scientific data.

Can an IND be submitted before NHP data are available?

This depends on the program. For certain lower-risk or localized AAV programs, a carefully justified package may be acceptable if supported by strong scientific rationale, suitable safety margins, and appropriate risk controls. For high-dose systemic AAV programs, submitting without relevant large-animal biodistribution and toxicology data can create significant regulatory risk, especially if there are liver, dorsal root ganglia, complement, or other toxicity signals.

Should liver enzyme elevation in NHP studies be managed only with stronger immunosuppression?

No. Immunosuppression can be an important risk-mitigation tool, but it should not be used to justify an otherwise unsafe dose. Dose-dependent liver enzyme elevation, especially when accompanied by bilirubin changes, coagulation abnormalities, complement activation, inflammatory markers, or histopathology, should be treated as a dose-window signal. The proposed clinical starting dose and escalation plan may need to be revised.

How should scAAV be handled in dose translation?

Self-complementary AAV can produce faster or stronger expression than single-stranded AAV in some contexts because it bypasses second-strand synthesis. However, the effect is product-specific and depends on the tissue, promoter, transgene, capsid, and model. It is not scientifically appropriate to apply a universal potency multiplier without direct comparative data.

Key Takeaways for AAV Translational Development

AAV dose translation is not a simple mathematical conversion from mouse to human. It is a cross-functional decision that connects biology, pharmacology, toxicology, CMC, and regulatory strategy.

• Start with the required human biological effect, not the animal dose.
• Use mouse data to support mechanism and proof of concept, but do not treat mouse vg/kg as a direct human dose.
• Match the scaling logic to the route of administration and target tissue.
• For systemic AAV, consider both vg/kg and total vector burden.
• Use NHP or other relevant large-animal data as a translational anchor when systemic exposure and safety risk are high.
• Make product potency, quality, and comparability part of the dose rationale.
• Document assumptions clearly before pre-IND discussions.

Many IND delays are not caused by the absence of a perfect formula. They are caused by unclear assumptions, incomplete comparability, weak potency justification, or a dose rationale that does not connect animal data to human biology. Writing the dose rationale as a concise internal memo before formal regulatory engagement can help align the development team, CRO partners, toxicology experts, CMC leads, and regulatory advisors around the same assumptions.

How PackGene Supports AAV Dose Translation and IND-Enabling Development

AAV dose selection depends not only on biology, but also on vector quality, potency, and process consistency. PackGene supports AAV programs from early research through IND-enabling and GMP manufacturing with integrated capabilities across vector design, plasmid optimization, AAV packaging, process development, analytical characterization, and regulatory-supportive documentation.

For teams preparing for pre-IND discussions or clinical translation, a robust AAV development plan should align dose strategy with manufacturing strategy from the beginning. High-quality vector production, reliable titer methods, potency assays, impurity control, biodistribution support, and comparability planning all strengthen the scientific foundation for first-in-human dose justification.

• AAV vector design and plasmid optimization for research and translational programs
• Research-grade, preclinical-grade, and GMP-aligned AAV production support
• Process development and scale-up strategies for consistent vector quality
• Analytical characterization, including titer, purity, genome integrity, empty/full assessment, and impurity testing
• CMC documentation support to help teams prepare for IND-enabling development and regulatory discussions

AAV dose translation is ultimately a risk-based, evidence-driven decision. Getting the dose strategy right early can help reduce development risk and move gene therapy programs more efficiently from animal studies toward clinical application.

References

1. S. Food and Drug Administration. Human Gene Therapy for Rare Diseases: Guidance for Industry. January 2020.
2. European Medicines Agency. Guideline on the quality, non-clinical and clinical aspects of gene therapy medicinal products. EMA/CAT/80183/2014.
3. Burr A, Erickson P, Bento R, Shama K, Roth C, Parekkadan B. Allometric-like scaling of AAV gene therapy for systemic protein delivery. Mol Ther Methods Clin Dev. 2022;27:368-379. doi:10.1016/j.omtm.2022.10.011.
4. Hinderer C, Katz N, Buza EL, et al. Severe toxicity in nonhuman primates and piglets following high-dose intravenous administration of an adeno-associated virus vector expressing human SMN. Hum Gene Ther. 2018;29(3):285-298. doi:10.1089/hum.2018.015.
5. Mendell JR, Al-Zaidy S, Shell R, et al. Single-dose gene-replacement therapy for spinal muscular atrophy. N Engl J Med. 2017;377:1713-1722. doi:10.1056/NEJMoa1706198.
6. Zolgensma (onasemnogene abeparvovec-xioi) prescribing information. Novartis Gene Therapies. https://www.novartis.com/us-en/sites/novartis_us/files/zolgensma.pdf
7. Francois A, Bouzelha M, Lecomte E, et al. Accurate titration of infectious AAV particles requires measurement of biologically active vector genomes and suitable controls. Mol Ther Methods Clin Dev. 2018;10:223-236. doi:10.1016/j.omtm.2018.07.004.