The same problem, solved in opposite directions
Ex-vivo brings the cell to the editor; in-vivo sends the editor to find the cell. Each pays for that choice in a different currency of loss.
Every barrier is a yield term
Ex-vivo and in-vivo gene medicine solve the same problem, getting an editor into the right cell in working condition, through completely different yield chains.
Ex-vivo
- Yield chain
- Collection → enrichment → editor delivery → viability → expansion → release
- Reported figure
- Mean transduction efficiency around 66% in published CAR-T manufacturing data; viability commonly exceeds 95% after expansion
- Trade-off
- Reduces immune mismatch but creates patient-specific scheduling and starting-material variability
Binding constraint: Living product cannot be destructively inspected in full: release testing must infer potency from a sample.
Four stages shared by both branches
Payload, delivery, editing, and release apply to both routes: only the mechanism inside each stage differs.
Payload and formulation
Editor, guide, template or transgene, vector chemistry, dose, route, and stability define the administered material.
- Measure
- Identity · activity · stability
- Failure boundary
- Payload and manufacturing limits can cap dose before biology ever becomes the constraint.
Where the frontier moves
Chemistries that improve stability and payload capacity without adding immunogenicity.
Biological delivery
Transfection or transduction ex vivo; circulation, tissue access, cell targeting, entry, escape, and trafficking in vivo.
- Measure
- Functional target-cell exposure
- Failure boundary
- Losses and off-target distribution mean bulk uptake overstates real delivery.
Where the frontier moves
Targeting chemistries that raise cell-type specificity beyond passive organ accumulation.
Editing and cell function
On-target change, editor duration, chromosomal integrity, viability, phenotype, expression, and potency determine useful action.
- Measure
- Potent correctly edited cells
- Failure boundary
- Off-target and heterogeneous response can undermine an otherwise successful delivery event.
Where the frontier moves
Transient, controllable editor activity that limits exposure window without sacrificing effect.
Release and clinical delivery
Expansion or fill, formulation, sterility, potency assays, logistics, conditioning, administration, monitoring, and redosing complete the medicine.
- Measure
- Released doses · therapeutic window
- Failure boundary
- Testing, immunity, and care burden can strand a biologically successful product from patients.
Where the frontier moves
Platform release processes and assays reusable across related products.
Specificity and manufacturing yield are both ratios
Targeting rarely sends every particle to one cell type, and living products cannot be destructively inspected in full. Therapeutic design must establish a window between effective target exposure and harmful off-target exposure while preserving enough product through every operation.
Programmable payloads move scarcity into tissue access and evidence
As editors become easier to design, each target tissue and product still needs characterized delivery, toxicology, genomic safety, potency, manufacture, comparability, clinical workflow, and long-term follow-up.
Match branch to disease
Choose ex vivo when cells can be collected and returned; choose in vivo when anatomy and scale justify direct delivery.
Limit active exposure
Use transient editors and controlled expression when durability is not required for the therapeutic effect.
Design for escape and specificity
Measure target-cell function, not bulk uptake, and suppress activity in off-target cells.
Build a release platform
Reuse closed processing, analytics, potency assays, and delivery knowledge across related products.
Who is building what
Ionizable lipid nanoparticles, base editing, prime editing, engineered AAV capsids, and virus-like particles solve the in vivo delivery bottleneck. Search the record, or filter by delivery architecture.
Intellia TherapeuticsNTLA-2001 & NTLA-2002Systemic intravenous delivery of Cas9 mRNA and sgRNA encapsulated in liver-trophic lipid nanoparticles, targeting TTR and KLKB1
- Reported evidence
- Pioneered first systemic in vivo CRISPR editing in humans; clinical trials document durable >90% reduction in serum TTR protein after a single dose.
- Announced next step
- Advancing pivotal Phase III trials toward first FDA approval of an in vivo systemic CRISPR therapeutic.
- Unresolved risk
- Acute infusion-related immune reactions to lipid nanoparticles, liver enzyme elevations, and lifetime durability of edits in hepatocytes.
Verve TherapeuticsVERVE-101 & VERVE-102In vivo adenine base editing (ABE) delivered via GalNAc-functionalized or ionizable LNPs to inactivate PCSK9 and ANGPTL3 in the liver
- Reported evidence
- Clinical trials demonstrate profound and durable lowering of LDL cholesterol in patients with heterozygous familial hypercholesterolemia.
- Announced next step
- Once-and-done permanent cardiovascular disease prevention replacing daily statins and recurring antibody injections.
- Unresolved risk
- Transient liver transaminase spikes, off-target base deamination across the transcriptome (RNA off-targets), and long-term surveillance.
Beam TherapeuticsBEAM-101 & Engineered LNPsPrecision base editing altering single nucleotides without double-strand DNA breaks, paired with extrahepatic LNP formulations
- Reported evidence
- Clinical data in sickle cell disease demonstrates high target editing efficiency; initiated non-human primate studies for lung and CNS delivery.
- Announced next step
- Expanding in vivo delivery beyond the liver using selective organ targeting (SORT) lipid nanoparticle chemistries.
- Unresolved risk
- Complex multi-component lipid formulation stability and maintaining prime/base editor expression kinetics without immune recognition.
Prime MedicinePrime Editing PlatformSearch-and-replace prime editing (Cas9 nickase fused to reverse transcriptase with pegRNA) correcting all twelve base-to-base transitions, insertions, and deletions
- Reported evidence
- Demonstrated precise gene correction in non-human primates for chronic granulomatous disease (CGD) and Wilson's disease.
- Announced next step
- Clinical entry for in vivo hepatic and ocular prime editing programs.
- Unresolved risk
- Large molecular cargo size (exceeding standard viral packaging limits) requiring split-AAV or highly optimized LNP delivery vectors.
CRISPR TherapeuticsCasgevy & In Vivo PipelineEx vivo electroporation of patient hematopoietic stem cells (Casgevy), paired with preclinical in vivo targeted LNP delivery discovery
- Reported evidence
- Casgevy received historical first regulatory approvals (FDA, MHRA, EMA) for sickle cell disease and transfusion-dependent beta-thalassemia.
- Announced next step
- Developing gentler, non-toxic conditioning regimens and transitioning ex vivo workflows toward direct in vivo stem cell editing.
- Unresolved risk
- High healthcare price tag ($2.2M per patient), harsh chemotherapy myeloablation conditioning, and specialized hospital infrastructure limits.
Dyno TherapeuticsCapsid.ai PlatformGenerative machine learning models designing synthetic adeno-associated virus (AAV) capsid variants with enhanced tissue tropism and immune evasion
- Reported evidence
- Partnered with Astellas, Roche, and Sarepta; identified novel capsids with dramatically improved muscle and CNS transduction in non-human primates.
- Announced next step
- Next-generation gene therapy vectors delivering high therapeutic payloads at 10- to 100-fold lower systemic viral doses.
- Unresolved risk
- High-dose AAV systemic toxicity (dorsal root ganglion toxicity, microangiopathy) and pre-existing neutralizing antibodies in human populations.
Broad Institute / Liu LabEngineered Virus-Like Particles (eVLPs)Retroviral protein shells engineered to package and deliver gene-editing ribonucleoprotein (RNP) complexes without transferring viral genetic material
- Reported evidence
- Demonstrated efficient in vivo base and prime editing in animal models of genetic blindness and metabolic disorders with zero off-target DNA integration.
- Announced next step
- Transient in vivo editing vehicles combining the delivery efficiency of viral capsids with the safety of non-viral RNP delivery.
- Unresolved risk
- Large-scale biomanufacturing yields and cell-specific surface targeting ligand engineering.
NIH Somatic Cell Genome Editing (SCGE)Consortium Delivery ToolkitPublic research consortium developing and independently validating in vivo delivery technologies, reporter animal models, and safety assays
- Reported evidence
- Published comprehensive comparative biodistribution benchmarks across dozens of viral and non-viral delivery platforms.
- Announced next step
- Accelerating translation of delivery platforms across non-liver organs (brain, muscle, eye, lung).
- Unresolved risk
- Bridging mouse model delivery findings to human clinical trials where vascular architecture and immune biology differ significantly.
Preclinical liver editing rates exceed 90% in animal models; clinical translation outside the liver (CNS, muscle, lung) is constrained by vascular endothelial barriers and systemic immunogenicity.
The optimistic view, with conditions
Gene medicine becomes a library of validated delivery platforms
The highest leverage comes from tissue and cell platforms whose tropism, intracellular release, editor compatibility, manufacture, safety, assays, and clinical workflow can support families of payloads.
Standardize the manufacturing platform
Reusable processing and potency assays reduce the patient-specific variability that drives cost and delay.
Raise specificity, not just uptake
Organ-level accumulation is the easy number; cell-type specificity and escape are the ones that matter.
Build reusable delivery platforms
A validated tropism or manufacturing process should support families of payloads, not one product at a time.
What turns an edit into a medicine
- Matched delivery branchEx-vivo or in-vivo chosen for the disease and tissue, not by default.
- Functional, not bulk, measurementTarget-cell function tracked instead of organ-level uptake or edit percentage alone.
- Controlled exposureEditor activity limited to what the therapeutic window actually requires.
- Reusable delivery platformsTropism, manufacturing, and assays validated once and reused across payloads.
- A real release pathwayPotency, sterility, logistics, and redosing solved alongside the biology, not after it.
The platform advances faster than coverage
WHO’s 2025 rare-disease resolution cites more than 7,000 rare diseases affecting over 300 million people. That is a rare-disease total, not a count of single-gene disorders or a tally of people eligible for editing. FDA’s approved cell and gene therapy list contains a growing but much smaller set of products and indications. The mismatch is why an editor that can change one sequence has not produced cures for every disease.
In 2025, clinicians reported a patient-specific in-vivo base edit for one infant with CPS1 deficiency. It is an important proof of personalized design, manufacture and regulatory review, but one treated infant cannot establish population safety or a repeatable price for thousands of variants. At the other end of the evidence chain, FDA expanded Casgevy to children aged two and older in July 2026 for specified sickle-cell and beta-thalassemia indications, showing that approved editing has begun to broaden while remaining disease-specific.
Safety sets a hard translation constraint. After reports of fatal acute liver failure in non-ambulatory Duchenne patients treated with Elevidys, FDA added a boxed warning and narrowed its indication in 2025. The vector, dose, tissue, patient age and immune response matter as much as editing accuracy. Any claim of cure needs durable clinical benefit and adverse-event follow-up, not only a corrected sequence or early biomarker.
Regulation and safety are moving targets
FDA's February 2026 plausible-mechanism guidance is a draft for individualized therapies, not an approved shortcut or evidence that each variant can use one license. Lilly completed its acquisition of Verve in July 2025, so Verve is no longer an independent public developer. Intellia said FDA lifted its MAGNITUDE Phase 3 hold in March 2026; the pause and resumption should both be recorded rather than treating either as a final safety verdict. There is no comparable public per-patient ex-vivo manufacturing-cost or long-term in-vivo editing-efficiency series across diseases, so list prices and isolated editing percentages cannot establish a general cure cost curve.
Sources, method, and boundaries
Regulatory distinctions and safety considerations follow FDA guidance. Quantitative delivery claims retain their experimental boundaries. The yield-chain framing combines ex-vivo manufacturing and in-vivo biodistribution without treating them as interchangeable processes.
- Endosomal escape
- The fraction of internalized nanoparticle cargo that reaches the cytoplasm rather than being degraded.
- Transduction efficiency
- The share of collected cells that successfully receive the genetic payload during ex-vivo manufacturing.
- Tropism
- A delivery vehicle's tendency to accumulate in or enter particular tissues or cell types.
