The FDA’s 2024 guidance on human gene therapy products incorporating genome editing addresses two fundamentally different types of programs under a single document. Ex vivo genome editing — in which cells are removed from the patient, modified outside the body, and reinfused — and in vivo genome editing — in which the editing machinery is administered directly to the patient and acts on cells within the body — share an underlying molecular biology but diverge sharply in their regulatory classification, CMC requirements, manufacturing framework, and nonclinical safety assessment strategy.
Understanding where the two frameworks diverge is not an academic exercise. The classification of genome editing components as drug substance or as critical manufacturing component is one of the most consequential regulatory determinations in a gene editing program’s early life. It determines what must be characterized, what must be released against defined specifications, what the IND must contain, and what the BLA manufacturing section must demonstrate. Getting it wrong early means rebuilding the regulatory foundation of the program later.
The Core Regulatory Distinction
The FDA’s guidance establishes the dividing line clearly: for in vivo genome editing products, the genome editing components — the guide RNA, the nuclease, the donor template if used, and the delivery vehicle — are the active pharmaceutical ingredients. They are administered directly to the patient. They are the drug substance. Their final formulation, including the delivery vehicle, is the drug product. The entire CMC framework — characterization, release testing, stability, potency assays — applies to the editing components and their formulated product.
For ex vivo genome editing products, the classification is different. The genome editing components are not the drug product — they are critical components used to manufacture the drug product. The drug product is the final edited cell population that is administered to the patient. The editing components — guide RNA, Cas9 protein or mRNA, delivery vehicle for ex vivo transduction — must be manufactured, characterized, and tested, but they are characterized as manufacturing inputs rather than as drug substances in the regulatory sense. What must be fully characterized and released as a drug product is the edited cells.
This distinction has cascading implications for what the IND must contain, what GMP standards apply, and what analytical testing is required at each stage of development.
What In Vivo Programs Must Demonstrate
For in vivo genome editing, the editing components and their delivery vehicle constitute the drug substance and drug product respectively. This means the full suite of drug substance characterization requirements applies: molecular structure and genetic sequence, manufacturing process description, in-process controls, raw material specifications, cell banking if applicable, drug substance release testing for identity, purity, potency and safety, and stability data.
The delivery vehicle — whether a viral vector such as AAV, a lipid nanoparticle for mRNA delivery, or another platform — carries the same regulatory weight as any other drug product delivery system. For AAV-delivered CRISPR components, the AAV manufacturing CMC requirements apply in full: three-plasmid or alternative production system, full capsid characterization, full/empty capsid ratio, vector genome titer, identity, potency, and all associated stability data. For LNP-delivered base editor or prime editor mRNA, the nanoparticle formulation must be characterized for particle size, encapsulation efficiency, mRNA integrity, and potency.
Potency assessment for in vivo genome editing drug products is particularly demanding because the guidance specifies that potency assays should measure multiple aspects of activity: the ability of the editing components to perform the intended genetic sequence modification, and the downstream biological effects in the relevant target cells or tissues. A potency assay that measures only guide RNA expression or Cas9 protein concentration does not meet this standard. The assay must demonstrate that the delivered components actually edit the intended target sequence at a meaningful frequency in the relevant cell context.
What Ex Vivo Programs Must Demonstrate
For ex vivo genome editing, the regulatory complexity shifts from the editing components to the edited cell product. The editing components — Cas9 protein or mRNA, guide RNA, donor template, electroporation parameters or other delivery method — must be described and characterized as manufacturing inputs, but the primary CMC burden falls on demonstrating that the manufacturing process produces a consistent, characterizable, potent edited cell population.
Release testing of ex vivo editing products must evaluate on-target editing efficiency, the total number of edited cells, and additional characterization of editing events. The guidance specifies that off-target editing frequency, intrachromosomal and interchromosomal rearrangements, and residual editing components must be assessed based on nonclinical study outcomes. This means that the nonclinical off-target characterization program must be completed before the release testing specifications for the clinical product can be finalized — the two programs are sequentially dependent, not parallel.
Potency testing for ex vivo edited cell products must measure both the cellular properties of the final product and the intended downstream biological modification. In early-phase studies, confirming the desired genetic sequence modification may be sufficient to characterize potency. For marketing applications, the guidance expects potency assays that demonstrate the downstream biological effect — not merely the presence of the edit. For a sickle cell disease program, a potency assay for early development might confirm BCL11A enhancer editing efficiency. A BLA-level potency assay should demonstrate that the editing has produced the intended increase in fetal hemoglobin expression.
The timing of genome editing steps within the ex vivo manufacturing process must be described in the IND, and in-process controls and testing must be established for steps with significant impact on editing efficiency and specificity. This is a more granular CMC requirement than is typical for conventional biologics, because the editing step itself — electroporation, transduction with viral delivery of editing components, ribonucleoprotein delivery — is a critical manufacturing step whose execution directly determines the quality attributes of the final product.
Delivery Method and Its Regulatory Implications
The choice of delivery method for ex vivo editing introduces additional regulatory considerations that differ depending on whether genetic material is integrated into the cell or remains transient.
When editing components are delivered as DNA via plasmid or viral vector, there is a risk that the editing machinery sequences — guide RNA expression cassette, Cas9 coding sequence — could integrate into the cell genome alongside the intended edit. Integrated Cas9 sequences in the final product represent a safety concern: continued Cas9 expression after infusion could introduce additional editing events in vivo. The guidance requires characterization of residual editing components in the final product, which for DNA-based delivery means assessment of integrated vector sequences beyond the intended edit.
When editing components are delivered as RNA or ribonucleoprotein complexes — mRNA encoding Cas9, or pre-assembled Cas9-guide RNA complexes — they are transient. They do not integrate, they are degraded by cellular ribonucleases over a period of hours to days, and the editing machinery is no longer present in the final cell product. This transient delivery approach substantially simplifies the residual editing component characterization burden and reduces the risk of ongoing editing in vivo. It is the approach used in Casgevy, the first approved CRISPR-based therapy, where Cas9 protein and guide RNA are delivered as a ribonucleoprotein complex into hematopoietic stem cells by electroporation.
Nonclinical Strategy Diverges by Mode
The nonclinical safety assessment strategy for in vivo and ex vivo programs diverges in ways that reflect the different nature of the patient exposure in each case.
For in vivo programs, biodistribution of the delivery vehicle — where editing components go in the body after administration, how long they persist, and whether editing occurs in non-target tissues — is a primary safety concern. The guidance requires assessment of editing activity in target and non-target tissues and evaluation of the potential for unintended germline modification. This last point is particularly significant for in vivo approaches that administer editing components systemically: if editing components reach the gonads, there is a theoretical risk of editing germline cells, which would create heritable modifications. The FDA’s guidance treats germline modification risk as a required element of the biodistribution safety assessment for in vivo programs.
For ex vivo programs, the cells themselves are characterized for editing outcomes before infusion. Biodistribution of the editing components after infusion is not a concern because the components are no longer present in the final product (for transient delivery approaches). The nonclinical safety assessment focuses instead on the safety of the edited cell population — uncontrolled proliferation, aberrant growth, chromosomal abnormalities, and immunogenicity of the edited cells and any expressed transgenes.
The Strategic Decision
The choice between in vivo and ex vivo genome editing for a given indication is primarily driven by biology: which cell type needs to be edited, whether those cells can be collected and returned ex vivo, and what the editing efficiency requirements are in each context. But the regulatory implications of that choice are substantial enough that they should be part of the strategic analysis at program inception.
An ex vivo program requires building a cell manufacturing capability — leukapheresis logistics, cell processing, GMP manufacturing, cryopreservation, chain of identity — that is operationally complex and capital-intensive, but that allows the editing to occur in a controlled environment with full characterization before patient exposure. An in vivo program requires building a delivery vehicle manufacturing capability — AAV or LNP production, formulation, and characterization — and accepting that the editing will occur in the patient’s body, in a biological environment that cannot be fully controlled or observed.
Both pathways are viable. Both have produced clinical results. But they are not interchangeable, and the regulatory infrastructure required for each is sufficiently different that switching from one to the other after clinical development has begun is not a minor adjustment. The regulatory classification decision — editing components as drug substance or as manufacturing input — should be made with full understanding of its downstream implications before the IND is filed.