One indication, many guides: designing a CRISPR/Cas9 program as a platform, from the first IND to a single BLA

Monogenic diseases are rarely caused by a single pathogenic variant. A gene may carry dozens of disease-causing mutations distributed across its patient population, each requiring a guide RNA of its own if the therapeutic strategy is nuclease-mediated disruption or correction at the variant. For a CRISPR/Cas9 program, that fact has a regulatory consequence: a product with a different guide is, in the Agency's eyes, a different product. The editor, the lipid nanoparticle, the formulation, the route of administration, and the manufacturing process may be identical, but the guide determines the target site, the editing efficiency, and the off-target profile, and the Agency's assessment of safety and activity attaches to the guide.

A sponsor that approaches this one guide at a time, as a sequence of independent programs, will spend a decade and a great deal of capital reaching a fraction of the patient population. A sponsor that designs the program as a platform from the outset can reach the whole population under one development program and, ultimately, one license. Two CBER documents provide the architecture: the 2022 guidance on studying multiple versions of a cellular or gene therapy product in a single early-phase trial, read together with the June 2026 draft on leveraging prior knowledge, governs development; the October 2026 draft on individualized human genome-editing products governs the destination. The program should be designed backward from the BLA that the individualized-products draft describes, using the IND structure and prior-knowledge principles that the development guidances provide.

Define the platform before anything else

Everything that follows depends on a precise statement of what is fixed and what varies. The fixed elements constitute the platform: the Cas9 editor and its expression format, the lipid nanoparticle composition and formulation, the route and method of administration, the manufacturing process and its control strategy, and the dose range. The variable element is the guide RNA, and with it the target locus, the edit achieved, the on-target efficiency, and the off-target profile. Nothing else should vary. A program that also changes the editor, the delivery vehicle, or the route between products is no longer a platform in the sense the Agency will accept, and the leverage described below disappears.

The definition must be written down, submitted, and defended at the first Agency meeting. It is the foundation for the division of the nonclinical and CMC packages into platform and guide-specific components, for the structure of the INDs, and, years later, for the definition of the individualized feature space in the BLA.

The development architecture: one primary IND, one umbrella protocol

CBER's 2022 guidance on studying multiple versions of a product in an early-phase trial addresses exactly this situation. Its recommended structure is a single primary IND that carries the umbrella clinical protocol and the data common to all versions, with each additional product version submitted as its own IND that cross-references the primary. Each secondary IND contains what is specific to its version and relies on the primary for the rest. New guides enter the program as new secondary INDs; the umbrella protocol is amended to add the corresponding cohort.

The June 2026 draft on leveraging prior knowledge supplies the principle that makes the cross-referencing substantive rather than formal: data generated on a platform may support a subsequent product that uses the same platform, provided the sponsor documents why the prior data are relevant and what remains product-specific. For a guide-only-variant program the relevance argument is as strong as it can be, because the platform is identical by construction. The sponsor's obligation is to show that it is identical, which is a CMC and documentation task, and to be rigorous about what the guide changes.

Dividing the nonclinical package

The platform package is generated once. It comprises the biodistribution and toxicology of the lipid nanoparticle delivering Cas9 in a relevant species, conducted with a representative guide or, where no guide has a target in the species, with a surrogate that engages the species' orthologous locus; the immunogenicity assessment of the editor; and the delivery-route safety studies appropriate to the anatomy, for a subretinal product the ocular toxicology and the assessment of the injection procedure. These studies address the questions that do not depend on which human sequence the guide targets, and the affirmative argument for not repeating them per guide should be made explicitly in the primary IND and at the Pre-IND meeting, not left for the reviewer to infer.

The guide-specific package is generated for every product. It comprises on-target editing efficiency and the nature of the edits in patient-derived cells carrying the variant; the off-target assessment, which under the April 2026 draft on NGS-based safety assessment means unbiased empirical discovery followed by targeted confirmation in the clinically relevant cell type with the clinical-grade editor and guide, together with an assessment of large deletions, rearrangements, and chromosomal integrity at the on-target site; and the product-specific proof of concept, which for a disease without an animal model carrying the variant rests on patient-derived cellular systems. The guide-specific package is smaller than the platform package but it is not optional for any guide, and its scope should be fixed as a template in the primary IND so that each subsequent guide is assessed to the same standard.

Dividing the CMC package

The manufacturing process is a platform process: the same unit operations, the same critical process parameters, the same in-process controls, the same lipid nanoparticle specifications, regardless of guide. The platform-level control strategy, process characterization, and, eventually, process validation are established once and cited in each secondary IND's Module 3 by cross-reference rather than reproduced. The May 2026 CMC Flexibilities guidance recognizes platform analytical procedures explicitly, allowing a method validated on the platform to be verified rather than re-validated for each product.

What remains guide-specific at release is the identity of the guide (sequence confirmation), its purity and the impurity profile of the synthetic oligonucleotide, the guide-to-editor ratio in the formulated product, and potency. Potency deserves particular attention: under the November 2026 Potency Assurance guidance, the sponsor needs a potency assurance strategy for the platform and a potency assay whose readout is the edit itself, applied with a guide-specific acceptance criterion for each product. Comparability between guides is largely a question of demonstrating that the platform process delivers the same particle with a different cargo; the data to show it should be designed into the first product's characterization.

The umbrella trial under a master protocol

The clinical program is one trial with one master protocol and a cohort for each guide. The master protocol fixes what the cohorts share: the eligibility framework (the disease, the clinical stage, the ocular or systemic criteria), the dose levels and the dose-escalation rules, the safety monitoring plan and the stopping rules, the endpoints and their timing, and the long-term follow-up obligations that attach to any genome-editing product. Each cohort adds a guide-specific appendix: the pathogenic variant that defines eligibility for that cohort, the natural-history benchmark against which that variant's outcome will be judged, and the product-specific potency and off-target acceptance criteria that must be met before the cohort opens.

FDA's 2022 master-protocols guidance was written for oncology, but its principles, a single infrastructure, a common data standard, prospective rules for adding and closing arms, and a governance structure spanning the arms, transfer directly. The 2022 multiple-versions guidance is the CGT-specific authority for the structure.

Learning forward from the first guide

The point of the umbrella is not administrative convenience. It is that what the first product teaches is applied, prospectively and by rule, to every product that follows. The first guide's dose-escalation outcome sets the starting dose for subsequent guides, so that later cohorts do not re-explore dose levels the platform has already shown to be safe; the protocol should state this rule in advance rather than negotiate it cohort by cohort. Each new guide nevertheless opens with a safety run-in, a small number of patients at the established dose with a defined observation period, because the guide-specific risks, principally off-target activity, are not inherited from the platform. One data monitoring committee oversees all cohorts, so that a signal in one cohort is assessed against the experience of all. Where the trial relies on natural-history or external controls, the control is shared, with variant-specific adjustment where the natural history differs.

Statistical borrowing across cohorts is defensible for safety, where the platform's contribution dominates, and is not defensible for efficacy, where each guide's editing efficiency at its own locus is a product-specific question. A Bayesian design that borrows safety information across cohorts while keeping efficacy inference cohort-specific is consistent with the September 2025 draft on innovative designs for small populations and is the appropriate frame for a program in which each variant cohort may enroll only a handful of patients.

Staggered entry and the handling of a signal

Guides enter the trial in a stated sequence, ordered by the size of the variant population and the maturity of each guide's nonclinical package. Before a guide enters, its secondary IND must be in effect, its guide-specific package complete to the template, and its product released against the guide-specific specifications. A safety signal in one cohort triggers a platform-level review: the committee determines whether the event is attributable to the platform, in which case enrollment pauses across cohorts, or to the guide, in which case the affected cohort pauses while the others continue. The rule is written into the master protocol so that the determination is made against predefined criteria rather than under the pressure of the event.

The destination: one BLA, a defined feature space, and variants added after approval

Until October 2026 the endpoint of such a program was uncertain: a sponsor could develop many guides under one umbrella but faced the prospect of licensing them one at a time. CBER's draft guidance on individualized human genome-editing products, issued on October 27, 2026, is scoped to this case and resolves it. The draft applies to products for monogenic diseases caused by known pathogenic variants, where every product variant is designed to achieve the same editing outcome, and it proposes a single BLA covering multiple variants. Its central construct is the individualized feature space: the defined limits within which the genome-editing components may vary while remaining within the scope of one license. The draft expects clinical data for the variants included at the time of the original BLA and proposes that new variants be added after approval through a postapproval change management protocol, supported by in vitro editing and safety data rather than new clinical trials.

The consequence for program design is direct. The umbrella trial should be built so that its cohorts, its dose rules, and its potency and off-target acceptance criteria map onto the feature space the BLA will later define. The first guide carries the clinical burden: it establishes the dose, the safety profile of the platform, and the clinical effect of the editing outcome. Later guides, if the program has been designed for it, inherit that evidence and need show only that they fall within the feature space and meet the guide-specific criteria. A program that reaches its BLA with an umbrella trial already structured in those terms will find the feature space easy to define; one that reaches it with a series of unrelated cohorts will have to reconstruct the argument.

Variants too rare to enroll

Every monogenic disease has a long tail of variants carried by a handful of patients, or one. The February 2026 draft on the Plausible Mechanism Framework addresses individualized therapies for exactly these patients, where conventional evidence generation is infeasible and the case rests on a well-characterized mechanism, a validated platform, and product-specific in vitro evidence. A guide-only-variant platform with an established feature space is the strongest possible position from which to invoke it: the platform's safety is established clinically, the editing outcome is established across variants, and what remains for the new guide is the in vitro package the framework contemplates. The Rare Disease Evidence Principles, announced in 2025, provide the corresponding process for programs that meet their criteria.

After licensure: platform designation

The Platform Technology Designation Program, created by section 506K of the Federal Food, Drug, and Cosmetic Act and described in the May 2024 draft guidance, is available only for a platform already incorporated in an approved product. It is therefore not an instrument for the development phase, and sponsors should not plan on it before the first license. Once the first guide-specific product is licensed, however, a designation request for the fixed Cas9-LNP platform becomes available, and if granted it allows subsequent applications incorporating the platform to be reviewed with the efficiencies the program contemplates. The prior-knowledge draft governs the road to the first approval; the designation program governs the road after it.

Engaging the Agency

The platform definition and the division between platform and guide-specific data are the questions to place before CBER at INTERACT, and to confirm at Pre-IND. They should be posed as positions for concurrence: that the specified elements constitute the platform; that the platform nonclinical package will be generated once with a representative guide and cross-referenced; that each guide's specific package will comprise the stated elements; and that the clinical program will proceed under a single umbrella protocol with the stated rules for dose inheritance, safety run-in, and cohort entry. Alignment on these points at the outset determines the cost and duration of the entire program. The sponsor should also ask, now that the October 2026 draft exists, how the Agency would expect the umbrella trial's structure to relate to the eventual feature space, since the answer shapes the trial's design from its first cohort.

Crawling Stone's role

Crawling Stone advises genome-editing programs on exactly this architecture: the platform definition, the division of the nonclinical and CMC packages, the IND structure, and the Agency questions that secure alignment on each. For a program in inherited retinal disease, a platform approach to products differing only by guide RNA was among the questions Crawling Stone placed before CBER at INTERACT. Sponsors contemplating a multi-variant program should settle the platform definition before the first guide's IND-enabling studies are designed; the decisions made at that point are the ones the October 2026 draft will later reward or penalize. Engage Crawling Stone to define them.

This article is general in nature and does not constitute regulatory advice for any specific program. Several of the guidances discussed are drafts and may change on finalization; the October 2026 draft on individualized genome-editing products is open for comment until November 27, 2026. Consult the current documents on the Regulatory resources page.