CRISPR Delivery CMC — LNP, AAV, and RNP Delivery Systems for Genome Editing Drug Products

Every CRISPR delivery decision is presented as a biology question: which system delivers the editing machinery to the target tissue with the highest efficiency and lowest immunogenicity? It is equally a CMC regulatory question, and the CMC architecture for LNP-delivered CRISPR mRNA, AAV-delivered CRISPR vectors, and protein-delivered RNP complexes are not interchangeable. Each delivery modality creates a distinct drug substance classification challenge, a distinct Module 3 CTD architecture, and a distinct set of analytical requirements that the IND CMC section must address — before CBER or CDER can evaluate whether the manufacturing controls are adequate for the patient population in the proposed clinical study.

CRISPR delivery CMC programs fail at CBER/CDER review not because the editing system is scientifically unsound, but because the IND CMC package is built from the editing technology’s perspective rather than the delivery modality’s regulatory classification perspective — leaving the reviewer without the drug substance classification rationale for each molecular component, the delivery-modality-specific manufacturing characterization, and the off-target editing safety characterization that demonstrates the non-clinical safety basis for the proposed clinical dose.

CRISPR Delivery Modality CMC Classification — Why LNP, AAV, and RNP Delivery Systems Each Require a Different Module 3 CTD Architecture

LNP-delivered CRISPR — Cas9 mRNA (approximately 4,200–4,500 nt, approximately 1.4 MDa) co-encapsulated with a synthetic guide RNA (sgRNA, approximately 100 nt) — creates a two-drug-substance classification: the Cas9 mRNA requires a 3.2.S section covering sequence characterization, cap structure (cap 1 efficiency ≥90%), and poly(A) tail distribution, while the sgRNA requires its own 3.2.S section covering sequence verification, 2′-O-methyl and phosphorothioate backbone modification confirmation by LC-MS/MS, and HPLC purity. FDA classification precedent from pre-IND interactions consistently treats both RNA components as co-formulated drug substances, not as a drug substance plus a formulation excipient — a distinction with direct consequences for the Module 3 architecture. AAV-delivered CRISPR faces a different constraint entirely: the AAV packaging limit of approximately 4.7 kb means full-length SpCas9 (cDNA approximately 4.2 kb plus promoter and polyA, at or near the packaging limit) cannot be co-packaged with a guide RNA expression cassette in a single vector, forcing a dual-vector architecture — one AAV delivering SpCas9, a second delivering the guide RNA and any donor template — with two independent 3.2.S drug substance sections, independent vector genome titer specifications by ddPCR (≤30% CV precision), and independent empty:full ratio specifications. Smaller Cas9 orthologs such as SaCas9 or CjCas9, at approximately 3.2 kb, enable single-vector delivery and avoid this architectural split. RNP delivery — Cas9 protein (approximately 160 kDa) pre-complexed with sgRNA — introduces a distinct CMC burden again: the Cas9 protein is a large recombinant protein drug substance requiring ICH Q6B-equivalent characterization (HCP, host cell DNA, endotoxin control) if produced in a mammalian or insect expression system, and the pre-formed RNP complex may be classified as drug product or co-formulated drug substance complex depending on the delivery vehicle.

AAV Dual-Vector CRISPR and LNP Co-Encapsulated CRISPR — Drug Substance Classification Pitfalls That Determine Whether Your IND CMC Passes CBER Review

The most consequential classification error in CRISPR-LNP programs is treating the sgRNA as a drug product excipient rather than a co-drug substance. The sgRNA is pharmacologically active — it directs the Cas9 protein to the specific genomic target site — and is not pharmacologically inert in the way a stabilizing excipient is; classifying it in the 3.2.P.4 excipient section rather than in its own 3.2.S section leaves the characterization package without the sequence verification, chemical modification confirmation, and purity specification that a pharmacologically active nucleic acid component requires. The equivalent pitfall in AAV dual-vector programs is a potency assay that measures only the SpCas9-carrying vector’s transduction efficiency without confirming that the guide-RNA-carrying vector actually delivers a functional guide RNA directing SpCas9 to the correct genomic target — because the drug product’s genome editing activity requires both vectors to function together at the intended co-administration dose ratio, a potency assay built around one vector’s transduction efficiency alone does not demonstrate the drug product’s actual mechanism of action. The regulatorily sound design is a cell-based co-transduction assay using both AAV vectors at the proposed clinical dose ratio, measuring on-target editing frequency by amplicon sequencing as the potency endpoint.

Off-Target Editing Safety Characterization — The GUIDE-seq and Targeted Deep Sequencing Program That CBER Expects Before the IND Can Proceed

Off-target genome editing is a safety-relevant CMC characterization requirement, not a research question to be resolved after IND clearance. The regulatorily accepted approach begins with GUIDE-seq analysis — a genome-wide, unbiased method identifying double-strand break sites — performed in a cell line relevant to the primary target tissue, typically identifying somewhere between 5 and 30 off-target sites for a well-designed sgRNA. That genome-wide screen is followed by targeted deep sequencing (sequencing depth ≥10,000× per site) at the top 15 GUIDE-seq-identified off-target sites in primary human cells relevant to the therapeutic application — CD34+ hematopoietic stem cells for a hematological indication, or hepatocytes for a liver-targeted program — with a non-clinical safety benchmark of off-target editing frequency ≤0.1% at any characterized off-target site relative to background, consistent with the safety thresholds applied in published CRISPR clinical program data. In silico prediction tools (such as sequence-mismatch-based algorithms) are not an acceptable substitute for this empirical characterization, because computational off-target prediction carries a known false-negative rate for off-target sites with three or more sequence mismatches — an IND submission relying solely on in silico prediction, without empirical genome-wide off-target detection, has not demonstrated the non-clinical safety basis CBER requires before Phase 1 can proceed.

The XGene CRISPR Delivery CMC Architecture

1. Delivery-modality-specific drug substance classification — a documented rationale for each molecular component (Cas9 mRNA, sgRNA, Cas9 protein, or AAV vector) matched to the delivery modality’s regulatory classification precedent. 2. CTD Module 3 architecture design — the correct number of 3.2.S sections and the 3.2.P structure for LNP co-encapsulation, AAV dual-vector, or RNP co-formulation. 3. Modality-specific potency assay design — on-target genome editing efficiency as the primary endpoint, using a co-transduction or co-delivery format that reflects how the drug product actually functions, not a single-component transduction or expression measurement. 4. Off-target editing characterization program — genome-wide GUIDE-seq (or equivalent) followed by targeted deep sequencing at the top 15 off-target sites, applying the ≤0.1% safety threshold. 5. Pre-IND CBER classification interaction — a documented pre-IND request confirming CBER’s position on drug substance classification for the specific delivery modality before the Module 3 CTD is finalized.

The CRISPR delivery CMC package that clears CBER/CDER review is not the one built from the editing technology outward — it is the one built from the delivery modality’s regulatory classification requirements, with every pharmacologically active molecular component classified as a drug substance, every potency assay measuring the drug product’s actual mechanism, and every off-target characterization program grounded in empirical genome-wide detection rather than computational prediction alone.

For your CRISPR delivery program, can you identify today whether your IND CMC section includes a drug substance classification rationale addressing each molecular component of the editing system — Cas9 mRNA, sgRNA, Cas9 protein, or AAV vectors — and whether your potency assay demonstrates on-target genome editing efficiency at the proposed clinical dose in a cell-based system relevant to the target tissue?