GT Drug Substance and Product Specifications — Building a Release Package for BLA

Every gene therapy lot that goes into a patient was released against a specification. Whether that specification can be defended at BLA — across all lots, all timepoints, all analytical method versions — is a CMC decision made at IND.

The specification table for a gene therapy drug substance is not an administrative checkpoint. It is the scientific record against which every clinical lot was qualified, and CBER reviewers will evaluate it at BLA with exactly that framing: were the acceptance criteria used to release your Phase I, Phase II, and Phase III material scientifically justified at the time they were set, and do they still hold up now that you have manufacturing history? The programs that reach BLA with clean specification packages built that foundation at IND. The programs that arrive at BLA with “report result” entries and missing numerical limits are the ones that consume entire CMC cycles in deficiency response.

The regulatory consequence of a weak specification architecture is not cosmetic. Under FDA’s Chemistry, Manufacturing, and Controls (CMC) Information for Human Gene Therapy INDs guidance (2020), CBER expects that even early-phase IND submissions include interim acceptance criteria for all critical quality attributes — not placeholders, not ranges to be determined after characterization. The standard is clear: the absence of a numerical limit for a release attribute at IND is not a provisional posture; it is a deficiency that will follow the program through every subsequent submission.

The Specification Architecture for Gene Therapy: What Tests Are Required at Drug Substance and Drug Product

A gene therapy drug substance specification table carries obligations that do not map cleanly onto small molecule or traditional biologic frameworks. ICH Q6B establishes the general principle that specifications must include tests for strength, purity, identity, and potency — but the GT-specific implementation of those categories is defined by product complexity, assay maturity, and the biological nature of the vector. For an AAV drug substance, the minimum release panel that CBER expects to see includes appearance (clear to slightly opalescent, colorless to light yellow, essentially free of visible particles), pH measured potentiometrically with an acceptance criterion of 7.0–7.5, and osmolality by freezing point depression with a limit of 270–340 mOsm/kg. These physicochemical attributes are among the most defensible in the specification table precisely because their methods are compendial, their acceptance criteria are mechanistically justified by formulation and stability science, and their historical performance across lots can be documented without sophisticated platform-specific assays.

Vector genome titer — measured by ddPCR using ITR-targeting primers — carries a quantitative lower acceptance criterion expressed in vg/mL. The selection of ITR-targeting for ddPCR is not arbitrary: ITRs are unique to the packaged AAV genome, present in two copies per full capsid, and refractory to amplification from unpackaged plasmid DNA that would confound SYBR-based quantification methods. Full/empty capsid ratio, assessed by analytical ultracentrifugation sedimentation velocity (AUC-SV), carries a specification of ≥90% full capsids in programs where the manufacturing process has been optimized to achieve that benchmark, with AUC-SV sedimentation coefficients for full versus empty AAV capsids differing by approximately 30 S units — resolution sufficient to support a quantitative specification rather than a characterization-only report.

The critical distinction CBER enforces — and that most early-phase CMC packages blur — is the line between lot release specifications with defined acceptance criteria and extended characterization testing that is informational only. Capsid titer measured by ELISA (such as the Progen PRATV ELISA for AAV) is frequently used in characterization to report the ratio of genome-containing capsids relative to total capsid protein, but it does not always carry an acceptance criterion at early phase. That is a legitimate regulatory posture only if the attribute is formally documented as characterization with explicit justification for why it is not release-tested; CBER will ask at BLA whether that designation was appropriate, and programs that cannot answer have a specification architecture problem they cannot fix retrospectively.

Potency, Identity, and Safety Testing in GT Specifications: The Minimum Required Package for BLA

Potency is the most scrutinized attribute in a gene therapy specification table, and FDA’s Potency Tests for Cellular and Gene Therapy Products guidance (2011) establishes the baseline expectation: each product must have at least one cell-based functional assay that measures the relevant biological activity of the vector. The acceptance criterion must be expressed as a percentage relative to a qualified reference standard — typically ≥50% relative to the reference standard — not as a raw transduction unit count or reporter gene signal without an anchor. Programs that list potency as “report result” at IND and carry that posture forward through Phase II are building toward a BLA deficiency that cannot be resolved without retrospective lot data reanalysis, and the reference standard itself must be qualified under a documented program before it is used to anchor any release decision.

Identity testing for an AAV drug substance requires confirmation of capsid serotype and expression cassette integrity. Capsid identity can be confirmed by serotype-specific ELISA or by PCR and sequencing targeting the cap gene region. Expression cassette integrity — confirmation that the packaged genome contains the correct sequence — requires either Sanger sequencing of the packaged vector genome after DNase digestion and proteinase K treatment, or NGS for programs where the transgene length or structural complexity demands higher-resolution characterization. USP <1047> provides guidance on gene therapy quality testing and reinforces that identity testing must be specific enough to distinguish the intended product from related vectors or process contaminants that could arise from shared manufacturing platform components.

Safety testing for host cell impurities in an AAV drug substance is specific and quantitative. Host cell protein (HCP) is tested by ELISA with a product-specific acceptance criterion typically in the range of ≤100–500 ng/mL depending on clinical dose and patient population; host cell DNA (HCD) is measured by qPCR using CHO- or HEK293-specific genomic DNA assays with an acceptance criterion of ≤10 ng/dose; and residual Benzonase, used in the manufacturing process for nucleic acid digestion, is controlled by sandwich ELISA at ≤1 ng/mL. Residual plasmid DNA carries its own qPCR-based specification — a test that is frequently absent from early-phase specification tables and that CBER identifies routinely in IND deficiency letters.

Acceptance Criterion Justification for GT Specifications: Where CBER Expects Scientific Rationale

Setting a specification limit is not the hard part. Justifying it is. CBER’s expectation at BLA — consistent with ICH Q6B and the Stage 3 continued process verification principles articulated in FDA’s Process Validation Guidance (2011) applied to lot release — is that every numerical acceptance criterion in the specification table has documented scientific rationale. That rationale takes one of three forms: a compendial or regulatory precedent (e.g., endotoxin ≤5 EU/mL by LAL, or subvisible particles per USP <787> with ≤6000 particles per container ≥10 μm and ≤600 particles per container ≥25 μm), a mechanistic link to clinical safety or efficacy (e.g., HCP limit derived from immunogenicity risk assessment at the clinical dose), or a statistical derivation from the manufacturing history of clinical lots.

The third basis — statistical derivation from clinical lot history — is precisely where Phase I specifications create BLA problems. At Phase I, there may be two or three manufacturing runs available when the IND is filed, providing no statistical basis for a tightly derived acceptance criterion; the limit must instead be set as an interim criterion anchored to manufacturing capability and safety reasoning, with an explicit commitment to tighten the specification as process understanding develops. CBER does not penalize programs for interim criteria at IND, but it does penalize programs for Phase III or BLA specifications that remain as broad as the Phase I interim limits with no evidence the acceptance criterion was re-evaluated as manufacturing data accumulated — a distinction that must be documented in the specification development history.

The most common operational failure in this space involves potency. A program sets potency as an absolute transduction unit titer measured by a research-grade flow cytometry assay, with no defined acceptance criterion or with a criterion expressed as a single-point count rather than a percentage relative to a reference standard. When CBER asks at BLA how the potency specification was justified, the answer must demonstrate that the acceptance criterion was set relative to a qualified reference standard, that clinical lots were released at ≥X% relative potency, and that clinical outcome data support the adequacy of that criterion. Programs that cannot reconstruct that chain of argument across all clinical lots — because the reference standard was changed mid-program without bridging data, or because the assay format shifted from flow to luminescence without requalification — are facing a specification architecture failure that no amount of retrospective data analysis can fully repair.

Building GT Specifications That Survive Both the BLA Review and the Commercial Lot Release Program

The XGene GT Specification Justification Architecture is a stage-gated methodology for gene therapy programs that maps every release attribute from IND to BLA, linking each test to its analytical method maturity, regulatory basis, clinical data support, and the statistical derivation of its acceptance criterion.

1. Attribute Classification and Release/Characterization Designation: For each analytical test in the DS and DP specification table, document whether it carries a numerical acceptance criterion (lot release) or is informational only (extended characterization), and record the regulatory and scientific basis for that designation — this step forces programs to resolve “report result” entries before IND submission, not during BLA review when the clinical lot history is already fixed.

2. Acceptance Criterion Anchoring: For each numerical limit in the specification, document its justification basis — compendial/regulatory precedent, mechanistic safety rationale, or statistical derivation from manufacturing history — and identify which basis applies at each clinical phase, preventing the common failure mode where a Phase I interim limit is carried unchanged to BLA without a documented re-evaluation decision.

3. Potency Specification Architecture: Document the reference standard qualification record, the relative potency acceptance criterion in percentage units, the cell-based assay validation status at each phase, and the retrospective lot-by-lot potency data that supports the acceptance criterion used for clinical lot release — this step addresses the single most common source of BLA CMC deficiencies in GT programs and cannot be reconstructed after the fact from raw assay data alone.

4. Specification Tightening Roadmap: For each attribute where the acceptance criterion was set as an interim limit at Phase I, document the planned tightening decision — the data threshold, the lot count, and the statistical method that will be used to derive the Phase III and BLA acceptance criterion — committing the program to a specification development path before CBER asks for it during a pre-BLA meeting.

The output of the XGene GT Specification Justification Architecture is a specification development dossier that maps every attribute in CTD Sections 3.2.S.4.1 and 3.2.P.5.1 to its acceptance criterion, its justification basis, and its stage-gate history — not a gap list, but a complete, audit-ready record that supports both BLA submission and pre-approval inspection without improvisation.

Gene therapy programs that treat specification-setting as a Phase III problem discover at BLA that the clinical lots released in Phase I and Phase II were qualified against criteria that cannot be retrospectively justified. CBER does not accept clinical performance as a substitute for documented specification rationale; the lot release history is a scientific record, not a clinical outcomes argument. The cost of rebuilding a specification justification architecture after Phase III is complete — while simultaneously managing BLA review timelines and pre-approval inspection readiness — is a CMC execution risk that compounds with every week of delay. The specification decisions made at IND create the evidentiary foundation on which the entire lot release history is judged.

In your current IND CMC specification table for your GT drug substance, can you identify which attributes have defined numerical acceptance criteria with documented justification, which are listed as “report result” with no limit, and whether your potency specification is expressed as an absolute unit or a relative comparison to a reference standard?