Microbiome Therapeutics — CMC Regulatory Framework for Live Biotherapeutic Products Under FDA

Vowst received FDA approval in April 2023 — the first orally administered microbiome therapeutic for C. difficile recurrence prevention. Rebyota was approved in November 2022 as the first fecal microbiota product for rectal administration. The regulatory precedent now exists. What has not changed is the pattern of IND CMC deficiencies that delays every LBP program treating the microbiome therapeutic as if it were a small molecule oral drug: no strain master bank, no whole-genome-sequencing-based strain identity documentation, no antibiotic resistance transfer risk assessment, and a viable cell count specification derived from manufacturing capability rather than a clinical dose-response relationship.

The approvals of Vowst and Rebyota did not lower the CMC bar for this product class — they defined it.

Strain Characterization by WGS — Identity, Antibiotic Resistance Profiling, and the HGT Risk Assessment CBER Requires Before Phase 1

Species-level identification by 16S rRNA sequencing tells you which bacterial species is present, but for a defined-consortium live biotherapeutic product, that is not the identity question CBER actually needs answered — because different strains within the same species can have dramatically different colonization behavior, immunomodulatory activity, and therapeutic properties, meaning strain-level resolution is the regulatory minimum, not an enhancement. 16S rRNA sequencing of the V3-V4 region against a reference type strain, commonly requiring 99% or greater identity, establishes species-level confirmation, but whole-genome sequencing with complete assembly, combining short-read and long-read sequencing to resolve both chromosome and plasmid content, is what actually confirms strain identity at the sub-species level through average nucleotide identity comparison against the reference genome. That same whole-genome dataset does double duty as the antibiotic resistance and virulence screening tool, since screening the assembled genome against resistance and virulence gene databases identifies genetic determinants that phenotypic testing alone cannot localize. Phenotypic susceptibility testing by broth microdilution against a panel of clinically relevant antibiotics, referenced against established epidemiological cutoffs, classifies each strain as wild-type or non-wild-type for each antibiotic tested; a strain flagged as non-wild-type requires the genomic data to determine whether the underlying resistance mechanism is chromosomally intrinsic and non-transferable, or carried on a mobile genetic element capable of horizontal transfer — and where a mobile element is confirmed, an in vitro conjugation transfer assay assessing actual transfer frequency to a recipient organism becomes a necessary risk characterization step. An IND CMC section that documents strain identity only by genus and species name, without 16S rRNA data, whole-genome sequencing, and antibiotic resistance profiling for each strain, is the specific gap that has triggered clinical holds — CBER cannot confirm the safety profile of live bacteria intended for human administration without this characterization in hand.

Anaerobic Manufacturing, Viable Cell Count Specifications, and the Distinction Between a Viability Specification and a Functional Potency Assay

Manufacturing a defined bacterial consortium — particularly one combining spore-forming and non-spore-forming species — requires anaerobic fermentation conditions maintained essentially free of dissolved oxygen throughout, followed by downstream processing choices, lyophilization for spore-forming strains and spray drying for vegetative non-spore-formers, that preserve each strain’s viability through the drug substance’s shelf life. Viable cell count by colony-forming unit assay, measured on selective anaerobic media, is the resulting viability specification, commonly set at a substantial minimum count per strain per dose at time of manufacture with a somewhat lower shelf-life floor that accounts for the expected viability decline over the product’s stated shelf life; for spore-forming strains, a parallel heat-resistant spore count specifically characterizes the fraction capable of surviving gastric transit to reach the colon. The critical distinction CBER’s 2022 draft LBP guidance draws explicitly is that this CFU-based viability measurement is not the same thing as a potency specification, because 21 CFR 610.10’s potency requirement calls for demonstrating the product’s specific ability to produce a defined biological result — competitive exclusion of a pathogen, immunomodulatory cytokine effect, or another therapeutic mechanism specific to the product’s claimed activity — which a colony count cannot demonstrate no matter how precisely it’s measured. A functional potency assay validated to standard analytical method requirements, producing a relative potency value against a characterized reference standard suitable for batch release, is what CBER expects to see alongside the viability specification at BLA stage. A drug substance specification table listing viable cell count as the only quantitative release attribute, without a validated functional potency assay behind it, has confirmed the bacteria are alive without confirming they do what the product claims to do therapeutically.

Vowst, Rebyota, and the BLA CMC Precedent That Defines What CBER Now Expects for All LBP Programs

The regulatory record for Vowst, approved as the first orally administered defined-consortium microbiome therapeutic, establishes the CMC precedent for spore-based products built from a defined set of characterized bacterial strains — including the whole-genome-sequencing-based strain identity documentation and spore count specification this article describes. Rebyota’s approval as the first fecal microbiota product for rectal administration establishes a parallel but structurally different precedent, since Rebyota is manufactured from an undefined donor-derived microbiota community rather than a defined consortium of individually characterized strains, requiring donor screening protocols and metagenomic composition analysis rather than strain-by-strain whole-genome characterization. The contrast between these two approved products is instructive precisely because it shows CBER accommodating two structurally different manufacturing philosophies — defined consortium versus undefined donor material — while holding both to the same underlying safety and identity principles: know exactly what organisms are in the product, and demonstrate the product does what it claims to do beyond simply containing live bacteria. A program modeling its CMC strategy on either precedent without first determining whether its own product is genuinely a defined consortium or an undefined community has picked the wrong template, since the strain characterization burden, the antibiotic resistance profiling scope, and even the composition analysis method differ substantially between the two regulatory paths these two approved products represent.

The XGene Live Biotherapeutic Product CMC Architecture — Strain Characterization, MCB/WCB, Manufacturing Controls, Potency Assay, and CBER BLA Documentation

The XGene Live Biotherapeutic Product CMC Architecture is a structured CMC development framework built around the recognition that live biotherapeutics require strain-level genomic characterization and a functional potency assay distinct from viability testing — attributes a small-molecule or food-supplement mental model will not anticipate.

1. Strain Characterization Cascade — Build 16S rRNA species confirmation, whole-genome sequencing for strain-level identity, and antibiotic resistance/virulence gene profiling into a single integrated characterization package for every strain in the consortium. 2. Master and Working Bank Establishment — Document strain banks with passage stability data confirming no drift in resistance genes, virulence genes, or 16S rRNA sequence across manufacturing. 3. Anaerobic Manufacturing Process Controls — Specify fermentation atmosphere, downstream processing method by strain type, and viability/spore count specifications with defensible shelf-life justification. 4. Functional Potency Assay Development — Build a validated, clinically relevant potency assay distinct from CFU viability testing, with a relative potency specification appropriate to the product’s claimed mechanism. 5. Defined-Versus-Undefined Product Classification — Determine explicitly whether the product is a defined consortium or an undefined donor-derived community before selecting the applicable CMC precedent and characterization scope.

The output is the LBP CMC package built for the biological product regulatory framework CBER actually applies, with strain-level genomic rigor and a genuine potency specification rather than a viability count standing in for one.

The regulatory record for Vowst (fecal microbiota spores, live-brpk, Seres Therapeutics, BLA 125757, approved April 26, 2023) established the CMC precedent for defined-consortium spore-based microbiome therapeutics, including whole-genome-sequencing-based strain identity and spore count specification requirements. The regulatory record for Rebyota (fecal microbiota, live-jslm, Ferring Pharmaceuticals, BLA 125739, approved November 30, 2022) established the parallel CMC framework for undefined donor-derived microbiota products, emphasizing donor screening and metagenomic composition analysis. FDA’s 2022 Draft Guidance for Industry: Quality Considerations for Live Biotherapeutic Products represents CBER’s most complete statement of the genetic stability, antibiotic resistance, and functional potency expectations this article addresses.

For your live biotherapeutic product IND or BLA CMC package, can you confirm today that your strain characterization includes whole-genome sequencing with antibiotic resistance profiling for each individual strain, and that your drug substance specification includes a functional potency assay distinct from viable cell count, bridged to your product’s specific claimed mechanism of action?