Process Characterization

Process Characterization, PPQ & CPV for Microbial Platforms

Design → GMP, without detours. A full-lifecycle programme that turns microbial processes into predictable, auditable, and continuously improving manufacturing systems—grounded in scale-down fidelity, clean CPP↔CQA maps, multicolumn capture (PCC), three-batch PPQ, and live CPV dashboards.

Executive overview

A microbial process is only as good as its ability to behave at scale, on schedule, under audit. Characterisation that reads like a thesis but cannot predict tomorrow’s lot is theatre. What you need is disciplined scale-down models that replicate mixing, oxygen, shear, residence time, and hold-time realities; CPP/CQA maps that name the few levers that move the few attributes that matter; DoE that finds robust islands rather than delicate peaks; PCC and membrane trains that hold productivity while consuming less resin and energy; PPQ that proves the argument with three consecutive lots; and CPV dashboards that watch the right tags and warn before CQAs drift.

Mika Biologics is microbial-first. We characterise E. coli, Pichia/yeasts, filamentous fungi, phage/lysins/bacteriocins, VLPs/OMVs/EVs, and LBPs (engineered probiotics) with the same philosophy: make the model true, measure what moves, and simplify the path to validation.

Interlock this service with Analytical & QC for Microbial Biologics (methods and COAs), Formulation & Aseptic Fill-Finish (Grade A Isolators) (hold times, CCIT, lyo), Yeast & Fungal Expression Systems, Exosomes & OMVs, Engineered Probiotics (LBP) GMP, and Advanced Yeast Glycoengineering (Pichia).

Mike Biologics, Petri dish graphic

What we deliver

  • Authoritative CPP↔CQA mapping anchored to QTPP and dossier claims (IND/IMPD → BLA/MAA); risk-ranked and experiment-proven.
  • Scale-down models that reproduce oxygen transfer, mixing time, shear spectrum, feed/methanol dynamics (Pichia), envelope stress (OMVs), infection kinetics (phage), and TFF/column histories.
  • Design spaces via DoE (classical and Bayesian) and MVDA, written so regulators can follow the logic in minutes.
  • Multicolumn capture (PCC) and membrane trains for high productivity at lower resin usage; resin life and cleanability trending built-in.
  • Process playbooks (USP/DSP/Formulation/Hold/Aseptic) with golden-batch traces and alarm bands.
  • Three-batch PPQ (with qualified scale-down) that closes cleanly and defends worst-case holds and ranges.
  • CPV dashboards that bind historian tags to CQAs/CQPs and trigger timely action; annual product reviews that write themselves.
  • Change control & comparability with pre-agreed acceptance bands and decision trees.

QTPP → CQA → CPP

We start with a single page everyone can hold in their head:

  • QTPP (examples): route/dose, DS/DP presentation (vial/PFS/capsule), storage (2–8 °C or lyo), potency mechanism (ADCC, enzyme unit rate, PFU, particle uptake), particle envelope (for VLP/EV), endotoxin limit (route-dependent).
  • CQAs (microbial-centric): identity (LC–MS/genome/markers), potency (PFU/ADCC/enzymatic units), aggregation (SEC-MALS), glycan states (afucose/sialylation/bisecting) for Pichia, residual DNA/HCP/LPS, particle size & count (NTA/DLS/EM), CFU & payload for LBPs, sterility/bioburden, pH/osmolality/appearance.
  • CPPs that move them: kLa/OTR, P/V, tip speed, ORP and headspace O₂ (anaerobes), methanol and redox (AOX1), feed composition and osmolality, induction timing and temperature ramps, nuclease exposure, TFF flux/ΔP and shear, column residence time/load/conductivity, buffer pH and ionic strength, lyo shelf temperature/Pirani profiles, hold-time and temperature at each node.

We prove the map with screening DoE → focused DoE → confirmation runs and quantify effect sizes. Anything that doesn’t move a CQA is removed from the “critical” vocabulary.

Scale-down models: fidelity first

A scale-down model is not a small bioreactor; it is a physics replica. For each platform we match the quantities that matter:

Upstream (USP)

  • E. coli: maintain kLa, P/V, mixing time, and tip speed envelopes; oxygen transfer matched to OUR; antifoam policies that do not cripple kLa; induction regimes (IPTG/arabinose/temperature) or infection kinetics (phage MOI and host OD).
  • Pichia/yeasts: methanol strategy (AOX1) or methanol-free (GAP) with redox management; temperature ramps for folding; co-factor and trace metal hygiene for glyco-transferases; redox (GSH/GSSG) boundaries when glycosylation is a CQA.
  • Anaerobes/LBPs: ORP and headspace O₂ as regulated impurities; pre-reduced media; closed loop gas handling; growth without payload expression (to reduce selection pressure) unless physiology requires it.
  • OMV producers: envelope stress bands that induce vesiculation without yield hysteresis; divalent ions and temperature set-points.

Downstream (DSP)

  • Clarification: shear spectrum (nozzle tips/flow fields) matched so floc size distribution is comparable; centrifuge g-profiles translated to bench flows.
  • TFF: same membrane chemistry/MWCO; matched flux and wall shear; realistic concentration factors and hold times; nuclease dose and timing replicated when DNA is a CQA.
  • Chromatography: identical resins/ligands; residence time, load density, pH and conductivity trajectories matched; gradient shape reproduced; CIP recipes validated at bench for fouling history; PCC logic prototyped with equivalent inter-column dynamics.
  • Filtration/polish: filter area per volume and fouling index; hold times bracketing worst-case shop-floor realities.
  • Formulation & lyo: DF buffers, excipients, Tm/Tg; lyo cycle with shelf temp / Pirani / capacitance signatures and acceptance bands that can be seen in CPV.

We document model qualification: what it replicates (and doesn’t), bias vs scale, and acceptance windows that keep PPQ honest.

Design of Experiments (DoE) and MVDA: robust islands, not sharp peaks

We run DoE to find robustness, not to chase vanity optima:

  • Screening (Plackett–Burman/definitive screening) to kill variables that don’t matter.
  • Response surfaces for the few that do (pH–conductivity–residence time for IEX; flux–TMP–CF for TFF; methanol–DO–temperature for AOX1).
  • Mixture designs for buffers/excipients.
  • Blocking for day/analyst/lot effects.
  • MVDA/latent variable models (PCA/PLS) to connect historian tags to CQAs and detect multi-collinearity and hidden drifts.

Outputs are written as design spaces with plain English guardrails: “Within pH X–Y and conductivity A–B, at residence time R±Δ, aggregate fraction remains ≤Z%.” No ambiguity, no acrobatics.

Multicolumn capture (PCC): productivity without pain

For microbial proteins (enzymes, Fc fusions from Pichia, certain cytokines) and high-volume campaigns, PCC enables near-continuous capture with higher resin utilisation:

  • Dynamic binding capacity measured under realistic feed windows (viscosity, HCP, DNA);
  • Resin life trended by pressure rise and carry-over;
  • Column cycling with real-time UV and conductivity controls;
  • Buffer & WFI consumption reduced and documented;
  • Membrane adsorbers considered where flow and fouling favour them.

We design PCC so it is explainable in an audit and transferable to single-column operation when the site requires it.

Platform-specific characterisation patterns

Phage therapeutics

Infection MOI vs host growth phase; lysis timing; nuclease conditioning; AEX/TFF residual DNA and LPS removal; titer loss vs ionic strength/excipients; DP shear sensitivity; 0.22 µm infeasibility narratives → aseptic closed processing; stability on PFU and genome integrity.

Lysin & depolymerase enzymes

Inclusion body strategy with refold windows (redox, urea/arginine, temperature); HIC/SEC polish vs activity retention; endotoxin control; potency drift under light/agitation; lyo cycle that protects active site.

Bacteriocins & lantibiotics

Secretion vs acid extraction; IEX/HIC stack; AU/MIC potency; salt legacies; ingredient vs drug CQA splits; thermotolerance at acidic pH.

VLPs/OMVs/EVs

Clarification that preserves vesicles; TFF (MWCO 300–500 kDa); SEC polish; NTA–DLS–EM convergence; endotoxin and host-component removal; particle stability vs ionic strength; aseptic filling of non-filterable bulk.

Pichia biobetters

Glycoform control (afucose, α2,6-sialylation, bisecting); redox/feed/temperature levers; IEX/HIC trains that protect sialic acids; linkage-specific analytics and ADCC/FcRn ties; comparability across campaigns.

Engineered probiotics (LBP)

ORP/DO and pre-reduced media; closed harvest; TFF and cryo-protectants; enteric capsules / microencapsulation and colon release; CFU and payload potency; kill-switch challenge assays; package O₂/moisture budgets; environmental risk narratives.

Hold times, transfers, and aseptic nodes

We specify maximum allowable hold times and temperatures at every node with data. For any non-filterable bulk (phage, vesicles), we define closed aseptic processing, media fills where appropriate, and EM trending ranges. Transfer lines, connectors, and filter integrity policies are documented with acceptance criteria that can be read by an inspector in a single view.

The PPQ plan

Three consecutive commercial-scale lots, at planned (or proven worst-case) ranges, with:

  • Qualified scale-down model and pre-PPQ comparability to development scale;
  • MBRs/eBRs locked; sampling plans that reflect variability;
  • CPP envelopes bracketing credible extremes (e.g., residence time upper bound that still meets purity);
  • Hold-time challenges folded into PPQ or supporting studies;
  • Aseptic demonstration: EM and media fills aligned;
  • Release panel that is phase-appropriate but BLA-grade;
  • Acceptance criteria that reflect real capability (Cp/Cpk) instead of hope.

The PPQ report tells a simple story: the model predicts; the lots perform; the system is controllable; the product is reproducible.

CPV dashboards: early warning, not forensic archaeology

We wire historian tags to CQAs and push them to a live dashboard:

  • USP: DO/ORP, kLa surrogates, base addition profiles, methanol flows, OUR/CER, redox, temperature ramps.
  • DSP: TFF flux/ΔP/fouling index, conductivity/pH traces through capture/polish, column ΔP and UV traces, filter capacities, DF volumes, lyo shelf temperature/Pirani/capacitance deltas.
  • Formulation & fill: bulk age/temperature, hold-time clocks, fill temperature, CCIT outcomes.
  • Analytics: key release attributes and stability drift.

On top we run:

  • Control charts with Nelson/WECO rules;
  • Golden-batch overlays (with tolerance bands);
  • Capability indices (Cp, Cpk) and rolling Ppk;
  • Signals when patterns predict CQA movement (PCA/PLS residuals);
  • One-click APR/PQR exports.

The rulebook is pre-agreed: what triggers a deviation, what triggers enhanced sampling, what constitutes process improvement vs. change control.

Documentation spine (so audits are short)

  • Characterisation report: models & DoE, CPP/CQA map, design space.
  • Control strategy: narrative and table; batch recipes by unit op; alarms and responses.
  • PPQ protocol & report: lots, ranges, acceptance, outcomes.
  • CPV plan: tags, charts, rules, responsibilities.
  • Comparability protocol: assays, bands, decision tree for scale/site/process changes.
  • Cleaning validation & resin life: CIP recipes, pressure/plate height trends, carry-over study.
  • Data integrity: LIMS/eBatch with ALCOA+; audit trails; training records.

30-day onboarding deliverables (phase-appropriate)

  • Day 10: QTPP → CQA → CPP one-pager; gap list; draft sampling plan.
  • Day 20: Scale-down qualification plan; initial DoE matrix; CPV tag shortlist.
  • Day 30: Draft control strategy; PPQ outline with proposed acceptance criteria; CPV dashboard prototype populated with historic data.

Illustrative acceptance bands

  • IEX capture: pH 7.6–8.0; conductivity 6–9 mS/cm; residence 4.0 ± 0.5 min → purity ≥ 95%, aggregates ≤ 1.5%.
  • TFF DF: flux 30–45 LMH; ΔP ≤ 1.2 bar; CF 5–8× → DNA ≤ 10 ng/mL; enzyme activity ≥ 90% of T0.
  • Phage AEX: load ≤ 1.0 × 10¹³ PFU/L; conductivity 3–5 mS/cm → endotoxin ≤ acceptance; PFU recovery ≥ 70%.
  • Pichia glyco-window: temp 20–24 °C; redox band X–Y; feed ratio A/B → afucose ≤ 1.0%, α2,6-Sia ≥ 10%, bisecting ≥ 15%.
  • LBP lyo: residual moisture 1.5–3.0%; package O₂ ≤ 1.0%; CFU decay ≤ 0.5 log/12 months.

Change control and comparability (no surprises)

Every foreseeable change has a lane:

  • Raw material (resin lot, membrane) → bridging;
  • Scale/site → comparability package with assays/bands;
  • Process (buffer pH/conductivity, residence time) → within design space = manage; outside = comparability;
  • Analytical → method transfer and cross-validation plan.

Decision trees are included in the protocol. Reviewers love them because they eliminate improvisation.

Build faster, ferment Smarter, Mika Biologics

Why Mika for process characterisation, PPQ & CPV

  • Microbial native: we know oxygen transfer, methanol/redox, vesicle shear, phage infection, and anaerobic ORP because we live there.
  • Audit logic: documents read like decisions, not diaries.
  • Through-line: analytics, formulation, and fill-finish teams co-design guardrails—no handoff friction.
  • PCC & membranes that are productive and defensible.
  • Dashboards that reduce meetings and shrink APR time.

Inter-page guidance

  • Pair with Analytical & QC for Microbial Biologics for release and stability method depth.
  • Use Formulation & Aseptic Fill-Finish (Grade A Isolators) for hold-time, lyo, CCIT, and fill logic.
  • For vesicles, cross-reference Exosomes & OMVs — Bioprocessing.
  • For glyco-tuned proteins, see Advanced Yeast Glycoengineering (Pichia).
  • For live products, see Engineered Probiotics (LBP) GMP.

FAQ

Do we need a qualified scale-down model before PPQ?
Yes. PPQ without a trustworthy small-scale surrogate invites surprises and weakens your lifecycle control narrative. We qualify what the model reproduces (oxygen, shear, residence, holds) and declare its limits.

Can PCC be used for ion-exchange as well as affinity?
Yes. PCC logic applies to IEX and mixed-mode where residence times, loads, and selectivity are well understood. We design for explainability and provide single-column fallbacks when sites require it.

How do you choose CPPs among dozens of variables?
By proof. We run screening DoE and MVDA on historian data, quantify effect sizes on CQAs, and keep only levers that move attributes. Everything else becomes “key” or “supporting,” not “critical.”

What if our product cannot be 0.22 µm filtered?
We characterise and validate closed aseptic processing, prove bioburden control, and align EM/media fills with the presentation. Release panels and rationales are written to satisfy reviewers.

How do CPV dashboards avoid alarm fatigue?
We bind only the few historian tags that predict CQAs, set tolerance bands and rules (Nelson/WECO, PCA residuals), and assign responses. Teams see fewer, more meaningful alerts.

Can you integrate our sponsor analytics into PPQ/CPV?
Yes. We transfer or bridge methods under SLA, qualify/validate as needed, and surface results in the same dashboards and annual reports.

Will regulators accept rFC for endotoxin in PPQ/CPV?
Where appropriate and method suitability is demonstrated, yes. We justify rFC or LAL in the control strategy and keep spike-recovery evidence current.

How do you write comparability when changing scale or site?
We define assay panels and acceptance bands up front (e.g., purity, aggregates, potency, particle metrics, glycan fingerprint), specify statistical tests, and include decision trees—so changes proceed without improvisation.

What’s the minimum for early-phase characterisation?
A lean QTPP→CQA→CPP map, a basic but honest scale-down, and targeted DoE on the two or three levers most likely to fail. You can add dimensionality later; do not skip fidelity now.

Can CPV feed directly into APR/PQR?
Yes. Dashboards export the plots and statistics your QA needs. APR ceases to be a reconstruction exercise.