Figure 1. Flowchart demonstrating general method of endotoxin removal. Solid boxes and lines denote mandatory steps while dotted lines are optional but recommended steps. (Lowe et al., 2012)
1. Define the Endotoxin Goal and Protein Constraints
Specify intended use, sample volume, protein concentration, required endotoxin endpoint, recovery target, activity requirement, and final buffer. Appropriate limits depend on use and applicable standards; do not apply one universal numerical threshold to every research or manufacturing sample.
Characterize protein size, charge, isoelectric behavior, oligomerization, hydrophobicity, cofactors, and stability. Endotoxin can associate through hydrophobic and electrostatic interactions, so the same buffer change that improves removal can unfold or precipitate the protein.
Use a meaningful reporting basis
Report endotoxin concentration and total endotoxin load, together with units per milligram or another justified product basis. Dilution can lower concentration without removing total endotoxin. Define accepted protein recovery using the functional, quality-qualified pool.
2. Establish a Matrix-Compatible Endotoxin Assay
Protein, salts, detergents, chelators, solvents, color, turbidity, and pH can inhibit or enhance endotoxin assays. Determine a suitable dilution range and demonstrate spike recovery or other required controls. Use endotoxin-free consumables and controlled water to avoid introducing contamination during measurement.
Retain sample identity, dilution, instrument, reagent lot, and calculation. A result below a method threshold should retain the qualifier; it is not zero. When comparing fractions with different matrices, confirm suitability for each or use a justified buffer-exchange step.
Assay variability can be large near the decision limit. Use replicate strategy and predefined acceptance appropriate to the method. Investigate unexpected nonlinearity across dilution rather than selecting the most favorable result.
3. Control Endotoxin Introduction before Purification
Reduce burden at the source through clean water, buffers, vessels, tubing, and handling. Expression host and lysis can create a high initial load, but later contamination from reusable components or raw materials can reverse earlier removal.
Track endotoxin across major operations to locate introduction or poor clearance. Sample blanks and buffers when a sudden increase appears. A polishing column cannot reliably compensate for uncontrolled contamination elsewhere in the process.
Minimize prolonged contact between protein and high-endotoxin matrices when association increases over time. Clarify promptly and maintain stable temperature and pH. Avoid assuming that filtration removes soluble or small endotoxin aggregates.
4. Understand Endotoxin Association and Aggregation
Endotoxin is amphiphilic and can form aggregates whose size and charge depend on salt, divalent cations, detergents, pH, and concentration. It may associate with hydrophobic or positively charged protein regions. These behaviors explain why one method works for one protein and fails for another.
Use small condition screens to test whether association can be weakened without damaging the protein. Vary pH, ionic strength, compatible additives, and concentration deliberately. Include no-media controls to detect precipitation, adsorption, or assay changes.
Do not infer removal mechanism solely from a lower assay response. The condition may mask endotoxin detection. Reassess fractions after suitable dilution or buffer exchange and verify protein quality.
5. Use Anion Exchange for Selective Endotoxin Binding
At suitable pH and ionic strength, negatively charged endotoxin can bind an anion exchanger while a protein with sufficiently different charge behavior flows through. Alternatively, both may bind and require selective wash or elution. Predict from pI only as a starting point; actual surface charge and association matter.
Screen pH and conductivity within the protein's stable region using ion exchange chromatography media. Measure target recovery, activity, and endotoxin in staged fractions. Define impurity capacity because late breakthrough can contaminate a large composite pool.
If the protein also binds strongly, a gradient can reveal whether a separation window exists. Avoid forcing release with conditions that compromise structure. Consider a different mechanism when the window is narrow or irreproducible.
6. Compare Affinity, Hydrophobic, and Phase-Based Options
Endotoxin-selective affinity interactions can provide strong binding, but capacity, specificity in the sample matrix, ligand leakage, regeneration, and protein adsorption must be evaluated. Hydrophobic methods may interact with endotoxin aggregates and the protein, creating recovery risk.
Detergent-assisted phase separation or other partitioning approaches can disrupt endotoxin–protein association, but residual reagents and protein stability require control. Ultrafiltration alone may not separate endotoxin reliably when aggregates and proteins overlap in effective size.
| Method concept | Potential advantage | Primary risk to check |
|---|---|---|
| Anion exchange | Charge-based flow-through or elution | Protein co-binding and impurity breakthrough |
| Endotoxin-selective affinity | Direct capture of endotoxin | Capacity, nonspecific protein loss, leakage |
| Hydrophobic interaction | Different surface association | Protein retention or aggregation |
| Phase or detergent method | Can disrupt association | Residual reagent and structural damage |
| Size-based separation | Useful for some aggregate differences | Overlapping sizes and dilution |
Use orthogonal approaches when one method cannot meet both recovery and endotoxin goals, but count cumulative yield and added process components.
7. Protect Protein Recovery and Function
Measure target mass, total protein, aggregation, concentration, and activity or binding after treatment. Endotoxin removal can produce a clear, low-protein supernatant because the product precipitated or adsorbed. Close mass balance across media, filters, phases, and containers.
Study concentration effects. Low protein concentration can increase nonspecific surface loss, while high concentration can promote endotoxin association or aggregation. Include a carrier only when compatible with intended use and analytics.
Use quality-based pooling. The lowest-endotoxin fraction may contain too little active protein; broader pooling may exceed the endpoint. Map the tradeoff and choose a robust boundary rather than reporting only the best individual fraction.
8. Place Endotoxin Removal in the Right Process Position
Early removal reduces downstream exposure but faces a complex feed and large endotoxin load. Late polishing treats a cleaner, smaller pool but endotoxin may be more tightly associated with concentrated product. Compare burden, volume, association, and available selectivity.
A capture step can remove much host material before targeted endotoxin polishing. Conversely, a well-designed anion exchange flow-through step may remove DNA and endotoxin together. Demonstrate both impurities rather than assuming a shared negative charge guarantees clearance.
Ensure the removal pool is compatible with concentration, sterile filtration, storage, or formulation. Add buffer exchange and residual-reagent control to the workflow when needed.
9. Challenge Robustness and Reuse
Challenge pH, conductivity, load, protein concentration, endotoxin burden, feed lot, and resin cycle where they can affect removal. Use staged samples to detect breakthrough. Define an operating range with margin for assay and process variability.
Develop cleaning and carryover controls for reusable media. Endotoxin can persist on systems, and a negative blank requires matrix-suitable assay recovery. Trend pressure, capacity, protein loss, and endotoxin clearance across cycles.
Scale by preserving residence time, normalized loads, inlet conditions, and collection logic while rechecking mixing and system hold-up. Use endotoxin-controlled buffers and equipment during scale studies so new contamination does not obscure chromatographic performance.
10. Distinguish True Endotoxin Removal from Dilution and Assay Masking
A protein sample reports a tenfold lower endotoxin concentration after a chromatography trial, but the product pool is also tenfold more dilute. Total endotoxin load is nearly unchanged. When results are normalized to recovered active protein, the apparent clearance disappears. The condition does not meet the process goal.
A second condition reports very low endotoxin without major dilution, yet spike recovery in the assay fails. The fraction contains a detergent that suppresses the readout. After compatible dilution and buffer exchange, endotoxin becomes measurable. The early result reflected assay masking rather than removal.
Map protein and endotoxin across all fractions
The team screens anion exchange conditions where endotoxin is expected to bind and the protein to flow through. At low conductivity, target recovery falls because the protein also binds. At higher conductivity, recovery improves but endotoxin breaks through early. An intermediate condition creates a usable window only at a limited impurity load.
Staged flow-through samples show that a large composite pool would exceed the endpoint near the end of loading. The operating load is set by endotoxin breakthrough, and a monitoring sample is taken before the predicted boundary. A later strip confirms the retained endotoxin load and supports mass balance.
A selective adsorption option gives greater endotoxin reduction but loses protein nonspecifically at low concentration. Adding a compatible stabilizing salt improves recovery without blocking endotoxin binding. A no-media control confirms that the protein is not simply precipitating during incubation.
Both options are compared by active protein recovered, total endotoxin, endotoxin per milligram of accepted protein, pool volume, time, and residual process additives. The anion exchange route is chosen for scale because it provides a broader recovery margin and integrates with DNA removal.
At scale, the team controls water and buffer endotoxin, feed hold, pH, conductivity, load, residence time, and system cleanliness. Assay suitability is demonstrated for feed, pool, and monitoring fractions. Blank cycles are sampled using the same containers and dilution strategy.
The final method does not claim universal endotoxin removal. It defines the tested protein, concentration, buffer, starting burden, operating range, assay controls, and maximum load. Changes to formulation or protein concentration trigger reassessment because association and recovery can shift.
11. Select a Method with a Recovery–Clearance Map
Compare methods at the same protein input and assay basis. Plot accepted protein recovery against residual endotoxin, activity, aggregate level, process time, pool volume, and residual reagents. Eliminate conditions that meet the endotoxin goal only through dilution or protein loss.
Confirm the leading method with independently prepared sample and deliberate parameter edges. Preserve raw assay controls and dilution data. Document what the result supports for the intended use without extending the conclusion to unrelated proteins or matrices.
For tailored support, share protein properties, starting endotoxin, required endpoint, current buffer, volume, concentration, stability, and assay controls through protein purification services.
Use endotoxin-spiked development samples cautiously. Added material may not associate with the protein in the same way as endotoxin carried through expression and purification. Compare spiked and native burden where feasible, and avoid basing the entire process on an easily removed spike.
Evaluate contact time for batch adsorbents or flow devices. Longer exposure can improve impurity binding but also increase protein adsorption, denaturation, or endotoxin redistribution. Separate equilibrium effects from practical processing time and microbial risk.
Check whether the removal step changes protein concentration or formulation enough to alter the assay dilution required. Maintain a sample-specific matrix-control plan after every significant method change. An apparently improved clearance should not be accepted until assay suitability remains demonstrated.
When the endpoint is near the analytical limit, define how censored results are reported and compared. Do not substitute zero for below-limit values. Use conservative calculations where needed and retain the method threshold in the final record.
Integrate endotoxin control with resin and consumable qualification. A low-endotoxin process can be disrupted by a new tubing lot, filter, raw material, or storage container. Investigation should cover the complete contact path, not only the removal column.
Define a response when a routine sample fails endotoxin criteria. Confirm assay suitability and sample identity, review dilution and controls, then investigate process source and removal performance. Avoid automatically reprocessing through the same method without understanding whether capacity, contamination, or association caused the failure.
If reprocessing is allowed, evaluate its effect on protein recovery, aggregation, activity, hold time, and cumulative exposure. Include reprocessed material in traceability and do not merge it silently with first-pass product. Prevention and robust loading limits are preferable to routine rescue.
Retain representative feed and product samples under qualified conditions when possible. Paired retesting can support an investigation into association, assay interference, or newly introduced contamination. Record all containers, buffers, operators, sampling points, dates, equipment states, and transfers because a low-level failure can originate after the chromatography step.
Balance endotoxin clearance with active protein recovery
Share the protein, buffer, starting endotoxin, target endpoint, stability limits, and assay controls to compare practical removal options.
References
- FDA. Pyrogen and Endotoxins Testing: Questions and Answers.
- Carta G, Jungbauer A. Protein Chromatography: Process Development and Scale-Up. 2nd edition. Wiley-VCH; 2020. doi:10.1002/9783527824045
- Lowe AJ, Bardliving CL, Batt CA. Methods for chromatographic removal of endotoxin. In: Voynov V, Caravella JA, eds. Therapeutic Proteins. Vol 899. Humana Press; 2012:265-275. doi:10.1007/978-1-61779-921-1_17
