Endotoxin Testing in Cell Culture

Endotoxin is lipopolysaccharide (LPS) from the outer membrane of Gram-negative bacteria. It is shed during growth and released in bulk when the cell lyses — which means that killing the bacteria does not remove the endotoxin. A solution can be perfectly sterile and heavily contaminated with endotoxin at the same time.

Three properties make it a persistent problem in laboratory water and reagents:

• It is heat-stable. Standard autoclaving at 121 °C does not destroy it. Dry-heat depyrogenation is required, and the commonly validated minimum for heat-stable materials such as glassware is 30 minutes at 250 °C.
• It passes sterilising filters. The LPS monomer is a few kilodaltons; even as aggregates it passes 0.2 µm membranes freely. Sterile filtration is not a depyrogenation step. Removal requires ultrafiltration (membranes of roughly 10 kDa cut-off retain LPS aggregates), affinity or charged-membrane adsorption, or distillation.
• It adsorbs to surfaces and it accumulates. Water systems that sit stagnant grow biofilm, and biofilm sheds endotoxin continuously. A water system that passes a microbial specification can still fail an endotoxin one.

The practical conclusion is that endotoxin has to be designed out at the raw-material and water-system level. There is no filter you can put on the end of the process to fix it.

Endotoxin is reported in endotoxin units (EU) rather than in mass, because different LPS chemotypes differ substantially in biological potency. The unit is defined against an international reference standard endotoxin from E. coli, and the widely used conversion is that 1 EU corresponds to approximately 0.1–0.2 ng of reference-standard endotoxin. Treat that as an order-of-magnitude conversion, not an exact one: it holds for the reference material and only approximately for whatever LPS is actually in your sample. Where you see EU/ng quoted the other way round, the same reference standard is typically assigned about 5–10 EU per ng.

You will meet three denominators:

• EU/mL for liquids — water, media, sera, buffers. This is the cell-culture unit.
• EU/mg for solids and purified proteins — normalised to the mass of the substance rather than the volume of solution.
• EU/device or EU/unit for finished articles, which is a manufacturing release unit rather than a cell-culture one.

For manufactured parenteral products, limits are derived from a threshold pyrogenic dose expressed per kilogram of body mass, with the general threshold of 5 EU/kg per hour and a much lower figure of 0.2 EU/kg for intrathecally administered products. These are product-release specifications used in manufacturing quality control; they are not cell-culture limits and they do not translate to a flask.

Endotoxin is not simply an inert impurity. In cells that express the TLR4–MD-2–CD14 receptor complex it is a potent, specific agonist, active at picogram-per-millilitre concentrations. What follows depends entirely on which cells you are growing.

Highly sensitive: monocytes and macrophages, dendritic cells, whole blood and PBMC preparations, primary immune cells generally. In these systems endotoxin drives cytokine release directly, and a contaminated serum lot or water source will produce a result that looks like a real biological effect. Any experiment reporting a cytokine response to a novel stimulus needs an endotoxin control on the stimulus preparation — this is one of the commonest confounders in the immunology literature.

Moderately sensitive: mesenchymal stromal cells, pluripotent and other stem cells, primary endothelial cells, and hepatocytes. Reported effects include altered proliferation, altered differentiation trajectory, and changes in surface marker expression.

Relatively insensitive: established immortalised lines such as CHO, HEK293 or HeLa tolerate endotoxin loads that would confound a primary immune culture. This is why a lab that has only ever grown CHO cells may never have thought about endotoxin, and why moving into primary cell work is often the point at which it suddenly appears as an unexplained variable.

For manufacture — cell therapy, vaccine and biologics production — endotoxin becomes a controlled attribute of the process rather than a nuisance, because it carries through into the product and must meet a defined release specification. That drives the demand for low-endotoxin raw materials with lot-specific certificates of analysis.

In rough order of how much trouble each causes:

1. Water. The largest single contributor by volume in any media preparation. Purified water that has stagnated, a deionisation bed that has been in service too long, or a distribution loop with a dead leg will all shed endotoxin from biofilm. Water for cell culture use should be specified on endotoxin, not just on resistivity.
2. Sera. Fetal bovine serum has historically been the main culprit because of its biological origin and collection conditions. Modern serum is filtered and tested, and grades are differentiated explicitly on endotoxin content.
3. Raw materials and powders. Amino acids, salts, glucose, hydrolysates and particularly protein supplements such as albumin or transferrin. Plant and yeast hydrolysates are a known source. Low-endotoxin grades exist for exactly this reason.
4. Labware and containers. Glassware washed but not depyrogenated. Non-tissue-culture-grade plastics. Reused bottles. Silicone tubing.
5. The lab itself. Water baths, humidified incubator water pans, and any warm standing water. These grow Gram-negative organisms readily, and while the culture may never be contaminated, aerosols and handling transfer endotoxin.

Notice what is absent from this list: your cells. Endotoxin arrives with what you add, not from what you grow.

All current endotoxin assays exploit the same biochemistry — a proteolytic cascade in horseshoe crab haemolymph that is triggered by LPS binding to Factor C, the first enzyme in the chain.

LAL (limulus amebocyte lysate) uses the whole natural cascade from Limulus polyphemus or Tachypleus tridentatus amebocytes. It exists in three compendial formats under USP <85>:

• Gel-clot — the cascade coagulates the lysate. The readout is binary at a defined lysate sensitivity (λ), and quantitation comes from a dilution series. Simple, cheap, no instrument, and still the referee method when a result is disputed. Semi-quantitative and labour-intensive.
• Turbidimetric — measures the increase in turbidity as coagulin forms. The kinetic version reads time-to-onset against a standard curve and is fully quantitative.
• Chromogenic — the activated enzyme cleaves a synthetic peptide–chromophore substrate, releasing p-nitroaniline measured at 405 nm. Kinetic chromogenic is the most widely used quantitative format and reaches down to about 0.001 EU/mL.

rFC (recombinant Factor C) replaces crab-derived lysate with recombinantly expressed Factor C and a fluorogenic substrate. It measures the same endotoxin activity and gives equivalent results, with a typical quantitative range of about 0.005–5 EU/mL. Because it uses only the first enzyme of the cascade, it does not respond to (1→3)-β-D-glucan, which is a genuine advantage: β-glucan from cellulose filters and some media components causes false positives in LAL via the Factor G side-branch. rFC has its own compendial chapter, USP <86>, and its adoption is driven both by that specificity and by removing dependence on horseshoe crab harvesting.

Monocyte activation test (MAT) is a different animal — an in vitro assay that measures cytokine release from human monocytic cells and therefore detects non-endotoxin pyrogens as well. It is a manufacturing QC method, not a routine cell-culture reagent test.

Method validation matters more than method choice. Cell culture media, sera and buffers frequently interfere with the cascade, either inhibiting it (divalent cations, chelators, pH outside 6–8, serum proteases) or enhancing it. Every sample type must be qualified by an inhibition/enhancement test: spike a known endotoxin concentration into the sample as a positive product control (PPC) and confirm recovery within 50–200%. If recovery falls outside that window, dilute the sample — up to the maximum valid dilution (MVD) calculated from the specification limit and the assay sensitivity — or treat it to remove the interference. An untested sample matrix produces numbers that mean nothing.

There is no single regulatory limit for cell culture materials as there is for injectables. What exists is a set of well-established industry conventions plus whatever your own application demands.

• Water. The pharmacopoeial limit for water for injection is 0.25 EU/mL, and cell-culture-grade water is commonly specified to the same figure. LAL reagent water — used for the assay itself, not for culture — must be <0.005 EU/mL.
• Fetal bovine serum. The general industry standard is <10 EU/mL, which is adequate for most established cell lines. Low-endotoxin grades are specified at <1 EU/mL and ultra-low endotoxin grades below 0.1 EU/mL. For primary immune cells, stem cells or manufacturing use, the difference is worth paying for; for routine CHO or HEK293 work it usually is not.
• Buffers, salts and supplements. ≤1 EU/mL is the usual cell-culture-grade specification for liquid reagents such as HEPES or PBS.
• Selection agents and small-molecule solutions. Commonly ≤1 to ≤5 EU/mL, reflecting the fact that they are used at high dilution into the final medium.

That last point is the one to reason from. What matters is the endotoxin concentration in the final medium, not in the bottle. A supplement at 5 EU/mL added at 1% v/v contributes 0.05 EU/mL to the finished medium; serum at 10 EU/mL added at 10% contributes 1 EU/mL and is the dominant term. Add up the contributions across every component before deciding which one to upgrade — buying ultra-low endotoxin water while adding standard serum at 10% is money spent in the wrong place.

Water. Production combines several orthogonal steps: reverse osmosis to remove the bulk of dissolved solids and macromolecules, distillation and/or ultrafiltration through a membrane of roughly 10 kDa cut-off to retain LPS aggregates, and continuous recirculation with UV treatment and periodic sanitisation to stop biofilm re-establishing. The distribution system matters as much as the purification: dead legs, stagnant sample points and inadequately sanitised fittings are where endotoxin comes back. Final container filling is done into depyrogenated containers, and the finished lot is tested by LAL or rFC and released against a stated specification.

Sera. Endotoxin is controlled first by collection and handling — hygienic collection, rapid chilling and a maintained cold chain limit Gram-negative growth in the raw material, which is far more effective than anything done downstream. Serial filtration through progressively finer sterilising-grade membranes reduces particulate and microbial load. Some low-endotoxin grades add an ultrafiltration or adsorption step, at the cost of removing some of the growth-promoting components that make serum useful in the first place, which is the trade-off behind the price difference between grades. Every lot is tested and the result appears on the certificate of analysis.

What to ask a supplier. A stated specification is necessary but not sufficient. Ask for the lot-specific certificate of analysis with the measured value, not just the specification limit; the method used (LAL gel-clot, kinetic chromogenic, or rFC); and whether the test was validated for that matrix with a positive product control. “Low endotoxin” without a number and a method on a lot-specific document is marketing, not data.

In your own lab. Depyrogenate glassware by dry heat at 250 °C for 30 minutes rather than relying on autoclaving. Use single-use, certified-endotoxin-free plasticware and pipette tips for endotoxin-sensitive work. Do not store purified water — use it fresh, because stored water grows biofilm. Prepare and filter media promptly rather than leaving intermediate solutions standing at room temperature. And keep an endotoxin control in any experiment where the readout is an immune response.