The Endotoxin Problem No One Talks About

Team Dynamic Matrices | 8 September 2025

Reproducibility checklists for organoid work cover cell line authentication, mycoplasma, passage number and medium composition. Almost none of them cover endotoxin. That is a gap, because the amount of lipopolysaccharide it takes to change an immune-relevant readout is far below anything you would notice by eye, and animal-derived matrices are a variable, largely unreported source of it.

How little it takes

Endotoxin is the LPS from Gram-negative bacterial cell walls. It signals through TLR4, and the sensitivity of that pathway is the whole problem. Human monocytes and macrophages respond to LPS at 10 pg/mL, roughly 0.1 EU/mL (reviewed in Beutler and Rietschel, Nature Reviews Immunology, 2003). Dendritic cells mature at similar levels. The FDA limit for parenteral drugs is 0.5 EU/mL for most products and 0.2 EU/kg for intrathecal ones, and cell culture reagents are often specified only as "under 1 EU/mL" or not at all. Many labs would be surprised to learn that "endotoxin-tested" on a data sheet often means "below a limit that would still activate a macrophage".

Why 3D systems are more exposed than 2D

In a monolayer, medium is changed every day or two and the surface is washed. In an embedded 3D culture, the matrix holds whatever it came with in direct contact with cells for weeks. Matrices are not washed, and endotoxin is heat-stable, so it survives every sterilisation step the material saw. Once TLR4 fires in one population, paracrine cytokines (IL-6, TNF, IL-1β) propagate the response to epithelial, stromal and immune neighbours. The result is a shifted transcriptome, an altered cytokine profile, and in tumour organoid co-cultures a changed drug or immunotherapy response, none of which is visible under the microscope.

The matrix is the largest unreported source

Basement membrane extracts are produced from mouse tumours and carry a bacterial-byproduct load that varies between lots. Manufacturers report an endotoxin value per lot, and it is usually low. But it is a bulk average from a diluted sample, not the local concentration at the surface of a dome where organoids sit, and the reported value is rarely carried into publications. Decellularised ECM is worse: tissue processing introduces endotoxin and most protocols do not depyrogenate. Collagen and fibrin from animal sources vary too.

Synthetic matrices remove the largest biological source but do not guarantee endotoxin-free conditions on their own, because water, buffers, peptides and labware all contribute. We recommend LAL testing as a routine practice for any matrix used in immune-relevant work.

What we recommend

1. Make LAL testing a release criterion for any matrix used with immune cells, tumour-immune co-cultures, inflammatory models or vaccine work. Kinetic chromogenic LAL or recombinant Factor C assays both work and cost a few euros per sample.
2. Set a threshold for your assay, not the supplier's. For macrophage or dendritic cell work we suggest under 0.1 EU/mL in the final gel. For epithelial-only organoids, 0.5 EU/mL is usually acceptable.
3. Use a chemically defined, xeno-free matrix where the biology allows it. It removes the one input you cannot control.
4. Report endotoxin load in methods alongside lot number and mechanical properties. It costs one line and it may explain a future replication failure.
5. Include an LPS-spiked positive control and a polymyxin B-treated negative control in immune assays. If polymyxin changes your baseline, you had a contamination problem.