Low Endotoxin Recovery: An FDA Perspective

The Code of Federal Regulations, 21 CFR 211.167(a), requires that any drug product claiming to be sterile and non-pyrogenic be tested for conformance for sterility and endotoxin. In addition, 21 CFR 610.13(b) requires that any biological product intended for use by injection be tested for pyrogenic substances by intravenous injection into rabbits. The rabbit pyrogen test requirement may be waived if a method is shown to be equivalent in accordance with 21 CFR 610.9. A “Low Endotoxin Recovery” (LER) phenomenon has been reported for undiluted biological products in certain commonly used formulations in which compendial methods cannot estimate endotoxin5, 6. Recent results from endotoxin spiking studies submitted to the FDA for several Biologics License Applications (BLAs) have confirmed that certain polysorbate/excipient combinations exhibit LER. However, results from spiking studies have been confounding, inconsistent, and difficult to interpret. This paper presents some of the regulatory challenges in assessing conformance to CFR pyrogen testing requirements.

Sponsors of BLA submissions have reported unacceptable time-dependent recovery of endotoxin spiked into undiluted drug product using the Limulus Amebocyte Lysate (LAL) USP <85> methods for endotoxin1-3, 5, 6. Chen and Vinther6 referred to this phenomenon as Low Endotoxin Recovery (LER). Chen and Vinther also reported a case in which spiked endotoxin in a biological product was undetectable using compendial methods but was pyrogenic in the USP <151> rabbit pyrogen test (RPT), indicating an endotoxin masking effect. LER or masking has been observed with several other biological products, suggesting that drug product release assays may underreport endotoxin levels in certain situations. The FDA has considered this a potential safety issue due to the possibility that products contaminated with endotoxin not detected by the compendial USP <85> methods may be pyrogenic in humans12, 13.
Endotoxin contamination of biological products may occur during manufacturing with a particular production process (e.g., a Gram-negative bacterial expression system) and is not adequately removed during purification; or when Gram-negative bacteria and/or endotoxin from raw materials, water, manufacturing equipment, etc. are introduced during processing18. Because endotoxins have strong adverse effects in humans, they must be reduced to levels that do not produce pyrogenic effects in humans. To ensure that the final drug product is non-pyrogenic, the manufacturing process must be microbiologically controlled and appropriate testing must be implemented for in-process intermediates and the final drug product.
Bacterial lipopolysaccharide (LPS) is the major component of the outer wall of all Gram-negative bacteria. LPS consists of a hydrophilic polysaccharide portion and a hydrophobic lipid A portion10, 11. Lipid A is responsible for the toxicity of Gram-negative bacteria. The most active form of lipid A contains six fatty acyl groups and is found in pathogenic bacteria such as E. coli and Salmonella species21, 22. Lipid A is negatively charged and can bind to positively charged drug products26.
LPS forms aggregates in solution. The molecular weight of the aggregates is about 300,000 to 1,000,000 Da while the molecular weight of the LPS monomer is about 10,000 to 20,000 Da and varies among various microorganisms18. The pyrogenic response varies for LPS preparations from various organisms14. The degree of LPS aggregation can affect the pyrogenicity and LAL reactivity28, 29. It has been suggested that dissociation of LPS occurs in the presence of citrate or phosphate which destabilizes the LPS structure by chelating the divalent cations5, 6, 25, 31. Mueller et al.16 suggested that aggregates are the active units of endotoxin. TNF production from human mononuclear cells and LAL activity were reported with aggregates but not with monomers obtained by dialysis treatment of samples23, 31. However, there is some controversy regarding the active form of LPS in vivo. Monomers and aggregates may be in equilibrium in vivo. How this affects biological activity is not well understood. Various factors, such as net charge of protein, hydrophobic proteins, pH, and ionic strength influence the aggregation state of LPS28, 29. Sasaki and White have suggested that LPS monomers and multimers are likely to be active units for the immune system27.
In vivo, LPS is recognized by the Toll-like receptor 4 (TLR4) which can interact with three different extracellular proteins (LPS binding protein [LBP], CD-14, and myeloid differentiation protein 2 [MD-2]) to induce a signaling cascade leading to the production of pro-inflammatory cytokines21, 22. The length and number of acyl chains in natural LPS vary among bacterial species and appear to be strongly correlated with agonist strength. For example, underacylated lipid A containing four or five fatty acids has been shown to inhibit host defense responses by blocking LPS-dependent activation of TLR4. Underacylated lipid A can compete for the same binding site on MD-2 and can, in a dose-dependent manner, inhibit the strong endotoxic response triggered by hexacylated LPS15. Antagonist LPS have been isolated from Rhodobacter sphaeroides, Porphyromonas gingivalis, and a mutant E.coli7.
Williams has suggested that protein binding may play a more significant role in LER than the disaggregation of endotoxin36. Endotoxins are known to interact with proteins, particularly with positively charged proteins17, 18. Petsch et al. were able to recover and quantify endotoxin bound to cationic lysozyme by digestion with protease K18. However, the use of proteases to de-mask protein-bound endotoxin has not been pursued so far as an option to address the LER issues.

FDA findings in BLA Submissions
In light of public reports on the inability of the compendial methods to detect endotoxin in some biological drug products, the FDA began to request spiking studies during the review of new BLAs regulated by the Center for Drug Evaluation and Research (CDER). LER results were summarized at the PDA 9th Annual Global Conference on Pharmaceutical Microbiology in 201412. LER continues to be associated with certain biological drug products, either monoclonal antibodies or therapeutic proteins. Results from spiking studies with either Reference Standard Endotoxin (RSE) or Control Standard Endotoxin (CSE) have indicated that endotoxin recovery is affected by the formulation excipients, protein, test method (Kinetic Chromogenic Assay [KCA], Kinetic Turbidimetric Assay [KTA] or Gel Clot Assay [GCA]), reagents from different vendors, hold temperature, and time. The greatest effects on spiked endotoxin recoverability have been observed with certain formulation excipients and proteins; in addition, the type of spike material (RSE, CSE, or Naturally Occurring Endotoxin [NOE]) has also influenced the outcome of the spiking studies35. LER is time-dependent; it can occur over the course of one week or very rapidly (within 4 hours) in the presence of citrate or phosphate plus polysorbate. More rapid LER effects have been reported when holding at 20-25˚C than at 2-8˚C. Data from BLA submissions is consistent with recently published reports1-3, 5, 6, 18, 19, 24-26, 33-36.
Studies reported in multiple BLA submissions indicate that correlation between the LAL tests and RPT is not always predictable. As shown in Table 1, various outcomes have been observed with CSE or NOE spiked into undiluted biological products. The lack of correlation between the LAL and the RPT results indicates that spiked endotoxin can be masked in the LAL test but remains potent in vivo; the safety concern relates to products that exhibit LER in the LAL test but are pyrogenic in the RPT.


BLA data indicates that detectability of spiked endotoxin from formulations containing polysorbate and either citrate or phosphate can disappear rapidly (within 4 hours) or occur over time (3-7 days). The majority of formulations with L-histidine and polysorbate are not affected by LER. However, certain proteins formulated with L-histidine and polysorbate do exhibit LER over time. In one study (Table 2), a monoclonal antibody formulated with L-histidine and polysorbate was spiked with 13 EU CSE per mL and held at 2-8˚C for seven days. Unacceptable recovery was observed after five days. Based on the pI of the protein and the pH specification for the product, the protein is expected to have a net positive charge in the formulated product. The ability of endotoxin to bind to basic proteins and become undetectable in LAL assays has been reported in the literature26, 35, 36. Therefore, endotoxin binding to the protein is one possible mechanism for LER in this case. The sponsor implemented the RPT as a release test for the product.


In another BLA submission (Table 3), the sponsor challenged a monoclonal antibody product formulated with citrate and polysorbate with 5 EU of CSE per mL. LER was observed when the KCA method was used, but spike recovery was acceptable when the GCA method was used. The GCA method was ultimately approved to release finished drug product.


Endotoxin recovery data from five additional BLA submissions also indicate that different LAL methods may react differently to spiked endotoxin standards. Table 4 presents a spectrum of LER results from challenge studies.


Product A exhibited LER with the three LAL methods when spiked with either CSE or NOE, while Product B exhibited LER with the three LAL methods when spiked with CSE but not when spiked with NOE. Therefore, the theory that NOEs are less prone to LER effects is supported by results from Product B but not from Product A. Other products exhibited LER with only two of the LAL methods; for example, similar to the case presented in Table 3, Product C exhibited LER with the KCA and KTA methods but not with the GCA method. Product D exhibited no LER with the GCA method. Product E exhibited no LER with the KTA method when spiked with CSE, but LER was observed when the challenge was RSE.
The use of NOE has been promoted by some sponsors because they appear to be less prone to LER than RSE or CSE4; however, NOE spiking studies have shown inconsistent results between the LAL detection method and the RPT. For example, in a study submitted by a sponsor, a therapeutic protein in a citrate and polysorbate formulation spiked with lipopolysaccharide (LPS) exhibited unacceptable endotoxin recovery after one day (Table 5).


No pyrogenic response was observed in rabbits injected with the product spiked with 10-13 EU/kg LPS (Table 6), which was consistent with the LAL test results. In the same study, the sponsor spiked the product with NOE prepared from the same E. coli strain as the purified LPS, and LER was not observed after 23 days (Table 5). However, the product spiked with 28-31 EU/kg NOE did not trigger a pyrogenic response in rabbits (Table 6), which was not consistent with the LAL test results. The positive control samples (LAL reagent water spiked with LPS or NOE) showed acceptable endotoxin recovery and were pyrogenic in rabbits (Tables 5 and 6).


Other factors to consider in the NOE studies is the variability of endotoxin potency of Gram-negative microorganisms (from < 50 EU/mL to > 58,000 EU/mL among various cultured bacteria), the effect of the LPS moiety structure8, 36, and the potential for agonistic and antagonistic effects from different lipid A structures derived from different LPS sources30. LER or masking can occur with either NOE or CSE24. It is not likely that studies with NOE will clarify and resolve these issues.

USP LAL compendial methods require the use of diluted samples in suitability tests, but LER or masking effects are not overcome by dilution. Therefore, additional validation studies should be conducted with undiluted protein products. Endotoxin recovery is influenced by protein concentration, overall protein charge, formulation constituents, and test methods. BLA sponsors should be cognizant of the limitations of LAL test methods for assessing the quality of drug substance and drug product. Different compendial LAL endotoxin detection methods (e.g., GCA, KCA, and KTA) may yield different results in LER studies. The compendial RPT should be used to release product when the LAL test methods are unable to detect spiked endotoxin. This is generally regarded an interim measure until a more suitable in vitro assay can be developed. Extensive historical data indicates that the RPT and the LAL correlate well, although the LAL test is regarded as 3 to 300 times more sensitive than the RPT32. However, this correlation is not as strong with certain protein products with a positively charged protein and/or formulated with polysorbate and citrate/phosphate excipients.
To mitigate the impact of masking or LER effects, sponsors should implement comprehensive microbial control strategies for drug substance and drug product manufacturing. The strategies should consider microbial ingress from raw materials, equipment, personnel, and the facility. The control strategy may include (but is not limited to) the use of qualified equipment, use of bioburden reduction filtration at key steps, limiting open operations, and use of closed systems for sampling. Manufacturing processes should be monitored using appropriate in-process tests to ensure that bioburden and endotoxin are controlled throughout the process. Finally, microbial release specifications should be linked to relevant quality attributes and endotoxin detection should be conducted using a valid method (21 CFR 211.167(a), 610.13(b), 610.9).
Thus far, a path forward for approval of BLA products regulated by CDER has been found in all cases in which masking or LER has been observed. Approval may involve a risk assessment for microbial control, additional process controls, implementation of the RPT for drug product release testing, and the use or development of alternative test methods, as appropriate. Additional controls for a product formulated with polysorbate may include setting an endotoxin specification (reported on the Certificate of Analysis) for an in-process material tested immediately prior to polysorbate addition and setting an endotoxin limit for the polysorbate solution. Use of the RPT for drug product release testing is considered an interim measure until a more suitable in vitro method is developed and shown to be suitable for testing the product. To determine whether LER in vitro corresponds to non-pyrogenicity in vivo, sponsors are typically requested to conduct studies in which the product is spiked with an endotoxin standard, held, and then tested at various time points by LAL and RPT methods in parallel. If the product exhibiting LER is pyrogenic in rabbits or if the study results can only be reviewed post-marketing, the BLA may be approved with RPT as a drug product release test. Regardless of whether or not the RPT is implemented for release testing, the sponsor will be asked to conduct post-marketing studies to address the LER issue and to elucidate the mechanism of LER in the product. Based on the results of these studies, the sponsor should investigate and/or develop alternative test methods.

1.Bolden, J.S. 2013. An Approach for Successful Endotoxin Challenge Study Design to Mitigate Low Endotoxin Recovery Concerns. PDA 8th Global Conference on Pharmaceutical Microbiology, Bethesda Maryland
2.Bolden, J.S. 2014. Overcoming Interference in the Bacterial Endotoxins Test: Through the Looking Glass. Bacterial Endotoxin Summit, Philadelphia, PA.
3.Bolden, J.S. et al. 2014. Evidence against a Bacterial Endotoxin Masking Effect in Biologic Drug Products by Limulus Amebocyte Lysate Detection. J. Parent Sci Tech 68(5): 472-477.
4.Bowers, K. and Johnson, L. 2013. Creation of an In-house Naturally Occurring Endotoxin Preparation for Use in Endotoxin Spiking Studies and LAL Sample Hold Time Analysis. American Pharmaceutical Review, 14 (Endotoxin Suppl),
5.Chen, J. 2014. Risk Management on Common Biologics Affected by Low Endotoxin Recovery. Bacterial Endotoxin Summit, Philadelphia, PA.
6.Chen, J. and Vinther, A. 2013. Low Endotoxin Recovery in Common Biologics Products. PDA Annual Meeting, Orlando Florida.
7.Coats, S.R. et al. 2005. MD-2 Mediates the Ability of Tetra-acylated and Penta-cylated Liposaccharide to Antagonize Escherichia coli Lipopolysaccharide at the TLR4 Signaling complexes. J. Immunol. 175(7): 4490-9.
8.Dubczak, J. 2014. LER: A Comparative LAL/MAT/Ligand Assessment. Presented at the Pharmaceutical Microbiology Forum Bacterial Endotoxins Summit Meeting, Philadelphia, PA.
9.FDA Guidance for Industry. Pyrogen and Endotoxin Testing: Questions and Answers. June 2012.
10.Galanos C. and Luderritz O. 1984. Lipopolysaccharide: Properties of an Amphipathic Molecule. In: Rietschel ET, ed. Handbook of Endotoxin. New York: Elsevier.
11.Hannecart-Pokorni, E. et al. 1973. Macromolecular Structure of Liposaccharide from Gram-negative Bacteria. Eur. J. Biochem. 38: 6-13.
12.Hughes, P. 2014. Endotoxin Challenges: A Regulatory Perspective. PDA 9th Annual Conference on Pharmaceutical Microbiology.
13.Hughes, P. 2015. FDA Regulations and Endotoxin Detection. PDA Europe Conference Pharmaceutical Microbiology.
14.Keene, W.R. et al. 1961. Observation on the Pyrogenic Response and Its Application to the Bioassay of Endotoxin. J. Clin. Invest. 40(2):295-301.
15.Martirosyan, A. et al. 2013. Lipopolysaccharides with Acylation Defects potentiate TLR4 Signaling and Shape T Cell Responses. PloS One 8(2): e55117.
16.Mueller, M. et al. 2004. Aggregates are the Biologically Active Units of Endotoxin. J. Biol. Chem. 279: 26307-26313.
17.Petsch, D., Deckwer, F.B. Anspach, F.B. 1998. Proteinase K Digestion Improves Detection of Bacterial Endotoxins by the Limulus Amebocyte Lysate Assay: Application for Endotoxin Removal from Cationic Proteins. Analytical Biochemistry, 259: 42-47.
18.Petsch, D. Anspach, F.B. 2000. Endotoxin Removal from Protein Solutions. J. Biotechnol. 76: 97-119.
19.Platco, C. 2014. Lab Experiences: Low Endotoxin Recovery. Bacterial Endotoxin Summit, Philadelphia, PA
20.Platco, C. 2014. Low Liposaccharide Recovery versus Low Endotoxin Recovery in Common Biological Product Matrices. American Pharmaceutical Review, Endotoxin Detection Supplement Part II.
21.Raetz, C.R. et al. 2007. Lipid A Modification Systems in Gram-Negative Bacteria. Annual Review of Biochemistry 76:295-329.
22.Raetz, C.R. et al. 2009 Discovery of a New Biosynthetic Pathway: the Lipid A Story. J. of Lipid Research: S103-S108.
23.Ribi, E. et al., 1966. Reaction of Endotoxin and Surfactants. J. Bacteriol. 92:1493-1509.
24.Reich, J. and Grallert, H. 2014. Low Endotoxin Recovery in Bio-Pharmaceuticals: Comparison of Naturally Occurring Endotoxins (NOE) and Commercial Standards. Poster Presentation at the PDA Annual Meeting, San Antonio, TX.
25.Reich, J. 2013. An Advanced Endotoxin Assay: Insights and Strategies for Overcoming Low Endotoxin Recovery in Complex Formulations. Presented at the Pharmaceutical Microbiology Forum Bacterial Endotoxin Summit Meeting, Philadelphia, PA.
26.Reich, J. 2014. Reliability of Endotoxin-Detection: Mechanistical Principles of Endotoxin-Masking and Strategies for De-Masking. Presentation at the PDA Conference in Berlin, Germany.
27.Sasaki, H. and White, S.H. 2008. Aggregation Behavior of an Ultra-Pure Liposaccharide that Stimulates TLR-4 Receptors. Biophys. J. 95: 986-993.
28.Schromm, A.B., et al. 1998. The Charge of Endotoxin Molecules Influences Their Conformation and IL-6 Inducing Capacity. J. Immunol.161: 5464-5471.
29.Schromm A.B., et al. 2000. Biological Activities of Lipopolysaccharides are Determined by the Shape of their Lipid A Portion. Eur. J. Biochem. 267: 2008-2013.
30.Teghanemt, A. et al. 2005. Molecular Basis of Reduced Potency of Underacylated Endotoxins. J. Immunol. 175(7):4669-76.
31.Tsuchiya, M. 2014. Possible Mechanism of Low Endotoxin Recovery. American Pharmaceutical Review, Volume 17, Issue 7.
32.Wachtel, R. Tsuji, K. 1977. Comparison of Limulus Amebocyte Lysates and Correlation with the United States Pharmacopeial Pyrogen Test. Appl Environ. Microbiology 33:1265-1269.
33.Williams, K.L. 2013. Low Endotoxin Recovery (LER) in Drug Products. PDA Letter, Volume XLIX, Issue 7.
34.Williams, K.L. and Chen, J. 2013. Follow-up on Low Endotoxin Recovery in Biologics. PDA Letter, Volume XLIX, Issue 9.
35.Williams, K.L. 2013. Endotoxin Test Concerns in Biologics. American Pharmaceutical Review 16 (Endotoxin Detection Suppl.): 4-9.
36.Williams, K.L. 2014. Endotoxin Aggregation & Binding Properties: Recovering Endotoxin Spikes from Product and Containers-closures. Presentation at the PDA Conference in Berlin, Germany.

Author Biographies

Patricia F. Hughes, Ph.D., has been with the FDA for more than 16 years and is a Team Leader in FDA/CDER in the Office of Pharmaceutical Quality, Division of Microbiology Assessment where she is responsible for the assessment of BLA submissions from a product quality microbiology perspective. She earned her Ph.D. in Microbiology from Georgetown University in Washington, D.C.

Colleen Thomas, Ph.D., has been with the FDA for more than 5 years. She is a reviewer in the FDA/CDER Office of Pharmaceutical Quality, Division of Microbiology Assessment. She is responsible for the assessment of BLA submissions from a product quality microbiology perspective. She earned her Ph.D. in Microbiology from Texas A&M University in College Station, TX and performed post-doctoral research at the USDA-ARS Southeast Poultry Research Laboratory.

Kalavati Suvarna, Ph.D. is a Senior Clinical Microbiology Reviewer in the Office of Antimicrobial Products within the FDA/CDER’s Office of New Drugs. She received a B.S. in Microbiology in 1985 and M.S. in BioPhysics in 1989 from the University of Bombay, and a Ph.D. in Biological Sciences from Northern Illinois University in 1994. Previously, she was a Consumer Safety Officer (Senior Policy Advisor in CDER’s Office of Compliance, Office of Manufacturing and Product Quality, Biotech Manufacturing Assessment Branch.

Bo Chi, Ph.D., is currently in FDA/CDER, Office of Pharmaceutical Quality, Division of Microbiology Assessment. Bo has been with the FDA for more than 12 years. She earned her Ph.D. degree in Microbiology from State University New York at Buffalo.

Reyes Candau-Chacon, Ph.D. has been at the FDA for over four years as a Microbial Quality reviewer of Biotech products. She holds a degree in Biology by the University of Sevilla, Spain and conducted research on transcriptional regulation during her postdoc in Marburg, Germany and at the University of Pennsylvania. She held positions as Senior Scientist and Director of Research and Development at Edge BioSystems and Transgenomic, two Biotech companies focused in Molecular Biology Reagents and Mutation Detection.

Candace Gomez-Broughton, Ph.D. is a Microbiologist in the Office of Pharmaceutical Quality of the FDA. Candace earned her Ph.D. in Microbiology and Immunology from the University of Illinois, Chicago where she focused her studies on the Human Immunodeficiency Virus assembly process. She earned her B.S. Degree in Biology from Spelman College in Atlanta, Georgia. Candace received her post-doctoral training at the University of North Carolina, Chapel Hill Lineberger Cancer Center.

Lakshmi Rani Narasimhan is a Microbiologist in the FDA/CDER Office of Pharmaceutical Quality, Division of Microbiology Assessment. She holds a Ph.D. in Industrial Microbiology from the University of Madras, India and pursued post-doctoral research at the Center for Biotechnology, University of Nebraska, Lincoln and at the Dept. of Biochemistry, Michigan State University. Prior to joining the Agency, she worked as a scientist in pharmaceutical companies and microbiological testing laboratories.

Subscribe to our FREE newsletter