Endotoxin Test Concerns of Biologics: The Role of Endotoxin as a Quality Indicator in Biologic Manufacturing Processes

July 1, 2015
Kevin Williams

BioPharm International, BioPharm International-07-01-2015, Volume 28, Issue 7
Page Number: 28–33

Low endotoxin recovery represents an opportunity to add value to the characterization of biologic drug products.


AbstractLow endotoxin recovery represents an opportunity to add value to the characterization of biologic drug products.

"Lipid A is … an information-rich molecule, with many possibilities for specific recognition by prokaryotic and eukaryotic proteins,” says Christian Raetz (1). Recent Parenteral Drug Association (PDA) conferences in Bethesda (October 2014) and Berlin (February 2015) highlight the ongoing disagreement among drug manufacturers, Limulus amebocyte lysate (LAL) test manufacturers, and regulators as to the significance of the low endotoxin recovery (LER) phenomenon. The controversy on endotoxin detection can be summarized by two different views or philosophies of microbiological contaminant control (Figure 1). The first philosophy would maintain that only “biologically active” components of bacteria are relevant to microbiological control. The second would seek to control microbiological process ingress by monitoring important markers or quality indicators such as endotoxin. Here, endotoxin and lipopolysaccharide (LPS) are used interchangeably. The latter philosophy does not need to answer the question, “Is the microbiological artifact biologically active?” to use it as a gauge as to the purity or contamination status of a given aseptic biologics process. In fact, the demonstration of “biological activity” is not as clear-cut as some would currently maintain.

Consider that the GMP context for product contamination is the concept of “adulteration,” which does not distinguish biologically active contaminants from those that are not biologically active. The mere proximity of “filth” is the traditional criteria for adulteration. The “old time” verbiage of the Code of Federal Regulations (CFR) is, “(1) if it consists in whole or in part of any filthy, putrid, or decomposed substance; or (2)(A) if it has been prepared, packed, or held under insanitary conditions whereby it may have been contaminated with filth,” and the GMP concept of “adulteration” via contamination portray product/process contamination in the context of probability, proximity, and severity of its occurrence (2). A valid end-product United States Pharmacopeia (USP) Bacterial Endotoxin Test (BET) <85> test of a finished drug or active pharmaceutical ingredient is a legal requirement.

Endotoxin in its many forms
Some have sought to solve the LER conundrum by substituting naturally occurring endotoxins (NOEs) that are not purified standards and that often show better recovery from spiked biologics subject to LER. However, when viewed by the second philosophy, that of choosing a quality indicator, it can be seen that only switching out the positive control (reference standard endotoxin [RSE] or control standard endotoxin [CSE] for NOE) does not provide any additional characterization of the drug process itself. A NOE spike may recover better than a standard CSE or RSE spike, but the recovered positive control does not represent the endotoxin content of the non-spiked sample. Rather, it assumes that the non-spiked sample contains monomers that are not biologically active with the LAL reagent and thus, concludes that the sample is not contaminated. The basic theory of LER is that LPS disaggregation to monomers occurs via the chelation of ions needed by LPS to maintain its aggregate form in solution by buffer (citrate or phosphate). The disaggregation is followed by the subsequent coating of LPS monomers with abundant polysorbate molecules, thereby forming a non-LAL-reactive or masked endotoxin-drug solution that does not allow LPS to react with LAL.

The assumption that resides in the first philosophy is that “monomers don’t matter,” because they are not active with LAL (although the monomers are sometimes active and sometimes inactive with the rabbit pyrogen test)(3) and thus, are of no concern to patients. However, the effect of monomers in the mammalian body is still unknown. What is known is that the monomer is the active endotoxic principle for all endotoxins (RSE, CSE, NOE; aggregated and disaggregated). This fact, that the sub-monomer Lipid A is the active principle of endotoxin response, has been established beyond doubt by various studies of the toll-like receptor 4 (TLR4) and associated co-receptor myeloid differentiation protein (MD-2), which show how the monomer fits into the hydrophobic pocket of MD-2 and how MD-2 with LPS fits into the TLR4 dimer to bring about the transmembrane signaling event that instructs the cell nucleus to produce cytokines (4).

MD-2, as the co-receptor of TLR4, holds the prototypically configured endotoxic Lipid A (hexa-acylated) in a hydrophobic pocket. Five of the six acyl group fingers rest inside the MD-2 hydrophobic pocket (glove) and the sixth finger (in the prototypical agonistic Escherichia coli LPS) must stick out to attach to a second TLR4 to help form the dimer. There are 10 functional TLRs in humans (TLR11, 12, and 13 have been lost from the human genome) (5), which are mixed and matched in like or different dimers (homodimers and heterodimers, respectively) to detect dozens of microbial artifacts. The TLR structure is based upon the leucine-rich repeat (LRR)-type sequence of alternating loops of hydrophobic and basic amino acids. Lipid A antagonists can prevent the activation of TLR4 by displacing the active monomer in the dimer structure and thus, preventing fulfillment of the conditions for activation. It is the fine structural detail of the Lipid A molecule that determines the endotoxin response, with a wide degree of variants demonstrating a wide spectrum of mammalian host responses (from agonistic to antagonistic).

The first TLR was discovered by the knockout of a receptor in Drosophila that subsequently allowed the fly to be invaded and overgrown by fungal hyphae (6), thereby beginning the quest for various additional TLRs and the various associated microbial artifacts that activate them. It should be noted that monomer structure of Lipid IVa (a precursor to Lipid A in bacterial Lipid A metabolism) and Eritoran are antagonists and do not fit into MD-2 as well as the E. coli LPS, and thus, cause blockage of the TLR4 receptor and its activity. The drug candidate Eritoran was hoped to be used to treat sepsis, but the drug failed in Phase III clinical trials. This shows that while the knowledge of what happens to endotoxin within the body is advanced, it is not yet complete. Also, the blocking of the TLR4 pathway does not negate the potential for the activation of the complement-coagulation system. 




The BLA and the unique role of LPS as a quality indicator
The biological license application (BLA) submission has become the focal point for the demonstration of time-dependent LPS spike recovery as a quality indicator in biologic products and constituents subject to LER. According to FDA, “Sponsors of BLA submissions have reported unacceptable time-dependent recovery of endotoxin spiked into undiluted drug product using the LAL USP <85> methods for endotoxin” (7). Manufacturers that submit BLAs are now being required to perform follow-up studies to demonstrate recovery of LER-prone samples and to perform rabbit pyrogen testing in lieu of LAL testing for such samples on an interim basis, until such a time as an alternative detection protocol may be developed.

In its simplest form, a quality indicator such as LPS provides information on a manufacturing process. It answers critical questions, such as:

  • Was a microbial contaminant present?

  • Where did it come from?

  • Which type of microbial contaminant was it?

  • How much of it was present?

  • How might its presence (and the quantity present) have affected product quality?

Therefore, as a quality indicator of a manufacturing process, it is not always the biological activity of the indicator itself that tells the story, but rather, the mere presence of the artifact. For example, the presence of coliforms indicates poor sanitary conditions (i.e., from waste water) in certain food processing environments or in nonsterile drug manufacturing. Coliforms are defined as Gram-negative rods that ferment lactose and produce acid and gas at 35 °C, while fecal coliforms are further distinguished by their ability to ferment lactose at higher temperatures (44.5–45.5 °C). Whether a specific coliform is a disease- or illness-causing organism does not negate its use as a quality indicator. Likewise, endotoxin found in a manufacturing process (in various forms) should still tell a story specific to its own historical occurrence in the process. It will be difficult to detect contaminants of a process when the analyte (LPS) of such testing is masked. Losing the visibility of what has occurred in a specific drug manufacturing process means losing the ability to assure characterization of the product itself. The ability to gauge LPS content via LAL testing, therefore, is diminished. Again, an indicator is significant not only for what it is (i.e., its own biological activity), but for what it means (i.e., microbial ingress has occurred and thus biological activity at the time of occurrence could have adversely affected the product). This nuanced distinction doesn’t seem widely appreciated. If monomers can be re-aggregated and detected, then the meaning of such a measurement can serve its historical purpose, namely, that of indicating the occurrence of Gram-negative bacterial ingress.

Therefore, LPS, even as a monomer, is significant both in its mere presence and, more importantly, as a microbial artifact, given its historical complexity-eons of history in prokaryotic and metazoan interactions, including shaping the genome of man. Nobel Prize-winner Bruce Beutler, who discovered TLR4, said, “Particularly because they strike down so many people before or during reproductive age, microbes constitute the strongest selective pressure with which our species must contend, and we may assume that microbes have shaped the human genome more than any other selective pressure in recent times” (8). The LPS monomer is not a depyrogenated or incinerated fragment of LPS, but rather, the entire bacterial pathogen-associated molecular pattern (PAMP, now largely called a MAMP or microbe-associated molecular pattern) that has served for an estimated one billion years (beginning in plants) as a target of metazoan immune systems in terms of recognition and response to prokaryotic invasion. If LPS were to be lost as a quality indicator during drug processing, it is hard to envision what might take its place. A risk-assessment philosophy (in terms of process testing points using alternate methods) of some sort would have to be used to justify the improbability of microbial ingress during processing of LPS-masked solutions; this represents a less-than-desirable situation. What is undesirable here is that contamination is a time-linked occurrence, and by its very nature, is unpredictable. At present, there is no good substitute for endotoxin detection.

Biologic molecules and manufacturing processes
Some have argued that the presence of monomers causes no harm to the patient. While this has not been scientifically proven, the harm envisioned from LER is not necessarily directly derived from drug solutions containing monomers of LPS, but rather, from poorly or under-characterized biologics as end-products. Biologics are typically large and complex molecules. The molecular weight of a monoclonal antibody is approximately 150,000 Daltons; compare that with an early therapeutic protein, such as r-human insulin, at approximately 5800 Daltons. Therefore, the characterization of the process that produces these biologics is paramount. A review of life-threatening adverse events that have occurred from first-in-human (FIH) doses of modern biologics (mAbs and therapeutic proteins) shows some tragic instances that arose from less-than-perfectly characterized drug products (Table I). In other words, some slight change in the manufacturing process produced a product that was slightly different at the molecular level. These molecules appeared to be characterized at the time, but had somehow, during the manufacturing process, accrued slight changes in fine structure (e.g., 3D structure, glycosylation, charge, hydrophobicity, folding, heterogeneity, bioactivity, truncation, oxidation, deamination, or aggregation) (9).

The clinical manifestation of endotoxin monomers vs. aggregates cannot be resolved
The two questions presented as a duality of competing hypotheses from Figure 1: “Monomers don’t matter” versus “Monomers matter”, in terms of biological activity when presented in drug products, may seem critical to the ongoing LER debate, but this issue cannot be easily solved. For every determination that “only” aggregates are biologically active, there are two other studies indicating that under specific physiological conditions, monomers are indeed the active physical unit of endotoxin (12, 13, 14). Levin showed one such example where, in the presence of hemoglobin, LPS disaggregation produced increasing LAL gelation times (15, 16). Aside from hemoglobin, “there are more than 200 acute phase proteins (APP) in mammals responding to endotoxin” (17). Short of producing and giving endotoxin monomers to compromised patients, such as those with sepsis, disseminated intravascular coagulation (DIC), or via the intrathecal route-something which would never be done-the harm or lack thereof of monomers cannot be established with any certainty. It is worth highlighting that there are challenges in determining “biological activity” given that the endotoxin response is partitioned in the body amongst different tissue types-for example, the gut tolerates a huge load of endotoxin-bearing Gram-negative bacteria, whereas the spinal column is very sensitive to extremely minute levels of endotoxin. Moreover, the body’s response may vary with different disease states such as inflammation, infection or DIC, for example.




Additionally, some small molecules that mimic Lipid A but are of different (sometimes widely unrelated) structures have been found to ameliorate the host response to LPS, hence making it difficult to rationalize that only the aggregate is active during the initial recognition of LPS. If only the aggregate were the active configuration of LPS, one wonders how small molecules could find their way to TLR4/MD-2 to mimic the LPS behavior in a disruptive manner. The corresponding aggregate behavior of small molecules seems unlikely to mimic that of LPS. Neal et al. used a small molecule (MW of ~390 versus ~1700 for Lipid A) to inhibit the host reaction to LPS both in vitro and in vivo (18). Slivka et al. used a 17-residue peptide (MD2-I), synthesized to reproduce the TLR4-binding region of the MD2 protein that contains all the critical interacting residues, and showed evidence that this sequence targets TLR4 directly as an antagonist (19). At any rate, the biological activity question cannot yet be answered definitively and does not affect the use of LPS as a biologics manufacturing process quality indicator.

LPS is viewed as a fever-causing substance first and foremost; however, as seen by F. Bang in 1955, it was the gelation of the Limulus blood-the coagulation dysfunction caused by endotoxin-that prompted the development of LAL as a detection method (gel clot). Therefore, careful consideration should be given to blood dysfunction possibilities that may come through the complement-activation pathway via monomeric endotoxin, as well as singularly focusing on the fever-causing activation of TLR4 (20). It is the TLR4 receptor-activated route that brings about the production of fever-causing cytokines and is the route that many presume requires an aggregated endotoxin presentation to activate. While rabbit pyrogen test can detect fever, it cannot detect blood-clotting dysfunction, nor can it be relied upon to consistently predict the occurrence of fever from masked or disassociated endotoxin, which may process differently and along a different timeline compared with traditional drug injections into rabbits.

Drug regulators, along with each therapeutic protein manufacturer, will have to determine the safeguards that will constitute cGMP activities for LER-prone solutions, be it increased process controls (increased microbial monitoring, hazard analysis, or validation of process hold-times) and/or a form of pretreatment for BET samples to re-aggregate disassociated endotoxin. It should at the very least be agreed that philosophically viewpoint 2 of Figure 1 would be the more appropriate paradigm for drug manufacturing control of contaminants.

Regulators, drug manufacturers, and LAL providers are still adjusting to LAL testing subsequent to the knowledge of the occurrence of LER. The BLA is the choke point for efforts to commercialize a biologic drug relative to LER, as BLA approval requires the demonstration of valid hold-time endotoxin spike studies that may demonstrate masked recovery due to LER.

LER is most relevant when viewed from a drug product characterization perspective. Important questions from a risk-analysis perspective regarding LER include:

  • What is the time-course profile of the process constituents in regard to microbial ingress? Does it lend itself to transient contamination events from a microbial control strategy vantage (21)?

  • Could the product be contaminated with endotoxin and could this contamination be subsequently masked by an LER or LER-like effect?

  • How could a contamination event of live Gram-negative bacteria, as evidenced by the detection of LPS, affect a drug’s properties as they relate to patient safety?

Each specific drug manufacturing process flow puts LER in context relative to risk management given the formulation constituents, hold times, temperatures, and container types. The “no patients are being harmed” argument associated with philosophy 1 seems premature. Even if this assumption turns out to be valid, there does not appear to be a valid substitute or replacement for a longstanding quality indicator such as LPS.

The risk to patients from poorly characterized biologics is not negligible. It is well known that the presence of both Gram-negative bacteria and endotoxin can induce changes to drug molecules during expression and purification processes. The new LER paradigm is reminiscent of the early mycoplasma debate, wherein the effect of mycoplasma on cell cultures used to produce drug products was up for debate (22, 23). Subsequently, as better methods of detection were developed, product characterization improved, and today mycoplasma testing is viewed as a critical routine quality indicator. Analogously, new sample treatments are needed to reveal the presence of endotoxin regardless of the aggregation state. LER represents an opportunity to add value to the characterization of biologic drug products.


Article submitted: Apr. 22, 2015.
Article accepted: May 12,  2015.

1. C. Raetz, J. Bacteriol. 175 (18), pp. 5745–5753 (Sept. 1993).
2. CFR Title 21, Part 351, (Government Printing Office, Washington, DC), pp. 114.
3. P. Hughes, “FDA Regulations and Endotoxin Detection,” presentation at the PDA Conference (Bethesda, MD, Oct. 21, 2014).
4. T. Scior et al, Comput. Struct. Biotechnol. J. 7 (9), pp. 1–11 (May 2013).
5. T. Kawai and S. Akira, Nature Immunol. 11 (5), pp. 373–384 (2010).
6. B. Lemaitre et al., Cell 86 (6), pp. 973–983 (September 20, 1996).
7. P. F. Hughes et al., “Low Endotoxin Recovery: An FDA Perspective,” BioPharm Asia, accessed June 12, 2015.
8. B. Beutler, “How Mammals Sense Infection: From Endotoxin to the Toll-like Receptors,” Nobel lecture presentation at the Center for Genetics of Host Defense, (The University of Texas Southwestern Medical Center, Dallas, TX, Dec. 7, 2011).
9. E. Narke, “Journey in the Development of Biologics Through End of Phase III,” slide presentation for raps.org, accessed June 12, 2015.
10. E. Tranter et al., Br. J. Clin. Pharmacol. 76 (2), pp. 164–172.
11. G. Suntharalingam, et al., N. Engl. J. Med. 355, pp. 1018–1028 (September 7, 2006).
12. M. Mueller et al., J. of Bio. Chem. 297, pp. 26307–26313 (2004).
13. H. Sasaki and S.H. White, Biophys. J. 95 (2), pp. 986–993 (2008).
14. K. Takayama et al., J. of Bio. Chem. 269, pp. 2241–2244 (1994).
15. R.J. Roth and J. Levin, “Effects of Human Hemoglobin on Bacterial Endotoxin In Vitro and In Vivo,” in Endotoxin in Health and Disease, H. Brade et al., Eds., (Marcel Dekker, Inc., New York, NY, 1999), pp. 389–402.
16. J. Levin, “The Original Description of the Limulus Amebocyte Lysate (LAL) Test: From Past to Present,” presentation at the PDA 9th Annual Global Conference on Pharmaceutical Microbiology (Bethesda, MD, 2014).
17. S. Chemonges, J.P. Tung, and J.F. Fraser, Proteome Sci. 12 (1), pp. 12 (2014).
18. M.D. Neal et al., PLoS ONE, 8 (6), pp. e65779 (June 2013).
19. P.F. Slivka, et al., Chembiochem. 10 (4), pp. 645–649 (Mar. 2, 2009).
20. P. Libby and D.I. Simon, Circulation 103 (13), pp. 1718–1720 (Apr. 3, 2001).
21. K. Suvarna et al., American Pharm. Rev. 14 (1) (January/February 2011), accessed June 12, 2015.
22. D.K.F Chandler, D.V. Volokhov, and V.E. Chizhikov, American Pharm. Rev., 14 (4) (May/June 2011), accessed June 12, 2015.
23. R.J. Geraghty et al., Br. J. Cancer 111, pp. 1021–1046 (2014).

About the Author
Kevin L. Williams is senior scientist of R&D at Lonza.

Article DetailsBioPharm International
Vol. 28, No. 7
Pages: 28–33

When referring to this article, please cite it as K.L. Williams, "Endotoxin Test Concerns of Biologics: The Role of Endotoxin as a Quality: Indicator in Biologic Manufacturing Processes," BioPharm International, 28 (6) 2015.