TR82 Low Endotoxin Recovery

Licensed to Goyal, Prateek/Lupin Ltd: Copying and Distribution Prohibited.
Technical Report No. 82

Low Endotoxin Recovery
PDA Low Endotoxin Recovery Technical Report Team
Dayue Chen, PhD, Eli Lilly and Company, Co-leader

Friedrich von Wintzingerode, PhD, Genentech, A Member of the Roche Group, Co-leader
Julie Barlasov-Brown, Merck & Co.
Lindsey Brown, PhD, U.S. Food and Drug Administration
Allen Burgenson, Lonza Group Ltd.
Joseph Chen, PhD, Ultragenix Pharmaceutical, Inc.
Monica Commerford, PhD, U.S. Food and Drug Administration
Gregory Devulder, PhD, bioMerieux, Inc.
Jennifer Farrington, PhD, Associates of Cape Cod, Inc.
Jessica Hankins, PhD, U.S. Food and Drug Administration
Patricia Hughes, PhD, U.S. Food and Drug Administration
Stefan Ishak, Novartis
Chris Knutsen, PhD, Bristol-Myers Squibb Inc.
Jack Levin, MD, University of California, San Francisco, School of Medicine
Jeanne Mateffy, Amgen Inc.
Ned Mozier, PhD, Pfizer Inc.
Scott Nichols, PhD, U.S. Food and Drug Administration
Cheryl Platco, Merck & Co. (retired)
Johannes Reich, PhD, Microcoat Biotechnologie GmbH
Stijn Seels, Sanofi
Anders Thorn, Novo Nordisk A/S
Masakazu Tsuchiya, PhD, Charles River Laboratories, Inc.
René Ørving, Biogen
Foreword
PDA is at its heart a science-based organization. Our activities and processes have been implemented
to promote the free exchange of ideas and promote inclusion in participation, regardless of the organi-
zation to which a member may belong. We also strive to achieve consensus among the subject matter
experts and to inform our members (and others) of the best practice (and in some cases, the regulatory
requirements). Our aim is always to make recommendations that advance patient safety.
PDA has also implemented a multistage review process for technical reports and position papers,
which includes not only the task force itself and the relevant advisory board, but also peer reviewers
who offer independent input on the subject matter. Finally, of course, the Board of Directors reviews
and approves all these documents. In every step, we document the resolution of all comments, and that
resolution is communicated to the next level of review. This process, although cumbersome, is how we
are working to ensure that we are exercising due diligence in positions adopted by the PDA.
Specifically, regarding this Technical Report, low endotoxin recovery (LER) has been discussed in open
fora many times over the last several years. The task force includes subject matter experts from several
organizations, as well as U.S. FDA. We increased the number of peer reviewers on this draft report to
30 individuals from 23 different industry, academic, and regulatory authority organizations.
The reliable and sensitive detection of bacterial endotoxin is a key test underpinning patient safety
in the manufacturing of parenteral drugs. The globally harmonized compendial tests linked to well-
characterized reference standards have provided this assurance to industry and regulators for decades
despite the complex nature of both the assays and the reference materials being. This complexity has
been demonstrated each time new assay approaches or reference materials have been proposed. Over
the past decade, biologics manufacturing has exploded, and new products, formulations, and delivery
systems continue to push the boundaries of modern manufacturing. In this context, it should not be
surprising that a phenomenon like LER arises to once again challenge our current understanding of
assay systems, reagents, and our own manufacturing process.
To that end, the PDA task force commissioned with this technical report went to the greatest lengths
possible to present as complete a picture of the current LER situation. This includes the historical and
mechanistic aspects of the endotoxin measurement challenges, as well as a standard protocol for devel-
oping product-specific hold studies, supported and informed by actual industry case studies. Both the
technical report team and the approving Biopharmaceutical Advisory Board understand that this is a
very complex and still evolving area of science that is not without controversy. Extensive peer review
comments were addressed, and the findings were widely presented in public fora before the comple-
tion of this report. PDA believes it is vitally important to make this information available to further the
scientific dialog and progress in this area and remains committed to revising and updating this material
as new discoveries and conclusions are made.
Richard M. Johnson
President & CEO, PDA
Low Endotoxin Recovery

ISBN: 978-1-945584-07-7
© 2019 Parenteral Drug Association, Inc.
All rights reserved.
Bethesda Towers
4350 East West Highway
Suite 200
Bethesda, MD 20814 USA
Tel: 1 (301) 656-5900
Fax: 1 (301) 986-0296
E-mail: [email protected]
Web site: www.pda.org
Table of Contents
1.0 Introduction��1 6.2 Microbial Control During Manufacturing and
1.1 Purpose�� 1 Product Quality�� 21
1.2 Scope�� 1 6.3 Endotoxin Structure�� 21
6.4 Limulus Amebocyte Lysate (LAL) Assay�� 22
2.0 Glossary of Terms��2 6.5 Endotoxin Reference Standards�� 24
6.5.1 U.S. FDA Development of Reference
2.1 Abbreviations�� 3 Standard Endotoxins�� 24
6.5.2 Harmonization of International
3.0 LER Hold-Time Studies��3 Reference Standard Endotoxins�� 26
3.1 Test Method�� 4 6.6 Natural Occurring Endotoxins (NOEs)�� 27
3.1.1 Reagents and Materials�� 4 6.7 Immune Response �� 27
3.1.2 Spiking of Undiluted Sample�� 4
3.1.3 Storage Containers�� 5 7.0 References��30
3.1.4 Batch Requirements�� 5
3.1.5 Storage Time and Temperature�� 5 8.0 Appendix: Case Studies of LER Occurrences��37
3.1.6 Controls�� 6
8.1 Case Study 1: Low Concentration PS20 Caused
3.1.7 Equipment�� 6 LER Reaction in the Presence of Monoclonal
3.1.8 Personnel�� 6 Antibody Product�� 37
3.1.9 Data Analysis and Interpretation�� 6 8.1.1 Introduction�� 37
8.1.2 Materials and Methods�� 37
4.0 Proposed Mechanisms of LER ��7 8.1.3 Results and Discussion�� 38
4.1 Differentiation of LER and Test Interference�� 8 8.1.4 Conclusion�� 39
4.2 Investigation of LER Causes�� 8 8.1.5 References�� 40
4.3 Proposed Two-Step Reaction Model of LER�� 9 8.2 Case Study 2: Mapping LER Effect in In-Process
4.4 Factors Influencing Reaction Kinetics�� 10 Stages of Purification�� 41
4.4.1 Energy Input (Extrinsic)�� 10 8.2.1 Introduction�� 41
4.4.2 Masking Capability of a Sample�� 11 8.2.2 Materials and Methods�� 42
4.4.3 Masking Susceptibility of Endotoxins�� 12 8.2.3 Results and Discussion�� 42
4.5 Potential Aggregation States of Detectable 8.2.4 Discussion�� 44
and Non-detectable Endotoxin�� 13 8.2.5 Key Lessons Learned�� 44
4.5.1 LER and Limited LPS Interaction with 8.2.6 Conclusion�� 44
Factor C�� 14 8.2.7 Recommendations�� 45
4.5.2 Activity of Monomeric and Aggregated LPS�� 14 8.3 Case Study 3: Use of Purified and Nonpurified
4.6 Summary �� 15 Endotoxins in Hold-time Studies�� 46
8.3.1 Introduction�� 46
5.0 Mitigation of LER��15 8.3.2 Materials and Methods�� 47
5.1 Mitigation through Sample Treatment�� 15 8.3.3 Results and Discussion�� 47
5.1.1 Evaluation of the Product�� 16 8.3.4 Molecular Basis of LER Susceptibility�� 52
5.1.2 Addition of Dispersants to Samples�� 16 8.3.5 Conclusion�� 54
5.1.3 Addition of Excess Divalent Cations�� 16 8.3.6 References�� 54
5.1.4 Other Sample Treatments�� 16 8.4 Case Study 4: Factors Affecting Low
5.1.5 Evaluation of LAL Assay Reagents�� 16 Endotoxin Recovery�� 55
5.1.6 Non-LAL Endotoxin Testing�� 17 8.4.1 Introduction�� 55
5.2 Evaluation through a Biological System�� 17 8.4.2 Materials and Methods�� 55
5.2.1 Endotoxin Dosage�� 17 8.4.3 Results and Discussion�� 57
5.2.2 Product Dosage�� 18 8.4.4 Effects of Temperature, pH, and Salt
5.2.3 Monocyte Activation Test �� 19 Concentration on LER�� 58
5.3 Interpretation of Results�� 20 8.4.5 Effects of Matrix Components on LER at
5.4 Summary�� 20 Relevant Concentrations�� 60
8.4.6 Effects of Dilution Methods on Recovery of
6.0 History and Rationale for LER Studies��20 RSE in LER Solutions�� 61
8.4.7 Conditions for Magnesium Dilution Method�� 62
6.1 Regulatory Framework�� 20
8.4.8 Conclusion�� 63
8.4.9 References�� 63 8.8.5 Conclusion�� 93

8.5 Case Study 5: Comparison of Different 8.8.6 References�� 93
Methods Used for Calculation of Percentile 8.9 Case Study 9: LER Case Study for a
Endotoxin Recovery from LPS-Spiked Monoclonal Antibody�� 94
Biological Therapeutic Products��65 8.9.1 Introduction�� 94
8.5.1 Introduction�� 65 8.9.2 Materials and Methods�� 94
8.5.2 Materials and Methods�� 65 8.9.3 Results and Discussion�� 97
8.5.3 Results�� 66 8.9.4 Conclusion�� 100
8.5.4 Discussion�� 68 8.9.5 References�� 102
8.5.5 Conclusion�� 70 8.10 Case Study 10: Endotoxin Recovery Studies
8.5.6 References�� 71 with a Biologic Formulated with Citrate and
8.6 Case Study 6: Lipopolysaccharide Masking and Polysorbate 20�� 103
Resolution of Low Endotoxin/Lipopolysaccharide 8.10.1 Introduction�� 103
Recovery (LER/LLR) in a Citrate Buffer Monoclonal 8.10.2 Materials�� 103
Antibody containing Polysorbate and Chelating 8.10.3 Methods�� 104
Agent�� 72 8.10.4 Results and Discussion�� 105
8.6.1 Introduction�� 72 8.10.5 Conclusion�� 108
8.6.2 Materials and Methods�� 72 8.10.6 References�� 109
8.6.3 Results and Discussion�� 74 8.11 Case Study 11: No LER Effect in a Sodium
8.6.4 Conclusion�� 78 Phosphate, Polysorbate 80 Formulation Matrix� 110
8.6.5 References�� 78 8.11.1 Introduction�� 110
8.7 Case Study 7: Evaluation of an Endotoxin 8.11.2 Materials and Methods�� 110
Demasking Protocol�� 79 8.11.3 Results and Discussion�� 111
8.7.1 Introduction�� 79 8.11.4 Conclusion�� 112
8.7.2 Materials �� 79 8.11.5 References�� 112
8.7.3 Methods�� 80 8.12 Case Study 12: Lack of LER Phenomenon in
8.7.4 Results and Discussion�� 80 A Therapeutic Protein Biologic with Known
8.7.5 Conclusion�� 84 LER-Causing Ingredients�� 113
8.7.6 References�� 84 8.12.1 Introduction�� 113
8.8 Case Study 8: Strategy to Investigate and 8.12.2 Materials�� 113
Overcome LER Driven by Protein ��86 8.12.3 Methods�� 114
8.8.1 Introduction�� 86 8.12.4 Results and Discussion�� 114
8.8.2 Materials and Methods�� 87 8.12.5 Conclusion�� 117
8.8.3 Results�� 88 8.12.6 References�� 117
8.8.4 Discussion�� 92
Figures and Tables Index
Table 3.1.5.1-1 Example for Identification of LER Table 6.5.2-1 Endotoxin standards and their
Process-Relevant Steps�� 6 assigned potency�� 26
Table 4.2-1 Investigation of LER Driving Forces in Figure 6.7-1 Structural Basis of Lipopolysaccharide
Biopharmaceutical Drug Products�� 9 Recognition by TLR4-MD-2 Complex�� 28
Figure 4.3-1 Schematic of Masking/Demasking Figure 6.7-2 TLR Recognition of Microbial
Phenomenon�� 10 Components�� 29
Figure 4.4.3-1 Covalent Modifications of Kdo2-lipid A Table 8.1.3-1 Recovery of CSE from Spiked DS and
in E. coli K-12 and Salmonella��13 DP Samples�� 38
Figure 6.3-1 E. coli K-12 Kdo-Lipid A Structure�� 22 Table 8.1.3-2 Recovery of NOE from Spiked DS
Figure 6.3-2 Degree of Lipid A Acylation Impacts and DP Samples�� 39
IL-I Response�� 23 Table 8.1.3-3 Recovery of CSE from Spiked DP
Table 6.5.1-1 U.S. Reference Standard Endotoxins Samples Stored at Below-freezing
development history�� 25 Temperatures (-30 °C)�� 39
Figure 8.2.1-1 Testing Setup Overview�� 41 Figure 8.4.6-3 Effects of Sorts of Chelating Agents
Figure 8.3.3.1-1 Masking of Four Model Bacterial and Detergents on LER�� 62
NOEs in LER solutions at Room Table 8.5.3-1 Endotoxin Activity Detected from
Temperature �� 48 LPS-spiked BTP-1 Over Time �� 66
Figure 8.3.3.1-2 Masking of Different Bacterial NOE in Table 8.5.3-2 Comparison of Recovery % Calculated
LER Solutions at Room Temperature�� 48 by Different Methods for BTP-1�� 66
Figure 8.3.3.2-1 Masking of Bacterial NOE and Purified Table8.5.3-3 Endotoxin Activity Detected from
Endotoxin in LER Solutions at Room LPS-Spiked BTP-2 Over Time�� 67
Temperature�� 49 Table 8.5.3-4 Comparison of Recovery % Calculated
Figure 8.3.3.3-1 Recovery of P. aeruginosa NOE as by Different Methods for BTP-2�� 67
a Function of Nutrient Supply and Table 8.5.3-5 Endotoxin Activity Detected from
Temperature�� 50 LPS-Spiked BTP-3 Over Time�� 67
Figure 8.3.3.4-1 Recovery of E. coli O113:H10 NOE Table 8.5.3-6 Comparison of Recovery% Calculated
as a Function of Bivalent Cation by Different Methods for BTP-3�� 68
Supply a pH�� 51
Table 8.5.3-7 Endotoxin Activity Detected from
Table 8.3.3.4-1 Varied Growth Conditions for E. coli� 51 LPS-Spiked BTP-4 Over Time�� 68
Table 8.3.3.4-2 Varied Growth Conditions for Table 8.5.3-8 Comparison of Recovery% Calculated
P. aeruginosa�� 51 by Different Methods for BTP-4�� 68
Figure 8.3.3.4-2 Recovery of P. aeruginosa NOE
Table 8.5.4-1 Effect of Pyrosperse™ on LPS Activities
as a Function of Bivalent Cation by USP <85> BET Method�� 69
Supply and pH�� 52
Table 8.5.5-1 Endotoxin Activities Recovered from
Figure 8.3.4-1 Lipid A Modifications in E. coli��53 LPS Spiked LRW Controls Over Time� 71
Figure 8.3.4-2 Comparison of MALDI-TOF MS Spectra Figure 8.6.2.1-1 Schematic of test vial preparations
from NOE and Purified Endotoxin�� 53
for conducting the hold study�� 73
Table 8.4.3.1-1 Kinetic Parameters for RSE Activity Figure 8.6.3.1-1 Pyrogen-Free Water (PFW) with
Change in LER Solutions Containing
Control Standard Endotoxin (CSE) Charles
Citrate �� 57
River Endosafe® Kinetic Turbidimetric
Table 8.4.3.1-2 Kinetic parameters for RSE activity Control after 3 days at 2-8 °C diluted with
change in LER solutions containing PFW and 100 mM Tris, 50 mM MgSO4
phosphate�� 58 (1/10 Strength Cation Buffer)��75
Figure 8.4.4-1 Effect of Temperature on Half-Life of Figure 8.6.3.1-2 Drug product with Control Standard
RSE Activity in LER Solutions�� 59 Endotoxin (CSE) Charles River
Figure 8.4.4-2 Effect of pH on Half-Life of RSE Activity Endosafe® Kinetic Turbidimetric
in LER Solutions�� 59 Control after 3 days at 2-8 °C diluted with
PFW and 100mM TRIS, 50mM MgSO4
Figure 8.4.4-3 Effect of Salt Concentrations on Half- (1/10 strength Cation buffer)�� 75
Life of RSE Activity in LER Solutions� 59
Figure 8.6.3.1-3 Pyrogen-Free Water (PFW) with Control
Figure 8.4.5-1 Effect of Citrate Concentrations on Half- Standard Endotoxin (CSE) Lonza KQCL
Life of RSE Activity in LER Solutions�� 60 Kinetic Chromogenic Control after 3
Figure 8.4.5-2 Effect of Phosphate Concentrations on Half- days at 2-8 °C diluted with PFW and
Life of RSE Activity in LER Solutions�� 60 25 mM Tris, 12.5 mM MgSO4 (1/40
Figure 8.4.5-3 Effect of Polysorbate 20 Concentrations Strength Cation Buffer)�� 75
on Half-Life of RSE Activity in LER Figure 8.6.3.1-4 Drug Product with CSE Lonza KQCL
Solutions�� 60 Kinetic Chromogenic Control after 3
Figure 8.4.6-1 RSE Activity Change in LER Solution days at 2-8 °C diluted with PFW and
at 25 °C with 3 Different Dilution 25 mM Tris, 12.5 mM MgSO4 (1/40
Methods�� 61 Strength Cation Buffer)�� 75
Figure 8.4.6-2 RSE Activity Change in LER Figure 8.6.3.1-5 Reactivation of CSE (LPS) Contaminated
Solution at 2 °C–5 °C with 3 into mAb/Citrate/PS Product 10 EU/
Different Dilution Methods�� 61 mL After 3 Days at 2-8°C Charles River
Endosafe® Kinetic Turbidimetric�� 76
Figure 8.6.3.1-6 Kinetic turbidimetric testing of E. Figure 8.9.3-1 Percent Recovery of Six Preparations
cloacae NOE�� 76 of LPS in LRW or Protein Sample at
Figure 8.6.3.1-7 Kinetic chromogenic testing of E. Room Temperature�� 98
cloacae NOE�� 76 Table 8.9.3-2 Recovery of CSE in Various Levels of PS80
Figure 8.6.3.1-8 Monoclonal Antibody in PFW and 100 in 20 mM Histidine buffer, pH 5.5�� 99
mM Tris, 50 mM MgSO4 (1/10 Strength Figure 8.9.3-2 Recovery of CSE Spiked into Protein
Cation Buffer) Kinetic Turbidimetric vs. LRW Held for 3 days at Ambient
(CRE) with CSE on Day 7�� 76 Temperature�� 99
Figure 8.6.3.1-9 Monoclonal Antibody in PFW and 100 Table 8.9.3-3 CSE Spiked at 1700 EU/mL in ~150
mM Tris, 50 mM MgSO4 (1/10 Strength mg/mL Product�� 100
Cation Buffer) Kinetic Turbidimetric Figure 8.9.3-3 LER Hold Time Spiked Protein Sample
(CRE) with NOE on Day 7�� 77 Study Calculated vs LRW and vs
Figure 8.6.3.1-10 Monoclonal Antibody in PFW and 25 Theoretical�� 101
mM Tris, 12.5 mM MgSO4 Figure 8.9.3-4 Comparison of Endotoxin Assay Kits
(1/40 Strength Cation Buffer) in a Hold Time LER Study at Ambient
Kinetic Chromogenic (Lonza) with Temperature�� 101
CSE on Day 7�� 77
Figure 8.10.2.1-1 Manufacturing Steps for Biotech
Figure 8.6.3.1-11 Monoclonal Antibody in PFW and 25 Product X Drug Substance�� 103
mM Tris, 12.5 mM MgSO4
(1/40 Strength Cation Buffer) Table 8.10.3.1-1 Endotoxin Spiking and
Kinetic Chromogenic (Lonza) with Container Types�� 104
NOE on Day 7�� 77 Table 8.10.4.1-1 Product X DS Spiked with 5 EU/mL
Figure 8.7.4.2-1 Reverse Hold/Time Study and CSE, BET with Dilution 1:8�� 106
Demasking Study of Endotoxin�� 82 Table 8.10.4.1-2 Product X DS Spiked with 5 EU/mL
Figure 8.7.4.4-1 Reproducibility and Robustness of the CSE, BET with Dilution 1:16�� 106
Developed Demasking Protocol�� 83 Table 8.10.4.1-3 Product X DS Spiked with 5 EU/mL
Figure 8.7.4.4-2 Robustness of Sample Preparation R. pickettii, BET with Dilution 1:8�� 106
Protocol�� 84 Table 8.10.4.1-4 Product X DS spiked with 5 EU/mL
Figure 8.7.4.4-3 Comparison of Different lots of R. pickettii, BET with Dilution 1:16� 107
LAL Reagent, Endo-RS Components, Table 8.10.4.3-1 Formulation Buffer Spiked with 5
and Sample Tested�� 85 EU/mL CSE, BET with Dilution 1:8�� 107
Figure 8.7.4.6-1 Detection of Endotoxins from Table 8.10.4.3-2 Formulation Buffer Spiked with 5 EU/
Different Species and Sensitivity�� 85 mL CSE, BET with Dilution 1:16�� 108
Figure 8.8.1-1 Reverse Hold Time Study Setup�� 87 Table 8.10.4.5-1 Unconditioned Bulk Spiked with 5 EU/
Figure 8.8.3.1-1 14-Day Hold Time Study for Drug mL CSE, BET with Dilution 1:8�� 108
Product at 2-8 °C with RSE�� 89 Figure 8.11.3-1 Endotoxin Recovery �� 112
Figure 8.8.3.1-2 7/3-Day Hold Time Study for Drug Figure 8.11.3-2 Average Relative Endotoxin
Product at 20-25 °C with RSE�� 90 Recovery Results �� 112
Figure 8.8.3.2-1 8-Day Hold Time Study for a Drug Table 8.12.4.1-1 LAL Assay Validity Criteria Summary� 115
Product at 2-8 °C with High-Potency
Table 8.12.4.2-1 Endotoxin Recovery Results from the
Endotoxin, O55:B5 (Lonza)�� 91
Drug Product and LRW Control�� 115
Figure 8.8.3.2-2 14-Day Hold Time Study for
Figure 8.12.4.2-1 Endotoxin Recovery Curves from the
Intermediate of a Drug Substance at
Test Article and LRW Control�� 116
2-8 °C with High-Potency Endotoxin,
O55:B5 (Lonza)�� 91 Table 8.12.4.2-2 Endotoxin Recovery Results from Two
Additional Drug Product Lots�� 116
Table 8.9.3-1 Recovery of CSE Spiked into 20 mM
Histidine +/- 83 mg/mL Trehalose, Figure 8.12.4.2-2 Endotoxin Recovery Curves from Two
pH 5.5�� 97 Test Articles and LRW Control�� 116
1.0 Introduction
The term “low endotoxin recovery” (LER) describes the failure to detect spiked endotoxin in some
finished sterile biological drug products when tested using the internationally harmonized compendial
Limulus amebocyte lysate (LAL) assay to detect bacterial endotoxin (1–6). Regulators and developers
became concerned that compendial methods required by international regulations were unable to con-
firm the absence of endotoxin in certain drug products, especially troubling considering the inherent
risks posed by accidental endotoxin contamination.
The failure to recover spiked endotoxins from finished drug products suggested that endotoxin con-
tamination from the manufacturing processes may not be detected at release, thus, pyrogenic products
could be distributed for commercial use in patients. This task force, however, searched peer-reviewed,
scientific literature, public data on recalls and adverse events, and available copies of the Centers for
Disease Control Morbidity and Mortality Weekly Reports and found no instances where a bacterial
endotoxin test (BET) failed to detect pyrogenic levels of endotoxin in samples of products released to
the market. Some investigational or marketed products were found to be pyrogenic, but those cases
were due to non-endotoxin pyrogens or an improperly performed BET (7). Hence, published histori-
cal data provide no direct evidence of LAL failing to detect pyrogenic products.
In 2013, the U.S. FDA began to request applicants submitting biologics licensing applications (BLA)
to provide evidence that endotoxin spiked into a drug product could be detected using the USP <85>
BET. Referring to confidential data from spiking studies submitted in BLAs, the FDA reported in 2014
and 2015 that poor endotoxin recoveries were observed in some cases, which confirmed the initial
findings of Chen and Vinther. (1,8,9) The data from endotoxin spike/hold studies presented in BLAs
were confounding and difficult to interpret, prompting the FDA to request more information and/or
additional studies for clarification. But spike/hold study protocols varied greatly from study to study
and company to company. The detection of endotoxin from spiked samples was influenced by sample
preparation methods, including pretreatment and vortexing time. The type of spiking standards, the
spike amount, hold-time and temperature of the spike sample, endotoxin recovery calculations, and test
execution all contributed significant variability to results, making them difficult to interpret and com-
pare. Certain proteins with isoelectric points (pI) above 8.0 or with positively charged conformational
regions in the protein structure were thought to be binding spiked endotoxin. Yet, certain protein-free
excipient solutions also appeared to affect the detectability of endotoxin by the LAL test.
Due to the complexity of biological drug formats and materials, historical data do not necessarily
provide an adequate basis for a safety assessment of new drugs, supporting the need for LER studies in
BLAs for specified biologics regulated by CDER-FDA. Subsequently, PDA recruited from its mem-
bers, academia, FDA, the biopharmaceutical industry, reagent suppliers, testing, and consulting firms
to form the LER Task Force in 2016 to address these perplexing, contradictory results.
1.1 Purpose
Members of the PDA LER Task Force prepared this technical report to describe the underlying mecha-
nisms and contributing factors of LER, summarize the potential clinical impact of the LER phenom-
enon, present guidelines for developing LER hold-time study design, and provide strategies for the
mitigation of LER.
1.2 Scope
Focusing primarily on protein drug products, this technical report provides an overview of the regula-
tory expectations for sterile and/or pyrogen-free drug products and recommend how these require-
ments can be met through the implementation of microbial process control and testing strategies. It
includes a standardized LER hold-time protocol that can be used to determine if a biological drug
product is affected by LER. Thirteen case studies from pharmaceutical companies, vendors, and con-
sulting firms comprise Section 8.0, offering a snapshot of the complexity and diversity of LER occur-
rences across various types of products and formulations. Based on these case studies, Section 4.0 seeks
to provide a mechanistic understanding of the LER phenomenon wherein the endotoxin is masked.
Technical Report No. 82 © 2019 Parenteral Drug Association, Inc. 1
Section 4.3 describes a two-step reaction model of LER for formulations containing a chelating agent
(e.g., citrate) and a surfactant (e.g., polysorbate). As proteins may also mask endotoxin resulting in
LER, the reversibility of the LER reaction and its effect on endotoxin detectability in the LAL assay
is discussed. Examples provide approaches used to detect or unmask endotoxin. Successful mitigation
strategies allow for the use of validated in vitro endotoxin testing methods for drug product release and
avoid the use of the pyrogen test using rabbits.
2.0 Glossary of Terms

To understand elements of endotoxin testing and issues that may be encountered, establishing a com-
mon terminology is critical. Definitions of these routinely used terms are adopted based on the current
understanding of the science and applications encountered in endotoxin recovery.
Activity Endotoxin
Ability of endotoxin (LPS) to initiate the LAL The major constituent of the outer membrane of
cascade in the compendial bacterial endotoxins Gram-negative bacteria is composed of lipid A,
test (BET) assay, or the ability to elicit a pyrogen- the core polysaccharide, and the O-antigen poly-
ic response in a compendial pyrogen test (2,10). saccharide; endotoxin is also known as lipopoly-
Activity can be measured by other assays such as saccharide (LPS) (14).
the monocyte activation test (MAT) or recombi-
nant Factor C tests (rFc), if such tests have been Reference Standard Endotoxin (RSE)
validated, to demonstrate that decisions made The primary standard from USP, EDQM, JP and
from the results are comparable to or superior to WHO for use in the harmonized compendial bac-
the compendial assay. terial endotoxins test (BET). The current 3rd Inter-
national Standard (WHO), USP, and EDQM RS
Activity is measured in endotoxin units (EU). In are lyophilized formulation that contains highly
terms of activity, one EU = one IU, regardless purified LPS that is chemically extracted and puri-
of the source. Activity is generally expressed as a fied from E. coli strain O113:H10:K(-) and further
concentration, usually EU/mL. formulated with stabilizers and excipients.
Control Standard Endotoxin (CSE) Lipopolysaccharide (LPS)
Endotoxin preparations other than the international See Endotoxin.
or national reference standards that are traceable in
their calibration to the international endotoxin ref- Low Endotoxin Recovery (LER)
erence standard. A CSE is a secondary or tertiary The inability to recover ≥50% activity over time
standard, commonly purified from Escherichia coli, when known amount of endotoxin is added to
and is usually manufactured and certified by an an undiluted product. LER cannot be overcome
LAL reagent manufacturer for use with a specific lot by dilution.
of reagent under defined assay conditions (11,12). Masking
Drug Product (DP) A type of interference that may result in low en-
The dosage form in the final immediate packag- dotoxin recovery.
ing intended for marketing (13). Measured Values
Drug Substance (DS) Those values where activity is confirmed by in-
Any substance or mixture of substances intended terpolation from a reference standard curve.
to be used in the manufacture of a drug (medici- Naturally Occurring Endotoxin (NOE)
nal) product and, when used in the production Endotoxin prepared from Gram-negative bacte-
of a drug, becomes an active ingredient of the ria produced under defined conditions and with
drug product (13). minimal nonchemical processing, e.g., centrifu-
gation and filtration (15).
2 © 2019 Parenteral Drug Association, Inc. Technical Report No. 82

Nominal Value Potency

The assumed activity of an endotoxin prepara- An expression of the activity of a secondary cali-
tion, dilution, or “spike” based on label-claim bration standard to relate units of weight (ng/
information. vial or ng/mL) to units of activity (EU/ng) in a
preparation.
2.1 Abbreviations
BET Bacterial endotoxin test LRW LAL reagent water
CDER Center for Drug Evaluation and MAT Monocyte activation test
Research
MVD Maximum valid dilution
CSE Control standard endotoxin
NOE Naturally occurring endotoxin
DP Drug product
PAMP Pathogen-associated molecular patterns
DS Drug substance
PPC Positive product control
EU Endotoxin unit
RPT Rabbit pyrogen test
LAL Limulus amebocyte lysate
RSE Reference standard endotoxin
LER Low endotoxin recovery
TLR Toll-like receptor
LPS Lipopolysaccharide
3.0 LER Hold-Time Studies

While LER hold-time studies are not part of compendial endotoxin method suitability testing, they
are defined as supplemental studies and are mandatory for LER detection. Variations in study design,
experimental procedures, and/or type of endotoxin used can produce confounding and sometimes
contradictory results regarding the LER phenomenon. In contrast, harmonized LER hold-time study
protocols that consider BET best practices have been shown to provide consistent results in many
different laboratories (16,17). Section 3.0 provides points to consider for planning and conducting
scientifically sound and harmonized LER hold-time studies. Alternative approaches may be used if
adequately justified. LER hold time studies for a BLA should be conducted according to a protocol
that has been preapproved by the quality unit. The protocol should include clearly defined acceptance
criteria. It may be necessary to conduct investigative testing for LER prior to any validation studies.
Those studies do not necessarily need the approval of the quality.
Validation of maximum quality control (QC) sample storage time, that is, the maximum allowable time
between sampling and testing, is not in the scope of this technical report and should be conducted sepa-
rately (11,12). Many of the principles presented below may also applied to the validation of QC sample
storage time (especially Section 4.1.2). However, there are three main reasons for this separation.
• LER hold-time studies should determine whether a product has LER under conditions which
are relevant for the manufacturing process. For parenteral products the manufacturing pro-
cess usually ends with filling into the final container (e.g., vials or pre-filled syringes).
• QC sample hold time refers to all sample types and is a general GMP requirement. Whereas
LER testing is specific to drug product and is currently requested for BLAs (11,18,19).

• In principle storage of QC samples can occur under conditions, which are different from
manufacturing conditions (e.g., samples can be diluted and/or frozen prior to testing)
3.1 Test Method

For LER hold-time studies, the validated BET method should be the same as the one used for routine
testing of the respective sample. In addition to applying the same method, the following criteria should
be the same:
• Vendor for lysate
• Vendor for endotoxin standard
• Procedure for sample and lysate preparation (e.g., storage of product samples prior testing (e.g.,
refrigerated or frozen), vortexing time and speed of samples, sample dilution scheme, or dilution
tubes, etc.)
• Data analysis method
• Sample dilution must not exceed maximum valid dilution (MVD)
If the validated dilution from product-specific method suitability testing fails to produce valid results
in the context of the LER hold-time studies, validated higher dilutions not exceeding the MVD may
be considered.
3.1.1 Reagents and Materials

The test should be performed with reagent sets from qualified vendors. Each lot of LAL reagent must
be qualified according to the compendial requirements prior to use. LAL reagent water (LRW), dis-
posables, and all buffers must be demonstrated to be free of detectable endotoxin.
3.1.2 Spiking of Undiluted Sample

LER studies should be done with drug product samples. If the drug product is lyophilized, it is recom-
mended to reconstitute the product following the product-specific instructions. Spiking of the undi-
luted sample solution should be done using control standard endotoxin (CSE) or reference standard en-
dotoxin (RSE) as the preferred standard. Additionally, naturally occurring (native) endotoxins (NOEs)
may be used as a supporting study. Spiking additional samples for purposes of investigation, if needed,
is recommended. If failure occurs, the same sample should be tested as well as an extra sample. The vol-
ume of the spike solution should not exceed 10% of the total volume of the spiked sample (e.g., 1 mL
of a 50 EU/mL spike solution is added to 9 mL of sample, resulting in 10 mL of spiked sample with an
endotoxin content of 5 EU/mL).
The spiking level should account for validated sample dilution and midpoint of the standard curve for
testing if the study is going to be performed using a kinetic method (a spiking level of 5 EU/mL is a
realistic value to detect LER). If the study is to be performed using a gel clot method, then the range
of dilutions to cover the range of expected recoveries to demonstrate 50% to 200% must be set up.
Example:
• Undiluted sample is spiked to a final concentration of 5 EU/mL
• BET method: Kinetic chromogenic assay
• Validated sample dilution for the BET: 1:50
• Four-point standard curve (0.005, 0.05, 0.5, and 5.0 EU/mL)
• Level of spiked endotoxin in 1:50 diluted sample: 0.1 EU/mL
Conclusion: 0.1 EU/mL is well within the range of the standard curve

A multiple-aliquot approach is recommended to avoid sampling repeatedly from a single container

(see Section 8.5). Two approaches for this are:
• A bulk sample may be spiked and time-point samples aliquoted, stored appropriately, and then,
assayed at the specified time point
• To conduct a reverse assay approach, aliquots may be prepared and spiked with endotoxin in a
reversed-time fashion; for example, Day 7 samples are spiked first, and Day 0 samples are spiked
last. All time-point samples are assayed at the same time.
3.1.3 Storage Containers

LER hold-time studies should be conducted using containers appropriate for endotoxin studies (i.e.,
no adsorption, no interference).
3.1.4 Batch Requirements

For representativeness of the testing results, samples should be taken from three batches to ensure that
variability among processes is captured. Each company and laboratory should determine the appro-
priate number of batches to use for validation depending on the product’s batch size, formulation, or
other considerations (20,21).
3.1.5 Storage Time and Temperature

LER studies should be conducted at process-relevant temperatures and times. The term “process
relevant” refers to process steps that are most likely to have an impact on LER (e.g., addition of poly-
sorbate plus chelator, hold times, open or closed process steps)
While process steps may be conducted at a variety of temperatures and times, for comparability
among tests and to facilitate study execution, process-relevant steps can be categorized into the fol-
lowing groups:
• If process-relevant steps are conducted at refrigerated temperatures, studies should be conducted
at 2-8 °C for seven days.
• If process-relevant steps are conducted at room temperature, studies should be conducted at 20–
25 °C for a process-relevant time determined by a risk-based evaluation (see Section 3.1.5.1).
• If there is a process-relevant step conducted at 2-8 °C, and one or more process-relevant steps
performed at room temperature, the studies should be conducted at room temperature of 20-25
°C for a process-relevant time determined by a risk-based evaluation (see Section 3.1.5.1).
If the hold-time study generates inconclusive results, or if additional information is needed, the hold-
time study may be extended by adding more time-points.
3.1.5.1 Example for a Risk-Based Evaluation to Determine LER Hold-Time Study Temperature and
Time Parameters
This example provides a risk assessment of a hypothetical manufacturing process, identifying critical
steps related to the LER phenomenon, and determining the process-relevant temperature and time
required for the LER hold-time study (Table 3.1.5.1-1). Each company is expected to perform a
similar risk assessment on their own manufacturing process with production and QC subject matter
experts identifying the process-relevant temperature and time parameters required for their specific
LER hold-time study.

Table 3.1.5.1-1 Example for Identification of LER Process-Relevant Steps
LER Process-
Process Step Comment
Relevant
Drug substance (drug + citrate) LER risk is low because polysorbate has not
1 No
stored at 2–8 °C been added
Addition of polysorbate (PS) LER risk is high because both citrate and
2 Yes
(processing time is 5 hours at RT) polysorbate are present
Addition of component(s) to drug + citrate
+ PS (processing step is performed in LER risk is high because endotoxin ingress is
3 Yes
an open vessel, transferred to a closed possible via open process step
system, and then held for 5 days at 2–8 °C)
Sterile filtration and filling (conducted
LER risk is high because of hold time at room
4 at room temperature for a validated Yes
temperature
maximum time of 30 hours)
Drug product sampling and release testing Validation of maximum QC sample storage
5 (validated QC sample storage time is 5 N/A time is not LER-process-relevant and should
days at 2-8 °C) be conducted separately
In the example:
• Three process steps were determined to be relevant for LER
• If a process-relevant step is conducted at 2–8 °C and another process-relevant step performed at room
temperature, the LER hold-time study should be performed at room temperature (20–25 °C)
• In this example, the LER hold-time study should be conducted at room temperature for the combined
longest process- and hold-times (Process step #2 + Process step #4 = 5 hours plus 30 hours = 35
hours). Process step #3 is not considered because LER would be detected during the room tem-
perature hold-time study (CSE/RSE is used as a worst-case analyte).
3.1.6 Controls
A spiked water control LAL reagent water (LRW) spiked with the same endotoxin and diluted in the
same way as the product sample) must be run in parallel to the spiked product sample in the same
container.
3.1.7 Equipment
Equipment requirements for LER hold-time studies are equivalent to those for routine BET.
3.1.8 Personnel
Technicians must be qualified to perform the BET and to execute the LER hold-time study protocol.
3.1.9 Data Analysis and Interpretation

To determine if LER is present, the endotoxin recovery is calculated by comparing the amount of
endotoxin in the sample against the amount of endotoxin in the LRW reference sample.
A minimum of four time-points must be conducted to ensure valid and accurate results. Two consecu-
tive data points falling below 50% recovery indicates LER. If the last time-point results in a recovery
of less than 50%, additional time-points should be tested to aid in the analysis of LER endotoxin-
masking and to ensure that the LER result is accurate.
Data from all batches must be plotted and interpreted separately; no averaging for the mean is allowed.
Each of the batches must independently fulfill the acceptance criteria. If any batch has a confirmed
failure (i.e., <50% endotoxin is recovered), the product or process step is considered to be exhibiting in-
terference with the recovery of endotoxin activity. The sample “time zero” endotoxin values should not
be used as a reference as LER may occur within minutes, which would lead to a highly biased reference.

3.1.9.1 2–8 °C Studies

For 2–8 °C studies, recovery must be calculated against the measured LRW reference sample if a high
concentration endotoxin preparation (e.g., >1,000,000 EU/mL) is used as an analyte.
• Rationale: No exact Certificate of Analysis (CoA) value is given for high-concentration endotox-
in preparations, and multiple dilution steps are often needed to obtain an appropriate concentra-
tion, which could result in error.
• For CSE, the CoA provides the exact endotoxin potency and can be used as reference. Few,
if any, dilution steps are required to achieve the appropriate concentration, thus reducing the
likelihood of producing an error. When using CSE, the theoretical value is also an appropriate
reference for calculation of recovery.
3.1.9.2 20–25 °C Studies

For 20–25 °C studies, recovery must be calculated against the measured spiked water control (see
Section 4.1.6) because endotoxin activity might be reduced over time if the endotoxin is stored at
room temperature (20–25 °C).
• Rationale: Only limited data on endotoxin storage at room temperature are available.
4.0 Proposed Mechanisms of LER

The likely molecular mechanisms causing LER in biopharmaceutical drug products are of interest in
that they suggest mitigation strategies, which are discussed further in Section 4.2.
As described in Section 6.3, detectable LAL activity can be variable, depending on the primary and
supramolecular structure of (lipopolysaccharide) LPS. In addition, measured biological activities can
vary depending on the detection method. Analysis of LER-affected samples comparing different test
methods (e.g., LAL vs. MAT vs. RPT) can sometimes lead to different results. Due to the general
molecular structure of endotoxin, including poly- or oligo-saccharide regions (core region and O-an-
tigen), LPS possesses an amphiphilic nature. In an aqueous environment, and above a specific concen-
tration, amphiphilic molecules such as LPS form supramolecular structures. The type of aggregation
has been shown to depend, in part, on the chemical structure of the aggregate-forming molecules (22).
The aggregation state of LPS has also been shown to have an effect on biological activity (22–24).
LER mechanisms are proposed based on the hypothesis that the supramolecular state of the endotoxin
is altered in a way that makes it less detectable (i.e., masked). Endotoxins found during contamination
are known to exist in a vast array of chemical forms. One of the most potent forms of endotoxin is a
molecule containing two D-gluco-configurated hexosamine residues, two phosphoryl groups, and six
fatty acids, including 3-acyloxyacyl groups with a defined chain length and in a distinct location. LPS
with these structures, such as CSE and RSE, are obtained from Gram-negative bacteria (e.g., E. coli)
(25). Reportedly, the organism and a host of environmental factors influence the chemical structure
of LPS and its supramolecular state, resulting in various masking susceptibilities of LPS. For LPS sus-
ceptible to LER, a two-step mechanism has been proposed (chelation of divalent cations + change of
aggregation state) (26,27). CSE and RSE preparations are standardized LPS that have been shown to
be susceptible to LER and, thus, are commonly used as the source of LPS for LER hold-time studies.
Although the two-step mechanism is believed to cause LER due to destabilization of salt bridges; other
endotoxins with different substituent groups (e.g., aminoarabinose) are not destabilized in the same
manner (28).
Section 4.5 describes a potential mechanism by which some LPS exhibit LER in certain formulations
and other types are less likely to do so. Different results are reported in hold-time studies depending

on the LPS used, so understanding these distinctions is important. If the LPS has an altered supramo-
lecular structure (i.e., the aggregation state), this can potentially result in a state that does not activate
Factor C and thus not detectable by LAL (29). LPS has been shown to form mixed aggregates with
a variety of matrix components (30). Moreover, such a transition from detectable to non-detectable
endotoxin is time- and temperature-dependent (31,30). The reaction kinetics can be influenced by
the experimental setup and execution as well as by the composition of the components in the spiked
sample. Alteration of the aggregation state is associated with a shift of the equilibrium state. To this
end, a new equilibrium state of endotoxin is formed under LER conditions.
Section 4.0 provides information on the potential mechanism and driving forces of LER. Differ-
ences between test results in the LER case studies (Section 8.0) are discussed in light of potential
mechanisms. Understanding the mechanism of LER allows better interpretation of results from LER
hold-time studies, which is the basis for development of solution-oriented approaches to minimize or
eliminate LER.
4.1 Differentiation of LER and Test Interference

LER should be clearly distinguished from the well-known inhibition in the BET test in order to be
understood (32–34). LAL test interference is described in all major pharmacopoeias (2,35,36). In the
case of BET interference, the reaction of the Limulus-based detection system is believed to be primar-
ily inhibited or enhanced. Such test interference can be caused by suboptimal pH, unbalanced cation
concentration, unspecific enzyme reactions, and so forth (32,37). As these interferences are often
concentration-dependent, dilution of the sample is usually the best approach to overcome this type
of interference (inhibition) (32). An internal control, positive product control (PPC), ensures that
the interference has been overcome. The PPC comprises a defined amount of endotoxin spiked into
diluted sample and the test is considered valid if 50% to 200% of the theoretical spike is recovered.
In order to determine the minimum dilution at which a valid test is achieved, a series of dilutions of
product are tested in this way. The dilution needed to achieve a valid assay must be less than or equal
to the MVD to ensure sufficient sensitivity of the test for a specific product. This is the traditional BET
approach and is still required.
In contrast, LER is a time-dependent and dilution-independent lack of recovery (see Sections 8.1–
8.10). To evaluate LER, endotoxin is spiked into the undiluted product. The spiked product is incu-
bated for a period of time (hold-time) before analysis, because the LER effect often occurs over time
(see Section 3.0). For analysis, the spiked sample is diluted using the qualified dilution of the sample
to exclude classical BET test interference. In the case of LER, valid PPCs are typically achieved, but
the spike in undiluted product shows low recovery, which clearly distinguishes LER from classical test
interference (i.e., inhibition). Therefore, if a sample is prone to LER, proper design of LER hold-time
study protocols is essential. The classical BET test is well described by international pharmacopoeias,
but the procedures for evaluation of LER are not defined in a regulatory document.
4.2 Investigation of LER Causes

LER has been reported in biopharmaceutical drug products, some of which contain high protein con-
centrations (e.g., monoclonal antibodies). Some are formulated with chelators and/or phosphate buf-
fer systems and polysorbates (38). As discussed in Section 3.0, LER hold-time studies are performed
to identify if a product causes LER. If LER is observed in a specific product, hold-time studies on the
drug alone and placebo (formulation without drug) may provide elucidation of which components
cause LER. Potential scenarios are given in Table 4.2-1.

Table 4.2-1 Investigation of LER Driving Forces in Biopharmaceutical Drug Products
Scenario Composition LER Root Cause

Drug product Yes
I Formulation without drug substance Yes LER caused by formulation
Drug without LER-inducing components No
Drug product Yes
II Formulation without drug substance No LER caused by drug substance
Drug without LER-inducing components Yes
Drug product Yes
LER caused by drug substance
III Formulation without drug substance Yes
and formulation
Drug without LER-inducing components Yes
LER may be caused by the formulation components alone or in combination with the protein drug
substance (30) (see Case Studies 8.2–8.10). It can also be caused by the drug substance itself, e.g.,
monoclonal antibody (33) (see Case Studies 8.1 and 8.8). Identification of the root-cause is helpful
in understanding the underlying mechanism of LER and the subsequent development of mitigation
strategies. If, for example, formulation excipients causing LER are identified (Table 3.1.5.1-1, Sce-
nario I), one approach would be to avoid them by improving or modifying the formulation of a given
drug product. This is often not possible, however, if the product is already in clinical development or
requires these components for stability. If avoidance of LER-causing components is not possible, ad-
ditional mitigation strategies are recommended (see Section 5.0).
4.3 Proposed Two-Step Reaction Model of LER

As described in Section 4.2, LER can be caused by the formulation components of a drug product
and/or by the drug substance itself. To study the impact of formulation, any model developed should
be based on the components whose chemical and physical attributes are known.
Although the properties of proteins may be known, their interactions with LPS are less well under-
stood, depending on 3D structure, hydrophobic/charged patches, etc. Typical formulation compo-
nents causing LER, such as buffers and surfactants, have been used to investigate the LER mechanism.
For example, a combination of chelator and nonionic surfactant has been shown to commonly induce
LER; whereas, the presence of only one of the components, chelator or surfactant, has been less likely
to induce LER (30).
Chelators, citrate for example, are known to form a complex with divalent cations. And surfactants
like polysorbate, e.g., Tween 20® and Tween 80®, are also commonly used to stabilize protein drugs.
Therefore, in case studies examining the LER phenomenon, a system containing citrate and polysor-
bate 20 was used (see Sections 8.3, 8.4). Based on these findings, a model is presented to describe LER
in molecular terms. By using various concentrations of the chelator, research has shown that LER is a
kinetically controlled process that follows a two-step mechanism (30,31).
In the first step, during mixing of the sample with the chelator, the salt bridges between LPS and di-
valent cations (e.g., Mg2+, Ca2+) may be destabilized, leading to reduced rigidity of the LPS aggregate.
In the second step, the surfactant may intercalate among LPS molecules and change the initial supra-
molecular structure by formation of mixed aggregates, e.g., micelles, lamellar, or hexagonal structures.
The equilibrium state of the LPS is shifted, in some cases, to become non-detectable; in other words,
the endotoxin is masked (P-LPS). Figure 4.3-1 illustrates this model, where the structure presents as
micellar, but is known to be much more diverse, including lamellar, hexagonal, and inverted cubic
structures stabilized by hydrophobic interaction (24). Further, Galanos has shown that the surface
residues in particular are stabilized by salt bridges (39).

=Endotoxin =Divalent cation =Chelator =Surfactant
Figure 4.3-1 Schematic of Masking/Demasking Phenomenon
In LPS stabilized by salt bridges, LER may thus be induced by the interplay of a chelator (e.g., citrate)
and a surfactant (e.g., polysorbate), but is not limited to these substances. Other molecules with the
ability to interfere with the salt bridge between LPS molecules and alter the aggregation state can cause
masking as well, for example, the protein itself. As the interaction between the endotoxin and the pro-
tein can either be hydrophobic or ionic, the protein on its own can also be responsible for alteration
of aggregation state of LPS (33). In such a case, a complex-forming buffer or surfactant would not be
required to induce LER.
Regardless of the molecular cause of masking, manifesting as LER, a new LPS aggregation state will
probably be formed. Destruction of the molecular structure of LPS is highly unlikely since it is known
to be very stable where only harsh conditions, such as dry heat > 250 °C, can destroy it (40). In some
cases, recovery of endotoxin activity has been demonstrated, suggesting that the LPS molecules can be
reconstituted in a manner where detection is re-established (see Section 8.7). To investigate the su-
pramolecular structures of masked endotoxin, studies have been performed; some potential states are
described in Section 4.5 (30,41). The results of these studies are preliminary and may be confined to
particular conditions (e.g., hold-time and temperature, test methods, sample preparations and compo-
sitions, source of the endotoxin), consequently these findings cannot be generalized as an explanation
for as the sole mechanism(s) of LER.
4.4 Factors Influencing Reaction Kinetics

In contrast to BET inhibition/enhancement, LER is a time-dependent phenomenon, where the ratio
of product to endotoxin is much higher than for PPC analysis, in which the diluted sample is spiked
(minimal dilution of spiked endotoxin). The transition from a detectable to a non-detectable state (less
than 50% recovery) may be due to a change of LPS aggregation state and may vary over time. LER
can occur within minutes or take several days, weeks, or even months (42). In some rare instances, no
masking kinetics can be determined, because the kinetics are faster than the time needed to perform
an endotoxin test. Basically, the rate of transformation of the supramolecular state depends on the
energy of the transition state relative to the initial (starting) state. The reaction time is a function of
the temperature and the relative transition state. In general, to start a reaction, the activation energy
must be available to a chemical system. If not preexisting, this energy can be extrinsically provided by
adding, for example, thermal or mechanical energy. The chemical system, that is, sample composition,
intrinsically determines the transition state and, therefore, the activation energy.
4.4.1 Energy Input (Extrinsic)

The reaction kinetics can be accelerated and decelerated by increasing or decreasing the energy input
(see Sections 8.3, 8.8) (30,31,43). The input of energy can be manipulated by changing the incuba-

tion temperature (thermal energy) during an LER hold-time study. LER is a time-dependent, kineti-
cally driven phenomenon; that the reaction kinetics take longer at 4 °C compared to 25 °C has been
demonstrated and described in detail (see Sections 8.3, 8.8) (30). As the reaction is spontaneous, but
kinetically hindered, it is also reaction-surface-dependent. This parameter can be influenced by sample
handling, for example, vortexing. After spiking a sample for the hold-time study, or before prepar-
ing dilutions, a sample is vortexed. The length and intensity of the vortexing influences the masking
behavior of the endotoxin (43). Further side-by-side comparison studies have indicated that the LER
effect is substantially enhanced by excessive vortex mixing (41). Thus, the experimental design can
substantially impact the reaction kinetics of the LER effect. Consequently, these parameters need to be
considered and controlled when planning LER hold-time studies (see Section 3.0).
4.4.2 Masking Capability of a Sample

The masking behavior of a given sample can be influenced by extrinsic factors as described in Sec-
tion 4.4. Intrinsic factors (concentration of formulation components, fillers included in endotoxin
standard formulations, and drug) also determine the masking capability of a sample which, in turn,
determines the reaction kinetics. The analysis of similar but slightly different sample compositions
(e.g., concentrations, pH, etc.) have shown that the kinetics of LER can be very different (see Sections
8.2, 8.4, 8.9).
4.4.2.1 Formulation Components

Various combinations of proteins and/or formulation components also can produce LER. Buffer sys-
tems containing both a chelating agent and surfactant have been shown to impact LER kinetics (30).
Buffers most likely to cause LER through chelation are those with divalent earth metal ions like Ca2+
and Mg2+. This chelation binds divalent ions leaving them unavailable to stabilize the LPS aggregates
via salt bridges. Histidine buffer does not appear to cause LER (see Section 8.12), possibly because
it reacts repulsively to cations in its protonated form (44). Consequently, the stronger the interaction
between divalent cations and the complex-forming agent, the faster the reaction kinetics. When using
LPS that is susceptible to the chelation effect, these effects would likely be exacerbated, less so for those
LPS preparations in which divalent cations are less prominent (39):
EDTA > Citrate > Phosphate > Histidine
The stability of a complex depends on a combination of sample pH and the chelating agent pKa. To
this end, the complex formation is typically stronger at higher pH values (see Section 8.4) (30). For ex-
ample, a phosphate buffered system with a pH of 3 is less likely to cause LER than a system with a pH of
7.5. In turn, the higher the pH of a sample, the more likely to cause LER due to faster reaction kinetics.
In addition to the complex-forming agent in the buffer system, the presence of a surfactant has been
shown to exacerbate LER. In biopharmaceutical formulations, polysorbates are commonly used, the
most prominent of which are polysorbate 20 and polysorbate 80. Polysorbate 20 shows faster kinet-
ics than polysorbate 80, hence a greater probability of causing LER (see Section 8.3). The fact that
polysorbate 20 contains no unsaturated fatty acids and has a higher diffusivity compared to polysor-
bate 80, might be the explanation for this observation (45). Thus, the formation of mixed aggregates
between polysorbate 20 and LPS is more favorable and suggests why it has faster reaction kinetics than
polysorbate 80.
4.4.2.2 Proteins
Proteins, such as monoclonal antibodies, have a molecular weight of approximately 150 kDa, whereas
surfactants have molecular weights of approximately 1 kDa. Molecular weight, charge distribution,
and hydrophobicity of proteins is unique to each protein based on its amino acid sequence. Since
proteins exist as three-dimensional structures, only surface-exposed residues and motifs can interact
with the LPS, which makes prediction of the propensity for interaction with LPS challenging. In
addition, proteins can interact with each other, forming higher molecular weight oligomers or multi-

subunit complexes, which further limits the accessible surface of the protein. Cationic proteins (overall
positively charged) are prone to bind to LPS, as LPS possesses a negative charge via its phosphate
groups and acidic sugars, enabling ionic interactions with positively charged patches on a protein
(33). Hydrophobic patches on a protein allow interactions with the lipid A of LPS via Van-der-Waals
interactions. Additionally, negatively charged patches of the protein may form complexes with divalent
cations which, in turn, induce destabilization of salt bridges between LPS molecules. Taken together,
the mechanistic interactions between a protein and LPS are clearly more difficult to predict without
empirical data.
In cases of masking induced by the protein component, the reaction kinetics can be influenced by
several parameters. The protein concentration can affect the kinetics; the higher the protein concentra-
tion, the faster the masking kinetics has been observed (see Section 8.2). This can be relevant while
investigating masking effects where protein concentrations vary. In addition, proteins are commonly
formulated with buffer components and surfactants as discussed above, which have been shown to
cause LER. Surfactants and LPS are amphiphilic. Thus, surfactants can compete with LPS for binding
to the hydrophobic patches of a protein. As a result, the surfactant can prevent the LPS from bind-
ing to the protein, preventing masking in the LAL assay. Alternatively, the binding of surfactants to
protein can limit their interactions with LPS and cause masking. Therefore, in some cases, LER is
minimized by the presence of surfactant and protein components (see Section 8.9). Such contradic-
tory effects of surfactants, which are common components of a biopharmaceutical drug product, make
the predictability of the LER effect challenging.
4.4.3 Masking Susceptibility of Endotoxins

Although LPS share a common structure, they are notable for their heterogeneity of structure, which
can affect biological activity (46). This structural diversity explains why different LPS preparations
show variable recovery in hold-time studies (see Sections 8.3, 8.10). The previously proposed two-
step reaction mechanism, for example, is confined to those LPS fractions with particular salt bridges.
In other cases, the function of the divalent cation is replaced by charged substituents. Consequently,
the presence of complex-forming agents may not destabilize the LPS–LPS interactions for those LPS
preparations, and complex-forming agents like citrate are less likely to reduce the rigidity of LPS ag-
gregates. Likewise, surfactants are less likely to intercalate between LPS molecules. Thus, modified
LPS molecules containing a positively charged substituent would reduce susceptibility to LER. Such
substitutions depend on the species of bacteria and the specific growth conditions, of course. Sub-
stituents can be attached to the lipid A portion as well as to the core region of the saccharide fraction
(Figure 4.4.3-1) which, in turn, affects the structure of LPS. The level of substitution depends on
the individual source of endotoxin, meaning that the substituents (e.g., addition of positively charged
amino groups) may not necessarily be equally distributed. This explains the wide range of masking
susceptibilities of endotoxins from different sources (see Section 8.3).
Plotting the time-dependency of the LER effect as a function of temperature, an Arrhenius correlation
can be obtained that indicates the kinetic process driving the masking process. Considering again the
activation energy model (see Section 4.4.1), the substitution by charged substituents (e.g., aminoar-
abinoses) increases the energy necessary to change the aggregation state. Investigation of the masking
kinetics of endotoxins from different origins has shown that the amount of energy needed for masking
varies (31). As the energy input for hold-time studies is determined by process and test conditions,
as well as formulation, the masking capacity of a sample can only be evaluated if an endotoxin that is
susceptible to masking is used.
An endotoxin susceptible to LER is masked by a chelator, like citrate, in combination with a surfactant,
like polysorbate 20. The supramolecular structure of such an endotoxin is stabilized by divalent cations.
Endotoxin preparations from E. coli, which have been shown to be prone to LER, are known to be
primarily stabilized by divalent cations and, therefore, are the recommended spike for hold-time studies.

Figure 4.4.3-1 Covalent Modifications of Kdo2-lipid A in E. coli K-12 and Salmonella
4.5 Potential Aggregation States of Detectable and Non-detectable

Endotoxin
LPS monomers are not expected to exist in nature as LPS is composed of aggregates due to the hy-
drophobic interactions between the lipid A molecules (24). Additional ionic interactions are formed
between LPS molecules. Further, LPS molecules are a diverse mixture of molecular species. They dif-
fer not only in the lipid A and core region, but also differ widely in molecular weight (MW) due to
their heterogeneity in the O-antigen. Consequently, a broad distribution of the hydrophilic–lipophilic
balance (HLB) can result in diverse supramolecular states. This heterogeneity is predictable and well
known; for example, cubic, lamellar, and hexagonal structures can be formed (24). A change in the
aggregation state (supramolecular structure) is expected during the transition from detectable to non-
detectable.
Depending on the sample conditions (i.e., matrix components, ionic strength, polarity, pH, etc.), dif-
ferent equilibrium states of the endotoxin aggregate are favored. The aggregation state, in turn, impacts
the detectability. For example, a variety of buffers are commonly used during different manufacturing
steps (e.g., isolation, purification, capture, polishing, filtration, chemical modification, and crystalliza-
tion). Thus, there are situations where a new equilibrium state is established, a changed supramolecular
structure of endotoxin that impacts endotoxin detectability.
To reverse the LER effect, the equilibrium state of the supramolecular endotoxin must be changed
back to a detectable state. The equilibrium state of a sample can be influenced passively by several fac-
tors. Some factors influence the equilibrium state early in the production process, such as the change
of a buffer or the temperature, which might induce spontaneous change from non-detectable to de-
tectable endotoxin. In addition to passive manipulation in order to improve detection, the conditions
can be actively manipulated by sample treatments for testing. This process is called demasking. In such
a case, the sample conditions are adjusted so that the detectable state of endotoxin is favored again
(see Section 8.7). As sample conditions (pH, ionic strength, protein content, etc.) vary from sample
matrix to sample matrix, no universal demasking treatment is available. A sample treatment must be
developed and adjusted case by case. Some mitigation strategies for overcoming LER are described in
more detail in Section 5.0.

4.5.1 LER and Limited LPS Interaction with Factor C

To further understand the limited interactions of masked endotoxin with Factor C of the Limulus-
based assays, knowledge of structure–function relationships is required (47,48). Most likely, the su-
pramolecular structure of the endotoxin has been altered. The exact supramolecular structures of LPS
responsible for activation of Factor C currently are unknown, in part, due to the incredible heteroge-
neity and lack of analytical methods. However, when LPS is masked (i.e., not detected by LAL, not
recognized by Factor C), its core chemical structure is assumed to be still present in the sample.
Changes in the state of endotoxins, for example, from higher order to monomeric structure, are ex-
pected in the presence of substances acting as surfactants, like detergent or protein. Electron micros-
copy data showing disaggregation of LPS using the detergent deoxycholate supports this (49). Further,
Kobayashi, et al., have postulated that Triton X-100 interferes with Factor C activity in the presence of
LPS (50). It is also plausible that this non-ionic detergent could affect the supramolecular structure of
LPS. Moreover, surfactant concentrations used in regular formulations are likely present in molar ex-
cess relative to LPS in hold-time studies. In such a case, the lipid A portion of LPS would probably be
embedded into a micelle and not be solvent-accessible. The endotoxin may not be detectable in LAL
because the LPS is disaggregated and the lipid A moiety is unable to bind Factor C in the lysate, pos-
sibly due to steric reasons. Another possible alteration of the aggregation state could be that only the
surface LPS molecules of aggregates are replaced with the detergents. The “inside” LPS of aggregates
are not changed because divalent cations are probably not removed from the inside; it would be dif-
ficult to remove all divalent cations, even by electrodialysis, (51). A chelating agent is probably unable
to remove all divalent cations from LPS aggregates, as some of the cations may be maintained in the
inside core of the LPS aggregates. In such a case, LPS activity is decreased by the replacement of LPS
molecules on the surface of the LPS aggregates with detergent molecules. In some cases, low levels of
non-ionic surfactants have even been shown to prevent LER in samples spiked with RSE (see Section
8.9). In general, however, the ultimate reaction pathway from the presence of a multimeric LPS to the
interaction of an individual LPS molecule with the Factor C is still unknown.
4.5.2 Activity of Monomeric and Aggregated LPS

Resolution of the fundamental question of whether monomeric or aggregated LPS represents the
active state of LPS is challenging, as conflicting positions have been published. One argument pro-
poses the active form is represented by monomeric endotoxin, which interacts directly with biological
receptors (52–54). The opposing argument asserts that only supramolecular, higher-order endotoxin
represents the active form of endotoxin (22,24,55). This is also feasible, because LPS is an amphiphilic
molecule and mainly presents in its aggregated form.
Because the literature is inconsistent, it remains possible that the LAL-detectable or biologically ac-
tive state of LPS is due to either structure—monomeric or supramolecular; although, in an aqueous
environment, free, unbound monomeric LPS would be thermodynamically unlikely. In that case,
the molecular structure of monomers and the formulation would drive the supramolecular structure,
which would then define the potential activity. In this way, the accessibility of lipid A is determined by
the core structure and the aggregation state of LPS.
The aggregation state of LPS, therefore, plays an important role, and many characterization meth-
ods, including NMR, cryo-TEM, and DLS, have been applied (24,29,56,57). However, results are
difficult to interpret with regard to which structures are responsible for Factor C-binding and LAL
activity. Aggregation states are also concentration-dependent, and endotoxin concentrations used for
structural analysis are mainly in the micro-molar range. In contrast, LAL methods are substantially
more sensitive in the pico- to femto-molar range. Therefore, characterization methods most likely do
not reflect the actual LPS structures present in the LAL methods used for LER studies. Though chal-
lenging, knowing the active state of an endotoxin will be necessary for a more complete understanding
of endotoxin detection and the LER phenomenon.

4.6 Summary
Section 4.0 presented current knowledge regarding some of the mechanistic factors involved in LPS
under various conditions and has described how to differentiate the BET (LAL) test interference from
the LER phenomenon. Because dilutions of drug product (or other sample type) are spiked for PPC
recovery in the pharmacopeial BET, the supramolecular state of LPS is less likely to be altered than in
LER studies, where undiluted product is spiked, and material is held (typically for days) at specified
temperatures. Hence, the traditional method of diluting the product to achieve recovery (i.e., PPC) in
the BET may not be as effective to achieve recovery of the endotoxin spike in the undiluted product
in LER hold-time studies. A model proposed for masking that seeks to explain why combinations of
chelator and surfactant can cause LER may also be a plausible mechanism for LPS that have structures
stabilized by divalent cations. The RSE—and, likely, the CSE, which is often derived in the same man-
ner—is known to have these structures, which may be why, in some cases, it is more likely to exhibit
LER compared to endotoxins stabilized by other means. Endotoxin preparations derived from other
organisms, or prepared with alternative growth conditions (media, salts, etc.), have been shown to be
less susceptible to LER. The susceptibility of RSEs/CSEs to LER is one reason these two sources of
LPS are considered to provide the “worst case” conditions which result in LER. While Section 4.0
provides a basis of understanding for why certain endotoxins are more prone to LER in certain prod-
uct formulations, Section 5.0 describes how this knowledge can be applied as a potential means for
resolving LER.
5.0 Mitigation of LER

The LER phenomenon occurs when a formulated product fails to meet the recovery requirements in
hold-time studies, even when recovery has been achieved using the test for interfering factors described
in major pharmacopeias (2-4) where the diluted product is spiked. LER occurs when two consecutive
time points fail to achieve 50% or greater recovery in a hold-time study. Therefore, the most direct
mitigation may be to ensure a sufficient number of samples are included, such that a single time point,
from which recovery is less than 50% at the end of a study, can be evaluated by an additional time
point to determine if LER has occurred. Another method for direct mitigation may be to implement
in-process samples at relevant steps in the process and test for endotoxin at those steps. Simple, tradi-
tional approaches for overcoming inhibition or enhancement are widely published and may also prove
useful in resolving LER (e.g., when inhibitors are present). The mitigation efforts described in this
section are predicated on the assumptions that LER is caused by 1) the protein product itself binding
to or disrupting the physical (supramolecular) state of spiked endotoxin; and/or 2) a particular combi-
nation of excipients where a chelating buffer (e.g., citrate) and a surfactant (e.g., polysorbate) are both
present. As discussed in Section 4.0, a possible cause of LER is the alteration of the supramolecular
state of endotoxin, which leads to an inadequate detection, especially for those stabilized by divalent
cations (including CSE and RSE). Mitigation processes for reversing LER are therefore categorized as
1) sample treatments or assay changes to recover spiked endotoxin, and 2) additional, more complex
procedures to compensate for situations where, despite laboratory efforts, LER has still not been over-
come.
5.1 Mitigation through Sample Treatment

Treatments traditionally applied for resolving inhibition (when PPC recovery is < 50% in the test for
interfering factors) in the BET should be examined first (e.g., sample pH, avoiding known inhibitors,
adsorption to plasticware, inadequate mixing). Many LER issues can be resolved by following exist-
ing guidance for proper study design and by maintaining careful attention to sample handling and
typical assay issues as described in Section 3.0. In addition, understanding the potential influences of
the product itself in combination with the formulation buffers and temperature on LPS recovery is

essential. It is recommended that the following approaches be performed in sequence, so that the most
practical and straightforward of the potential solutions are studied first. If this approach is not success-
ful, a combination of the individual strategies may be considered.
5.1.1 Evaluation of the Product

Low endotoxin recovery due to product binding may be considered in cases where the structure of the
protein (active ingredient in the product) is substantially hydrophobic or electropositive (e.g., has re-
gions of positive charge that can bind to the negatively charged endotoxin). The isoelectric point (pI)
of a protein is not a reliable indicator of LER propensity, since all proteins are zwitterionic and net
charges, if known, that might be an issue are specific to regions on a given protein. For this reason, it
is not usually beneficial to adjust the pH of samples to other than the pH range recommended by the
LAL reagent manufacturer (typically pH 6-8 for the reaction mixture of the LAL reagent and the di-
luted sample) as required for the pharmacopeial BET test. However, obtaining information regarding
protein net charge and hydrophobicity is useful before embarking on the LER remediation steps be-
low. Section 4.0 suggests studies to separate the effects of product from formulation components. The
literature can be consulted for potential solutions when LPS binding to the product is suspected (33).
5.1.2 Addition of Dispersants to Samples

In some instances, the addition of dispersants to samples can resolve recovery issues in the BET; thus,
dispersants are potentially able to overcome LER. These reagents are available commercially (e.g., PY-
ROSPERSE™, Endo-RS®, CRL BD100®) and, although their compositions are proprietary, these reagents
are simple to evaluate experimentally using a variety of volumetric combinations, times, temperatures,
and mixing protocols to determine if they overcome LER. If successful, the use of proprietary dispersants
is relatively straightforward to add to the BET method validation by the QC product testing laboratory.
5.1.3 Addition of Excess Divalent Cations

In some cases, excess Mg+2 has resolved LER (58,59). Another approach that has exhibited limited
success is to lower the pH in samples containing citrate to weaken the chelation effect. Based on the
pKa(s) of the functional groups, phosphate buffers would not be expected to benefit by lowering pH.
Generally, situations known to be affected by chelating buffers in combination with surfactants may
benefit from knowledge of the buffer systems and their impact on LER.
5.1.4 Other Sample Treatments

For cases where highly lipophilic or positively charged LPS may be adsorbed to the product, recovery
can sometimes be improved by sample treatments other than pH and dispersants. For instance, in-
creasing the hydrophobicity of the solution with organic solvents (e.g., isopropanol) or adding other
common substances of various hydrophobicity (e.g., bovine serum albumin) in rare cases has been
shown to improve recovery. Proteolysis with proteinase K at optimal enzyme temperature incubation
has also been shown in very rare cases to improve recovery of spiked LPS (33). Additionally, while pro-
teins can be precipitated by any number of means, LPS can co-precipitate with the protein during the
treatment and not be recovered when associated in this manner. In all these scenarios, the lack of con-
taminating endotoxin in the added reagent needs to be ensured. These potential solutions to LER are
empirical; no single treatment is known to work for any specific category of products or formulations.
5.1.5 Evaluation of LAL Assay Reagents

For cases where sample treatments fail to mitigate LER, other commercial LAL technologies should be
evaluated. In some cases, proprietary lysate formulation components, reagents, and vendor materials
may solve recovery issues. Though empirical, this is potentially a quick solution. However, if a lysate or
standard other than that used for release-testing is needed to resolve LER, this may require additional
effort to qualify LAL reagents (see Section 8.9).

Commercial assay kits (e.g., kinetic/endpoint, turbidimetric/chromogenic, gel clot) from various es-
tablished vendors may also be considered; these kits should be evaluated early, though, especially if
the sample treatments in Section 5.1.4 are not effective in resolving LER. As with the BET test, using
only materials calibrated to RSE with appropriate CSE/lysate combinations is important. If use of a
different LAL reagent is the solution to LER, the same CSE/lysate combinations should be used for the
BET drug product release testing, which would require full validation with the new reagents.
5.1.6 Non-LAL Endotoxin Testing

In some cases, technologies other than LAL have been evaluated in LER studies. In one example,
samples exhibiting LER in hold-time studies were tested with recombinant Factor C (see Section 8.9).
Although recombinant Factor C did not improve recovery in this case, that option could be evaluated
along with the LAL methods (60).
5.2 Evaluation through a Biological System

If the mitigation efforts suggested in Section 5.1 have failed, the next phase of study may be to de-
termine if the spiked endotoxin is detected in a biological system. The compendial pyrogen test, with
samples that are spiked and held with endotoxin (Section 3.0), will produce two possible outcomes
after the LER hold-time treatment: pyrogenic or not pyrogenic.
When endotoxin-spiked or -held product samples are found to be pyrogenic, manufacturers should
implement the RPT as an interim QC release test until a suitable in vitro test method can be developed
and implemented. The development and implementation of a valid in vitro endotoxin test can occur
after BLA approval, as a post-marketing commitment. Given the insensitivity and the variability of the
RPT, firms are encouraged to implement an in vitro endotoxin test for the release of the drug product
in lieu of the RPT as soon as possible after BLA approval. Performing the endotoxin test could prove
especially beneficial in countries where animal use is strongly discouraged and contract facilities for
testing in rabbits are not available.
When endotoxin-spiked or -held product samples are not found to be pyrogenic, a pyrogen test for
product release may not be required. Firms may be asked to conduct additional in-process endotoxin
tests prior to the addition of the formulation component causing LER, if known. Endotoxin testing
of input raw materials is critical as those materials may contribute to the final endotoxin burden in the
finished product. Tight microbial control of all subsequent processing steps is very important to ensure
low endotoxin levels in the finished product.
Finally, a post-marketing commitment to evaluate LER mitigation strategies and to develop, validate,
and implement an in vitro endotoxin test not subject to LER may be recommended for the release of
the finished drug product after U.S. FDA approval. (One such example, the monocyte activation test,
is described in Ph. Eur. 2.6.30 and discussed in Section 5.2.3.)
Pyrogen testing in rabbits presents a significant challenge since the sensitivity of the test can vary great-
ly between studies depending on the source and quantity of LPS and the variability associated with
the rabbits. These differences, based on sample handling, source and dose level of endotoxin, genotype
and breeding of rabbit colonies, and animal handling, are to be expected. Section 5.3 below provides
designs and analysis protocols to make sense of pyrogen testing studies in rabbits, and suggests how to
evaluate alternative tests, such as the in vitro tests cited here in Section 5.2.
5.2.1 Endotoxin Dosage

The amount of endotoxin capable of eliciting a febrile response (failure to meet the requirements of a
compendial pyrogen test that has specific data processing algorithms for determining “pass” or “fail”
based on increases in body temperatures) must be determined for each preparation of endotoxin. The
most is known about RSE, a highly purified preparation. If other preparations of endotoxin are used,
they will need to be evaluated carefully in animals to define the pyrogenic dose. Although studies sug-

gest that endotoxin may be pyrogenic when administered at 5 EU/ kg within one hour, in practice,
this may fail to be pyrogenic (e.g., in rabbits) (61,62). Much higher levels (e.g., 40 EU/kg) may be
required to elicit a pyrogenic response in rabbits (63–65). Based on these findings, 35-40 EU/kg is rec-
ommended as the level of RSE to spike. Due to animal welfare concerns, gaining approval to conduct
extensive studies on other sources of endotoxin may not be possible. Therefore, the following guidance
uses RSE as the example of LPS under the assumption that at least 35 EU/kg is needed to reliably fail
the compendial pyrogen test. The difficulties associated with animal testing and the lack of sensitivity
of the RPT is why alternative in vitro tests are preferred, as is addressed throughout Section 5.2.
5.2.2 Product Dosage

Another major consideration in conducting animal or in vitro studies is the dose of the product tested.
For the traditional pyrogen test in rabbits with unspiked product, animals are administered a quantity
of drug equivalent to the maximum dose per kg of body weight of a human subject. For example, a
fixed dose of 1 mL of a 150 mg/mL drug product given subcutaneously to human subjects weighing
50-100 kg would be calculated as follows:
150 mg per 50 kg (Smallest Patient) = Max Exposure of 3 mg/kg
In compendial RPT methods, administration of very small-volume doses is neither recommended nor
practical, so dilution is required prior to treatment. This is complicated by the fact that LER samples
are expected to be spiked “neat,” that is, undiluted. A guide for conducting the LER hold-time and
preparation of dosing solutions for pyrogen test in rabbits, based on an assumed human dose of 3 mg/
kg active pharmaceutical ingredient containing a targeted 35 EU/kg of endotoxin, and matching per
kg human dosing, is as follows:
1. Use protein sample of 150 mg/mL
2. Reconstitute RSE (10,000 EU) with 1 mL LRW to achieve 10,000 EU/mL
3. Combine and mix 0.876 mL (876 μL) of stock RSE with 5 mL of protein sample
a. (10,000 EU/mL) * (0.876 mL/[0.876+5.0 mL]) = 1491 EU/mL
b. (150 mg/mL) * (5.0 mL/[5.0 mL+0.876mL]) = 128 mg/mL
4. If sample is to be held three (3) days, test these samples immediately (T0) and after hold time
a. Dilute 0.78 mL of spiked LRW or sample in 99 mL of PBS
b. Protein concentration = (128 mg/mL) / (0.78 mL/[0.78 mL + 99 mL]) = 1.0 mg/mL
c. RSE concentration = (1491 EU/mL) / (0.78 mL/[0.78 mL + 99 mL]) = 11.6 EU/mL
d. For a 3.0 kg rabbit, administer 9 mL of the above (4a-4d)
e. Check
i. 1.0 mg/mL protein * 9 mL = 9 mg protein
• 9 mg/3 kg = 3 mg/kg protein dose
ii. 11.6 EU/mL RSE * 9 mL = 105 EU

• 105 EU/3 kg = 35 EU/kg RSE dose
f. Based on individual rabbit weights, administer 3 mL/kg for the RPT
Before conducting LER studies in animals, ensure that the unspiked sample in the RPT is not inher-
ently pyrogenic. The same should be done before conducting the MAT design described in Section
5.2.3. Data analysis for both the RPT and MAT LER studies is discussed in Section 5.3.

5.2.3 Monocyte Activation Test

The monocyte activation test (MAT), which also requires product-dosing knowledge and careful at-
tention to study design, has been proposed as an in vitro alternative to the pyrogen test in rabbits. The
basic approach and design described in Ph. Eur. 2.6.30 is an important first step in choosing which
particular protocol is relevant. While some variation may be needed during an LER remediation study,
the development and controls recommended in Ph. Eur. 2.6.30 are, for the most part, an excellent
guide for MAT applications. The example provided for the RPT study design above (Section 5.2.2)
can also be used to illustrate the LER application. The exact source of the endotoxin to be used—hepa-
rinized human whole blood or peripheral blood monocyte preparations from human blood—should
be determined and employed to screen donors using a series of dilutions of endotoxin in LRW.
5.2.3.1 Donor Selection and Pooling

Human donors show a high degree of variability to the same preparation of endotoxin and different
sensitivities to other sources of endotoxin. For this reason, a pre-selected donor source (or pool of do-
nors) of sufficient quantity is recommended to allow development and testing in the LER evaluation.
Typically, a solution of about 1 EU/mL, when diluted 10X (e.g., 20 μL into 180 μL of blood per well
in a microtiter plate) should give a response of, at least, an order of magnitude above background.
Detectability is sometimes defined as three times the background controls, which can vary depending
on donor, so should be evaluated before testing. A time-course measuring expression of multiple cyto-
kines is needed to select the conditions and times required to produce pro-inflammatory cytokine lev-
els (typically IL-1β, IL-6 or IL-8) that are at least three times above background and have not reached
a maximum response within a defined time frame (previously determined in a time-course study).
Having results within range of the cytokine standard curve is important for monitoring recovery in the
LER study. The LER study can be designed once these conditions are demonstrated and reproducible
results can be obtained.
5.2.3.2 Sample Preparation

The sample in the product dosage example (1491 EU/mL) would be much too high to be useful for
the MAT so, in this case, a significant dilution would be required prior to adding to cells to achieve
the desired target endotoxin assay concentration. Following the RPT example in Section 5.2, a sample
that contains 1491 EU/mL, can be diluted by adding 68 μL into 10 mL of diluent to give a concen-
tration of 10.1 EU/mL, which is very close to the target of ~1 EU/mL when further diluted 10X in
blood or peripheral blood mononuclear cells (PBMCs). LER can be monitored in MAT by directly
comparing the induced cytokine in LRW to that spiked into neat product. Thus, the recovery of 50%
to 200% against the spiked LRW control is similar to that for the initial BET hold-time LER study.
5.2.3.3 Cell Line Selection

Cultured, stable cell lines, e.g., HEK-293/TLR4 that overexpress TLR4, may produce more consistent
results in the MAT than PBMCs or heparinized whole blood. The method development required
when using a cell line is similar to that described in Section 5.2.3.1. Since the donor variability is re-
moved, optimizing the dose levels and assay conditions with the identical endotoxin source to be used
for the LER study are the main assay development activities. Most commercially available cells include
the molecular components necessary to bind endotoxin (e.g., TLR4, MD2, and CD14), transduce the
signal, and generate a response; this should be considered when selecting the cell line. Ideally, cells that
express only one TLR should be used, as endogenous levels of other TLRs may cause a lack of specific-
ity for LPS (e.g., TLR5 is activated by flagellin). As with primary cells (e.g., whole blood or PBMCs),
the criteria confirming a lack of LER is 50% to 200% measured against an LRW control (66). These
cell lines often require antibiotics to maintain selective pressure to express the desired TLR.

5.3 Interpretation of Results

The two options presented (USP <151> Pyrogen Test and MAT) may be applied to samples treated
exactly as those for the LER hold-time studies where LER was observed. The exception is the spike
levels, which may need adjustment to account for animal-dosing in the RPT test. Otherwise, the RPT
data is interpreted as “pass” or “fail” based on the limits in the compendia. Regarding responses in the
MAT or with cell lines, LER is confirmed if recovery is outside the 50% to 200% range. In summary,
samples where spiked endotoxin in product passes as opposed to identically prepared spikes in LRW
fail is interpreted as a lack of concordance with LAL tests.
5.4 Summary
The goal of mitigation is to manage situations where LER is observed by BET in hold-time studies in a
manner that minimizes the risk of undetected endotoxin in drug products. In cases where the problem
has been thoroughly investigated and the LER problem persists, the recommended course of action is
to thoroughly test all materials (by BET) up to the point at which the components responsible for the
LER are added into the manufacturing process, as described in Section 5.2.
6.0 History and Rationale for LER Studies

The results of the first PDA LER workshop in March of 2016 in San Antonio, Texas, included several
agreed-upon recommendations which were published in the PDA Letter in September 2016: “PDA’s
LER Task Force Holds its First Workshop” (67).
Based on additional information obtained during task force discussions and related scientific findings,
this technical report addresses concerns regarding the suitability of the LAL assay to detect endotoxin
in certain formulations. This section describes the incorporation of the LAL test into U.S. regulations,
appropriate reference standards used for the LAL assay, microbial control in manufacturing processes,
and the impact of endotoxins on human health.
6.1 Regulatory Framework

U.S. FDA regulations require that each batch of drug product purporting to be sterile and/or pyrogen-free
be tested for sterility and pyrogens (21 CFR 211.167). These and other requirements, specifically 21 CFR
211.160(b), are intended to ensure that the finished products meet all scientifically sound and appropriate
quality and purity standards (68). Specific regulations for biologics (21 CFR 610.13(b)) state that “each lot
of final containers of any product intended for use by injection shall be tested for pyrogenic substances by
intravenous injection into rabbits….” (69). Biological drug products must be tested initially using either
USP <151>–Pyrogen Test (compendial) or 21 CFR 613(b)–Purity (regulatory) methods to verify that the
drug product molecule itself is not pyrogenic; this test is not required for toxins or toxoids, per 21 CFR
610.13(b). Once the molecule is shown to be nonpyrogenic, the USP BET methods, which are more sensi-
tive and more reliable, can be used in lieu of the USP <151> Pyrogen Test to release a biological product.
The harmonized U. S. Pharmacopeia (USP <85>), European Pharmacopoeia (Ph. Eur. 2.6.14), WHO
(International Pharmacopeia Chapter 3.4) and Japanese Pharmacopoeia (JP 4.01) compendial meth-
ods are often subject to test interferences (2–6). Test interferences are commonly mitigated through
sample dilution, up to the maximum valid dilution (MVD), or by other validated means of sample
preparation. When those methods cannot easily mitigate interferences, alternative methods in accord
with 21 CFR 610.9 can be developed and used (70). This regulation requires applicants to provide evi-
dence that the quality element of the new or modified standard methods (regulatory or compendial) is
equivalent to the standard method(s). This is consistent with section 6.30 of the USP General Notices
that enables the use of an alternative method or procedure, given that it must be fully validated and

must produce comparable results to the compendial method or procedure within allowable limits established
on a case-by-case basis (71). In cases where the modified or equivalent methods do not mitigate test
interferences, the rabbit pyrogen test or an equivalent test should be used (12).
In establishing final drug product specifications for endotoxin, the limit should be below the limit
determined to produce a pyrogenic response in humans (e.g., intravenous administration of 5 EU/
kg/hour or intrathecal administration of 0.2 EU/kg/hour). In addition, in establishing a final endo-
toxin specification, the maximum theoretical contribution from the primary packaging components,
formulation excipients, water-for-injection (WFI), and bulk drug substance should be considered to
ensure that theoretical endotoxin limits are not exceeded (72,73).
6.2 Microbial Control During Manufacturing and Product Quality

Most therapeutic protein products are derived from either natural or recombinant microbial or mam-
malian cell expression systems. These products and process intermediates are produced using complex,
multistep processes that are at risk from microbial contamination, because the raw materials, personnel,
and manufacturing environment are potential sources of bioburden, endotoxin, and other adventitious
agents. Contaminating microorganisms in the process and product intermediates may be capable of pro-
liferating and producing proteases and other by-products that can affect the safety, purity, and potency of
the finished drug product. For example, microbial proteases can clip and degrade a protein product dur-
ing production and shelf life, affecting product purity, immunogenicity, and potency. Other microbial
by-products, such as endotoxin and other pathogen-associated molecular patterns (PAMPs), can impact
safety (74). The manufacturing process must be under microbial control throughout the various steps to
ensure the final product meets established quality standards of safety, purity, and potency.
A holistic microbial control strategy is critical to manufacturing safe and efficacious products. A com-
prehensive strategic approach consists of conducting operations in facilities designed and maintained
to minimize contamination and cross-contamination risks. These facilities may contain classified areas
supplied with HEPA-filtered air, pressure differentials, and air exchanges that minimize contamina-
tion risks. The facilities should be appropriately cleaned and environmentally monitored to ensure
microbial control. Equipment should be designed and maintained to minimize microbial risks when
operated as closed systems and cleaned, sanitized, and/or sterilized in place. Additional controls should
be in place for operations that expose the product to the environment. Such controls may include
the use of local laminar flow cabinets, with personnel appropriately gowned, and frequent filtration
of process intermediates with 0.2 µm filters to remove incidental microbial contaminants. Processes
should be monitored for bioburden and endotoxin at the points in the process most vulnerable to
microbial contamination and monitoring results documented to ensure the state of microbial control.
Open, manually intensive operations or hold-times longer than 24 hours may be particularly at risk for
microbial contamination and should be very closely monitored and controlled. The presence of micro-
organisms can be detected and quantified using a bioburden test, which should provide information
related to the number and type of microorganism and the possible impact to the process and product.
In some cases, firms have used bioburden test results as a surrogate test for toxins and non-endotoxin
PAMPs on the final drug product (75). The use of alternative and rapid bioburden methods could be
very useful to monitor and control the process in real time. However, these methods have not yet been
implemented widely. In contrast, in-process testing of endotoxin can provide more timely information
about the state of microbial control as this test can be performed in under two hours. Finally, the en-
dotoxin test provides additional assurances that the total endotoxin load of the final product is below
the pyrogenic threshold.
6.3 Endotoxin Structure

Endotoxin, or lipopolysaccharide (LPS), is the major constituent of the outer membrane of Gram-negative
bacteria; it is composed of three domains—the O-antigen polysaccharide, core oligosaccharide, and lipid A
(76). Lipid A is the bioactive portion of LPS (endotoxin) that activates the innate immune system via toll-like

receptor 4/myeloid differentiation factor 2 (TLR4/MD2), which is expressed by various cell types including
monocytes, macrophages, and platelets (77,78). Additionally, lipid A is responsible for anchoring LPS within
the outer membrane of Gram-negative bacteria and is the conserved molecular component of LPS.
Lipid A biosynthesis is a well-conserved and ordered process among most Gram-negative bacteria;
however, some Gram-negative organisms lack this biosynthetic pathway (76,79,80). The laboratory
of Christian Raetz characterized and elucidated the E. coli lipid A biosynthetic pathway, which consists
of nine enzymatic steps. Ultimately, the endotoxin biosynthetic pathway in E. coli typically results in
a lipid A structure that has six fatty acyl chains linked to a sugar backbone in an asymmetrical pattern,
with two unmodified phosphate groups (76,80,81).
Many Gram-negative bacteria also possess enzymatic machinery which modifies their lipid A struc-
ture, resulting in prolific lipid A structural diversity, depending on the given organism. Addition-
ally, the structural heterogeneity of lipid A has been shown to depend on the growth and environmen-
tal conditions of the Gram-negative organism (e.g., pH, temperature, divalent cation concentration,
etc.). These modifications most often involve variations in acyl chain number, acyl chain length, and/or
modification of the phosphate groups (79). The hexa-acylated lipid A structure of E. coli (Figure 6.3-1)
is a known potent immune-modulator, and structural data have demonstrated that specific portions of the
lipid A moiety interact with TLR4/MD-2, escalating the innate immune response (77,82,83). Not all en-
dotoxin structures are capable of eliciting such an immunologic response. For example, the Gram-negative
organism Helicobacter pylori possesses an endotoxin structure with only four acyl chains (84,85). Pertinently,
the tetra-acylated lipid A structure of H. pylori has been shown to be up to 1,000 times less potent than the
lipid A of E. coli (86–89). Additionally, work by Loppnow et al., (see Section 6.7) demonstrated that In-
terleukin-1 (IL-1) production varies depending upon the structure of LPS (Figure 6.3-2) (90). Ultimately,
the diversity and heterogeneity of lipid A species among Gram-negative bacteria emphasize the importance
of using a well-characterized standard for the endotoxin testing of drug products, e.g., reference standard
endotoxin (RSE) isolated from E. coli. In addition, the lipid A structural diversity for various Gram-negative
organisms frequently identified in manufacturing environments has not been characterized.
6.4 Limulus Amebocyte Lysate (LAL) Assay

In 1956, Bang observed that infections with Gram-negative bacteria in Limulus polyphemus (horseshoe
crab) led to the coagulation of the hemolymph (blood) (91). Subsequently, Levin and Bang demon-
strated that the activation of coagulation was a direct and major response to bacterial endotoxin, the chief
component of the outer leaflet of the outer membrane of Gram-negative bacteria (92–94). The entire
coagulation mechanism is contained in the amebocytes, the only type of cell in the circulation of Limulus
Figure 6.3-1 E. coli K-12 Kdo-Lipid A Structure (courtesy of M. Stephen Trent)

Figure 6.3-2 Degree of Lipid A Acylation Impacts IL-I Response

(modified with permission by U. Zähringer)
(93). Additional efforts led to the preparation of a lysate of washed Limulus amebocytes that could be
used to assay the presence of minute concentrations of bacterial endotoxins (94). Observations that the
coagulation of amebocyte lysate by endotoxin was dependent on temperature and pH and was sensitive
to serine and cysteine protease inhibitors indicated that this coagulation system was enzymatic in nature
(95). Consequently, clinical use of the LAL assay to diagnose septicemia was evaluated (96,97).
Since LAL is a blood product, the U.S. FDA decided to write regulations pertaining to its prepara-
tion and use (98,99). On January 12, 1973, the Agency announced its intention to license LAL and
declared that the LAL assay could not be used as a final pyrogen test for drugs administered to humans
due to questions concerning lysate variability (100,101). Nevertheless, the test could be used for the
detection of endotoxin in drug manufacturing with in-process sampling for endotoxin. Additional
standards for the licensure of LAL were established on September 18, 1973, deeming that LAL could
not be licensed unless it had been tested for sensitivity against the reference endotoxin; specifically, the
lysate should produce a positive reaction to 0.1 ng of reference endotoxin and a negative reaction in
response to 0.0125 ng of reference endotoxin. On November 13, 1974, the Bureau of Biologics (now
the Center for Biologics Evaluation and Research or CBER) recognized that bacterial endotoxins were
unavoidable contaminants of the inactivated influenza vaccine. Therefore, a requirement was added
to 21 CFR 610.11 that each influenza vaccine needed to be assayed for endotoxin against a reference
standard preparation provided by the FDA (102). Although the LAL method was not specifically re-
quired, most manufacturers used LAL testing to assay for endotoxin concentration (103).
Following a comprehensive review of the manufacturing process, including inspection of the manu-
facturing facilities and practices, licensure of the first lysates was granted in 1977. In accordance with
licensure, every lot of lysate was tested to verify the claimed sensitivity (98). As production techniques
were standardized and improved, on November 4, 1977, the FDA announced the licensing of LAL
as an alternative for the rabbit pyrogen test (RPT) in biological products and medical devices (104).
A draft guidance concerning the FDA’s position on use of LAL in lieu of the RPT for drugs (not
biologicals) was made available on January 18, 1980 (105); another draft was released on March 29,
1983, that also included medical devices and biologics (106). In December 1987, the final guidance
was published to convey acceptable conditions for using the LAL test as an end-product endotoxin
test (in lieu of the RPT) for human and animal injectable drugs, biologics, and medical devices (107).
The 1987 guidance was withdrawn on June 2011 and replaced in June 2012 with the Guidance for
Industry: Pyrogen and Endotoxins Testing: Questions and Answers (12).

The active FDA-licensed manufacturers of LAL are1:

• Associates of Cape Cod, Inc. (License #0700)
• Baxter Healthcare Corporation (License #0140)
• Charles River Endosafe, a Division of Charles River Laboratories, Inc. (License #1197)
• Lonza Walkerville, Inc. (License #1775)
• Wako Chemicals USA, Inc. (License #1762)
6.5 Endotoxin Reference Standards

The difficulties and peculiarities of preparing and defining biological reference standards were understood
as early as a World Health Organization (WHO) meeting in 1950. In the early 1950s, desiccated bacteria
(usually Shigella dysenteriae) were used to standardize and control RPT. After four years of debate and
discussion, the WHO established that pyrogen reference standards would come in the form of purified
endotoxin (108). In doing so, the Organization began a trend of universalizing, simplifying, and stan-
dardizing biological reference standards based on purified, characterized macromolecules from a single
source, which is reflected in the way the three International Standards have been developed since 1989.
Since Limulus-based approaches are ultimately biological assays, the lysates were intrinsically variable.
Consequently, an effort to generate robust, reliable standard endotoxins was begun to quantify the sen-
sitivity of each lot of LAL. Early contributors in these discussions pointed out the ideal characteristics
of an RSE should be:
1. Well-known and pyrogenic in nature 4. Stable over a number of years
2. Well-characterized 5. Produced in sufficient quantity
3. Capable of producing a positive febrile 6. Reliable and reproducible
response in rabbits (for RPT correlation)
Thus, an RSE would be homogeneous in preparation, readily resuspended, and not easily absorbed to
containers or vessels used in dilution (109).
6.5.1 U.S. FDA Development of Reference Standard Endotoxins

The U.S. FDA began its efforts to create a reliable reference standard in 1972 where, in the laboratory
of Dr. D.B. Castor of the Bureau of Drugs, one lot of Klebsiella pneumoniae endotoxin (Lot 1b) was
freeze-dried (in 0.1% human serum albumin) and used for several years. Variations in pyrogen test
results with Lot 1b were observed among various laboratories. Although there was no direct evidence
that the endotoxin preparation itself was the cause of the variability (more likely, the variability was
due to the pyrogen test using rabbits), when it became necessary to replace Lot 1b, the decision was
made to use the more scientifically understood, Westphal-extracted LPS from E. coli O113:H10:K(-)
as an RSE, using the strain from Dr. J.A. Rudbach of the University of Montana (110,111). The re-
sultant lots were prepared by the FDA and designated as follows: EC-1 for the 1975 E. coli production
(200 vial-scale production), and EC-2 for the larger scale (1,500 vials) successor in 1976. At the urging
of endotoxin experts, endotoxin units (EU) were assigned to EC-2 such that 1.0 EU = 0.2 ng EC-2.
Thereby, future endotoxin preparations with different potencies could be directly correlated without
reliance on mass. Because “EU” is currently used to designate the sensitivity of LAL to endotoxin,
the actual choice of the source for endotoxin is not important, if the proposed endotoxin standard
is chemically purified and well-characterized, and its potency correlates to an RSE. FDA uses E. coli
(EC) endotoxin, which is distributed by the U.S. Pharmacopeial Convention (USP, see section below).
1
At the time this technical report was published.

Because human serum albumin (HSA) is known to bind purified LPS, the FDA attempted to produce its subsequent reference standards (EC-3 and EC-4) without
HSA, but these lots lacked sufficient potency (58). FDA then contracted Mallinckrodt Pharmaceuticals to prepare a large batch of RSE to replace EC-2, without
human serum albumin. Instead, lactose and polyethylene glycol were used as stabilizers in the final preparation. Thirty-thousand vials of this lot were produced: half
were given to FDA CBER (who refers to the standard as EC-5) and the other half given to the USP (who refers to it as Lot F, see section below). The current RSE
preparation (EC-6), developed at the National Institute for Biological Standards and Control in London, includes the same stabilizers as EC-5; it is distributed by
USP, JP, and the European Directorate for the Quality of Medicines (EDQM) as Lot G (Table 6.5.1-1 below).
Table 6.5.1-1 U.S. Reference Standard Endotoxins development history
U.S. Reference Standard Endotoxins

Name Organism Source Year produced Fillers/Stabilizers Prepared by Preparation notes
Klebsiella pnuemoniae DB Castor, U.S. FDA 0.1% human 31 g produced
1b 1972 U.S. FDA
ATCC 13883 Bureau of Biologics serum albumin lyophilized
Lactose and 200 vials produced (small pilot project)
Escherichia coli JA Rudbach,
EC-1 1975 polyethylene U.S. FDA lyophilized
0113:H10:K(-) University of Montana
glycol small pilot batch
1500 vials produced
Lactose and 1,000 µg/vial
EC-2 1976 polyethylene U.S. FDA lyophilized
glycol EU assigned as 0.2 ng EC-2 = 1 EU
5,000 EU/vial
EC-3 1980 none U.S. FDA Lacked sufficient potency; not used
EC-4 1980 none U.S. FDA Lacked sufficient potency; not used
30,000 vials produced
EC-5 Lactose and 10,000 EU/vial
(USP, 1981 polyethylene Mallinckrodt lyophilized
“lot F”) glycol Used in 1 and 2 international collaborative study
st nd
0.12 ng EC-6 = 1 EU = 1 IU
National Institute 75,000 vials produced
EC-6 1994 Lactose and
Escherichia coli JA Rudbach, for Biological 10,000 EU/vial
(USP, (implemented in polyethylene
0113:H10:K(-) University of Montana Standards and Lyophilized
“lot G”) 1997) glycol
Control (UK) Used in 2nd international collaborative study

6.5.2 Harmonization of International Reference Standard Endotoxins

The large number of different endotoxin standard preparations in use, especially internationally, create
challenges and confusion in the pharmaceutical industry. Considering the importance of endotoxin
testing as a key safety measure, globally harmonized reference materials are necessary in order to re-
duce the level of variability in a key test parameter. To this end, purified bulk material from strain
0113:H10:K(-) was donated to the USP by the U.S. FDA and has been used as the starting material
for the WHO International Standards for Endotoxin (1st International Standard in 1989, 2nd Interna-
tional Standard in 1996, and 3rd International Standard in 2012), as well as the Ph. Eur. BRP prepara-
tions 3 and 4, and the USP Reference Standards from Lot F onwards. The WHO International Stan-
dards are commissioned and approved by the WHO Expert Committee for Biological Standardization
(ECBS) based large international studies to establish an agreed potency value. The first IS contained
2 mcg of endotoxin (same bulk as for USP/FDA STD EC5) + 3 mg trehalose/ampoule. The 2nd and
3rd IS contained 1.2 mcg endotoxin/vial formulated with lactose and polyethylene glycol. The ECBS
calibrated the 1st IS based only on the gelation data and made it a standard only for gelation. It was
only with the second IS that there was an opportunity to attempt to bring the WHO, USP/FDA and
EDQM standards into better harmonization. Since the 2nd IS, both the European and US Pharma-
copeial standards originate from the same formulation and fill as the IS, and their potency has been
established in the same study. All materials have been formulated and filled at the National Institute
for Biological standards and Control (NIBSC) in the UK. Since the 2nd IS of 1996, unit equivalence
of 1IU = 1EU has been established (Table 6.5.2-1) (64,112,113).
The Japanese National Reference endotoxin has historically been based on material from different
bacterial strains and not harmonized with the materials described above. More recently, there has been
the desire to establish a material highly comparable to the existing International, European, and U.S.
reference standards.
Table 6.5.2-1 Endotoxin standards and their assigned potency
Standard Year Established/NIBSC WHO ref Potency Assigned

1 IS
st
1989, 84/650 14000 IU/vial
2 IS
nd
1996, 94/580 10000 IU/vial
3 IS
rd
2012, 10/178 10000 IU/vial
While this technical report focuses on endotoxin standards used for the testing of parenteral drugs,
the medical devices community uses an E. coli O55:B5 endotoxin (previously available from Difco
Laboratories) as an endotoxin standard. The Health Industry Manufacturing Association (HIMA)
has advocated for the adoption of the O55:B5 standard. Fortunately, a major collaborative study has
shown that the O55:B5 endotoxin preparation is as potent as the O113:H10:K(-) preparation when
tested with LAL (114). Today, the O55:B5 preparation is commonly used as a control standard en-
dotoxin (CSE).
Because the common means of detecting pyrogens (i.e., RPT) and endotoxins (i.e., LAL) are biological
in nature, they are subject to variability (laboratory-to-laboratory and lot-to-lot). Accordingly, consider-
able scientific discussion spanning four decades concluded that a purified, potent, and well-character-
ized LPS is the best analyte for the RSE. Since the availabiltiy of RSEs, which are a primary standard, are
a limited, they are not intended for day-to-day use in the pharmaceutical industry. Instead, CSEs, which
are secondary standards, are generally used in the daily operation of drug manufacturing. These CSEs
must be calibrated to the activity of the limited supply of RSE. The official compendia typically do not
provide guidance for the establishment of secondary or working standards due to the legal recognition
and official status associated with these standards that obligate the use of a pharmacopeial standard with
a referee procedure in the case of a dispute. Given the absence of this guidance, CSEs should adhere to
the same major requirements outlined for RSEs; namely that they are robust, reliable, stable, and from
a defined strain. These attributes are the culmination of nearly a century of experience with a variety of

preparations as well as biological and biochemical research into the structure and biochemical interac-
tions of LPS with the human innate immune system. Because the source of possible contamination in
drug manufacturing is impossible to know a priori, use of an endotoxin reference that would identify
the LER phenomenon is crucial. Purified LPS from E. coli has been identified as susceptible to LER (30;
41). At the time of publication, production of EC-7 has been scheduled and will be based on the same
source of bulk material, E. coli O113:H10:K(-), as EC-6.
6.6 Natural Occurring Endotoxins (NOEs)

The newly coined term “naturally occurring endotoxin” (NOE) refers to Gram-negative bacterial en-
dotoxin preparations using defined conditions with minimal nonchemical processing (e.g., centrifu-
gation and filtration). Endotoxins prepared in this manner are intended to reflect endotoxins as they
occur in the manufacturing environment. Schwarz and colleagues also reported that both NOEs and
chemically purified LPS endotoxin standards can be affected by LER (115). Refer to Section 5.0 of
the technical report for information regarding the mechanism of LER.
The susceptibility of endotoxin to undergo masking may be due, in part, to the structure of the lipid
A domain because this drives the thermodynamics of assembly of its biophysical state. As discussed in
Section 3.3, various Gram-negative organisms modify their lipid A structure depending upon their
growth and/or environmental conditions. Work described in Case Study #3, “Use of Purified and
Non-purified Endotoxins in Hold-time Studies” (Section 8.3), has shown that masking susceptibil-
ity depends on the bacterial species and that culture conditions do affect the masking susceptibility
of endotoxins. Additionally, this work demonstrated that isolation methods do not alter the primary
structure of the lipid A profiles (i.e., lipid A profiles from NOEs are similar to their highly purified
counterparts); however, the supramolecular structures of endotoxin may be influenced by the prepara-
tion method.
Since the methodologies and organisms used to generate NOEs may differ from laboratory to labo-
ratory, use of NOEs may further complicate the variability, which can be observed in the BET. For
example, Bowers and Tran used 0.45µm filtration to generate NOEs from various organisms (116).
Other laboratories generate NOEs by using alternative filtration methods or by autoclaving or other
heat treatment. The variety of techniques and endotoxins that may be chosen by any given laboratory
is reminiscent of the early days of LAL testing. For instance, E. coli endotoxin purchased from Difco
Laboratories (Detroit, MI) and crude Vibrio endotoxin were used by Levin and coworkers to initially
describe endotoxin-induced coagulation of the blood of Limulus and of Limulus amebocyte lysate,
though, these studies were performed prior to the establishment of the endotoxin unit (EU) (94). Be-
cause growth conditions also play an important role in endotoxin heterogeneity among various Gram-
negative bacterial species, using a characterized standard such as an RSE or calibrated CSE to perform
the BET eliminates the introduction of additional variations into the test method.
6.7 Immune Response

Bacterial endotoxins can produce a wide range of significant pathophysiological effects with implica-
tions for patient safety, e.g., blood coagulation and activation of complement and kallikrein/kinin
system cascades. Therefore, the biological effects of endotoxin need to be considered in addition to
its pyrogenicity related to recent advances regarding increased understanding of how endotoxins may
activate the immune system.
Based on the crystal structure, a monomer of lipopolysaccharide binds to MD-2. The LPS-MD-2 com-
plex then binds to the TLR4 receptor dimer (Figure 6.7-1) (82). Formation of this receptor-ligand
complex leads to the activation of signaling cascades that result in the production of pro-inflammatory
cytokines. These chemokines recruit immune cells to the location where the innate immune system
has been activated, inducing inflammation (117–121). Activated macrophages and neutrophils mi-
grate to the site of inflammation to kill the invading pathogens in a controlled fashion and end the
infection. However, overactivation of this pathway can lead to severe, if not fatal, consequences.
Figure 6.7-1 Structural Basis of Lipopolysaccharide Recognition by TLR4-MD-2 Complex (LPS = green; TLR4 = orange;
MD-2 = blue) Figured modified from original.
Small amounts of endotoxin can lead to an immune reaction; therefore, ensuring that manufactur-
ing processes minimize potential endotoxin contamination and, by proxy, bacterial contamination, is
imperative. There are multiple pathogen-associated molecules (PAMPs), such as flagellin, porins, and
nucleic acid, that can induce an immune response. In the presence of a pathogen or PAMP, the im-
mune system uses a robust approach to defend against infection. PAMPs bind receptors that, in turn,
activate the innate immune system, critical for protection against invading pathogens (Figure 6.7-2)
(122). The primary inflammatory response can be localized or systemic and can provide a beneficial
defense against pathogens. In addition to its function as a primary and rapid line of defense against
Gram-negative bacteria, activation of the TLR4 pathway by endotoxin can trigger adaptive immunity.
Thus, molecular interactions between endotoxin and the human proteins MD2 and TLR4 can initiate
both the innate and the adaptive systems.
When elevated concentrations of LPS are detected in the circulation, the TLR4-mediated innate im-
mune-signaling cascade is activated in epithelial and immune cells. The activated cells, in turn, pro-
duce pro-inflammatory chemokines and cytokines, such as IL-1β, IL-6, IL-18, and TNF-α. Studies
using mouse models have used doses ranging from 1 mg/kg to 25 mg/kg to replicate the physiologic
effects of high concentrations of endotoxin in the blood of humans or severe infections (123). In
contrast, studies in which human volunteers were intravenously administered doses of RSE (2–20
EU/kg) have shown that the response to a low level of endotoxin exposure is not as severe (124–127).
However, administration of endotoxin in this range has been shown to induce fever, malaise, increased
cytokine levels in the blood, and altered cardiac function (68–72).
Under normal conditions, gut flora produces LPS which can potentially enter the systemic circulation
due to gut permeability. LPS that enters the portal circulation from the bowel, however, is detoxified
by the liver (73–76). Levels of endotoxin in circulating blood range from 0-10 pg/mL as measured in
heat-treated plasma anticoagulated with endotoxin-free heparin (128,129). Levels of endotoxin range
from 0 to 12.4 pg/ml in heparinized heat-treated “serum,” and endotoxin levels are reported to range
from 0 to 0.2 ng/mL in plasma anticoagulated with EDTA (130,131). Accidental parenteral delivery

Figure 6.7-2 TLR Recognition of Microbial Components, adapted from Liu, et al.
of endotoxin circumvents this critical hepatic safeguard and can have multiple physiological effects.
In 1998, an outbreak of endotoxin-like reactions occurred in patients given intravenous gentamicin.
Within three hours of administration, patients developed symptoms that strongly suggested endotoxin
exposure, i.e., fever, shaking chills, increased heart rate, and decrease in blood pressure. Subsequent ep-
isodes occurred across the U.S. and all were linked to a single manufacturer (132). Tests revealed that
the gentamicin was positive for endotoxin by LAL. Although the concentration of endotoxin in the
gentamicin was below the USP limit, the off-label dosing resulted in parenteral delivery of endotoxin
above the USP limit of 5 EU/kg/hr. Because the endotoxin concentration was critical to understand
this adverse event, the inability to accurately measure endotoxin concentration as caused by LER can
place patients at substantial risk.
In addition to being a potent activator of innate immunity, endotoxins can stimulate adaptive immu-
nity (133). Cells involved in adaptive immunity (monocytes, dendritic cells, B-cells, T-cells) respond
in different ways to endotoxin. For example, monocytes differentiate into macrophages to phagocytose
pathogens and can function as antigen-presenting cells to induce an adaptive immune response in
the presence of endotoxin (134). Additionally, endotoxin-mediated induction of the TLR4 signaling
pathway in immature dendritic cells leads to dendritic cell maturation, during which the dendritic
cells phagocytose and process-putative antigens. Dendritic cells migrate from epithelial to lymphoid
tissue and act as antigen-presenting cells, which interface with multiple subtypes of T- and B-cells to
generate an adaptive immune response to the processed antigen (135). Furthermore, activation of the
TLR4 signaling pathway alters the function of T- and B-cells and helps dictate the adaptive immune
response to a pathogen by enhancing cell migration, cell attachment, clonal expansion, and antibody
production at endotoxin concentrations as low as picograms per milliliter (136,137).
Due to its potency in stimulating an adaptive immune response, endotoxin could function as an adju-
vant for vaccines; however, endotoxin toxicity prevents its use as a viable adjuvant. Monophosphoryl
lipid A (MPLA) is a lipid A preparation derived from Salmonella minnesota R595 LPS (138). MPLA
has the same basic lipid A structure as LPS and binds to the MD-2/TLR4 receptor complex. Unlike

LPS, MPLA lacks several modifications found on LPS and has a fraction of the potency of (139–141).
Despite the reduced potency, MPLA is immunostimulatory and is used as an adjuvant in commercial
applications to improve the immune response to vaccines (142,143).
Just as MPLA is useful as an adjuvant for vaccines, the unintended presence of endotoxin in low
concentrations may be detrimental to the therapeutic effect of parenteral drugs by the induction of
an unintended immune response. Low doses of contaminant PAMPs such as endotoxin can function
as an adjuvant, which results in accidental immunogenicity against protein-based drugs. A study in
macaques showed that low levels of endotoxin (10 pg dose, avg. weight 5.8 kg, 1.7 pg/kg) when de-
livered with a therapeutic protein led to the production of antibodies against the therapeutic protein.
Although the dose of endotoxin was not sufficient to induce a systemic inflammatory response and
below endotoxin limits for parenteral drugs, it was sufficient to activate not only a local innate immune
response, but also adaptive immunity (144).
Parenteral delivery of drugs bypasses normal physical barriers to infection and may cause an immune
response. Previous instances of drug contamination by either pathogens or endotoxin highlight the
importance of sensitive and robust endotoxin testing that is adequate to provide for patient safety.
Now that LER has been identified, it is important that endotoxin tests control for this phenomenon
to provide confidence that negative endotoxin results truly assure the absence of significant endotoxin
contamination. This is particularly important for drugs produced using formulations that contain
components known to contribute to LER. Other instances when endotoxin cannot be detected using
methods based on LAL, the pyrogen test is used. In the pyrogen test, a febrile response in rabbits acts
as proxy to indicate potential contamination by endotoxin and/or other pyrogens.
The current, incomplete knowledge of the LER phenomenon raises the possibility that microbial
control during manufacturing cannot be accurately assessed. Alternative methods are currently in
use or are being developed to detect endotoxins (e.g., TLR4 assays, monocyte activation test, and
biosensor-based endotoxin detection assays) (16,66,145–149). However, due to the complexity of
new biologic drug formulations and materials, historical data do not necessarily provide an adequate
basis for a safety assessment of new drugs. Therefore, BLA applicants should anticipate that LER
studies may be required.
7.0 References
1. Low Endotoxin Recovery in Common Biologics of Health, Labour, and Welfare. 2016. https://
Products. Chen, J. and Vinther, A. Orlando, www.mhlw.go.jp/file/06-Seisakujouhou-
Fla. : Parenteral Drug Association, April 2013. 11120000-Iyakushokuhinkyoku/JP17_
Presented at the 2013 PDA Annual Meeting. REV_1.pdf.
2. General Chapter <85> Bacterial Endotoxins Test. 6. Chapter 3.4 Test for Bacterial Endotoxins. World
USP 41–NF 36. U.S. Pharmacopeia. 2018. Health Organization. 2012. http://www.who.int/
www.usp.org. medicines/publications/pharmacopoeia/Bacterial-
3. Ministry of Health, Labour and Welfare. Gen- endotoxins_QAS11-452_FINAL_July12.pdf.
eral Tests, Process and Apparatus, Sect. 4.01: 7. Impact of extractables/leachables from filters on
Bacterial Endotoxins Test. Japanese Pharmaco- stability of protein formulations. Huang, M, et
poeia, 17th Ed. Tokyo : s.n., 2016, p. 110. al., 11, 2011, J Pharm Sci, Vol. 100, pp. 4617-
4. European Pharmacopoeia 2.6.14 Bacterial 4630.
Endotoxins Test. European Medicines Agency. 8. Low Endotoxin Recovery: An FDA Perspective.
2010. https://www.edqm.eu/ (accessed August Hughes, P, et al., April 10, 2015, BioPharma
11, 2014). Asia. https://biopharma-asia.com/magazine-
5. The Japanese Pharmacopoeia XVII General Test articles/low-endotoxin-recovery-an-fda-per-
4.01: Bacterial Endotoxins Test. Japan Ministry spective/ (accessed November 19, 2018).

9. Hughes, P. Endotoxin Challenges: A Regula- 21. Technical Report No. 60 Process Validation: A
tory Perspective. PDA 9th Annual Conference Lifecycle Approach. Parenteral Drug Associa-
on Pharmaceutical Microbiology. 2014. tion. 2013. www.pda.org/bookstore (accessed
10. General Chapter <1223> Validation of Alterna- September 11, 2013).
tive microbiological methods. USP 41/NF36. 22. Aggregates Are the Biologically Active Units of
U.S. Pharmacopeia. 2018. www.usp.org. Endotoxin. Mueller, M, et al., 2004, J. Biol.
11. EP 5.1.10 Guidelines for using the test for bacte- Chem., Vol. 279, pp. 26307-26313.
rial endotoxins. European Pharmacopeia 9.0. 23. Chemical structure, molecular conformation, and
2017. www.edqm.eu. bioactivity of endotoxins. Seydel, U, et al., 2000,
12. U.S. Food and Drug Administration. Guid- Chem. Immunol., Vol. 74, pp. 5-24.
ance for Industry: Pyrogen and Endotoxins 24. Self-Organisation, Thermotropic and Lyotropic
Testing: Questions and Answers. U.S. Food & Properties of Glycolipids Related to their Biologi-
Drug Administration; Guidance, Compliance cal Implications. Garidel, P. and al, et. 2015,
& Regulatory Information; Guidances (Drugs). Open Biochem J, Vol. 9, pp. 49-72.
[Online] Jun 2012. [Cited: Mar 28, 2018.] 25. Bacterial endotoxin: molecular relationships of
https://www.fda.gov/downloads/Drugs/Guid- structure to activity and function. Rietschel, E,
anceComplianceRegulatoryInformation/Guid- et al., 1994, FASEB J. Off. Publ. Fed. Am.
ances/UCM310098.pdf. https://www.fda.gov/ Soc. Exp. Biol., Vol. 8, pp. 217-225.
downloads/drugs/guidances/ucm310098.pdf.
26. Fortifying the barrier: the impact of lipid A
13. Quality Guideline Q7: Good Manufacturing Prac- remodelling on bacterial pathogenesis. Needham,
tice Guide for Active Pharmaceutical Ingredients. B and Trent, M. 7, 2013, Nat Rev Microbiol,
International Council on Harmonisation. 2000. Vol. 11, pp. 467-481.
http://www.ich.org/fileadmin/Public_Web_Site/
ICH_Products/Guidelines/Quality/Q7/Step4/ 27. Removal of the outer Kdo from Helicobacter
Q7_Guideline.pdf (accessed 16 Feb, 2017). pylori lipopolysaccharide and its impact on the
bacterial surface. Stead, C, et al., 4, 2010, Mol
14. Lipopolysaccharide Nomenclature – Past, Present Microbiol, Vol. 78, pp. 837-852.
and Future. Hitchcock, P, et al., 3, 1986, J.
Bacteriology, Vol. 166, pp. 699-705. 28. The Power of Asymmetry: Architecture and As-
sembly of the Gram-Negative Outer Membrane
15. General Chapter <1228.5> Endotoxin Indica- Lipid Bilayer. Henderson, J et al., 2016, Ann
tors. USP 41/NF36. U.S. Pharmacopeia. 2018. Rev Micro, Vol. 70, pp. 255-278.
www.usp.org.
29. Morphology, size distribution, and aggregate struc-
16. Detection of pyrogens using human whole blood. ture of lipopolysaccharide and lipid A dispersions
Hartung, T and Wendel, A. 70, 1995, ALTEX, from enterobacterial origin. Richter, W, et al., 5,
Vol. 12, pp. 70-75. 2011, Innate Immun, Vol. 17, pp. 427-438.
17. Results of a harmonized endotoxin recovery study 30. Masking of endotoxin in surfactant samples: Ef-
protocol evaluation by 14 BioPhorum Operations fects on Limulus-based detection systems. Reich, J,
Group (BPOG) member companies. Bolden, J, et al., 2016, Biologicals, Vol. 44, pp. 417-422.
et al., 2017, Biologicals, Vol. 48, pp. 74-81.
31. Factors Affecting Reduction of Reference Endotox-
18. Code of Federal Regulations Title 21 Part in Standard Activity Caused by Chelating Agent/
211.170: Reserve samples. U.S Food and Drug Detergent Matrices: Kinetic Analysis of Low En-
Administration. 2017. https://www.accessdata. dotoxin Recovery. Tsuchiya, M. 6, 2017, PDA J
fda.gov/scripts/cdrh/cfdocs/cfcfr/CFRSearch. Pharm Sci Technol, Vol. 71, pp. 478-487.
cfm?fr=211.170 (accessed June 26, 2018).
32. Resolving LAL Test interferences. Cooper, J. s.l. :
19. Code of Federal Regulations Title 21 Part 211.192: Parenteral Drug Association, 1990, PDA J.
Production record review. U.S. Food and Drug Parenteral Sci Tech, Vol. 44, pp. 13-15.
Administration. 2017. https://www.accessdata.
fda.gov/scripts/cdrh/cfdocs/cfcfr/CFRSearch. 33. Proteinase K digestion of proteins improves
cfm?fr=211.192 (accessed June 26, 2018). detection of bacterial endotoxins by the Limulus
amebocyte lysate assay: application for endotoxin
20. Guidance for Industry: Process Validation: Gener- removal from cationic proteins. Petsch, D, et al.,
al Principles and Practices. U.S. Food and Drug 1998, Anal. Biochem, Vol. 259, pp. 42-47.
Administration. 2011. http://www.fda.gov/
downloads/Drugs/Guidances/UCM070336. 34. Endotoxin contamination of parenteral drugs
pdf (accessed January 28, 2015). and radiopharmaceuticals as determined by the

limulus amebocyte lysate method. Twohy, C, et 48. High-affinity LPS binding domain(s) in recombi-
al., 1984, J Parenteral SciTech, Vol. 38, pp. nant factor C of a horseshoe crab neutralizes LPS-
190-201. induced lethality. Tan, N, et al., 2000, FASEB
35. EDQM Council of Europe. European Phar- J., Vol. 14, pp. 859-870.
macopoeia (Ph. Eur.) 9th edition. 2017. https:// 49. Macromolecular Structure of Lipopolysaccharides
www.edqm.eu/en/european-pharmacopoeia- from Gram-Negative Bacteria. Hannecort-
ph-eur-9th-edition. Pokorni, E, et al., 1973, Eur. J. Biochem, Vol.
36. Pharmaceuticals and Medical Devices Agency. 38, pp. 6-13.
The Japanese Pharmacopoeia 17th edition. 50. The N-terminal Arg Residue Is Essential for
2016. http://www.mhlw.go.jp/file/06-Sei- Autocatalytic Activation of a Lipopolysaccharide-
sakujouhou-11120000-Iyakushokuhinkyoku/ responsive Protease Zymogen. Kobayashi, Y, et
JP17_REV_1.pdf. al., 2014, JBC, Vol. 289, pp. 25987-25995.
37. McCullough, K. The bacterial endotoxins test: 51. Electrodialysis of Lipopolysaccharides and their
a practical approach. River Grove, IL : DHI Conversion to Uniform Salt Forms. Galanos, C
Publishing, LLC, 2011. and Luderitz, O. 1975, Eur J Biochem, Vol.
38. Formulation and delivery issues for monoclonal 54, pp. 603-610.
antibody therapeutics. Daugherty, A and Mrsny, 52. Aggregation Behavior of an Ultra-Pure Lipo-
R. 2006, Adv. Drug Deliv. Res, Vol. 58, pp. polysaccharide that Stimulates TLR-4 Receptors.
686-706. Sasaki, H and White, S. July 2008, Biophysical
39. The Role of the Physical State of Lipopolysaccha- Journal, Vol. 95, pp. 986-993.
rides in the Interaction with Complement. Gala- 53. Monomeric Re lipopolysaccharide from Esch-
nos, C and Lüderitz, O. 1976, Eur J Biochem, erichia coli is more active than the aggregated
Vol. 65, pp. 403-408. form in the Limulus amebocyte lysate assay and
40. Williams, Kevin, [ed.]. Endotoxins: Pyrogens, in inducing Egr-1 mRNA in murine peritoneal
LAL Testing and Depyrogenation. Third Edition. macrophages. Takayama, K, et al., 3, Jan 21,
Boca Raton : CRC Press, 2007. p. 440. Vol. 1994, J Biol Chem, Vol. 269, pp. 2241-2244.
167. ISBN 9780849393723. 54. Human hemoglobin increases the biological activ-
41. Endotoxin recovery using limulus amebocyte ly- ity of bacterial lipopolysaccharides in activation
sate (LAL) assay. Bolden, J, et al., 2016, Biol. J. of Limulus amebocyte lysate and stimulation of
Int. Assoc. Biol. Stand, Vol. 44, pp. 434-440. tissue factor production by endothelial cells in
vitro. Kaca, W, et al., 4, Dec 1, 1994, Innate
42. Kinetics of morphological changes in surfactant Immunity, Vol. 1, pp. 243-252.
systems. Gradzielski, M. 2003, Curr. Opin.
Colloid Interface Sci, Vol. 8, pp. 337-345. 55. Structural prerequisites for endotoxic activity in the
Limulus test as compared to cytokine production
43. Deutschmann, S, et al., Improved Bacterial En- in mononuclear cells. Gutsmann, T, et al., 1, Feb
dotoxin Test for the Determination of Endotoxins. 2010, Innate Immunity, Vol. 16, pp. 39-47.
EP 3 124 976 A1 Germany, July 28, 2015.
European Patent Application. 56. Aggregation behavior and size of lipopolysaccha-
ride from Escherichia coli O55:B5. Bergstrand,
44. The multiple roles of histidine in protein interac- A, et al., 2006, Colloids Surf. B Biointerfaces,
tions. Liao, S-M, et al., 44, 2013, Chem. Cent. Vol. 53, pp. 9-14.
J, Vol. 7.
57. Biophysical Analysis of Lipopolysaccharide Formu-
45. Interfacial and surface tensions of toluene/water lations for an Understanding of the Low Endotoxin
and air/water systems with nonionic surfactants Recovery (LER) Phenomenon. Correa, W, et al.,
Tween 20 and Tween 80. Bak, A and Podgór- 12, 2017, Int. J. Mol. Sci., Vol. 18, p. 2737.
ska, W. 2016, Colloids Surf. Physicochem.
Eng. Asp, Vol. 504, pp. 414-425. 58. Immunochemistry of O and R Antigens of Salmo-
nella and Related Enterobacteriaceae. Luderitz,
46. Structure of bacterial lipopolysaccharides. Caroff, O, et al., 1, s.l. : Bacteriological Reviews, 1966,
M and Karibian, D. 2003, Carbohydr. Res, Vol. 30. 221-231.
Vol. 338, pp. 2431-2447.
59. The dispersion of gram-negative lipopolysaccha-
47. Definition of endotoxin binding sites in horseshoe ride by deoxycholate. Subunit molecular weight.
crab Factor C recombinant sushi proteins and Shands, J and Chun, P. 3, 1980, J. Biol. Chem,
neutralization of endotoxin by sushi peptides. Tan, Vol. 255, pp. 1221-1226.
N. 2000, FASEB J., Vol. 14, pp. 1801-1813.

60. Lipopolysaccharide-sensitive serine-protease zymo- States Pharmacopoeia. 2013. https://www.

gen (factor C) found in Limulus hemocytes. Isola- uspnf.com/sites/default/files/usp_pdf/EN/
tion and characterization. Nakamura, T, et al., USPNF/usp-nf-notices/usp38_nf33_gn.pdf.
1986, Eur. J. Biochem., Vol. 154, pp. 511-521. 72. Guidance for Industry. Sterile Drug Products
61. Properties of US Standard Endotoxin (EC-5) in Produced By Aseptic Processing – Current Good
human male volunteers. Hochstein, H, et al., 1, Manufacturing Practice. U.S. Food and Drug
1994, Innate Immun, Vol. 1, pp. 52-56. Administration. 2004.
62. Stress-induced neuroinflammation: role of the 73. Guidance for Industry Pyrogens and Endotoxins
Toll-like receptor-4 pathway. Gárate, I, et al., 1, Testing Questions and Answers. U.S. Food and
2013, Biol Psychiatry, Vol. 73, pp. 32-43. Drug Administration. 2012. https://www.fda.gov/
63. The processing and collaborative assay of a refer- downloads/drugs/guidances/ucm310098.pdf.
ence endotoxin. Hochstein, H, et al., 1, 1983, J. 74. The bacterial cell envelope. Silhavy, TJ and al.,
Biol. Stand, Vol. 1, p. 251. et. 5, 2010, Cold Spring Harbor Perspectives
64. The International Standard for Endotoxin: evalu- in Biology, Vol. 2, p. Epub.
ation in an international collaborative study. 75. Biologics Production: Impact of Bioburden Con-
Poole, S and Mussett, M. 2, Apr 1989, J Biol taminations of Non-Sterile Process Intermediates
Stand, Vol. 17, pp. 161-171. on Patient Safety and Product Quality. von Win-
65. Development, Validation and Applications of tzingerode, F. April 2017, American Pharma-
the Monocyte Activation Test for Pyrogens Based ceutical Review, Vol. Endotoxin Supplement,
on Human Whole Blood. Schindler, S. et al., 4, pp. 10-19.
Dec 18, 2009, Altex, Vol. 26, pp. 265-277. 76. Lipopolysaccharide Endotoxins. Raetz, C and
66. Use of Toll-Like Receptor Assays To Detect and Whitfield, C. 2002, Annual Review of Bio-
Identify Microbial Contaminants in Biological chemistry, pp. 635-700.
Products. Huang, L, et al., 11, 2009, J Clin 77. The molecular basis of the host response to lipo-
Microbiol, Vol. 47, pp. 3427-3434. polysaccharide. Bryant, C, et al., 2010, Nature
67. Parenteral Drug Association. PDA News Reviews Microbiology, pp. 8-14.
Archive: PDA’s LER Task Force Holds its 78. Platelets express functional toll-like receptor-4.
First Workshop. Parenteral Drug Association. Andonegui, G, et al., 2005, Blood, Vol. 106,
[Online] August 29, 2016. [Cited: October pp. 2417-2423.
31, 2017.] https://www.pda.org/pda-europe/ 79. Diversity of endotoxin and its impact on
news-archive/full-story/2016/08/29/pda-s-ler- pathogenesis. Trent, M, et al., 2006, Journal of
task-force-holds-its-first-workshop. Endotoxin Research, pp. 205-223.
68. Code of Federal Regulations Title 21 Part 211: 80. Biosynthesis and Export of Bacterial Lipopolysac-
Current Good Manufacturing Practice for charides. Whitfield, C and Trent, M. 2014,
Finished Pharmaceuticals. U.S. Food and Drug Annual Review of Biochemistry, pp. 99-128.
Administration. Apr 2017.
81. Detailed characterization of the lipid A fraction
69. Code of Federal Regulations Title 21 Part from the nonpathogen Acinetobacter radioresistens
610.13: General Biological Products Standards: strain S13. Leone, S , et al., 2007, J. Lipid
Purity. U.S. Food and Drug Administration. Res., Vol. 48, pp. 1045-1051.
2005. https://www.ecfr.gov/cgi-bin/text-idx?S
ID=4d5a30970b7be387bbdb917f8901b158 82. The structural basis of lipopolysaccharide recogni-
&mc=true&node=se21.7.610_113&rgn=div8 tion by the TLR4-MD2 complex. Park, B, et al.,
(Accessed May 15, 2018). 2009, Nature, pp. 1191-1195.
70. Code of Federal Regulations Title 21 Part 610.9: 83. Chemistry of lipid A: at the heart of innate im-
General Biological Products Standards: Equiva- munity. Molinaro, A, et al., 2015, Chemistry,
lent Methods and Processes. U.S. Food and Drug Vol. 21, pp. 500-519.
Administration. 2005. https://www.ecfr.gov/ 84. Lipid A in Helicobacter pylori. Mattsby-Baltzer,
cgi-bin/retrieveECFR?gp=1&SID=b10cded I, et al., 1992, Infect. Immun., Vol. 60, pp.
dc657d81be748458b8d968e1e&ty=HTML 4383-4387.
&h=L&mc=true&r=PART&n=pt21.7.610#
85. Structural characterization of the lipid A compo-
se21.7.610_19 (accessed May 15, 2018).
nent of Helicobacter pylori rough- and smooth-
71. General Notices Section 6.30. Alternative and form lipopolysaccharides. Moran, A, et al., 1997,
Harmonized Methods and Procedures. United J. Bacteriol., Vol. 179, pp. 6453-6463.

86. Low biological activity of Helicobacter pylori 100. Status of Biological Substances Used for Detecting
lipopolysaccharide. Muotiala, A, et al., 1992, Bacterial Endotoxins. U.S. Food and Drug Ad-
Infection and Immunity, pp. 1714-1716. ministration. January 12, 1973, Fed Reg, Vol.
87. Immunobiological activities of chemically defined 38(8), p. 1404. https://www.gpo.gov/fdsys/
lipid A from Helicobacter pylori LPS in compari- pkg/FR-1973-01-12/pdf/FR-1973-01-12.pdf.
son with Porphyromonas gingivalis lipid A and 101. Limulus Amebocyte Lysate: Proposed Additional
Escherichia coli-type synthetic lipid A (compound Standards. U.S. Food and Drug Administra-
506). Ogawa, T, et al., 1997, Vaccine, pp. tion. Sept 18, 1973, Fed Reg, Vol. 38(180),
1598-1605. pp. 26130-26132. https://www.gpo.gov/fdsys/
88. Endotoxic and immunobiological activities of a pkg/FR-1973-09-18/pdf/FR-1973-09-18.pdf.
chemically synthesized lipid A of Helicobacter pylori 102. Inactivated Influenza Vaccine General Safety
strain 206-1. Ogawa, T, et al., 2003, FEMS Im- Test. U.S. Food and Drug Administration. Nov
munology and Medicial Microbiology, pp. 1-7. 13, 1974, Fed Reg, Vol. 39(220), p. 40016.
89. Colonize, evade, flourish: How glyco-conjugates https://www.gpo.gov/fdsys/pkg/FR-1974-11-
promote virulence of Helicobacter pylori. Rubin, E 13/pdf/FR-1974-11-13.pdf.
and Trent, M. 2013, Gut Microbes, pp. 439-453. 103. Statistical Determination of Endotoxin Con-
90. IL-1 induction-capacity of defined lipopolysac- tent in Influenza Virus Vaccine by the Limulus
charide partial structures. Loppnow, H, et al., 9, Amoebocyte Lysate Test. Rastogi, S, et al., 1977,
1989, J Immunol, Vol. 142, pp. 3229-3238. J Clin Microbiol, Vol. 6, pp. 144-148.
91. A Bacterial Disease of Limulus Polyphemus. 104. Licensing of Limulus Amebocyte Lysate: Use as an
Bang, F. 1956, Bull. Johns Hopkins Hosp, Vol. Alternative for Rabbit Pyrogen Test. U.S. Food
98, p. 325. and Drug Administration. Nov 4, 1977, Fed
Reg, Vol. 42(213), p. 57749. https://www.gpo.
92. The Role of Endotoxin in the Extracellular Co- gov/fdsys/pkg/FR-1977-11-04/pdf/FR-1977-
agulation of Limulus Blood. Levin, J and Bang, 11-04.pdf.
F. 1964, Bull. Johns Hopkins Hosp, Vol. 115,
pp. 265-274. 105. Human and Veterinary Drugs; Availability of
Draft Guideline for Use of Limulus Amebocyte
93. A description of cellular coagulation in the Limu- Lysate. U.S. Food and Drug Administration.
lus. Levin, J and Bang, F. 1964, Bull. Johns Jan 18, 1980, Fed Reg, Vol. 45(13), p. 3668.
Hopkins Hosp, Vol. 115, pp. 337-345. https://www.gpo.gov/fdsys/pkg/FR-1980-01-
94. Clottable Protein in Limulus: its Localization 18/pdf/FR-1980-01-18.pdf.
and Kinetics of its Coagulation by Endotoxin. 106. Human, Biological, and Animal Drugs and
Levin, J and Bang, F. 1968, Thrombos. Diath. Medical Devices; Availability of Draft Guidelines
Haemorrh, Vol. 19, pp. 186-197. for Use of the Limulus Amebocyte Lysate (LAL)
95. An invertebrate coagulation system activated by Test. U.S. Food and Drug Administration. Mar
endotoxin: evidence for enzymatic mediation. 29, 1983, Fed Reg, Vol. 48(61), p. 13906.
Young, N, et al., 7, Jul 1972, J Clinical Invest, https://www.gpo.gov/fdsys/pkg/FR-1983-03-
Vol. 51, pp. 1790-1797. 29/pdf/FR-1983-03-29.pdf.
96. Gram-Negative Sepsis: Detection of Endotoxemia 107. Guideline on Validation of the Limulus Ame-
with the Limulus Test. Levin, J, et al., 1972, bocyte Lysate Test as an End-Product Endotoxin
Ann. Intern. Med., Vol. 76, pp. 1-7. Test for Human and Animal Parenteral Drugs,
Biological Products, and Medical Devices. U.S.
97. Endotoxin and endotoxemia. Levin, J. 1973,
Food and Drug Administration. December
New Engl. J. Med, Vol. 288, p. 1297.
1987, Fed Reg, p. 54. http://www.gmpua.
98. Hochstein, H. Role of the FDA in Regulat- com/Validation/Method/LAL/FDAGuideline-
ing the Limulus Amoebocyte Lysate Test. [ed.] ForTheValidationA.pdf.
Richard B. Prior. Clinical Applications of the
108. The international pyrogen reference preparation.
Limulus Amebocyte Lysate test. Boston : CRC
Humphrey, J and Bangham, D. 1959, Bull
Press, 1990, pp. 37-50.
World Health Org, Vol. 20, pp. 1241-1244.
99. Hochstein, H and Novitsky, T. Limulus Amebo-
109. Sabey, G. International Acceptance of Refer-
cyte Lysate Test for Bacterial Endotoxins. [ed.]
ence Endotoxins. [ed.] Levin J, and Novitsky
Martin C. Easter. The Application and Acceptance
TJ Watson SW. Endotoxins and Their Detection
of Rapid Microbiological Methods in Pharmaceuti-
with the Limulus Amebocyte Lysate Test. New
cal Industry. s.l. : CRC Press, 2003, pp. 1-20.
York : Alan R. Liss, Inc., 1982, pp. 153-160.

110. Lipid A: Chemical Structure and Biological Ac- 124. The cardiovascular response of normal humans to
tivity. Westphal, O, et al., 1973, J Inf Diseases, the administration of endotoxin. Suffredini, A, et
Vol. 128 (suppl), pp. 17-29. al., 5, 1989, New England Journal of Medi-
111. Preparation and Properties of a National Refer- cine, Vol. 321, pp. 280-287.
nece Endotoxin. Rudbach, J, et al., 1, 1976, J 125. Human Responses to Bacterial Endotoxin. Burrell, R.
Clin Micro, Vol. 3, pp. 21-25. 1994, Circulatory Shock, Vol. 43, pp. 134-153.
112. WHO/BS/2012.2193 and working document 126. Dose-related inflammatory effects of intrave-
QAS/12.501: Report of an international collabora- nous endotoxin in humans: evaluation of a new
tive study to evaluate three preparations of endotoxin clinical lot of Escherichia coli O:113 endotoxin.
for their suitability to serve as the third interna- Suffredini, A, et al., 1999, Journal of Infectious
tional standard for bacterial endotoxin. World Diseases, Vol. 179, pp. 1278-1282.
Health Organization. 2012. http://apps.who. 127. Human endotoxin administration as an experi-
int/iris/bitstream/handle/10665/96606/WHO_ mental model in drug development. Suffredini, A
BS_2012.2193_eng.pdf;jsessionid=3A0A70A467 and Noveck, R. 2014, Clinical Pharmacology
75E73C8FEF3904624D2C7B?sequence=1. and Therapeutics, Vol. 96, pp. 418-422.
113. Second International Standard for endotoxin: 128. Detection of endotoxin in the plasma of patients
calibration in an international collaborative with gram-negative bacterial sepsis by the Limu-
study. Poole, S. et al., 3, 1997, J Endotoxin lus amoebocyte lysate assay. Pearson, F, et al., 6,
Res, Vol. 4, pp. 221-231. 1985, Journal of Clinical Microbiology, Vol.
114. Comparison of several control standard endotox- 21, pp. 865-868.
ins to the national reference standard endo- 129. Endotoxin levels measured by a chromogenic assay
toxin—an HIMA collaborative study. Pearson, F, in portal, hepatic and peripheral venous blood
et al., 1985, Appl.Environ.Microbiol, Vol. 50, in patients with cirrhosis. Lumsden, A, et al., 2,
pp. 91-93. 1988, Hepatology, Vol. 8, pp. 232-236.
115. Biological Activity of Masked Endotoxin. 130. Relationship between plasma levels of lipopoly-
Schwarz, H, et al., 2017, Scientific Reports, saccharide (LPS) and LPS-binding protein in
Vol. 7, p. 44750. patients with severe sepsis and septic shock. Opal,
116. Creation of an In-house Naturally Occurring S, et al., 5, 1999, Journal of Infectious Disease,
Endotoxin Preparation for Use in Endotoxin Vol. 180, pp. 1584-1589.
Spiking Studies and LAL Sample Hold Time 131. Endotoxin levels correlate positively with a seden-
Analysis. Bowers, K and Tran, L. 2011, Ameri- tary lifestyle and negatively with highly trained
can Pharmaceutical Review, Vol. 14, p. 6. subjects. Lira, F, et al., 2010, Lipids in Health
117. Fingerprinting of the TLR4-induced acute and Disease, Vol. 9, p. 82.
inflammatory response. Bosmann, M, et al., 3, 132. Endotoxin-Like Reactions Associated with Intravenous
2012, Exp Mol Pathol, Vol. 93, pp. 319-323. Gentamicin – California, 1998. Centers for Disease
118. The pathophysiology and treatment of sepsis. Control and Prevention. 41, October 23, 1998,
Hotchkiss, R and Karl, I. 2, 2003, N Engl J Morb Mortal Wkly Rep, Vol. 47, pp. 877-880.
Med, Vol. 348, pp. 138-150. https://www.cdc.gov/mmwr/preview/mmwrht-
119. Invited review: Bacterial lipopolysaccharides and ml/00055322.htm (accessed June 26, 2018).
innate immunity. Alexander, C and Rietschel, E 133. Studies on the O Antigen of Salmonella Typhosa
T. 2001, J. Endotoxin Res., Vol. 7, pp. 167-202. V. Enhancement of antibody response to protein
120. LPS/TLR4 signal transduction pathway. Lu, Y-C, antigens by the purified lipopolysaccharide. John-
et al., 2008, Cytokine, Vol. 42, pp. 145-151. son, A, et al., 2, 1956, Journal of Experimental
Medicine, Vol. 103, pp. 225-246.
121. Selected Toll-like receptor agonist combinations
synergistically trigger a T helper type 1-polarizing 134. Monocyte differentiation and antigen-presenting
program in dendritic cells. Napolitani, G, et al., 8, functions. Jakubzick, C, et al., 2017, Nature
2005, Nature Immunology, Vol. 6, pp. 769-776. Reviews Immunology.
122. Therapeutic implications of the TLR and VDR 135. Control of adaptive immunity by the innate immune
partnership. Liu, P and al, et. 3, 2007, Trends system. Iwasaki, A and Medzhitov, R. 4, 2015,
Mol Med, Vol. 13, pp. 117-124. Nature Immunology, Vol. 16, pp. 343-353.
123. Animal models of sepsis. Fink, M. 1, January 1, 136. Low concentrations of lipopolysaccharide syn-
2014, Virulence, Vol. 5, pp. 143-153. ergize with peptides to augment human T-cell

proliferation and can prevent the induction of 146. New era in pyrogen testing. Ding, J and Ho, B.
non-responsiveness by CTLA4-Ig. Goodier, M 2001, Trends in Biotechnol, pp. 277-281.
and Londei, M. 1, 2001, Immunology, Vol. 147. Endotoxin detection – From Limulus amebocyte
102, pp. 15-23. lysate to recombinant factor C. Ding, J and Ho,
137. The modulatory effects of lipopolysaccharide-stim- B. 2010, Sub-cellular Biochemistry, Vol. 53,
ulated B cells on differential T-cell polarization. pp. 187-208.
Xu, H, et al., 2, 2008, Immunology, Vol. 125, 148. More than 70 years of pyrogen detection: current
pp. 218-228. state and future perspectives. Fennrich, S, et al.,
138. Monophosphoryl lipid A obtained from lipopolysac- 2016, ATLA, Vol. 44, pp. 239-253.
charides of Salmonella minnesota R595. Purifica- 149. Methods of endotoxin detection. Su, W and
tion of the dimethyl derivative by high performance Ding, X. 4, 2015, Journal of Laboratory Auto-
liquid chromatography and complete structural mation, Vol. 20, pp. 354-364.
determination. N, Qureshi and al, et. 9, 1985, J
Biol Chem, Vol. 260, pp. 5271-5278. 150. Isolation of Lipopolysaccharide Endotoxin.
Tauber, H and Warfield, Garson. 1959, J. Biol.
139. Characterization of a nontoxic monophosphoryl Chem, Vol. 234, pp. 1391-1393.
lipid A. A, Johnson and al., et. Suppl 5, 1987,
Rev Infect Dis, Vol. 9, pp. S512-S516. 151. Blood-Borne Lipopolysaccharide Is Rapidly Eliminated
by Liver Sinusoidal Endothelial Cells via High-
140. The vaccine adjuvant monophosphoryl lipid A as a Density Lipoprotein. Yao, Z, et al., 6, 2016, Journal
TRIF-biased agonist of TLR4. Mata-Haro V; et al., of Immunology, Vol. 197, pp. 2390-2399.
5831, 2007, Science, Vol. 316, pp. 1628-1632.
152. Detection of circulating tumor necrosis factor af-
141. Putting endotoxin to work for us: monophos- ter endotoxin administration. Michie, H, et al.,
phoryl lipid A as a safe and effective vaccine 23, 1988, New England Journal of Medicine,
adjuvant. Casella, C and Mitchell, T. 20, 2008, Vol. 318, pp. 1481-1486.
Cell Mol Life Sci, Vol. 65, pp. 3231-3240.
153. Clearance of bacterial lipopolysaccharides and
142. New hepatitis B vaccine formulated with an lipid A by the liver and the role of argininosucci-
improved adjuvant system. Kundi, M. 2, 2007, nate synthase. Satoh, M, et al., 1, 2008, Innate
Expert Rev Vaccines, Vol. 6, pp. 133-140. Immunity, Vol. 14, pp. 51-60.
143. Liposomes containing lipid A: an effective, safe, 154. Low Lipopolysaccharide Recovery versus Low
generic adjuvant system for synthetic vaccines. al, Endotoxin Recovery in Common Biological Prod-
Alving C: et. 6, 2012, Expert Rev Vaccines, uct Matrices. Platco, C. September 1, 2014,
Vol. 11, pp. 733-744. American Pharmaceutical Review (Endotoxin
144. In Vivo Effect of Innate Immune Response Modu- Detection Suppl, Part II), pp. 4-7. http://www.
lating Impurities on the Skin Milieu Using a Ma- americanpharmaceuticalreview.com/Featured-
caque Model: Impact on Product Immunogenicity. Articles/167451-Low-Lipopolysaccharide-
Haile, L, et al., 3, 2017, Journal of Pharmaceu- Recovery-versus-Low-Endotoxin-Recovery-in-
tical Sciences, Vol. 106, pp. 751-760. Common-Biological-Product-Matrices/.
145. Recent advances in biosensor based endotoxin
detection. Das, A, et al., 2014, Biosensors and
Bioelectronics, Vol. 51, pp. 62-75.

8.0 Appendix: Case Studies of LER Occurrences

The following case studies were generously supplied by the participating LER Task Force members and
their respective companies in order to fully demonstrate the complexity and variability of endotoxin
testing for a variety of biopharmaceutical drug products. Each case study presents actual data gener-
ated from low endotoxin recovery (LER) hold studies conducted with actual, approved biopharmaceu-
tical products and intermediate materials.
8.1 Case Study 1: Low Concentration PS20 Caused LER Reaction in the
Presence of Monoclonal Antibody Product
Low concentrations of polysorbate 20 (PS20, 0.006%) can contribute to LER phenomenon in the
presence of monoclonal antibody product (mAb) protein when spiked with control standard endo-
toxin (CSE). The level of this endotoxin masking is shown to be proportional to the level of protein
concentration. Freezing of the test samples to 30 °C or below has been shown to reduce the LER
phenomenon. The LER phenomenon was not detected when naturally occurring endotoxin (NOE) is
used as the spiking analyte in the endotoxin hold study instead of CSE.
8.1.1 Introduction
Endotoxin spiking studies were performed on a new product introduction of a mAb in both the drug
product and drug substance forms to assess if LER was present. LER was exhibited in both the drug
product and drug substance post day 1 of the hold time study, with full LER shown on all samples at
Day 14. A series of experiments was performed using CSE and NOE to assess if the LER phenomenon
could be overcome, focusing on protein concentrations and storage temperature of samples.
8.1.2 Materials and Methods

• Product Matrix: Acetate, calcium, sucrose, 0.006% (w/v) polysorbate 20
• pH: 5.2
• Drug product (DP) concentration: 120 mg/mL
• Drug substance (DS) concentration: 90 mg/mL
• pI: 6.8
• Lipopolysaccharides (LPS) supplier: Charles River Laboratories (CRL)
• NOE: Enterobacter cloacae, CRL
• Method: Kinetic Chromogenic, CRL
• Spike time points: Days 0, 1, 2, 3, 7, 14
• Storage temperature:
– Studies 1 and 2: (2–8 °C) and endotoxin spiking concentrations (10 EU/mL)
– Study 3: (-30 °C or below and endotoxin spiking concentrations (10 EU/mL)
• Sample Dilutions: LAL reagent water (LRW) with 1:100 dilution factor
• Controls: LRW
• Recovery method: Calculated using T=0 spike check prepared in LRW
• Serial dilution: 1:10 dilution series was used with LRW as diluent
• Additional information: Spiked DS stored in polystyrene tubes; spiked DP stored in glass syringes.
All dilutions prepared using polystyrene tubes after testing. Endoscan V software used with kinetic
chromogenic BET method (CRL).

8.1.3 Results and Discussion

Drug substance (DS) and drug product (DP) of an mAb were spiked with CSE (CSE: ~1,000 EU/sample,
target concentration 10 EU/mL) and held in 2–8 °C for up to 14 days to determine endotoxin recovery over
time. DS and DP samples from three different manufacturing lots were used in this evaluation study. Ali-
quots of each sample were tested on Day 0 after mixing with spiked endotoxin for to establish baseline prior
to storing at refrigeration temperature storage. Endotoxin recovery from each spiked sample was assessed on
Days 1, 2, 3, 7, and 14 (Table 8.1.3-1). Recovery percentage of spiked undiluted sample was calculated by
comparing against the positive control, which was LRW spiked with 10 EU/mL of CSE on Day 0 without
hold time. Table 1 showed shows that both DS and DP exhibited LER phenomenon by Day 2 as the per-
centage of recovery of some samples fell below 50%. DS exhibited more pronounced LER phenomenon,
with the average recovery rate of three samples falling below 50%. From Day 3 onwards, LER was clearly
demonstrated in all DS samples, whereas the time-course evaluation of DP samples showed that it is not
until Day 14 did the average of endotoxin recovery failed to meet the 50% criterion.
Table 8.1.3-1 Recovery of CSE from Spiked DS and DP Samples
Recovery (%)
Original Data
Drug Product Drug Substance
(CSE)
Sample 1 Sample 2 Sample 3 Sample 1 Sample 2 Sample 3
Day 0 77 68 58 108 81 80
Day 1 55 66 55 77 61 57
Day 2 57 50 46 52 40 35
Day 3 56 50 46 37 30 21
Day 7 58 51 53 16 19 0
Day 14 43 42 42 12 11 0
The potential impact of PS20 on CSE recovery using placebo formulation buffer and ultrafiltration/
diafiltration intermediate samples that were without PS20 was evaluated. Hold time of the evaluation
was limited to Day 7. Recovery of CSE from all samples met the acceptance criterion with greater than
80% (results not shown here), indicating PS20 is a likely contributing factor to LER phenomenon
observed in the presence of this product. Interestingly, a placebo formulation buffer with PS20 did not
impact endotoxin recovery, suggesting that 0.006% PS20 itself does not cause detectable LER.
Based on the more pronounced LER phenomenon (~10% recovery) observed in CSE-spiked DS at
Day 14, however, the potential impact of high protein content contributing to LER phenomenon
became apparent since the formulation of DS and DP is exactly the same with the protein concentra-
tion, 120 mg/mL and 90 mg/mL, respectively, being the only exception. In this case, the high protein
content may be a strong contributor to LER.
For companies that have experienced the LER phenomenon with their products to determine if prod-
ucts actually contain endotoxin (unlikely in purified form but, rather, in a natural lipopolysaccharide
complex state) a common remediation is to gather additional data using NOE. Although NOE has
not been approved as a standard and is not the only option, recent NOE results obtained from indus-
try suggest that LER phenomenon observed with CSE might be overcome by switching to the use of
NOE in the hold time evaluation (1,2).
A second hold time evaluation study was performed using NOE to determine endotoxin recovery
over time. In this NOE study design, DS and DP lots, sample hold condition, and recovery calcula-
tion were exactly the same as the CSE hold time study with the exception of the source of endotoxin.
Results are shown in Table 8.1.3-2. All NOE-spiked DS and DP samples showed no less than 75%
recovery of NOE from samples held at 2-8 °C for 14 days.

Table 8.1.3-2 Recovery of NOE from Spiked DS and DP Samples
Recovery (%)
DP Data
Drug Product Drug Substance
(NOE)
Day 0 101 95 93 155 121 118
Day 1 96 75 75 142 125 122
Day 2 92 78 78 132 121 116
Day 3 97 84 84 127 114 108
Day 7 115 98 98 123 122 110
Day 14 124 104 104 107 103 96
The mechanism of diminishing recovery of endotoxin in this product is not completely elucidated.
High protein contents complexing with spiked CSE, a purified lipopolysaccharide, is thought to be
the main contributing factor. Combined with PS20, which seemed to augment the effect of LER, CSE
that was spiked in DS was barely recoverable.
Reich and Reich and Gallert have shown endotoxin masking can be temperature-dependent in that
the loss of detectable, spiked endotoxin occurs at a faster rate at room temperature than in 2-8 °C
environments (1,2). To challenge this theory, a third hold study was performed on DP, at ultralow
temperature (-30 °C or below) using a CSE spike, in an attempt to decrease the rate of endotoxin
masking. Results of the frozen study are shown in Table 8.1.3-3. The hold duration in this study was
limited to 7 days, instead of 14 days, mainly due to schedule constraints. All recovery was greater than
50%, with the exception of one DP sample from Day 1. The same sample met the recovery criterion
in subsequent time points; therefore, the assayable endotoxin content can be considered stable and can
be recoverable up to 7 days under this freezing storage conditions.
Table 8.1.3-3 Recovery of CSE from Spiked DP Samples Stored at Below-freezing Temperatures (-30 °C)
Recovery (%)
Lot 1 Lot 2
Day 0 92 97 99 95 91 79
Day 1 60 61 65 44 69 57
Day 2 79 69 82 75 75 71
Day 3 83 94 87 82 82 73
Day 7 77 97 73 74 74 70
8.1.4 Conclusion
Low concentration polysorbate 20 (PS20, 0.006%) does not cause LER in the absence of mAb protein;
however, it can contribute to LER phenomenon in the presence of protein. Protein in the absence of
PS20 does not exhibit LER phenomenon. The level of endotoxin masking is proportional to the level
of protein concentration. As demonstrated, a more pronounced LER was observed in DS (120 mg/
mL) than DP (90 mg/ML). Taken together, results obtained suggest that PS20 at the concentration
of 0.006% does not cause, but does contribute to, the LER phenomenon, and that this level of LER
is influenced by the level of protein concentration. This appeared to be a product-specific observation.
Improvement of endotoxin recovery can be achieved with this particular protein by decreasing the rate
of endotoxin masking by storing samples in ultralow temperatures (-30 °C or below). Furthermore,
the LER phenomenon was not detected when NOE, instead of CSE (lipopolysaccharide), was used in
the endotoxin hold study (3,4).

8.1.5 References
1. An advanced endotoxin assay insights and strategies for overcoming Low Endotoxin Recovery in complex formulations.
Reich J. Presented at the PDA Global Conference on Pharmaceutical Microbiology, Bethesda, MD, 2013.
2. Low Endotoxin Recovery in Bio-Pharmaceuticals: Comparison of Natural Occurring Endotoxins (NOE) and
Commercial Standards. Reich J. and Grallert, H. Poster presentation at the 2014 PDA Annual Meeting,
San Antonio, TX.
3. Low Lipopolysaccharide Recovery versus Low Endotoxin Recovery in Common Biological Product Matrices. Plat-
co, C. September 1, 2014, American Pharmaceutical Review (Endotoxin Detection Suppl, Part II), pp.
4-7. http://www.americanpharmaceuticalreview.com/Featured-Articles/167451-Low-Lipopolysaccharide-
Recovery-versus-Low-Endotoxin-Recovery-in-Common-Biological-Product-Matrices/.
4. Evidence against a bacterial endotoxin masking effect in biologic drug products by Limulus amebocyte lysate detec-
tion. Bolden, J; et al. 2014, J Parenter Sci Technol, Vol. 68 (5), pp. 472-477.

8.2 Case Study 2: Mapping LER Effect in In-Process Stages of Purification

Due to general concerns of regulatory authorities, the U.S. FDA requested that analytical sample hold
time studies be performed to determine whether endotoxin recovery is consistent over time for in-
process, bulk drug substance, and drug product matrices. A case study was conducted to evaluate the
effect of storage on LER, using a spike concentration of 10 EU/ml control standard endotoxin (CSE)
to quantify the reduction in recovery over time.
8.2.1 Introduction
Challenge studies were initiated on in-process steps for three batches of biosimilars at 50 mg/mL pro-
tein with Citrate or Tween 20. Undiluted samples were spiked with lipopolysaccharide (LPS) at the
concentration of 10 EU/mL, same as the CSE. This study was a requirement from a customer and the
U.S. FDA to show how our in-process samples acted regarding LER. The reverse approach was used.
For each batch, the following was prepared:

• Test article: A vial with 10 EU/ml drug substance spiked with endotoxin to a final concentra-
tion of 10 EU/ml
• Test article negative control: One vial with 10 ml unspiked drug substance
• Vial positive control: One vial of 2 ml LAL reagent water (LRW) spiked with endotoxin to a
final concentration of 10 EU/ml replacing in-process/drug substance
• Vial negative control: One vial of 2 ml unspiked LRW in place of the drug substance
Figure 8.2.1-1 depicts the testing setup.

An initial target spike concentration of 10 EU/ml was used to evaluate the effect of storage as 10 EU/
mL is sufficient to quantify reduction in recovery over time. The samples were held for 10 days at 2–8
°C, and a result for T=x samples was compared to the T=0 result for percent of recovery.
Calculations of the recovery for each product and were calculated using the equation:
(t = x)
Recovery [%]= 100
(t = 0 LRW)
Figure 8.2.1-1 Testing Setup Overview


Materials
• 50 mg/mL drug substance • Pyrogen-free consumables, lysate, control
• 2.0 mg/mL tris-sodium citrate dihydrate standard endotoxin (CSE), and laboratory
reagent grade water
• 0.68 mg/mL citric acid monohydrate
• Sterile, LAL reagent water or pyrogen-free water
• 1.2 mg/mL L-histidine
• Sterile, pyrogen-free container/vial (borosilicate)
• 10.8 mg/mL L-histidine monohydrochloride
monohydrate • Calibrated pipettes
• 25.0 mg/mL D-sorbitol • Three lots of in-process steps and drug
substance
• 0.8 mg/mL polysorbate 20
Method
The kinetic turbidimetric (KTA) method (Method C) was used to test Pyrogent-5000 (LONZA).

Results
The results presented are data from testing of the five samples taken from five different steps in the process.
Step 1 Data: In-process step PP1 shows an instant LER effect, with no recovery of the spiked
sample matrix (10 EU/mL)
Step 2 Data: In-process step PP2 shows a weak decline in the product with over 50% recovery of the 10 EU

Step 3 Data: In-process PP3 shows a weak decline in the product and still with over 50% recovery
of the 10 EU and a possible dilution fault
Step 4 Data: In-process PP4 shows no LER effect present in the product and still over 50% recovery
of the 10 EU
of the 10 EU

of the 10 EU
8.2.4 Discussion
Results of this study indicate that endotoxins can be recovered in most of the in-process and bulk drug
substance samples, regardless of hold time, when held at 2–8 °C. The first in-process step, however,
shows LER caused by the product. At T0 h, the recovery is 0% of the spike; 48 h and 72 h, the recov-
ery remains at 0%. This is considered a massive decline of endotoxin presence in the samples and an
indication of a LER effect.
As Step 1 is not a critical in-process step and is at the start of the purification process, most likely it is
the “nature” of the samples to show reduced recovery in that there are probably more interfering fac-
tors at this stage than downstream. In essence, the test for endotoxin impurities has been performed
on this sample prior to purification of the material.
8.2.5 Key Lessons Learned

• The following are three points learned by this evaluation:
• Protocol: Having a good protocol for studies (reverse study approach) is essential to remove as-
say to assay variability.
• Mapping Material: In this case study, Step 1 is different from the other in-process steps as it
contains Mycophenolic acid (MPA) and Kifunensine, which will be analyzed to see if it has an
LER effect.
• General Approach: As this was not the first time that an issue with in-process controls had
arisen, the researchers had created a general document “Risk Assessment for Observed Low En-
dotoxin Recovery for In-process, Bulk Drug Substance, and Drug Product Samples” to establish
testing procedures.
8.2.6 Conclusion
This Case Study describes how to perform a risk assessment if and when LER is observed upon perfor-
mance of the LER protocol for in-process, bulk drug substance, and drug product samples. The risk
assessment provides a means of defining risk categories and levels and suggests mitigation approaches for
the varying risk levels. The assessment can be used, along with any and all current understanding of LER
and internal and external requirements, to demonstrate assayable endotoxin stability and quality systems
pertaining to endotoxin sampling, establishing product quality attribute limits, and batch disposition.

8.2.7 Recommendations
Based on the findings of this case study, the authors recommend the following points as best practices:
• Implement a risk assessment tool/guideline for all Limulus amebocyte lysate (LAL) laboratories
• Work with authorities to align on testing only where it is relevant in order to minimize the risk
of conducting unnecessary tests
• Champion the use of naturally occurring endotoxin (NOE), which is more representative of
contaminants and is not subject to the same macromolecular masking phenomenon observed
with CSE and reference standard endotoxins.

8.3 Case Study 3: Use of Purified and Nonpurified Endotoxins in

Hold-time Studies
Endotoxin detection in parenterally administered drug products (including biologics) is a mandatory
requirement of current pharmacopoeias. In accordance with the U.S. FDA, hold time studies must
be performed to demonstrate the stability of assayable endotoxins in undiluted drug products. A
controversial debate about the choice of the right reference material (purified endotoxin versus crude
naturally occurring endotoxins) for spiking of these drugs products has been ongoing. As part of this
debate, some have suggested that the phenomenon of LER is an artifact of purified endotoxin rather
than a consequence of a contamination event during parenteral manufacturing processes. Therefore,
the susceptibility of crude naturally occurring endotoxins (NOEs) and purified endotoxin towards
LER in solutions containing common excipients for biologic drug formulations was compared. The
data presented herein demonstrate that LER is not solely a phenomenon of purified endotoxin, but
also affects crude NOE preparations. Moreover, susceptibility to LER seems to be determined by
bacterial strain and growth conditions rather than the endotoxin grade of purity. These observations
should be considered when preparing hold time studies.
8.3.1 Introduction
A defined reference standard (i.e., primary standard) is needed for compendial endotoxin testing. The
primary standard for Limulus amebocyte lysate (LAL)-based assays is the reference standard endotoxin
(RSE).1 In addition, secondary control standard endotoxins (CSE) are frequently used in LAL assays,
which are calibrated against the RSE. By definition, a reference standard should be of highest purity
and thoroughly characterized to ensure its identity, quality, purity, and potency. The phenomenon of
LER has been attributed by some to purified endotoxin (chemically, LPS) and has sometimes been
considered an artifact of the purified molecule liberated from its native environment. As a conse-
quence, the use of crude NOEs to spike undiluted drug products in hold time studies was discussed;
however, contamination in pharmaceutical manufacturing processes may be heterogeneous and is not
necessarily reflected by use of a single NOE, which must be considered. Moreover, the term “natural”
does not reflect that NOEs are also grown and processed under defined laboratory conditions and,
hence, do not represent a native contamination.
In order to unravel the relationship between purified versus natural endotoxin, regarding their suscep-
tibility to LER, a total of five independent hold time studies were performed. In addition, a sixth study
provides the first structural data enabling a basic understanding of the molecular basis that might ren-
der some endotoxin resistant to LER. These studies are designated as (a) NOEs from different species
including common water contaminants, (b) Purified endotoxin versus NOEs, (c) Effect of nutrient
supply and temperature on LER, (d) Effect of magnesium concentration and pH on LER, and (e)
Potential link of Lipid A modifications on LER susceptibility based on mass spectroscopy data. All
case studies were performed at room temperature to mimic specific manufacturing processes which
take place at ambient temperature (e.g., product filling and packaging procedures) and to demonstrate
worst-case conditions. Hold-time studies performed at defined temperatures may also be useful, how-
ever, to reflect specific manufacturing scenarios. The data included herein demonstrate that LER is not
only a phenomenon of purified endotoxin, but one that also affects some crude NOEs, whereas others
are resistant to LER. Moreover, susceptibility to LER seems to be determined primarily by bacterial
strain and growth conditions rather than by the endotoxin grade of purity. These observations should
be considered when preparing a hold time study to identify LER.
1
Also termed 3rd International Standard (World Health Organization), H0K354 (U.S. Pharmacopeia), BRP-5 (European Pharmacopeia), or EC-7 (FDA).


8.3.2.1 Preparation of Purified Endotoxin
Endotoxin was extracted from Escherichia coli O113:H10 by hot water/phenol method and further
purified according to the RSE protocol. Briefly, bacteria were grown in M9 minimal medium for 24
hours at 37 °C and harvested by centrifugation at 6,000 g for 10 minutes. Endotoxin was extracted
from cell pellet by hot water/phenol method according to Westphal procedure (1). The supernatant
was used to prepare crude NOEs (see 8.3.2.2). Crude endotoxin extracts were further purified by
ethanol precipitation and subsequent treatment with nucleases and dialysis according to Rudbach
procedure (2). Purity of endotoxin was confirmed by UV absorbance measurement at 260 nm (nucleic
acids) and 280 nm (proteins) and SDS-PAGE. Activity of purified endotoxin was determined by En-
doZyme (Hyglos) assay. Purified endotoxin was stored as 1 mg/ml stock solutions at 4 °C.
8.3.2.2 Preparation of Crude NOE

Bacteria used for NOE preparation were obtained from ATCC or DSMZ (Deutsche Sammlung von
Mikroorganismen und Zellkulturen) or were clinical isolates kindly provided by the University of Re-
gensburg. Generally, organisms were incubated with shaking at 150 rpm overnight in 100% lysogeny
broth (LB) medium at 37 °C, according to standard procedures. To test the influence of growth condi-
tions (Case Study C), bacteria were either grown in 1% or 100% LB medium at room temperature or
37 °C as indicated. To analyze the influence of magnesium on LER susceptibility (Case Studies D and
E), bacteria were grown overnight at 37 °C in M9 minimal medium with the indicated concentrations
of magnesium at pH 7.5 or 5.8. Cells were depleted from supernatant by sedimentation at 6000 g for 15
minutes. Supernatant samples were subsequently filtered using a 0.2 µm membrane filter and labeled as
“NOE”. Sterility of NOEs was confirmed by plating 100 µl of each NOE solution on CASO agar plates
and incubation overnight at 37 °C. To prevent growth promotion, 0.05% (v/v) sodium azide (Merck)
was added. Activity was determined by EndoZyme (Hyglos) assay (see Tables 8.3.3.4-1 and 8.3.3.4-2).
8.3.2.3 Hold-Time Studies

For hold-time studies, stock solution of purified endotoxin and crude NOEs with 1000 EU/ml were
prepared and diluted 1:100 into 1 ml of LER solution containing either 10 mM citrate or 10 mM
phosphate buffer with 0.05% surfactant (polysorbate 20 or 80). In addition, the same amount of en-
dotoxin was spiked into LAL reagent water (LRW) as water control samples. For some experiments,
samples were prepared in triplicate (n = 3) as indicated. Samples were prepared in reverse order (i.e.,
starting with the latest time point) and incubated at room temperature in the dark for the indicated
period. After indicated time points, activity was determined by EndoZyme (Hyglos) assay and the
recovery was calculated (see below). The citrate and phosphate buffers and 0.05% polysorbate 20/80
solutions were prepared from stocks purchased from Sigma-Aldrich (Ph. Eur. grade).
8.3.2.4 Analysis
For analysis at indicated time points, samples were diluted 1:10 and 1:100 in LRW and activity was mea-
sured using EndoZyme recombinant Factor C assay (Hyglos) on a BioTek FLx-800 instrument with Gen5
software. The recovery was calculated as the percentage of activity present after indicated time points in
LER buffer compared to the corresponding water control samples. If samples had been prepared in tripli-
cate, the mean recovery and the standard deviation (SD) was calculated. Recoveries are valid between 50%
and 200% of theoretical spike and recoveries below 50% are considered to demonstrate LER.

8.3.3.1 LER Susceptibility of Different Bacterial Species
In order to compare the masking susceptibility of NOEs from different bacteria, NOEs were prepared
from E. coli O113:H21, Enterobacter cloacae, Burkholderia cepacia, and Pseudomonas aeruginosa and spiked
into 10 mM citrate buffer containing 0.05% polysorbate 20. Recoveries were determined after 0, 1, 2, 5,

Figure 8.3.3.1-1 Masking of Four Model Bacterial NOEs in LER solutions at Room Temperature
NOEs were prepared from E. coli O113:H21 (blue), E. cloacae (red), B. cepacia (green), and P. aeruginosa (purple) and spiked into 10
mM citrate/0.05% polysorbate 20 solution. At indicated time points, the endotoxin recovery was determined by EndoZyme assay (n
= 3, Mean ± SD). The red dotted line represents the 50% cutoff (recoveries below are considered LER).
and 7 days of incubation at room temperature. At the initial time point (Day 0), recoveries ranged from
125% to 215% of the corresponding water control sample, representing the theoretical starting concen-
tration (Figure 8.3.3.1-1). Interestingly, after one day of incubation at room temperature, recoveries of
all NOEs, except B. cepacia, were below 50%, thus demonstrating strong LER. Only the NOE from B.
cepacia retained its activity within a 50% to 200% acceptance level over the course of the seven-day study.
These data show that NOEs prepared from four different bacteria display diverse susceptibilities to
LER. To verify this observation, the study was extended by analyzing more species including common
water contaminants. To this end, NOEs were prepared from E. coli O55:B5, E. coli O113:H21, S.
marcescens, P. aeruginosa, B. cepacia, S. maltophilia, E. cloacae, and R. pickettii. NOEs were spiked into
10 mM citrate buffer containing 0.05% polysorbate 20 and recoveries were determined after 0, 1, 5,
Figure 8.3.3.1-2 Masking of Different Bacterial NOE in LER Solutions at Room Temperature
NOEs were prepared from E. coli O55:B5 (red), E. coli O113:H10 (green), E. cloacae (purple), S. marcescens (blue), P. aeruginosa
(orange), B. cepacia (grey blue), S. maltophilia (light red), and R. pickettii (grey green) and spiked into LER solutions containing 10
mM citrate/0.05% polysorbate 20. At indicated time points, the endotoxin recovery was determined by EndoZyme assay (n =1, WC
= Water Control). The red dotted line represents the 50% cutoff (recoveries below are considered LER).

and 7 days at room temperature. After Day 1, recoveries for all NOEs, except B. cepacia and R. picket-
tii, were less than 50% and show severe LER (Figure 8.3.3.1-2). As shown before, only B. cepacia and
R. pickettii withstood the LER effect as far as the 7-day time point. Most affected NOEs were derived
from E. coli O113:H21, Salmonella maltophilia, E. cloacae, and E. coli O55:B5, respectively.
Together, under the given experimental conditions, NOEs from different bacteria behaved differently
with the majority of tested NOEs, showing severe susceptibility to LER with recoveries below 50%
after one day at room temperature. Only NOEs from B. cepacia and R. pickettii were insensitive to
LER under the applied conditions. Hence, masking susceptibility seems to be determined primarily by
the bacterial strain. In addition, single NOE preparations do not reflect the entirety of possible NOEs,
and no species is representative of another in regard to LER susceptibility.
8.3.3.2 Both Purified Endotoxin and Crude NOE are Affected by LER
It has been suggested that LER is only an artifact of purified endotoxin and does not affect NOE,
meaning that LER is mainly connected to the degree of purification (3). To directly compare their
LER susceptibility, NOEs and purified endotoxin were prepared from the same organism and tested in
parallel. Importantly, E. coli O113:H10 and E. coli O113:H21 have the same LPS structure (defined
by the O antigen) but differ in the composition of their flagella (defined by the H antigen) (4). In
mammals, flagellin (i.e., the protein that forms flagellum) is recognized by a specific innate immune
response receptor termed Toll-like receptor 5 (5). So far, however, there is no evidence that flagellin
affects bacterial endotoxin testing. For this case study, both crude NOEs as well as purified endotoxin
were spiked into 10 mM citrate buffer containing 0.05% polysorbate 80 and recoveries were deter-
mined after the time points indicated in Figure 8.3.3.2-1. Apart from initial variations, within the
first four hours, the loss in recovery is virtually identical for both purified endotoxin and crude NOE
from both serotypes. After 18 hours, all samples showed LER with recoveries below 50%.
Within the given test variability, there were no substantial differences between purified endotoxin and
crude NOE preparations of the same strain, indicating that purification does not render endotoxin
more susceptible to LER. Under the given test conditions, both NOE and purified endotoxin are
similarly affected by LER.
Figure 8.3.3.2-1 Masking of Bacterial NOE and Purified Endotoxin in LER Solutions at Room Temperature
NOEs and purified endotoxin were prepared from E. coli O113:H10 (red and green) and E. coli O113:H21 (dark and light blue) and
spiked into LER solutions containing 10 mM citrate/0.05% polysorbate 80. At indicated time points, the endotoxin recovery was
determined by EndoZyme assay (n =1). The red dotted line represents the 50% cutoff (recoveries below are considered LER).

8.3.3.3 Growth Conditions Influence LER Susceptibility

To investigate the influence of different growth conditions on LER susceptibility, P. aeruginosa was
grown under four different conditions (1% or 100% LB at room temperature or 37 °C). NOEs were
prepared in triplicate and spiked into 10 mM citrate buffer containing 0.05% polysorbate 80. Recov-
eries were determined after indicated time points (Figure 8.3.3.3-1). Initial recoveries at Day 0 ranged
from 100% to 190%. Interestingly, after two days, NOEs prepared from bacteria grown at 37 °C were
already affected by LER, whereas NOE from bacteria grown at room temperature showed no masking
susceptibility. The same was observed after three days but, after six days, NOEs derived from room
temperature cultivation were also slightly affected by LER. Interestingly, the nutrient supply (i.e., 1%
or 100% LB) had no impact on LER susceptibility under the given experimental conditions.
In this study, different temperatures and growth conditions were applied to reflect the majority of pro-
cessing steps during parenteral manufacturing. Since NOE from P. aeruginosa grown at room tempera-
ture are more resistant to LER, the higher temperature during bacterial growth resulted in NOEs more
susceptible to LER, irrespective of the nutritional supply. Whether this growth-dependent response
is typical for all Gram-negative bacteria has to be investigated in more detail. Nevertheless, it can be
concluded that growth conditions influence masking susceptibility of NOEs.
8.3.3.4 Effect of Magnesium Concentration and pH on LER

In order to obtain a deeper look into the influence that growth conditions have on LER susceptibility,
NOEs were prepared from E. coli O113:H10 and P. aeruginosa grown under different magnesium
concentrations and pH conditions in M9 minimal medium (Tables 8.3.3.4-1 and 8.3.3.4-2). NOEs
were spiked into either 10 mM citrate or phosphate buffer containing either 0.05% polysorbate 20
(T20) or polysorbate 80 (T80), and recoveries were determined after the time points indicated in
Figure 8.3.3.4-1 and Figure 8.3.3.4-2. For E. coli O113:H10, masking in citrate buffer is faster
compared to phosphate buffer, and polysorbate 20 represents a worse case than polysorbate 80 (Figure
8.3.3.4-1). Hence, differences were observed that are solely based on LER buffer compositions. More-
over, NOE from bacteria grown under high magnesium concentrations (i.e., 10 mM) are more stable
towards LER in T20/phosphate compared to bacteria grown under low magnesium concentrations
(i.e., 10 µM), irrespective of the pH of the medium (7.5 versus 5.8). The same holds true for masking
in T80/phosphate conditions. In contrast, in LER buffer containing either T20 or T80 in citrate buf-
fer, no substantial difference was observed in LER susceptibility of NOEs from bacteria grown under
varying magnesium concentrations. Thus, the magnesium concentration, in this study, had a strong
impact on LER susceptibility of E. coli O113:H10 NOEs in phosphate buffer, whereas no differences
were observed in citrate buffers. In addition, pH variations did not appear to cause major changes.
Figure 8.3.3.3-1 Recovery of P. aeruginosa NOE as a Function of Nutrient Supply and Temperature
NOEs were prepared from P. aeruginosa grown in 1% or 100% LB medium at either 37 °C or room temperature and spiked in triplicate
into LER solutions containing 10 mM citrate / 0.05% polysorbate 80. At indicated time points, the Endotoxin recovery was determined
by EndoZyme assay (n =3, Mean ± SD). The red line represents the 50% cutoff (recoveries below are considered as LER).

A B
C D
Figure 8.3.3.4-1 Recovery of E. coli O113:H10 NOE as a Function of Bivalent Cation Supply a pH
NOEs were prepared from E. coli O113:H10 grown under various growth conditions (see Table I) and spiked into LER solutions
containing (A) 10 mM citrate/0.05% polysorbate 20; (B) 10 mM phosphate/0.05% polysorbate 20; (C) 10 mM citrate/0.05% poly-
sorbate 80; and (D) 10 mM phosphate/0.05% polysorbate 80. At indicated time points, the endotoxin recovery was determined by
EndoZyme assay (n =1). The dotted red line represents the 50% cutoff (recoveries below that line are considered LER).
Table 8.3.3.4-1 Varied Growth Conditions for E. coli

(Values in brackets are endotoxin concentrations in NOE preparations)
Number Medium pH MgSO4 [mM] Endotoxin [EU/ml]

1 M9 7.5 0.01 433,715
2 M9 7.5 10 66,257
3 M9 5.8 0.01 90,962
4 M9 5.8 10 89,035
For P. aeruginosa, after six hours in T20/citrate, all recoveries were below 50% (Figure 8.3.3.4-2). By contrast,
in T20/phosphate, virtually all recoveries were still above 50% after 20 hours, but, after two days, NOEs de-
rived from all tested conditions showed LER. So, T20/citrate has a stronger LER effect for all isolates. More-
over, also in this particular buffer system, the growth conditions render endotoxin more or less stable towards
LER. Here, high magnesium concentrations (i.e., 10 mM) during bacterial growth resulted in NOEs more
susceptible to LER compared to NOEs derived from bacteria grown in low magnesium concentrations (i.e.,
10 µM). Importantly, this was only observed in citrate buffer, whereas no distinct differences were found in
phosphate buffer. Again, variations in pH of growth medium did not influence LER susceptibility.
Table 8.3.3.4-2 Varied Growth Conditions for P. aeruginosa
Values in brackets are endotoxin concentrations in NOE preparations. Note: The endotoxin concentration in NOE preparation 2
(derived from 10 mM Mg/pH 7.5) was below 1,000 EU/ml and, therefore, was not included.
Number Medium pH MgSO4 [mM] Endotoxin [EU/ml]

1 M9 7.5 0.01 62,043
2 M9 7.5 10 737
3 M9 5.8 0.01 9,312
4 M9 5.8 10 32,511

A B
C D
Figure 8.3.3.4-2 Recovery of P. aeruginosa NOE as a Function of Bivalent Cation Supply and pH
NOEs were prepared from P. aeruginosa grown under various growth conditions (see Table 8.3.3.4-2) and spiked into LER
solutions containing (A) 10 mM citrate/0.05% polysorbate 20; (B) 10 mM phosphate/0.05% polysorbate 20; (C) 10 mM
citrate/0.05% polysorbate 80; and (D) 10 mM phosphate/0.05% polysorbate 80. At indicated time points, the endotoxin recovery
was determined by EndoZyme assay (n =1). The dotted red line represents the 50% cutoff (recoveries below that line are
considered LER). Note: NOE preparation 2 is due to low endotoxin concentration not included.
Taken together, a complex interplay of growth conditions and buffer compositions strongly influences
the LER susceptibility of different bacteria to varying extents. Some bacteria (e.g., E. coli O113:H10)
are more stable towards LER in high magnesium concentrations, whereas others (e.g., P. aeruginosa)
are more susceptible to LER under the very same conditions. Therefore, the specific bacterial strain,
as well as growth and buffer conditions, must be considered when performing hold time studies. Tak-
ing these observations into account, the use of a defined purified endotoxin (i.e., RSE), representing a
worst-case scenario, would be a preferable alternative.
8.3.4 Molecular Basis of LER Susceptibility

That both magnesium concentration and pH variations in growth medium impact the Lipid A struc-
ture of endotoxin has already been demonstrated (6). These variations in growth conditions result in
covalent modifications of Lipid A with, e.g., aminoarabinose (Ara4N) at C4 phosphate or phospho-
ethanolamine at C1 phosphate (Figure 8.3.4-1). Moreover, the number of acyl chains and phosphate
groups at Lipid A is of critical importance for the proinflammatory effects of endotoxin (7). Hence, a
link between growth conditions, Lipid A structure, and LER susceptibility appears most likely.
In order to find possible explanations for the differences in the susceptibility of various organisms to
the LER phenomenon, LPS BioSciences (Paris, France) performed MALDI-TOF mass spectrom-
etry of purified endotoxin and crude NOE from E. coli O113:H21. Importantly, both structures are
the same, consisting of prototypical hexaacyl bisphosphorylated Lipid A with typical acyl side chain
length (Figure 8.3.4-2A). Hence, there was no difference in the Lipid A structure between purified
endotoxin and crude NOE from the very same bacterial strain. Thus, extraction and purification of
endotoxin do not affect the Lipid A structure.
In the next step, Lipid A structures of E. coli O113:H21, which is extremely sensitive to LER, and B.
cepacia, which is not susceptible to LER, were analyzed by mass spectrometry and compared (Figure
8.3.4-2A/B). Interestingly, the LER-resistant endotoxin of B. cepacia harbors an aminoarabinose group,
whereas this modification is absent in E. coli O113:H21 LPS. The formation of positively charged

Figure 8.3.4-1 Lipid A Modifications in E. coli

The Aminoarabinose (Ara4N) side chain (green) of Lipid A is covalently bound to C4 phosphate group. This modification is
catalyzed by ArnT enzyme in E. coli. The occurrence of aminoarabinose side chains is believed to be a bacterial adaptation to
overcome membrane dislocations occurring in some environments, for example, limiting concentrations of divalent cations or
fluctuations in pH (6). In addition, EptA catalyzes the addition of phosphoethanolamine (PEtn) at C1 phosphate group during
certain growth conditions (6).
A) E. coli O113:H21
B) B. cepacia
Figure 8.3.4-2 Comparison of MALDI-TOF MS Spectra from NOE and Purified Endotoxin
MALDI-TOF mass spectrometry data were collected from (A) E. coli O113:H21 and (B) B. cepacia for both NOE (bottom)
and purified endotoxin (top) and compared (m/z = mass divided by charge number; FA = Fatty Acid; P = Phosphate group;
Ara4N = 4-amino-4-deoxy-L-arabinose). Apart from differences in intensities, the spectra from NOE and purified endotoxin
represent similar Lipid A structures for each strain if grown under the same conditions.

aminoarabinose modifications of Lipid A may facilitate increased resistance to LER by reducing the
repulsive force of adjacent negatively charged phosphate groups under LER conditions, i.e., in the ab-
sence of bivalent cations, which usually bridge these phosphate groups. Whether these covalent Lipid
A modifications represent a general mechanism present in all Gram-negative bacteria remains to be
elucidated. However, these data support the assumption that variations in growth conditions result in
molecular modifications of Lipid A which, in turn, render endotoxin more or less susceptible to LER.
8.3.5 Conclusion
Endotoxin detection is a mandatory requirement of current pharmacopeias, and the U.S. FDA de-
mands hold-time studies from manufacturers to prove the “assayability” of endotoxin in undiluted drug
products. Hence, the endotoxin spike should reflect a potential contamination during parenteral manu-
facturing. The data presented herein demonstrate that different endotoxin show a heterogeneous LER
susceptibility, which is obviously influenced by a variety of different factors. These factors include, but
are not limited to, bacterial species and strain, growth medium and temperature, hold time, hold-time
temperature, LER solution configuration (i.e., polysorbate 20 or 80 and citrate or phosphate buffers),
method performance and, arguably, the side chain formation of LPS as organisms attempt to adjust to
various environmental growth conditions. Therefore, whether or not a potential contamination in a
drug product is detectable is not readily predictable. As a consequence, an endotoxin spike that is sus-
ceptible to LER is required. Otherwise, the implementation of a hold-time study is purposeless.
As shown here in several case studies, that LER is a function only of purified endotoxin cannot be con-
cluded as the majority of crude NOE preparations also suffered LER-based recovery loss. In instances
where NOE and corresponding purified endotoxin derived from the same bacterial strain were tested
side by side, those recoveries were generally very similar. Importantly, structural studies of NOE versus
purified endotoxin of E. coli O113:H21 were shown to consist of the same underlying Lipid A molecular
structure. Hence, there is no apparent difference between NOE and purified Endotoxin. Based on this,
endotoxin purification does not influence its susceptibility to LER. Taken together, these studies suggest a
mechanism by which, for a given strain, certain growth conditions lead to defined Lipid A modifications
which, in turn, render endotoxin, irrespective of the grade of purification, more or less susceptible to
LER. Due to the tremendous heterogeneity of NOEs derived from different bacteria, a well-defined and
standardized (i.e., purified) endotoxin should be used in hold-time studies. So far, the best characterized
and commonly used standard preparation is the reference standard endotoxin (RSE). Admittedly, the
RSE also might not be the ideal solution but, to date, it is the best available. Further studies are urgently
needed to unravel and understand the relationship between LPS structure and function in more detail.
All together, these observations should be taken into consideration when performing hold time studies.
8.3.6 References
1. Über die Extraktion von Bakterien mit Phenol/Wasser. Westphal, O; et al., 1952, Z Naturforsch, Vol. 7, pp. 148-155.
2. Preparation and properties of a national reference Endotoxin. Rudbach, J; et al., 1976, J Clin Microbiol.
Vol. 3(1), pp. 21-25.
3. Low Lipopolysaccharide Recovery versus Low Endotoxin Recovery in Common Biological Product Matrices.
Platco, C. September 1, 2014, American Pharmaceutical Review (Endotoxin Detection Suppl, Part II),
pp. 4-7. http://www.americanpharmaceuticalreview.com/Featured-Articles/167451-Low-Lipopolysaccha-
ride-Recovery-versus-Low-Endotoxin-Recovery-in-Common-Biological-Product-Matrices/.
4. Species-Wide Variation in the Escherichia coli Flagellin (H-Antigen) Gene. Wang, L; et al., 2003, J Bacte-
riol. Vol. 185(9), pp. 2936-2943.
5. The innate immune response to bacterial flagellin is mediated by Toll-like receptor 5. Hayashi, F; et al., Na-
ture. 2001. Apr; Vol. 410(6832), pp. 1099-1103.
6. Lipid A modification systems in gram-negative bacteria. Raetz, C; et al., 2007, Ann Rev Biochem. Vol. 76,
pp. 295-329.
7. Synthetic and natural Escherichia coli free lipid A express identical endotoxic activities. Galanos, C; et al.,
1985, Eur J Biochem. Vol. 148, pp. 15.

8.4 Case Study 4: Factors Affecting Low Endotoxin Recovery

Reduction of reference standard endotoxin activity was kinetically analyzed under LER conditions con-
taining a chelating agent and a detergent and was considered an apparent first-order reaction. Tempera-
ture, pH, and salt concentrations affected the reduction rates of reference standard endotoxin activity;
temperature appeared to be the most important factor affecting LER. Components of LER matrices, such
as citrate and polysorbate 20, showed similar LER effect at the concentrations commonly used. Phosphate
showed negative correlation against the half-life of reference standard endotoxin in LER solutions. Lower
temperature, lower pH, and a higher salt concentration are preferable to avoid LER in a hold-time study.
Effects of dilution methods were also evaluated and the dilution method with 2 mM magnesium was
considered to be suitable for the detection of reference standard endotoxin activity in LER solutions.
8.4.1 Introduction
LER-type phenomena have been observed before. For example, reference standard endotoxin (RSE) or
control standard endotoxin (CSE) activity was decreased by trace amounts of certain metal cations, such
as aluminum (1), iron (1,2), and gallium (1,3), and some antibiotics (3). Tsuchiya developed a modified
spiking method to observe the effect of a sample on endotoxin detection (1), and Fujita, et al., applied the
method to their products (3). They demonstrated that the addition of endotoxin, after mixing the sample,
and Limulus amebocyte lysate (LAL) avoided the effects of those substances on spiked endotoxin. The
modification of spiking methods suggested that those substances altered the potency of spiked endotoxin,
and that there was no inhibition of the substances to the LAL activation. In LER cases, the activity of the
spiked lipopolysaccharide (LPS) to undiluted products was not recovered by a common treatment, such
as simple dilution with water, even though the diluted sample did not show inhibition to the bacterial
endotoxins test (BET) (4). This suggests that LER is a change in endotoxin activity caused by the product
and is not inhibition of LAL activation by the product. Typical LER is caused by a matrix containing a che-
lating agent and a detergent (4,5); several mechanisms have been proposed (6,7). Some factors affecting
LER were reported (7,8), but no study has described a kinetic analysis on LER under different conditions.
Platco et al., reported that a non-interfering concentration of cation buffer unmasked CSE/LPS in
their monoclonal antibody products containing citrate and polysorbate (9). They considered that the
dilution of their products, with cation buffer containing magnesium, resurrected the CSE activity.
The results of this study confirmed the effectiveness of the magnesium dilution in the LER study and
found that spiked RSE activity in LER solutions was decreased during the dilution with water, even
though its activity was maintained in the original LER solutions for more than two weeks at 2 °C-5 °C
(10). This finding provided a new aspect of the mechanism of LER.
In this study, real-time activity changes of RSE were measured under LER conditions. Even though the
LER mechanism is considered to involve at least two steps, chelating reaction and detergent effect (6,7),
the kinetics of RSE activity change can be analyzed by an apparent first-order reaction, and the half-life of
RSE was calculated from the reaction rate constant k for each LER condition (11). The information on
the half-life of RSE provides the outcome of factors affecting LER, which is useful to design the hold-time
study conditions. Dilution methods were also evaluated with LER solutions. The direct assay method and
the magnesium dilution method were developed to measure the activity in original LER solutions, and the
magnesium dilution method was considered a suitable dilution method for LER solutions (10).

8.4.2.1 Sample Preparation
Stock solutions were prepared from reagent-grade chemicals and were then filtered with positively charged
0.2 mm filters (Acrodisc® Syringe Filter 0.2 mm Posidyne® Membrane, Pall Life Science, Ann Arbor, MI).
Endotoxin was not detected in the stock solutions by the LAL test. LAL Reagent Water (LRW, Charles
River, Charleston, SC) was used, and 5% sodium chloride was purchased from B/Braun (5% Sodium
Chloride Injection USP, Irvine, CA). A solution containing 111% of target concentrations of the com-
ponents were prepared from stock solutions and used as a base LER solution. The concentrations of the

components were adjusted at 100% by addition of one-tenth volume of an RSE solution for the initia-
tion of LER reaction. The USP reference standard endotoxin (RSE, 10,000 USP endotoxin units (EU)
per vial) was reconstituted with 5 ml LRW and diluted with LRW to preferable concentrations.
8.4.2.2 Sample Dilution Methods and Measurement of Endotoxin

Endotoxin activity was measured by the kinetic chromogenic (KCA) or the kinetic turbidimetric
(KTA) LAL method. LAL reagents used were Endochrome-KTM (Charles River) for KCA and KTA2TM
(Charles River) for KTA. The LAL reagents were reconstituted with the endotoxin-specific buffer
(BG120, Charles River) to eliminate possible glucan interference. Endotoxin standard curves were
established by logarithmic plotting of endotoxin concentrations versus onset times (reaction time to
obtain a 5% decrease of transmittance of a reaction mixture). Quadratic regression was used to calcu-
late endotoxin concentrations in samples. The dilution methods are as follows:
• Direct Assay Method: A 0.01 mL of sample was added to a reaction tube containing 0.1 mL LAL
and 0.1 mL LRW, and the reaction mixture was set on a tube reader (ToxinometerTM, Wako Pure
Chemical Industries, Ltd., Osaka, Japan) to perform the LAL test.
• Water Dilution Method: Samples were diluted with LAL Reagent Water (Charles River) in poly-
propylene test tubes (Falcon 352054, BD Biosciences, Bedford, MA). Endotoxin activity in the
diluted samples were measured by the LAL test.
• Magnesium Dilution Method: Samples were diluted with 2 mM magnesium sulfate solution in
polypropylene test tubes. Endotoxin activity in the diluted samples was measured by the LAL test.
8.4.2.3 Measurement of Reduction Rates of RSE Activity in LER Solutions

A base LER solution in a polystyrene test tube was put on an aluminum block heater set at a target
temperature and kept for 15 min to 1 h before starting the reaction. One-tenth of an RSE dilution was
added to the LER solution to start the reaction. The LER solution was vortexed for approximately 15
s, and a sample at T0 was taken for the initial assay. Endotoxin activity was measured by the direct assay
method with preferable timing. The direct assay method could technically be performed as frequently
as once per minute, but the timing was adjusted to the RSE activity reduction rate. In most of the cases,
assays were performed every 10 min. Reduction rates of RSE activity were obtained from slopes derived
by plotting incubation time versus the natural logarithm of reciprocals of endotoxin recovery.
8.4.2.4 Analytical Theory and Assumption

Though LER phenomenon is supposed to be two-step reaction (7), the RSE activity change was
simply assumed to be a first-order reaction from active state LPS (LPSa) to inactive state LPS (LPSi).
LPSa→ LPSi
The rate of reaction is expressed as:
v = -d [LPSa]/dt = k[LPSa]
where v is reaction velocity and k is reaction rate constant.
The above equation can be written equivalently as:

-d (c-x)/dt = dx/dt = k(c-x)
where c is the initial concentration of LPSa and x is a concentration of reacted LPSa at time t.
When t = 0 and x = 0, the integrated equation is:

k = (1/t )* ln(c/(c-x))
modify this equation:
ln (c/(c-x)) = kt (Eq. 1)
Where LPS activity is expressed as recovery ratio, c = 1 and c-x = endotoxin recovery at time t. There-
fore, k can be obtained as a slope of plotting t versus the natural logarithm of 1/endotoxin recovery. If
this assumption is not correct, this plotting should not be linear. The reaction rate constant k was used
as the reduction rate of RSE activity under the analyzed LER condition.
8.4.2.5 Calculation of Half-Life of RSE Activity in LER Solutions

The half-life (HL) of spiked endotoxin in an LER solution was defined as the reduction time required
at a given condition to reduce by 50% the spiked endotoxin activity. Time required to obtain 50%
activity was calculated from a reaction rate constant k and used as an HL value.
Modify Eq. 1:
t = (1/k) ln(c/(c-x))
HL is a time when c = 1 and x = 0.5:
HL = (1/k) ln(1/0.5)= (1/k) ln 2 = 0.693/k

8.4.3.1 Suitability of Analysis as First-Order Reaction
Table 8.4.3.1-1 and Table 8.4.3.1-2 show the results of calculation of the reduction rate k. Correlation
coefficients r were higher than 0.9 instead of three cases that showed less than 0.1 for k. Because most of
the correlation coefficients were higher than 0.9, good linearity was indicated and, therefore, the RSE
reduction was considered an apparent first-order reaction. The three cases that showed low correlation
coefficients had no impact on the conclusion, because they were considered to be caused by gradual slopes
of the regression curves. Suitability of the assumption allowed calculation of the reaction rate constants k
and the half-life of RSE activity. The magnitude of effects on the half-life of RSE activity in LER solutions
was compared to the factors affecting LER. Temperature, pH, salt concentrations, and the components in
LER matrices, i.e., citrate, phosphate, and polysorbate 20, were selected to evaluate their effects on LER.
Table 8.4.3.1-1 Kinetic Parameters for RSE Activity Change in LER Solutions Containing Citrate
Citrate (mM) Polysorbate 20 (%) pH Temperature (oC) NaCl (%) RSE (EU/mL) k Y-intercept r
0.5 0.05 7.4 23 0.8 5.0 0.98 -0.04 0.977
1 0.05 7.4 23 0.8 5.0 1.4 -0.09 0.970
2 0.05 7.4 23 0.8 5.0 2.3 -0.24 0.987
5 0.05 7.4 23 0.8 5.0 2.9 -0.06 0.995
10 0.05 7.4 2 0 0.2 0.048 -0.01 0.644
10 0.05 7.4 2 0 2.0 0.037 -0.07 0.657
10 0.05 7.4 23 0 5.0 4.6 -0.17 0.998
10 0.05 7.4 23 0.8 5.0 3.2 -0.15 0.998
10 0.2 7.4 23 0.8 5.0 3.4 -0.13 0.998
10 0.0001 7.4 23 0.8 5.0 0.077 -0.29 0.195
10 0.001 7.4 23 0.8 5.0 1.4 -0.17 0.981
10 0.01 7.4 23 0.8 5.0 3.2 -0.09 0.995
10 0.05 7.4 23 2.0 5.0 1.8 -0.07 0.984
10 0.05 7.4 23 3.0 5.0 0.71 -0.20 0.914
10 0.05 6.4 23 0 5.0 0.61 -0.31 0.930
10 0.05 6.9 23 0 5.0 0.93 -0.09 0.989
10 0.05 7.4 23 0 0.2 2.5 -0.15 0.990
10 0.05 7.4 23 0 0.2 3.5 -0.13 0.989
10 0.05 7.8 23 0 5.0 7.6 -0.04 0.996
10 0.05 7.4 30 0 2.0 9.3 -0.33 0.991
10 0.05 7.3 30 0 5.0 13 -0.31 0.993
10 0.05 7.4 36 0 2.0 32 -0.54 0.986
10 0.05 7.3 36 0 5.0 42 -0.11 0.999
25 0.05 7.4 23 0.7 5.0 2.9 -0.05 0.993
Plotting incubation time versus the natural logarithm of reciprocals of endotoxin recovery provided the kinetic parameters,
reaction rate constant k, y-intercept, and correlation coefficients r.

Table 8.4.3.1-2 Kinetic parameters for RSE activity change in LER solutions containing phosphate
Phosphate Polysorbate Temperature

pH RSE (EU/mL) k Y-intercept r
(mM) 20 (%) (oC)
2.5 0.05 7.1 24 5.0 0.38 -0.03 0.954
5 0.05 7.4 24 5.0 0.78 -0.06 0.991
10 0.05 7.3 24 5.0 1.5 -0.23 0.978
25 0.05 6.6 23 5.0 2.1 -0.12 0.996
25 0.05 7.3 23 5.0 4.2 -0.19 0.997
25 0.05 7.9 23 5.0 6.1 0.19 0.981
Plotting incubation time versus the natural logarithm of reciprocals of endotoxin recovery provided the kinetic parameters,
reaction rate constant k, y-intercept, and correlation coefficients r.
8.4.4 Effects of Temperature, pH, and Salt Concentration on LER

Figure 8.4.4-1 shows the effect of temperature on the half-life of RSE activity under an LER condi-
tion. Higher temperature produced a shorter half-life of RSE activity. A negative linear correlation was
observed between temperature and the logarithm of half-life of RSE activity. Reich, et al., pointed out
that the LER reaction was temperature-dependent (7). In this study, the temperature effect was kineti-
cally analyzed. Temperature showed a strong effect on the half-life of RSE activity in LER solutions.
The results agreed with the observation by Reich, et al., and provided an estimation of the difference of
the half-life of RSE activity caused by different storage temperatures. For example, the half-life of RSE
activity at 2 oC was about 100 times longer than that at 25 oC. This indicates that temperature should
be strictly controlled for the hold-time studies. Bolden, et al., reported that mixing time of LER solu-
tions affected endotoxin recovery, and a shorter mixing time (1 min) showed better endotoxin recovery
than a longer mixing time (15 min) (8). Their samples were stored at 2 oC-8 oC, but the temperature
was elevated during the mixing, which could explain lower endotoxin recovery in the experiments
(9). Because the temperature of LER solutions cannot be ignored in hold-time studies, previous data
should be reanalyzed considering the operating temperature. The conditions of LER studies should
be designed carefully, especially on the temperature control, because temperature strongly affects the
half-life of RSE activity under LER conditions.
As shown in Figure 8.4.4-2, higher pH produced a shorter half-life of RSE activity. The slope with ci-
trate buffer was slightly steeper than that with phosphate buffer, and both buffer components showed
a negative linear correlation between pH and the logarithm of half-life of RSE activity. It was reported
that the pH of the sample affected LER (7). The pH effect obtained in this study confirmed that
higher pH provided a faster decrease of RSE activity. This suggested that the chelating reaction is very
important for the LER process because pH affects the chelating reaction.
Lower sodium chloride concentrations produced a shorter half-life of RSE activity (Figure 8.4.4-3). A
positive linear correlation was observed between salt concentration and the logarithm of half-life RSE
activity. Higher salt concentrations increased the half-life of RSE activity. Biopharmaceutical products
usually contain saline, which can provide them less LER effects than water-based products.

Figure 8.4.4-1 Effect of Temperature on Half-Life of RSE Activity in LER Solutions

The LER solutions contained 10 mM sodium citrate, 0.05% polysorbate 20, and 0.2, 2.0, or 5.0 EU/mL RSE. The pH of the
solutions was between 7.3 and 7.4. The regression curve obtained was log(HL) = -0.0882 x (Temperature) + 1.41, and the
correlation coefficient was -0.996.
Figure 8.4.4-2 Effect of pH on Half-Life of RSE Activity in LER Solutions

The LER solutions with citrate (red circles) contained 10 mM sodium citrate, 0.05% polysorbate 20, and 5.0 EU/mL RSE. The pH
of the solutions was between 6.4 and 7.8. The regression curve obtained (long dashed red line) was log(HL) = -0.846 x (pH) +
5.57, and the correlation coefficient was -0.967. The LER solutions with phosphate (blue triangles) contained 25 mM phosphate
buffer, 0.05% polysorbate 20, and 5.0 EU/mL RSE. The pH of the solutions was between 6.6 and 7.9. The regression curve
obtained (short dashed blue line) was log(HL) = -0.363 x (pH) + 1.89, and the correlation coefficient was -0.994.
Figure 8.4.4-3 Effect of Salt Concentrations on Half-Life of RSE Activity in LER Solutions
The LER solutions contained 10 mM sodium citrate, 0.05% polysorbate 20, 5.0 EU/mL RSE, and 0-3% sodium chloride. The pH of the
solutions was 7.4. The regression curve obtained was log(HL) = 0.268 x (NaCl) – 0.873, and the correlation coefficient was 0.986.

8.4.5 Effects of Matrix Components on LER at Relevant Concentrations

Effects of matrix components on RSE activity in LER solutions were observed at concentrations com-
monly used in biopharmaceutical products. Sodium citrate (Figure 8.4.5-1), sodium phosphate (Figure
8.4.5-2), and polysorbate 20 (Figure 8.4.5-3) were chosen as the matrix components. The concentra-
tions commonly used for sodium citrate, phosphate buffer, and polysorbate 20 were 5-25 mM, 10-25
mM, and 0.01%-0.2%, respectively. Sodium citrate and polysorbate 20 showed a similar half-life of RSE
at the concentrations commonly used. The half-life of RSE activity was prolonged by further reduction of
the concentrations of these components. Even though longer half-life was observed at the concentrations
lower than that commonly used, these concentrations may not be relevant for biopharmaceutical formu-
lations. A linear correlation was observed in the logarithmic plotting of phosphate buffer concentration
versus half-life of RSE activity. Citrate produces a chelate effect and removes divalent cations from LPS.
Phosphate is not a chelating agent, but competitively removes divalent cations from LPS. Therefore, ci-
trate is more effective in removing divalent cations from LPS. The difference between the slopes may be
caused by the difference in the strength of the agents in divalent-cation removal from LPS.
Figure 8.4.5-1 Effect of Citrate Concentrations on Figure 8.4.5-2 Effect of Phosphate Concentrations
Half-Life of RSE Activity in LER Solutions on Half-Life of RSE Activity in LER Solutions
The LER solutions contained 0.5–25 mM sodium citrate, The LER solutions contained 2.5–25 mM phosphate buffer
0.05% polysorbate 20, 0.8% sodium chloride, and 5.0 EU/mL (PB), 0.05% polysorbate 20, and 5.0 EU/mL RSE. The pH of
RSE. The pH of the solutions was 7.4. the solutions was 7.1-7.4. A linear relationship was observed
logarithmic plotting between the phosphate concentrations
and the half-life of RSE activity. The regression curve obtained
was log(HL) = -1.03 x log(PB) – 0.674, and the correlation
coefficient was -1.00.
Figure 8.4.5-3 Effect of Polysorbate 20 Concentrations on Half-Life of RSE Activity in LER Solutions
The LER solutions contained 10 mM sodium citrate, 0.0001%–0.2% polysorbate 20, 0.8% sodium chloride, and 5.0 EU/mL RSE.
The pH of the solutions was 7.4.

8.4.6 Effects of Dilution Methods on Recovery of RSE in LER Solutions

RSE activity change in the LER solutions containing 10 mM sodium citrate, 0.05% polysorbate 20,
and saline was measured by three different dilution methods—the direct assay method, the water
dilution method, and the magnesium dilution method—at 25 oC and 2 oC-5 oC. The RSE activity
gradually decreased over time at 25 oC (Figure 8.4.6-1). The RSE activity was lower with the water
dilution method than with others after 10 min, even though the compositions of the reaction mixtures
were exactly the same as the direct assay and water dilution methods. Similar results were also observed
in the LER solution stored at 2 oC-5 oC after one day (Figure 8.4.6-2). These results suggested that
RSE aggregates were quickly altered during the dilution with water. Interestingly, the RSE activity
was maintained for four days at 2 oC-5 oC with the direct assay method and the magnesium dilution
method (Figure 8.4.6-2). RSE activity was also maintained for at least 15 days at 2 oC-5 oC in LER
solutions with different compositions (Figure 8.4.6-3). In the direct assay method, RSE aggregates
probably reacted with the LAL reagent before the aggregates were altered. The magnesium dilution
method probably prevented the alteration of RSE aggregates during the dilution. With the water di-
lution method, even though the RSE activity in the LER solution decreased in one day, the activity
Figure 8.4.6-1 RSE Activity Change in LER Solution at 25 °C with 3 Different Dilution Methods
The LER solutions used contained 10 mM sodium citrate, 0.05% polysorbate 20, and 20 EU/mL RSE, and stored at room
temperature (25 °C). The pH of the solutions was 7.1. Samples were taken from the test tube at 0, 10, 20, 30, and 60 minutes
after RSE was added to the LER solution and used for the direct assay method (Direct) and preparation of 10-fold dilutions with
water (Water) and 2 mM magnesium sulfate solution (Mg).
Figure 8.4.6-2 RSE Activity Change in LER Solution at 2 °C–5 °C with 3 Different Dilution Methods
The LER solutions used contained 10 mM sodium citrate, 0.05% polysorbate 20, and 20 EU/mL RSE, and was stored in a
refrigerator (2 °C–5 °C). The pH of the solutions was 7.1. Samples were taken from the test tube at 0.02 (30 minutes), 1, and 4
days after RSE was added to the LER solution and used for the direct assay method (Direct) and preparation of 10-fold dilutions
with water (Water) and 2 mM magnesium sulfate solution (Mg).

Figure 8.4.6-3 Effects of Sorts of Chelating Agents and Detergents on LER

Four kinds of LER solutions were used. LER solutions were prepared with a combination of a chelating agent (10 mM sodium
citrate (Na citrate) or 20 mM phosphate buffer (PB)) and a detergent (0.1% polysorbate 20 (PS20) or 0.1% polysorbate 80
(PS80)) and were spiked with 20 EU/mL RSE. The LER solutions were stored in a refrigerator (2 °C–5 °C) for 15 days. The pH
of the solutions was 7.0-7.1. Samples were taken from the test tube and used for the direct assay method (Direct), and for
preparation of 10-fold dilutions with water (Water) and 2 mM magnesium sulfate solution (Mg).
was maintained after 0.02 day (30 min) at low temperature (Figure 8.4.6-2). This indicates that the
original RSE aggregates can be maintained for a while at low temperature. Considering these results,
there are three phases of RSE in LER solutions:
• First phase: RSE activity is high with all three dilution methods.
• Second phase: RSE activity is kept in the original LER solutions and is decreased by the dilution
with water.
• Third Phase: RSE activity is decreased, and all three dilution methods show low activity.
Different compositions of the LER solutions showed different degrees of the reduction of the RSE activ-
ity in the LER solutions with the water dilution method (Figure 8.4.6-3). Polysorbate 80 showed less
LER effect than polysorbate 20. This suggested that the carbon chain length in the detergent is important
in LER effect because lauric acid in polysorbate 20 has a closer carbon chain length to the fatty acids
in Escherichia coli Lipid A than oleic acid in polysorbate 80. Citrate showed a stronger LER effect than
phosphate. This can be explained by the strength of the chelate effect, as discussed in Section 8.4.5.
8.4.7 Conditions for Magnesium Dilution Method

The magnesium dilution method showed comparable values to the direct assay method (Figures
8.4.6-1 – 8.4.6-3) indicating that the addition of magnesium prevents the change in RSE aggregation
status during the dilution. This method is probably suitable for the measurement of endotoxin activ-
ity in original LER solutions. Appropriate magnesium concentrations were examined for this method
by using an LER solution containing 10 mM sodium citrate, 0.05% polysorbate 20, and 20 EU/mL
RSE in saline stored at 2 oC–5 oC for three days. RSE recovery in 10-fold dilutions was consistently
obtained in the range between 100% and 140% with magnesium concentrations between 0.6 mM
and 10 mM. The RSE recovery was lower with magnesium concentrations below 0.6 mM, indicating
that the magnesium concentration should be equal to or higher than 0.6 mM to neutralize the citrate
effect in the diluted LER solutions. The rate for the first dilution with 2 mM magnesium sulfate was
compared between 5-fold and 50-fold using the same LER solution, and there was no significant dif-
ference in RSE recovery between dilution rates. The average recovery was 119% with the coefficient of
variation at 5.1%. The spiked RSE concentration did not affect the RSE recovery in the magnesium
dilutions of the LER solutions containing 10 mM sodium citrate and 0.05% polysorbate 20 in the
range between 2 EU/mL and 200 EU/mL. These results suggest that the magnesium dilution method
using 2 mM magnesium solution is useful to measure endotoxin activity in LER solutions. The results
with the magnesium dilution method 1) reproduced the results of Platco, et al., although the mag-
nesium concentrations were different, and 2) support the suggestion that the the storage temperature
of the samples at 2 oC–8 oC is important (21). Further, the results indicate CSE activity was already

decreased in the LER solutions because CSE activity was resurrected by dilution of magnesium sul-
fate in Tris buffer. However, the results in the Platco, et al., study indicated that the RSE activity was
maintained in the undiluted LER solutions, and that the magnesium dilution method prevented the
alteration of RSE aggregates by dilution.
8.4.8 Conclusion
LER was considered a temperature-dependent chemical reaction affected by pH, salt concentrations,
and components of LER matrices. Since the half-life of RSE activity in LER solutions showed a linear
relationship with temperature, pH, and salt concentration, information on the half-life is useful to
anticipate the LER effects in pharmaceutical products. Lower temperature, lower pH, and higher salt
concentrations are recommended to avoid LER.
Since the BET and the rabbit pyrogen test are bioactivity assays, they do not determine the physical
concentrations of LPS. Therefore, the physical weight of LPS in the samples cannot be determined
using the BET, based on the LAL test. The potency of RSE or CSE expressed in endotoxin units per
nanogram (EU/ng) is known, but the potency of contaminated endotoxin in samples is unknown;
only the endotoxin activity in the sample is known. There is an LPS aggregation status the potency of
which is very low under LER conditions. Because the biological activity and clinical risks of undetect-
able LPS under LER conditions are unknown, further studies are necessary. This subject is different
from the target of the BET, which measures only endotoxin activity in products. Therefore, the LER
issue is not the BET issue; it is a new matter to be investigated. Considering that LER is the change
in potency of RSE or CSE and does not reflect the stability of contaminated endotoxin in products,
the current design for hold-time studies using purified LPS may not be suitable. For the future inves-
tigation of LER, an understanding of LER mechanism will be important, and the information on the
kinetically analyzed factors affecting LER will be helpful.
There seem to be three phases in LER solutions. The first phase is a stage progressing chelation on the
LPS aggregates, though the LPS aggregates are still strong enough to keep the activity in the water
dilutions. In the second phase, the chelation on the surface LPS probably proceeds quickly, but the
LPS aggregates are still maintained in the solution. The activity is decreased by the water dilution, but
the magnesium dilution can maintain the LPS aggregates to show activity. The second phase can be
maintained for a long time at a low temperature. In the third phase, significant amounts of the surface
LPS molecules may be replaced with detergent molecules, and the biological activity is lost. Even the
direct assay and magnesium dilution methods cannot detect the original levels of RSE activity in the
LER solutions in this phase. Considering this mechanism, controlling the temperature of samples and
using the magnesium dilution method for hold-time studies is important.
Acknowledgements
The author thanks Dr. Jack Levin and Dr. James F. Cooper for their thoughtful suggestions.
8.4.9 References
1. Limulus Test and its Application (in Japanese). Tsuchiya, M. 1990, J Antibact Antifung Agents, Vol. 18, pp.
287-294.
2. Effects of Iron on Bacterial Endotoxin. Roth, R; et al., 1997, J Endotoxin Res., Vol. 4(4), pp. 273-278.
3. Saline and Buffers Minimize the Action of Interfering Factors in the Bacterial Endotoxins Test. Fujita, Y; et al.,
2011, Anal Biochem., Vol. 409(1), pp. 46-53.
4. The use of endotoxin as an analyte in biopharmaceutical product hold time study. Bolden, J; et al., 2015, Phar-
macopeial Forum, Vol. 41(5).
5. Low Endotoxin Recovery in Common Biologics Products. Chen, J and Vinther, A. Presented at the 2013 PDA
Annual Meeting, Orlando, FL.
6. Possible Mechanism of Low Endotoxin Recovery. Tsuchiya, M.2014, Am Pharm Rev, Vol. 17(7), pp. 18-23.

7. Masking of endotoxin in surfactant samples: Effects on Limulus-based detection systems. Reich, J, et al., 2016,
Biologicals, Vol. 44, pp. 417-422.
8. Endotoxin Recovery Using Limulus Amebocyte Lysate (LAL) Assay. Bolden, J; et al., 2016, Biologicals, Vol.
44(5), pp. 434-440.
9. Resolution of Low Endotoxin/Lipopolysaccharide Recovery (LER/LLR) Testing. Platco, C; et al., 2016, Am
Pharm Rev, Vol. 19(5), pp. 20-25.
10. Data Based Mechanism of Low Endotoxin Recovery (LER). Tsuchiya, M. Presented at the PDA 12th Annual
Global Conference on Pharmaceutical Microbiology, Bethesda, MD, October 2017.
11. Factors Affecting Reduction of Reference Endotoxin Standard Activity Caused by Chelating Agent/Detergent Ma-
trices: Kinetic Analysis of Low Endotoxin Recovery. Tsuchiya, M. 2017, PDA J Pharm Sci Technol, Vol. Nov/
Dec 71, pp. 478-487.

8.5 Case Study 5: Comparison of Different Methods Used for Calculation of

Percentile Endotoxin Recovery from LPS-Spiked Biological Therapeutic
Products
A phenomenon termed low endotoxin recovery (LER) was first described by Chen and Vinther in 2013 (1).
LER is characterized by a time-dependent loss of endotoxin activity measured by the limulus amebocyte lysate
(LAL) assay when lipopolysaccharides (LPS) are spiked into certain matrices of biological drug products. This
prompted the U.S. FDA to request firms to provide evidence that the LAL-based bacterial endotoxin test
(BET) method (i.e., USP <85>) can reliably recover endotoxin from spiked drug products over time in support
of Biologics License Applications (BLA) (2,3). A known amount of lipopolysaccharide (LPS) is spiked into
undiluted biological products; samples are then taken and tested at different time points to determine the per-
centile recovery over a period of storage time at specific temperatures. However, there is no regulatory guideline
or industry consensus as to how the percentile recovery should be calculated. As a result, different studies often
use different “baseline” endotoxin activity to calculate the recovery.
This case study reports on three different methods to calculate endotoxin recovery from multiple
spike/hold studies, compares the results, and discusses those results in the context of method harmo-
nization across the industry to ensure consistency.
8.5.1 Introduction
Recovery of less than 50% of the spiked amount of endotoxin is generally considered unacceptable,
which is referred to as LER. Three methods were employed in various spike/hold studies to calculate
the percentile recovery by using different denominators: 1) using the endotoxin activity at the initial
time point as the denominator, 2) using the theoretical endotoxin activity based on reference standard
endotoxin (RSE) titration, and 3) using the endotoxin activity recovered from LAL reagent water
(LRW) at the corresponding time point as the denominator. To assess the relationship between the
endotoxin percentile recoveries and calculation method, a systematic comparison of the three calcula-
tion methods using data from multiple spike/hold studies was conducted. The results indicate that
percentile recoveries could be influenced, sometimes significantly, by the calculation method.

8.5.2.1 Products and Purified LPS
All products used in these spike/hold studies were biological therapeutic products (BTP) produced in
mammalian cells in full compliance with cGMP requirements. These BTP were formulated in matri-
ces containing both citrate and polysorbate 80 (PS80). High-titer purified LPS used in the studies was
purchased from Lonza (Walkersville, Md.).
8.5.2.2 Endotoxin Assay for Routine Batch Release Testing

The compendial kinetic chromogenic method was verified in accordance with USP <85> to determine
the dilutions needed to avoid any matrix interference. Control standard endotoxin (CSE), chromo-
genic LAL, high-titer purified LPS, LRW, 50 mM Tris buffer, and dispersing agent Pyrosperse™ (a
metallo-modified polyanionic dispersant) were obtained from Lonza (Walkersville, Md.).
8.5.2.3 Spike/Hold Time Recovery Studies

BTP (monoclonal antibodies and Fc-fusion proteins, referred to as BTP-1, BTP-2, BTP-3, or BTP-
4) aliquots in pyrogen-free containers were spiked with purified LPS to targeted concentrations. The
spiked containers were then stored (hold) at 2-8 ºC. At each time point, a single container was removed
from storage and mixed for 1 min by vortex prior to testing. In parallel to each individual product,
LRW was spiked to the same final concentration and then dispensed into multiple individual contain-
ers. These LPS-spiked LRW were stored and tested in the same manner as the spiked BTP. Multiple
representative batches were evaluated by the spike/hold study for each of the individual BTPs.

8.5.2.4 LER Definition and Recovery % Calculations

LER is defined as less than 50% recovery calculated by the selected method for two consecutive time
points in the spike/hold study.
Calculation Method I
Ax
Recovery [%]= x 100
I
Where Ax is the amount of LPS (EU/mL) recovered from the spiked BTP at a given time point and I
is the LPS (EU/mL) recovered from the spiked BTP at the initial time point.
Calculation Method II
Ax
Recovery [%]= x 100
Wx
Where Ax is the amount of LPS (EU/mL) recovered from the spiked BTP at a given time point and Wx
is the LPS (EU/mL) recovered from the spiked LRW (positive control) at the corresponding time point.
Calculation Method III

Ax
Recovery [%]= x 100
T
Where Ax is the amount of LPS (EU/mL) experimentally recovered from the spiked BTP at a given
time point and T is the theoretical LPS (EU/mL) spiked into the BTP.
8.5.3 Results
The results from the BTP-1 spike/hold study are summarized in Table 8.5.3-1.
Table 8.5.3-1 Endotoxin Activity Detected from LPS-spiked BTP-1* Over Time
Theoretical Endotoxin Activity (EU/mL)

Material
Spiking (EU/mL) Initial Day 1 Day 3 Day 7 Day 14
BTP-1 batch A 20 24.6 28.6 27.0 24.8 27.6
BTP-1 batch B 20 26.6 29.8 29.2 24.2 30.0
BTP-1 batch C 20 26.0 32.2 26.4 26.6 26.6
LRW 20 20.8 25.8 21.8 24.2 22.2
* The matrix contains 20 mM sodium citrate, 200 mM NaCl, 0.03% PS80, and 80 mg/mL BTP-1, pH 5.7.
At the initial time point, the LPS activity measured in LRW was in good agreement with the theoretical
activity of 20 EU/mL. LPS activities detected in BTP-1 at the initial time point were approximately 30%
higher than the anticipated 20 EU/mL. LPS recovery at each time point was calculated using the three
different methods described in Section 8.5.2.4. The recovery is expressed in percentile as shown in Table
8.5.3-2. No LER was observed throughout the time course examined, and percentile recoveries calculated
by the three different methods varied by no more than 30% at any given time point.
Table 8.5.3-2 Comparison of Recovery % Calculated by Different Methods for BTP-1
Endotoxin Recovery% Calculated Using Different Methods (I, II, III)

Material Initial Day 1 Day 3 Day 7 Day 14
I II III I II III I II III I II III I II III
BTP-1 batch A 100 118 123 116 111 143 110 124 135 101 102 124 112 124 138
BTP-1 batch B 100 128 133 112 116 149 110 134 146 91 100 121 113 135 150
BTP-1 batch C 100 125 130 124 125 161 102 121 132 102 110 133 102 120 133
LRW 100 100 104 124 100 129 105 100 109 116 100 121 107 100 111

Table8.5.3-3 Endotoxin Activity Detected from LPS-Spiked BTP-2* Over Time

Material
Spiking (EU/mL) Initial Day 1 Day 3 Day 7 Day 14
BTP-2 batch A 19 22.8 38.8 21.2 9.2 4.9
BTP-2 batch B 19 23.3 39.4 27.5 13.0 8.8
BTP-2 batch C 19 23.0 40.2 28.9 14.8 8.9
LRW 19 26.1 26.7 20.1 22.0 27.6
* The matrix contains 10 mM citrate, 255 mM mannitol, 0.02% PS80, and 3 mg/mL BTP-2, pH 6.5.
At the initial time point, the LPS activities detected in LRW and BTP-2 were approximately 37% and
21% higher than the theoretical titer of 19 EU/mL, respectively. Table 8.5.3-4 shows the LPS percentile
recoveries calculated using the three different methods. Recoveries were all above 50% for up to three days
for all three BTP-2 batches regardless of the calculation methods. On Day 7, two of the three batches had
LPS recovery >50%, while the remaining batch showed only ~40% LPS recoveries. On Day 14, LER was
observed for all three BTP-2 batches as LPS recovery fell below 50%. Percentile recoveries calculated by
the different methods were in good agreement with regard to LER classification using the 50% cutoff.
Table 8.5.3-4 Comparison of Recovery % Calculated by Different Methods for BTP-2
Endotoxin Recovery% Calculated Using Different Methods (I, II, or III)

BTP-2 batch A 100 87 120 170 145 204 93 105 112 40 42 48 21 18 26
BTP-2 batch B 100 89 123 169 148 207 118 137 145 56 59 68 38 32 46
BTP-2 batch C 100 88 121 175 151 212 126 144 152 64 67 78 39 32 47
LRW 100 100 137 102 100 140 77 100 106 84 100 116 106 100 145
Table 8.5.3-5 Endotoxin Activity Detected from LPS-Spiked BTP-3* Over Time

Material Spiking
(EU/mL) Initial Day 1 Day 3 Day 7 Day 8
BTP-3, batch A 20 52.1 46.8 34.3 21.7 21.2
BTP-3, batch B 20 45.8 49.6 37.2 22.8 20.8
BTP-3, batch C 20 47.1 48.1 39.7 21.5 18.7
LRW 20 72.7 60.4 55.2 50.0 57.1
* The matrix contains 20 mM citrate, 200 mM NaCl, 0.02% PS80, and 160 mg/mL BTP-3, pH 5.7.
Endotoxin activities recovered from both BTP-3 and LRW at the initial time point were much higher
than the theoretical activity of 20 EU/mL. Recoveries at each time point were calculated using the
three different methods described in Section 8.5.2.4. Because of the higher-than-anticipated endotox-
in activity at the initial time point, the percentile recoveries obtained by Method III were significantly
greater than those based on Methods I and II as shown in Table 8.5.3-6. As a result, LPS recoveries
of >50% were only obtained up to Day 3 and the recoveries fell to ≤50% for all batches on the next
time point (Day 7), based on the calculation by Methods I and II. On the other hand, LPS recoveries
calculated by Method III remained greater than 100% at all time points for all three methods, except
the last time point for BTP-3, batch C, throughout the entire time course evaluated. The LPS recovery
from BTP-3 as well as LRW based on Method III often reached 200-300%. It became clear that LPS
recoveries can vary significantly depending on the calculation method as demonstrated in the spike/
hold study with BTP-3.

Table 8.5.3-6 Comparison of Recovery% Calculated by Different Methods for BTP-3

BTP-3 batch A 100 72 261 90 77 234 66 62 172 42 43 109 41 37 106
BTP-3 batch B 100 63 229 108 82 248 81 67 186 50* 46 114 45 36 104
BTP-3 batch C 100 65 236 102 80 241 84 72 199 46 43 125 40 33 94
LRW 100 100 364 83 100 302 76 100 276 69 100 250 79 100 286
* Rounded to 50% from 49.8%.
The results from spike/hold study BTP-4 are summarized in Table 8.5.3-7.
Table 8.5.3-7 Endotoxin Activity Detected from LPS-Spiked BTP-4* Over Time

Material Spiking
(EU/mL) Initial Day 1 Day 3 Day 4 Day 7
BTP-4, batch A 444 689 506 342 214 18.8
BTP-4, batch B 444 759 541 376 207 27.8
BTP-4, batch C 444 692 531 287 94.0 14.0
LRW 444 404 367 360 327 321
* The matrix contains 10 mM citrate, 0.02% PS80, and 40 mg/mL BTP-4, pH 6.5.
Activities recovered from the BTP-4 at the initial time point were significantly higher than the theo-
retical activity expected, as shown in Table 8.5.3-8. Because of the higher-than-anticipated endotoxin
activity at the initial time point were only seen with BTP-4, but not with LRW, the percentile recovery
obtained by Method I exhibited a faster decline over time relative to those determined by Methods II
and III. As a result, LPS recoveries of >50% were only obtained at the initial and Day 1 time points
per Method I. On the other hand, LPS recoveries of >50% sustained up to the Day 3 time point for
all three batches based on the calculations by Methods II and III (Table 8.5.3-8).
Table 8.5.3-8 Comparison of Recovery% Calculated by Different Methods for BTP-4

BTP-4 batch A 100 175 155 73 138 114 50 *
95 77 31 65 48 2.7 5.9 4.2
BTP-4 batch B 100 188 171 71 147 122 50* 104 85 27 63 47 3.7 8.7 6.3
BTP-4 batch C 100 171 156 77 145 120 41 80 65 14 29 21 2.0 4.4 3.2
LRW 100 100 91 91 100 83 89 100 81 81 100 74 79 100 72
* Rounded to 50% from 49.6% and 49.5% for batch A and batch B, respectively.
8.5.4 Discussion
That some products can inhibit the LAL assay (4,5), a phenomenon known as interference that can be
readily circumvented by dilutions using LRW, has long been recognized (6). The LER phenomenon is
characterized by the progressive and time-dependent loss of endotoxin activity. Unlike traditional in-
terference, LER cannot be overcome by simple dilutions with LRW or other diluents. To assess wheth-
er a particularly formulated product causes LER, a known amount of endotoxin is spiked into the test-
ing product; the spiked samples are then stored at appropriate temperatures and tested by LAL assay
repeatedly to determine the recoverability of the spiked endotoxin over time. Because the expectation
of having such LER data as part of the BLA is relatively recent (2,3), no experimental procedures for
conducting spike/hold studies have yet been established. And while ≤50% recovery is generally con-
sidered as the threshold for LER designation, no method for percentile recovery calculation has been

standardized or harmonized. As a result, various methods have been used to obtain endotoxin recovery
calculations by different investigators. Whether endotoxin recoveries under a specific condition could
potentially vary depending upon calculation methods is not known, however. This is problematic in
terms of setting a common criterion to assess the potential risk of LER. Therefore, careful evaluation
to determine the potential impact of calculation methods on LER classification based on a 50% cutoff
is necessary. From the comparative evaluation described here, using the same dataset from four indi-
vidual spike/hold studies, endotoxin recoveries measured in percentile clearly may diverge enough to
alter the LER classification when using different calculation methods.
Percentile recoveries from the spike/hold studies of BTP-1 and BTP-2 are comparable across all time
points regardless of the calculation methods (Table 8.5.3-2 and Table 8.5.3-4). In the case of BTP-1,
endotoxin recoveries did not fall below 50% throughout the 14-day hold duration for all three batches
(Table 8.5.3-2). While endotoxin recoveries showed steady decline, and eventually fell below 50%
across all three batches on Day 14 for BTP-2, the three methods showed comparable results through-
out all time points monitored.
In contrast, percentile LPS recoveries from the spiked BTP-3 varied significantly depending upon the
calculation methods as shown in Table 8.5.3-6. Noticeably, percentile recoveries determined by Method
III were significantly higher than those calculated by Methods I and II throughout the entire time period
(Table 8.5.3-6). The actual endotoxin activities determined by LAL method in both BTP-3 and the
LRW control were much higher than the targeted value of 20 EU/mL at the initial time point (Table
8.5.3-5). Endotoxin activities in the LRW control showed some reduction through Day 3 but stabi-
lized at ~50 EU/mL throughout the remainder of the study (Table 8.5.3-5). On the other hand, the
activities in all three batches of BTP-3 steadily declined over time to approximately 20 EU/mL on Day
8 (Table 8.5.3-5). LPS activities recovered from BTP-3 fell to below 50% on Day 7 per Method I or II
as shown in Table 8.5.3-6. On the other hand, the percentile recovery of LPS was nearly 100% on Day
8 according to Method III (Table 8.5.3-6). This clearly illustrates the importance of harmonizing the
calculation method used to judge whether a particular product/matrix could potentially cause LER. The
batch release method for BTP-3 was verified in accordance with USP <85> using Tris buffer containing
Pyrosperse™ as diluents to improve the positive product control recoveries. In order to mimic the batch
release BET method, the same Pyrosperse™-containing Tris buffer was used as a diluent in the BTP-3
spike/hold study instead of LRW. The substantially higher LPS activity observed, relative to what was
expected (theoretical value) in the BTP-3 spike/hold study, likely resulted from the use of Pyrosperse™ in
the diluent. This notion was supported by the result from an experiment in which LRW and Pyrosperse™-
containing buffer were compared side by side as diluents according to the USP <85> BET method. As
shown in Table 8.5.4-1, higher-than-anticipated LPS activity was observed in both BTP-1 and BTP-3
when Pyrosperse™ Tris buffer was used as diluent. On the other hand, LPS activity detected was at par
with the theoretical spiking level of 20 EU/mL. Apparently, Pyrosperase™ can significantly enhance the
LPS detectability by the USP <85> BET method. Because Method III uses the theoretical LPS activity as
the denominator, significantly higher LPS activity (due to enhancement by Pyrosperase), resulted in the
overestimation of the percentile recoveries as seen in the BTP-3 case.
Table 8.5.4-1 Effect of Pyrosperse™ on LPS Activities by USP <85> BET Method*
Theoretical Spiking
Material Diluent Dilution Fold LPS (EU/mL)#
(EU/mL)
20 LRW 100 20.2
BTP-3
20 Pyrosperse™ Tris Buffer 200 41.0
20 LRW 100 18.9
BTP-1
20 Pyrosperse™ Tris Buffer 200 45.6
* BTP-1 and BTP-3 were spiked with LPS at the 20 EU/mL targeting concentration.
#
LPS-spiked materials were held at 2 °C-8 °C for one day and then sampled for LPS detection by USP <85> BET method
using LRW- or Pyrosperse™-containing Tris buffer as diluents.

Similar to what was observed in the BTP-3 study, the percentile LPS recoveries from the spiked BTP-
4 were dependent on calculation methods (Table 8.5.3-8). However, Method II gave systematically
higher results than Methods I and III in the BTP-4 study (Table 8.5.3-8). While the measured LPS
activity in the LRW control matched closely the theoretical value (444 EU/mL) at the initial time
point, significantly higher activity was recovered from all three BTP-4 batches (Table 8.5.3-7), imply-
ing an enhancing effect of the BTP-4 matrix. LPS activity recovered from BTP-4 fell below 50% on
Day 4 for all three batches per Method I (14-31%) or III (21-48%) as shown in Table 8.5.3-8, con-
stituting LER. According to Method II, only one of three batches showed recovery below 50% (batch
C) on Day 4 (Table 8.5.3-8). While the difference in percentile recoveries calculated by different
methods was not as great as that in the BTP-3 case, it still impacted the timing of LER classification
based on the 50% cutoff.
The three different methods used for calculating endotoxin recoveries in an attempt to determine LER
designation based on the 50% cutoff criterion could potentially be impacted. Comparing the data
from these four spike/hold studies revealed two scenarios: First, there were no differences in endotoxin
recoveries among the three methods with regard to LER designation (BTP-1 and BTP-2; Table 8.5.3-
2 and Table 8.5.3-4). Second, the percentile endotoxin recoveries calculated by the three methods dif-
fered and impacted the LER designation as seen in the cases of BTP-3 and BTP-4 (Table 8.5.3-6 and
Table 8.5.3-8). When endotoxin activities recovered from testing at the initial time point and from
LRW controls are comparable to one another as well as to the targeted spiking amount (theoretical
level), the three methods should give very similar percentile recoveries. In the BTP1 and BTP-2 stud-
ies, LPS activities recovered from the product batches, as well as the LRW controls, at the initial time
point (Table 8.5.3-1 and Table 8.5.3-3) fell well within the 50-150% range of the theoretical spiking
LPS level and the percentile recoveries were similar (Table 8.5.3-2 and Table 8.5.3-4). By compari-
son, in the BTP-3 and BTP-4 studies and the LRW controls, LPS activities recovered from the product
batches at the same points deviated greatly from the theoretical (targeted) LPS level (Table 8.5.3-5 and
Table 8.5.3-7). For example, LPS activities recovered from BTP-3 batches and the LRW control at the
initial time point were 2.2 to 3.6 times that of the targeted spiking level of 20 EU/mL (Table 8.5.3-5).
Method I used LPS activities recovered from the spiked product at the initial time point as the de-
nominator, which makes sense. However, this method may encounter situations that need additional
criteria or run the risk of overestimating the loss of endotoxin activities. Certain product matrices may
cause instantaneous loss of endotoxin activity upon spiking, which may require an additional control,
such as spiked LRW, to ensure that the loss of endotoxin is due to human error. Alternatively, product
matrices could have a transient enhancing effect, resulting in higher-than-anticipated endotoxin ac-
tivity at the initial time point as seen in the case of BTP-4 (Table 8.5.3-7). In turn, this could cause
underestimation of the percentile recovery at a given time point.
Method III used theoretical LPS activities as the denominator to calculate the percentile recoveries.
When the measured endotoxin activities recovered from LPS-spiked products at the initial time point
or LRW are closely similar, Method III works as well as Methods I and II. However, if the measured
endotoxin activities are significantly higher than the theoretical spiking amount, as seen in both BTP-3
and BTP-4 studies (Table 8.5.3-5 and Table 8.5.3-7), Method III could systematically overestimate
the percentile recoveries, potentially misleading the conclusion. While not seen in these studies, the
measured endotoxin activities may also be significantly lower than the theoretical spiking amount,
resulting in an underestimation of the percentile recoveries.
8.5.5 Conclusion
Since the purpose of the spike/hold study is to determine whether endotoxin spiked into a given prod-
uct would lose activity over time, knowing the starting amount of endotoxin is critical. Thus, it can
be used as the denominator to accurately determine the percentile recoveries at different time points
throughout the study duration. The three methods compared in this study used different denomina-
tors. Based on the side-by-side comparison using data from four independent spike/hold studies,

Method II (using LPS recovered from the spiked LRW as the denominator) is superior to Method I
and Method III. As shown in Table 8.5.5-1, LPS spiked into LRW exhibited remarkable stability over
the duration of the hold time in all four studies.
Table 8.5.5-1 Endotoxin Activities Recovered from LPS Spiked LRW Controls Over Time
Endotoxin Activities (EU/mL) Recovered at Indicated Time Points Activity

Study
Ratio*
Initial Day 1 Day 3 Day 4 Day 7 Day 8 Day 14
BTP-1 20.8 25.8 21.8 NM #
24.2 NM *
22.2 1.08
BTP-2 26.1 26.7 20.1 NM *
22.0 NM *
27.6 1.06
BTP-3 72.7 60.4 55.2 50.0 NM* 57.1 NM* 0.79
BTP-4 404 367 360 327 321 NM* NM* 0.80
* Activity ratio = (EU/mL at the last time point) ÷ (EU/mL at the initial time point)
#
NM = not measured
In summary, the evidence provided indicates that Method II, which uses LPS recovered from the
spiked LRW controls as the denominator, is relatively more consistent and more reliable in estimating
whether a given product matrix causes the loss of detectable endotoxin over time.
8.5.6 References
2. Endotoxin challenge: A regulatory perspective. Hughes, P. Presented at the PDA 9th Annual Global Confer-
ence on Pharmaceutical Microbiology, Bethesda, MD, October 2014.
3. Low Endotoxin Recovery: An FDA Perspective. Hughes, P, et al., April 10, 2015, BioPharma Asia. https://
biopharma-asia.com/magazine-articles/low-endotoxin-recovery-an-fda-perspective/ (accessed November
19, 2018).
4. Endotoxin contamination of parenteral drugs and radiopharmaceuticals as determined by the limulus amebocyte
lysate method. Twohy, C; et al., 1984, J Parenter Sci Technol, Vol. 38, pp. 190-201.
5. Interference in the limulus amebocyte lysate assay for endotoxin determination in peritoneal dialysis fluids and
concentrates for hemodialysis. Bohrer, D; et al., 2001, J Pharm Biomed Anal. Vol. 26, pp. 811-818.
6. Resolving LAL interferences. Cooper, J. 1990, J Parenter Sci Technol, Vol. 44, pp. 13-15.

8.6 Case Study 6: Lipopolysaccharide Masking and Resolution of Low

Endotoxin/Lipopolysaccharide Recovery (LER/LLR) in a Citrate Buffer
Monoclonal Antibody containing Polysorbate and Chelating Agent
LER was observed for a monoclonal antibody product, composed of a monoclonal antibody at 25 mg/
mL in 20mM citrate, 150mM sodium chloride, 20uM DTPA, 0.025% polysorbate 80 buffer with a
pH of 6.0. Diethylenetriaminepentaacetic acid (DTPA) is a strong chelator of divalent cations. The
drug product, drug substance, and upstream process samples (does not contain polysorbate) had no
measurable endotoxin when tested using compendial kinetic methods. Purified lipopolysaccharide
(LPS) in the form of Control Standard Endotoxin (CSE) was used to artificially contaminate the prod-
uct to demonstrate the stability of assayable endotoxin as required by a U.S. FDA guidance document
(1). The addition of the CSE surrogate contaminant rapidly lost biological endotoxin activity when
added directly into these product matrices. Naturally occurring endotoxin (NOE) from numerous
bacterial sources used as a surrogate contaminant remained stable. Restoration or reactivation of the
biological activity of CSE was achieved by using an experimental design that reduces assay variability
and replacing divalent cations to the sample test matrix prior to limulus amebocyte lysate (LAL) test-
ing. The sample pretreatment had no effect on the NOE activity. NOE represents a true contamina-
tion event and is being proposed as the appropriate artificial surrogate for contamination (2).
8.6.1 Introduction
Experiments were performed to demonstrate the stability of assayable endotoxins using both CSE
and NOE as surrogate contaminants for endotoxin. CSE rapidly lost biological activity while NOE
from numerous sources of bacteria remained stable and biologically active. Additional experimenta-
tion shown in Figure 8.6.3.1-1 also revealed that the masking phenomenon is temperature-dependent
with higher temperatures masking CSE at accelerated rates in comparison to 2-8 ºC. Storage at refrig-
erated temperatures demonstrated masking by Day 3. The masked samples did retain some biological
activity which may reflect the purity of the LPS molecules used to prepare the vendor CSE. The results
were obtained using Charles River Endosafe® reagents for kinetic turbidimetric (CRE KT) and Lonza
reagents for kinetic chromogenic (KQCL) testing.
A sample preparation method employing Charles River Endosafe® cation replacement buffer (1 M Tris,
500 mM MgSO4) was developed to comply with the U.S. FDA request to demonstrate the stability of
assayable endotoxin using CSE as a surrogate contaminant. The sample preparation method would allow
the reaggregation of the disaggregated biologically inactive LPS molecules to a state of biologically active
detectable molecules. This procedure for artificial contamination followed by masking and subsequent
demasking that would then allow detection of biologically active CSE forms by any test method (5).

The enhancement/inhibitions properties of the product matrix must be well defined before any LER
studies can be conducted. The noninterfering dilution of the product is where the positive product
controls (PPC) are 50-200% of a theoretical value. The noninterfering dilution (or greater) is used for
LER studies on product contaminated with CSE.
A similar enhancement inhibition profile should be performed to determine the optimal noninterfer-
ing concentration of the stock divalent cation solution (1 M Tris, 500 mM MgSO4). The concentra-
tion determined via the inhibition profile should be noninhibitory with the test method and vendor
lysate in the absence of product.
The LER studies were conducted by contaminating the undilute product with an appropriate amount of
LPS/CSE and NOE. The resultant mixture, when diluted to the noninterfering test dilution, would result
in the theoretical value falling at approximately the midpoint of the standard curve used for the assay.
Studies were conducted using both pyrogen-free water (PFW) with no demasking capability and cation

replacement buffer (CB) with demasking capability as diluents for the product sample. PFW controls were
contaminated and assayed simultaneously with the product samples to negate the day-to-day variability of
the assay. Using an experimental design in which the contamination of the longest held sample is performed
first, working backwards from the longest time to T=0, and using individual containers for each test day, all
contaminated samples can be tested simultaneously which negates the day-to-day assay variability.
The case studies were performed by adding contaminants in concentrations that, when tested at 1:10
dilution (noninterfering test dilution for Charles River Endosafe® KT) and at 1:20 dilution (noninter-
fering test dilution for Lonza KQCL), would measure endotoxin at approximately the higher to middle-
of-the-standard curve range (5, 2, 0.5, 0.125, and 0.05 EU/mL). The inoculum level was kept small, in
mcL amounts. Contaminant stock concentrates should have endotoxin activity values high enough to
minimize dilution of the product (<1% of total volume of the sample) with the contaminant. The stock
CSE contaminant was approximately 1000 EU/mL. The CSE-added inoculum level was 50 mcL (for
5 mL product sample) or resultant 10 EU/mL. At a 1:10 test dilution, the theoretical concentration
would be approximately 1.0 EU/mL and, at a 1:20 test dilution, would be 0.5 EU/mL. The NOE sur-
rogate contaminant, Enterobacter cloacae endotoxin, measured approximately 800 EU/mL. The NOE-
added inoculum level was 50 mcL (for 5 mL product sample) or a resultant 8 EU/mL for Experiment
B. The NOE-added inoculum level was 65 mcL (for 5 mL product sample) or a resultant 10 EU/mL for
Experiment C. At a 1:10 test dilution, the theoretical concentrations would be approximately 0.8 (Exp-
B) and 1.0 EU/mL (Exp-C) and, at a 1:20 test dilution, would be 0.4 (Exp-B) and 0.5 EU/mL (Exp-C).
The pyrogen-free water controls should be used to determine percentage of activity recovery from the
samples. Theoretical values may not be accurate. The theoretical activity value of CSE will only be
valid if the samples with CSE have been contaminated from the same CSE/Lysate-paired lot used for
testing, since CSE is reconstituted to a stock value when standardized with the lysate reagents from
a given method and vendor. A value of not more than a 50% loss of activity when compared to the
water control has generally been accepted as evidence of the absence of LER, and a reactivation of at
least 50% CSE is satisfactory for reactivation of CSE biological activity.
8.6.2.1 Enhancement/Inhibition Testing of Monoclonal Drug Product and Optimization of the Cation
Replacement Buffer Concentration
• Kinetic Turbidimetric (KT) from Charles River Endosafe®: The minimum noninhibitory concen-
tration tested was a 1:10 dilution in both PFW and 1:10 in 1/10 strength cation buffer (100 mM
Tris, 50 mM MgSO4) as diluents.
• Kinetic Chromogenic (KC) from Lonza (KQCL): The minimum noninhibitory concentration test-
ed was a 1:20 dilution in both PFW and 1:20 in 1/40 strength cation buffer (25 mM Tris, 12.5
mM MgSO4) as diluents (Figure 8.6.2.1-1).
Vortex CSE and NOE (use same stock vial source) before contaminating samples
Store samples 2–8°C
Minimize vortexing contaminated samples until T=0 test day
CSE
Sample
NOE
CSE
PFW
NOE
Day 7 Day 5 Day 3 Day 2 Day 1 4 hour Time=0

Dec 01 Dec 03 Dec 05 Dec 06 Dec 07 Dec 08 Dec 08
Figure 8.6.2.1-1 Schematic of test vial preparations for conducting the hold study

• Contamination and Masking:

– Placed 5 mL of drug product samples into a vial for each test day
– Starting with the longest test day, added artificial contaminant to the undilute sample in μL amounts
to the desired concentration which, when diluted for testing, will fall approximately at the midpoint of
the standard curve; continued adding surrogate contaminant for each day until T=0
– Prepared a set of pyrogen-free water (PFW) controls in the same manner
– Stored samples at 2-8 ºC between contamination days until tested
• Pre-test: Sample treatment (demasking) and dilution

– Vortexed contaminated sample
– Prepared sample dilution to appropriate concentration for the test method using the optimized cation
buffer as diluent and pyrogen-free water as diluent for comparison
– Vortexed diluted samples and allowed diluted samples to equilibrate for at least 10 minutes at room
temperature, vortexing intermittently at least three times during this time period
– Proceeded to test all samples on same day
• Test Parameters Experiment A:

– Test intervals: Day 3 (after masking), Day 3 after demasking
– Storage between test intervals: 2-8 °C
– Surrogate Contaminant: Control Standard Endotoxin (1000 EU/mL stock)
– Theoretical contaminant concentrations to be tested: 2, 5, 10, 15, 20, and 30 EU/mL
– Test methods: CRE KT and Lonza KQCL
• Test parameters of Experiment B:

– Test intervals: Day 0, Day 3 for CSE (after masking), and Day 4 for NOE (after masking)
– Surrogate Contaminants: Control Standard Endotoxin (1000 EU/mL stock); Naturally
Occurring Endotoxin (800 EU/mL stock from E. cloacae)
– Theoretical contaminant concentrations to be tested: CSE 10 EU/mL, NOE 8 EU/mL
• Test parameters of Experiment C:

– Test intervals: Day 0, 4 hr, Day 1, 2, 3, 5, 7
– Surrogate Contaminants: Control Standard Endotoxin (1000 EU/mL stock); Naturally Oc-
curring Endotoxin (800 EU/mL stock from E. cloacae)
– Theoretical contaminant concentrations to be tested: CSE 10 EU/mL, NOE 10 EU/mL

8.6.3.1 Results
The results of Experiments A, B, and C are presented in Figures 8.6.3.1-1 to -4 (A), Figures 8.6.3.1-5
to -7 (B), and Figures 8.6.3.1-8 to -11* (C).
* Note: Samples are upstream processing with chelator but without polysorbate

Figure 8.6.3.1-1 Pyrogen-Free Water (PFW) with Control Figure 8.6.3.1-2 Drug product with Control Standard
Standard Endotoxin (CSE) Charles River Endosafe® Kinetic Endotoxin (CSE) Charles River Endosafe® Kinetic
Turbidimetric Control after 3 days at 2-8 °C diluted with PFW Turbidimetric Control after 3 days at 2-8 °C diluted with
and 100 mM Tris, 50 mM MgSO4 (1/10 Strength Cation PFW and 100mM TRIS, 50mM MgSO4 (1/10 strength
Buffer) Cation buffer)
Figure 8.6.3.1-3 Pyrogen-Free Water (PFW) with Control Standard Endotoxin (CSE) Lonza KQCL Kinetic Chromogenic
Control after 3 days at 2-8 °C diluted with PFW and 25 mM Tris, 12.5 mM MgSO4 (1/40 Strength Cation Buffer)
Figure 8.6.3.1-4 Drug Product with CSE Lonza KQCL Kinetic Chromogenic Control after 3 days at 2-8 °C diluted with
PFW and 25 mM Tris, 12.5 mM MgSO4 (1/40 Strength Cation Buffer)

Figure 8.6.3.1-5 Reactivation of CSE (LPS) Contaminated into mAb/Citrate/PS Product

10 EU/mL After 3 Days at 2-8°C Charles River Endosafe® Kinetic Turbidimetric
Figure 8.6.3.1-6 Kinetic turbidimetric testing of E. cloacae NOE
Figure 8.6.3.1-7 Kinetic chromogenic testing of E. cloacae NOE
Figure 8.6.3.1-8 Monoclonal Antibody in PFW and 100 mM Tris, 50 mM MgSO4 (1/10 Strength Cation Buffer)
Kinetic Turbidimetric (CRE) with CSE on Day 7

Figure 8.6.3.1-9 Monoclonal Antibody in PFW and 100 mM Tris, 50 mM MgSO4 (1/10 Strength Cation Buffer) Kinetic
Turbidimetric (CRE) with NOE on Day 7
Figure 8.6.3.1-10 Monoclonal Antibody in PFW and 25 mM Tris, 12.5 mM MgSO4 (1/40 Strength Cation Buffer) Kinetic
Chromogenic (Lonza) with CSE on Day 7
Figure 8.6.3.1-11 Monoclonal Antibody in PFW and 25 mM Tris, 12.5 mM MgSO4 (1/40 Strength Cation Buffer) Kinetic
Chromogenic (Lonza) with NOE on Day 7

8.6.3.2 Discussion
The studies demonstrate that the masking phenomena with CSE/LPS occurs at 2-8 ºC by Day 3, and
immediately at room temperature for this formulation of product. The data also demonstrate that the
phenomena is not specific to vendor, test method, or sample vessel with all test variations demonstrat-
ing a loss of CSE/LPS activity. The masking with CSE occurs with or without polysorbate present for
this formulation. The optimized cation buffer is capable of reversing the loss of CSE/LPS activity, al-
though the loss of CSE activity and subsequent reactivation using divalent cation replacement may not
be 100% complete. Masking is a temperature-dependent phenomenon which occurs faster at elevated
temperatures. The use of dispersing reagents such as Pyrosperse™ showed no significant effect on CSE/
LPS recovery. NOE did not exhibit the masking phenomenon with this product formulation and was
unaffected by the cation replacement treatment.
The LPS masking mechanism is caused by the chelation effect of the buffer components and the aggre-
gation state of the molecules (6). While each therapeutic monoclonal antibody formulation is unique
in terms of structure and chemistry, the data presented demonstrated success at resurrecting the CSE
activity by using commercially available cation replacement buffer. Stock cation buffer (0.5 M MgS04,
1 M Tris) was diluted in pyrogen-free water to be used as a sample preparation diluent. The concentra-
tion of the cation buffer had to be optimized for each lysate vendor, method/technique, and product
formulation prior to use.
8.6.4 Conclusion
The data demonstrate that satisfactory hold time recovery studies of CSE were achieved by adding
a sample pretreatment and dilution step using cation replacement buffer. The data also demonstrate
that the NOE is not subject to masking in this product formulation and is not affected by the cation
replacement buffer treatment.
The following sample treatments are satisfactory for recovery of masked CSE/LPS for the following products:
• Charles River Endosafe® Kinetic turbidimetric testing: Dilute samples 1:10 (or greater) in
100 mM Tris, 50 mM MgSO4, vortexed intermittently for 10 minutes prior to testing
• Lonza KQCL Kinetic chromogenic testing: Dilute samples 1:20 (or greater) in 25 mM Tris, 12.5
mM MgSO4, vortexed intermittently for 10 minutes prior to testing
8.6.5 References
1. Guidance for Industry Pyrogens and Endotoxins Testing Questions and Answers. U.S. Food and Drug Adminis-
tration. 2012. https://www.fda.gov/downloads/drugs/guidances/ucm310098.pdf.
2. The use of endotoxin as an analyte in biopharmaceutical product hold time study. Bolden, J; et al., 2015, Phar-
macopeial Forum, Vol. 41(5).
4. Understanding the LAL and Rabbit Pyrogen Tests for LER studies. Platco, C. 2015, Am Pharm Rev. Vol. 18(6),
pp. 108-112.
5. Resolution of Low Endotoxin/Lipopolysaccharide Recovery (LER/LLR) Testing. Platco, C; et al., 2016, Am
Pharm Rev, Vol. 19(5), pp. 20-25.

8.7 Case Study 7: Evaluation of an Endotoxin Demasking Protocol

The aim of this study was to develop a sample preparation protocol for the reliable detection of po-
tential endotoxin contaminants in a problematic sample matrix for bacterial endotoxin testing (BET).
The procedure developed is intended to demask a sample containing citrate and polysorbate formula-
tion components. Initially, the performance of a hold time study showed that this sample was affected
by LER. The recovery of spiked endotoxin was <20%, indicating that the endotoxin was masked.
Therefore, the potential for false-negative or nondetection of endotoxin-contaminated solutions exists.
In order to ensure the reliable detection of endotoxin in a masked sample, a protocol was developed to
demask the endotoxin. Accordingly, dedicated demasking components were mixed thoroughly with
the sample. After incubating for approximately one hour, the sample was diluted and analyzed using a
kinetic chromogenic Limulus amebocyte lysate (LAL) assay.
The data presented here clearly demonstrate that the protocol developed enables demasking of en-
dotoxin from a relevant BET sample. The overall results show that the mean recovery of endotoxin
in the demasked samples was approximately 90%–100%. Furthermore, the method showed reliable
endotoxin recovery under all test conditions. Taken together, the application of the dedicated sample
preparation protocol on the LER-effected sample ensures robust endotoxin recovery and eliminates
the risk of false-negative test results due to the LER effect using compendial LAL tests.
8.7.1 Introduction
Recently, inadequate endotoxin detection by LAL-based assays has been reported in various biologi-
cal pharmaceutical drug products after a certain holding time (1). Due to the highly regulated and
complex nature of the production of biopharmaceuticals, the effect of LER has become the focus of
activity for various companies and regulatory agencies (2,3). LER poses a potential risk to human
health, as masked endotoxin can induce the expression of pro-inflammatory cytokines and surface ac-
tivation markers even at low concentrations (4). Various efforts have been made to overcome LER and
to detect masked endotoxins, either by the use of newly developed techniques, like the EndoLISA® or
the Monocyte Activation Test (MAT), or by the use of different additives to induce a demasking effect
on a sample (e.g., drug product). The U.S. Food and Drug Administration requires hold time stud-
ies during the Biologics License Application (BLA) submission phase of drug development in which
the detection of a spiked endotoxin is demonstrated after a certain hold time, e.g., comparable with
the time of a production process (5). As the induction of endotoxin masking can depend on several
product- and process-specific parameters (6,7), the development of a demasking protocol is specific
for each drug product and must be established on a case-by-case basis.
A demasking protocol must be robust, reproducible, and reliable as it is used for in-process controls
or the final release of a drug product. Various other factors may also influence the test. For example,
the LAL, which is derived from a natural source, might differ from batch to batch. Additionally, the
quality and formulation of the control standard endotoxin (CSE) used might vary. Here, the improve-
ment of current LAL testing on an LER-effected sample relevant for BET is shown by application of
a dedicated sample preparation protocol. The results clearly demonstrate the validity of this protocol
according to robustness, reproducibility, and reliability criteria.
8.7.2 Materials
• Demasking: EndoGrade Glass Test tubes and Endo-RS components (Hyglos GmbH, Bernried,
Germany)
• Absolute ethanol for analysis: Ph. Eur. grade (Merck Chemicals GmbH, Darmstadt, Germany)
• LAL reagent water (LRW): Charles River Laboratories
• Kinetic chromogenic LAL tests (KQCL) and high-potency control endotoxin standard (CSE):
Derived from an Escherichia coli O55:B5 (Lonza Cologne GmbH, Köln, Germany)

• Reference standard endotoxin (RSE): (European Directorate for the Quality of Medicines,
Strasbourg, France)
• Naturally occurring endotoxin (NOE): Stock solutions (Hyglos GmbH, Bernried, Germany)
• Eppendorf 2 mL cups: Purchased from Eppendorf GmbH, Wesseling-Berzdorf, Germany
• All endotoxin-free pipette tips: Purchased from Sarstedt, Nümbrecht, Germany
• Relevant sample for BET investigation: Provided by the study sponsor
8.7.3 Methods
8.7.3.1 Preparation of Endotoxin Standards for Hold Time Study
The RSE was vortexed (1400 rpm) for 30 min in EndoGrade glass test tubes using endotoxin-free LRW
to obtain a stock concentration of 10,000 EU/mL. For the high potency endotoxin, a stock concentra-
tion of 1.3 x 106 EU/mL was prepared similar to the RSE. The NOE from Enterobacter cloacae, Burk-
holderia cepacia, and Pseudomonas aeruginosa were provided in solution. From the corresponding stock
solutions, 100-fold concentrated working solutions were prepared according to the specific endotoxin
limit of the sample. Between dilution steps, the endotoxin samples were vortexed (1400 rpm) for 2 min.
8.7.3.2 Reverse Mode for Spiking

The solid sample was dissolved in LRW and aliquoted into 1 mL sample containers. All samples were
stored at room temperature (RT). At each time point (0, 1, 2, 4, 6, 8, and 24 hours), an aliquot of
the sample and the LRW control (WC) were spiked with an endotoxin concentration at the sample-
specific endotoxin limit and stored at RT until measurement. For measurements, all samples were
diluted and measured using an LAL test according to test specifications provided.
8.7.3.3 Preparation of the Demasking Components and Demasking Protocol

All demasking components (A, B, D, and E) were taken from the Endo-RS kit, which was dedicated
for demasking. The components were adjusted to RT prior to use. While components A, B, and D were
diluted in LRW, component E was diluted in 70% (v/v) ethanol in 2 mL Eppendorf cups according to
the dedicated sample preparation protocol. All dilutions were vortexed (1400 rpm) for 30 seconds and
all dilution steps were performed in EndoGrade glass test tubes. Demasking components A, B, and D
were mixed stepwise with the sample and vortexed for 4 min at 1400 rpm after every transfer. Compo-
nent E was vortexed for 20 min after mixing. The samples were kept without vortexing at RT for 30 min
(demasking time). Afterwards, the demasked sample mix was vortexed for 2 min (1400 rpm), diluted in
LRW, and transferred to the LAL test plate. In parts of this study, the demasking time, the vortex time,
the temperature, and the endotoxin concentrations were varied as described in Section 8.7.4.
8.7.3.4 LAL Chromogenic Test and Validation Criteria

The diluted sample (100 µL) was transferred to a microtiter plate and an LAL test using the KQCL kit
was performed according to the manufacturer’s package insert instructions. A three-point standard curve
in the concentration range between 0.01 EU/mL and 1.0 EU/mL was used. All tests were performed
using a plate reader with a filter at 405 nm (BioTek Eon or a BioTek Elx808) and evaluated using the
WinKQCL software (Lonza). For all sample measurements, a positive product control (PPC 0.1 EU/
mL) was included. A measurement was considered to be valid when the PPC recovery was achieved with-
in the 50% to 200% recovery criteria. The recovery rates were calculated against the WC at time point 0.

To conduct a meaningful demasking study, a reliable, robust, and reproducible sample preparation pro-
tocol must be established. For this study, a dedicated protocol for a BET sample relevant to the pharma-
ceutical industry was evaluated. For investigation of LER in the given sample, a hold time study using the
reverse mode was used. The data showed that the sample was severely affected by LER (endotoxin mask-
ing) using the compendial LAL test. To demask endotoxin, the sample was pretreated using components
from Endo-RS (A, B, D, E). After demasking, the sample was measured using a compendial LAL test.

8.7.4.1 Hold Time Study Shows Low Endotoxin Recovery

To evaluate whether the sample is affected by low endotoxin recovery, a reverse hold time study was
performed over three days at RT. The reverse mode was chosen in order to avoid assay-to-assay varia-
tion as all time points were measured simultaneously in the same analytical run. The sample clearly
showed LER over time, as indicated in Figure 8.7.4.2-1a. After a hold time of only 5 h, the endotoxin
recovery in the sample dropped below 50%, indicating a strong masking effect in the sample.
8.7.4.2 Development of a Sample Preparation Protocol for Demasking

As standard procedures were not able to detect the masked endotoxin, alternative LAL assays were
tested; however, the use of alternative LAL assays or the use of a commercial dispersing agent could
not resolve the LER issue. Therefore, a more extensive demasking toolbox was used. The Endo-RS
toolbox contains several components that are dedicated for endotoxin demasking of complex samples
and formulations. As no universal demasking protocol is available, the required components and cor-
responding concentrations had to be determined experimentally.
For developmental purposes, the EndoLISA® assay was employed initially because it is less influenced
by interference from demasking solutions. With this method, the endotoxin is first captured by bind-
ing to an endotoxin-specific ligand and subsequently detected by a combination of a recombinant Fac-
tor C and a highly sensitive fluorescent detection method. As the EndoLISA® technology is currently
not a compendial test, the protocol developed was transferred afterwards to a compendial chromo-
genic LAL assay, a standard method in quality control testing of pharmaceutical products.
After EndoLISA® testing, a sample preparation protocol based on Endo-RS technology was combined
with a kinetic chromogenic LAL test method. The final demasking protocol in this sample contained
components A, B, D, and E; the optimal concentration range of the components was determined us-
ing titration experiments. In brief, the sample was mixed stepwise with the demasking components
and incubated at RT. To overcome test interference effects, a further dilution step was included before
transfer of the sample to the LAL assay.
8.7.4.3 Evaluation of the Sample Preparation Protocol on a Sample Relevant for BET
Using the dedicated sample preparation protocol, almost 100% of spiked endotoxin could be recov-
ered (Figure 8.7.4.2-1b). The recovery of the demasked sample was similar to the WC, while no
recovery was seen in the masking control (MC, sample before demasking; Figure 8.7.4.2-1b). As the
time point of a contamination event is not known, the developed protocol has to be independent of
the masking time. To evaluate this criterion, a demasking kinetic of over 72 hours was performed. Five
aliquots of the sample were masked simultaneously and demasked at the indicated time points. The
demasking was successful at all time points, as compared to the WC (Figure 8.7.4.2-1c).
8.7.4.4 Robustness and Reliability of Demasking Protocol

To evaluate the reproducibility of the given sample preparation protocol, the sample was repeated on
two different days simultaneously by two different operators. Additionally, the protocol was repeated
21 times to evaluate the performance over several runs. No difference was seen in runs performed by
different operators and the variance between runs was within a range of 20% (Figure 8.7.4.4-1).
To further evaluate the robustness of the protocol, variation in 1) the incubation time after demasking, 2)
the vortex time between the addition of the demasking components, and (c) different incubation tempera-
ture during demasking were examined (Figure 8.7.4.4-2). No difference was observed in the variation of
the demasking time between 20 min and 40 min (Figure 8.7.4.4-2a) or in a variation of the vortex times
between the addition of the demasking components (Figure 8.7.4.4-2b). It was concluded that variations
in the demasking times and the vortex times studied have no influence on the demasking efficiency.
To evaluate the effect of variations in temperature during the demasking procedure, the sample, de-
masking components, and demasking was performed at 18 °C, 22 °C, and 26 °C. The best recoveries

A 250
200
150
Recovery (%)
100
50
0
0 5 10 15 20 25 30
Masking time (h)
• Recovery to WC Tmp 0h (%)
B 200
150
Recovery (%)
100
50
0
Demasking MC WC Tmp WC Tmp 0h
demasking
C 200 n Demasking
n Masking control
150
Recovery (%)
100
50
0
2h 4h 24 h 48 h 72 h
Figure 8.7.4.2-1 Reverse Hold/Time Study and Demasking Study of Endotoxin

(a) Endotoxin recovery of the sample at RT over three days. The endotoxin recovery% is given against the masking time in hours [h]. The sample was spiked
with endotoxin (RSE) and incubated up to 24 hours. The recovery was calculated against the WC at time point 0 h. The upper and lower limit of 50% and
200% are shown as grey lines.
(b) Endotoxin recovery% after demasking (n=75) in comparison with the masking control (MC, n=15), the LRW control (WC) at the time point of demasking
(Tmp demasking, n=12) and the WC (n=7) at time point 0 h. All bars represent the mean of at least three independent measurements. Error bars represent
the standard error of the mean. No error bars are shown for the masking control, as the determined values are partially lower than the sensitivity of the assay.
(c) Endotoxin recovery% after demasking at different time points after masking. The recovery of the demasked sample is shown side-by-side with the
corresponding MC after 2 h, 4 h, 24 h, 48 h and 72 h. The recovery of the demasked sample is shown as the mean of three independent demasking replicates.
All bars represent the mean of at least three independent measurements. Error bars represent the standard error of the mean. No error bars are shown for the
masking control, as it was determined in duplicate.

200
150
Recovery (%)
100
50
0
mean (n=21) Operator 1, Operator 2, Operator 2,
day 1 (n=3) day 1 (n=15) day 2 (n=3)
Figure 8.7.4.4-1 Reproducibility and Robustness of the Developed Demasking Protocol

The reproducibility of the developed demasking protocol performed by two different operators (operator 1, n = 3; operator 2, n
= 18) at two different days (Day 1: n = 3; Day 2: n = 18) is shown. The endotoxin recovery is shown in percent [%]. The first
column shows the mean of 21 measurements performed by two operators on two different days. Error bars represent the standard
error of the mean.
were obtained at RT and elevated temperatures (Figure 8.7.4.4-2c). In this case, it was concluded
that the temperature is an important factor influencing the demasking efficiency. Thus, for this sample
matrix, performing a demasking study in a lab with a temperature above 21 °C is recommended.
Besides robustness of operational procedures, the use of different lots of components were compared.
The particular lot-to-lot comparisons were performed side-by-side to avoid introducing other vari-
ables. Therefore, three different lots of the KQCL test (Figure 8.7.4.4-3a), three different lots of the
Endo-RS demasking kit (Figure 8.7.4.4-3b), and three different lots of the sample under study (Fig-
ure 8.7.4.4-3c) were compared. No difference in the demasking efficiency of the different lots were
observed, indicating that the protocol developed is robust.
8.7.4.5 Detection of Endotoxins from Different Species

Within a production facility, a potential endotoxin contamination event is not predictable; therefore,
the source of contaminating bacterial species can vary. To have a reliable test system, the demask-
ing procedure should be able to detect endotoxins from different species. To evaluate this method, a
purified high-potency endotoxin from E. coli O55:B5 was used. To show broad applicability of the
demasking protocol, the RSE and three different NOEs from P. aeruginosa, B. cepacia, and E. cloacae
were incubated in the sample and subsequently demasked (Figure 8.7.4.6-1a). After incubation and
before demasking, the endotoxins showed varying degrees of masking, which indicates that different
endotoxins possess different masking susceptibilities. However, after application of the demasking
protocol, the recovery resulted in a valid range for all endotoxin spike types. These data show that the
sample preparation protocol can, independently of the source and degree of masking, have sufficient
demasking capabilities.
8.7.4.6 Sensitivity of Demasking Protocol

In order to meet sample-specific requirements regarding test sensitivity, the masking and demasking
experiments were performed with endotoxin concentrations at the product-specific endotoxin limit.
Additionally, four different endotoxin concentrations (1.2-fold, 1.0-fold, 0.6-fold, and 0.2-fold of the
regulatory endotoxin-limit (Figure 8.7.4.6-1b)) were demasked. For all endotoxin concentrations,
full recovery after demasking was achieved. It was concluded that the demasking protocol developed
is valid between 1.2-fold and 0.2-fold of the regulatory endotoxin-limit. To exclude false-positive de-
masking results, the demasking protocol was applied to a nonspiked sample. No unspecific activation
was observed in a kinetic chromogen LAL test.

A 200
B 200
Recovery (%)
150 150
Recovery (%)
100 100
50 50
0 0
20 min 30 min 40 min 2-2-2-15 min 4-4-4-20 min 6-6-6-25 min
200
C
150
Recovery (%)
100
50
0
18°C RT 26°C
Figure 8.7.4.4-2 Robustness of Sample Preparation Protocol

(a) Endotoxin recovery as dependent on incubation time after demasking. The recover% is given against the incubation time in minutes after
demasking[min]. All bars represent the mean of at least three independent measurements. Error bars represent the standard error of the mean.
(b) Endotoxin recovery as dependent on vortex times after addition of demasking components. The recovery% is given against the vortex times during
demasking in min after addition of demasking components [min]. The demasking components were added stepwise to the sample. After adding components
A, B, and D, the samples were vortexed for 2 min, 4 min, and 6 min, respectively. After adding component E, the samples were vortexed for 15 min, 20 min,
and 25 min, respectively. All bars represent the mean of at least three independent measurements. Error bars represent the standard error of the mean.
(c) Endotoxin recovery as dependent on incubation temperature during demasking. The recovery% is given against the incubation temperature during demasking
[°C]. Demasking was performed at 18 °C, 22 °C (RT), and 26 °C respectively. All bars represent the mean of at least three independent measurements. Error
bars represent the standard error of the mean.
8.7.5 Conclusion
The results demonstrate that the method developed enables demasking of endotoxin from a sample
relevant for BET. The overall results show that the mean recovery of the demasked sample was between
90% and 100%. The method showed reliable endotoxin recovery under all conditions tested. Taken
together, the application of the dedicated demasking method ensures robust endotoxin recovery, there-
by diminishing the risk of false-negative test results while performing compendial LAL assays.
8.7.6 References
2. Results of a harmonized endotoxin recovery study protocol evaluation by 14 BioPhorum Operations Group
(BPOG) member companies. Bolden, J, et al., 2017, Biologicals, Vol. 48, pp. 74-81.
3. Low Endotoxin Recovery: An FDA Perspective. Hughes, P, et al., April 10, 2015, BioPharma Asia. https://biophar-
ma-asia.com/magazine-articles/low-endotoxin-recovery-an-fda-perspective/ (accessed November 19, 2018).
4. Biological activity of masked endotoxins. Schwarz, H; et al., 2017, Sci Rep, Vol. Mar; 7(7), pp. 44750.
5. Guidance for Industry Pyrogens and Endotoxins Testing Questions and Answers. U.S. Food and Drug Adminis-
tration. 2012. https://www.fda.gov/downloads/drugs/guidances/ucm310098.pdf.
6. Proteinase K digestion of proteins improves detection of bacterial endotoxins by the Limulus amebocyte lysate assay: ap-
plication for endotoxin removal from cationic proteins. Petsch, D, et al., 1998, Anal. Biochem, Vol. 259, pp. 42-47.

A B
200 200
150 150
Recovery (%)
Recovery (%)
100 100
50 50
0 0
lot 1 lot 2 lot 3 lot 1 lot 2 lot 3
C 200
150
Recovery (%)
100
50
0
lot 1 lot 2 lot 3
Figure 8.7.4.4-3 Comparison of Different lots of LAL Reagent, Endo-RS Components, and Sample Tested
(a) Comparison of three different LAL reagent lots. The endotoxin recovery% is shown against the different LAL reagent lots. No difference is seen between
different LAL reagent lots in the demasking efficiency. All bars represent the mean of at least three independent measurements. Error bars represent the
standard error of the mean. No error bars are shown for the masking control, as it was determined in duplicate.
(b) Comparison of three different lots of the Endo-RS kit components. The endotoxin recovery% is shown against the different Endo-RS kit lots. All bars
represent the mean of at least three independent measurements. Error bars represent the standard error of the mean.
(c) Comparison of three different lots of sample were tested. The endotoxin recovery% is shown against the different sample lots. All bars represent the mean
of at least three independent measurements. Error bars represent the standard error of the mean.
B 200
A 200 n Demasking 150

n Masking control
Recovery (%)
150
Recovery (%)
100
100
50
50
0 0
RSE P. aeruginosa B. cepacia E. cloacae 1.2 x ET limit 1.0 x ET limit 0.6 x ET limit 0.2 x ET limit
Figure 8.7.4.6-1 Detection of Endotoxins from Different Species and Sensitivity

(a) Recovery of endotoxin from different sources. The gray bar shows the recovery after demasking; the black bar represents the masking control (MC). All
bars represent the mean of at least three independent measurements. Error bars represent the standard error of the mean.
(b) Detection limit and sensitivity of the sample preparation protocol was tested for four different concentrations of the endotoxin according to the specific
endotoxin limit of the sample. All endotoxin recoveries are given in % and all bars represent the mean of at least three independent measurements. Error bars
represent the standard error of the mean.

8.8 Case Study 8: Strategy to Investigate and Overcome LER Driven by Protein
This case describes challenges in designing a robust and reproducible experimental setup in order to
investigate drug products and drug substances for LER. The products in question initially were not
suspected of triggering LER as they do not contain any of the known LER triggers, such as polysor-
bate. After investigating different study designs, the reverse study design was determined to be the
most beneficial compared to the traditional study design. In a reverse study design, the samples are
spiked at different days making it possible to analyze all samples in the same assay, thus avoiding assay-
to-assay variation. A study of spiked water control was performed in parallel to ensure that any decline
in recovery was due to LER and not an artifact of the experimental setup. The recovery of endotoxin
was calculated relative to the spiked water control at time point 0, measured in three replicates. A
reproducible experimental design was developed for both drug products (liquid) and drug substances
(solid). For the drug product described in this case, temperature-dependent LER was observed and
could be mapped to a drug substance. The observed decline in LER was temperature-dependent, as
the spike could be fully recovered at 2 °C-8 °C during the duration of the study but declined to below
50% recovery after 50 hours when the samples were kept at room temperature.
The case also describes two solid drug substances that introduced several challenges to the experimental
design which were overcome. With the design in place, the outcome of the studies was that one drug
substance gave rise to LER while the other did not. The overall conclusion pointed to the importance
of careful experimental design to obtaining reproducible results with sufficient low variability to draw
meaningful conclusions. Furthermore, LER was uncovered although it could not have been predicted
from the composition of the products.
8.8.1 Introduction
During recent years, the subject of endotoxin hold time studies has been intensively discussed in both
literature and in the pharmaceutical industry (1,2). The LER issue has primarily been associated with
specific formulation components such as polysorbate. These discussions question the relevance, study
designs, and interpretation of results. Furthermore, the discussions among industry and regulators
reveal how each laboratory is faced with local pharmaceutical production parameters that may differ
from the examples used in the literature and at conferences.
To find a suitable strategy for performing endotoxin hold time studies for a pharmaceutical product,
the study design was carefully considered and appropriate controls were applied. This case study pres-
ents one strategy for the design of experiments and interpretation of results based on the conditions
under local circumstances. The design strategy was applied for three different sample types: one drug
product and two drug substances. The results were evaluated against control samples and a different
outcome was obtained for all three compounds.
Two examples of investigations conducted for drug substances are presented here. In both cases, the
drug substance was a solid material consisting of dry protein. To investigate endotoxin hold time in a
dry sample matrix, the first challenge in the experimental design was that, when adding a liquid spiked
solution to the solid materials, the sample was no longer solid. Therefore, a strategy was developed to
perform hold time studies for solid materials, and the decision was made to resuspend the sample to a
production-relevant concentration.
Another major decision was choosing between a standard hold time study design or a reverse study.
In the standard study design, all samples are spiked the same day, either in one pool or in individual
containers, and measured at different time points. In the reverse study, aliquots of undiluted sample
are spiked at different time points during the hold time study and all samples are analyzed on the last
day of the study. In this way, all samples are analyzed in the same assay, eliminating assay-to-assay varia-
tion (see Figure 8.8.1-1). Hughes discusses the benefits and limitations of the two strategies in further
detail (1). In this case, the reverse strategy was applied for all three compounds.

all time points

Aliqutation of
ay
11
10
13
da
td
ay
ay
ay
ay
ay
ay
ay
rst
as
nd
nd
nd
nd
nd
nd
nd
n fi
nl
do
do
do
do
do
do
do
do
do
ike
ike
ike
ike
ike
ike
ike
ike
ike
Sp
Sp
Sp
Sp
Sp
Sp
Sp
Sp
Sp
Sample
3x Assay of all
Day Day Day Day Day Day Day Day samples at T0
Day
14 10 8 6 5 3 2 1
0
Start Timeline for Hold Time Study End
Figure 8.8.1-1 Reverse Hold Time Study Setup

Aliquots of undiluted sample are spiked at the different time points during the hold time study and all samples are analyzed at
the last day of the study. All samples from the hold time study are analyzed in the same assay.
When calculating the final percent recovery, several options are available when choosing the reference
value: theoretical spike value, value measured in spiked water at time point 0, or value measured in
spiked water at the given time point. Theoretically, if measurements are accurate, the value chosen
should not matter. For these experiments, the measured value of spiked water at time point 0 was used.
To ensure additional confidence that this value was accurately representing the true reference value,
the control measurement was made in triplicate and the mean of the three replicates was used as the
reference value. Additional replicates were made, as the chosen value would substantially impact the
overall study. No matter the study design applied, the choice of controls is crucial to interpreting the
results without bias.
In this case study, outcomes presented compounds both with and without LER properties. When
encountering a compound where LER cannot be excluded, the challenge of mitigating the risk of
undetected endotoxin in the production arises. Consequently, this case study proposes circumventing
actions for compounds with LER in pharmaceutical quality control laboratories.

8.8.2.1 Sample Constitution
The drug product described in this case was protein in the concentration 1-20 mg/ml in an aqueous
formulation containing phenol, an isotonicity agent, and metals of ~pH 7. Neither the drug sub-
stances nor drug product contained citrate, histidine, phosphate, surfactants, or polysorbate. The two
drug substances both comprised solid proteins.
8.8.2.2 Spiking of Samples

Samples and water controls were spiked with reference standard endotoxin (RSE) or high-potency
endotoxin, both supplied by Lonza. No naturally occurring endotoxins (NOE) were investigated in
this case study. The reverse study design was applied, where samples were spiked on different days and
all measurements taken just after time point 0. The samples were spiked with no more than 1% of the
sample volume if possible. Solid samples were resuspended to a process-related concentration prior to
spiking, taking the spike volume into account. The spike level selected was mid-curve and related to
the endotoxin limits.

The drug product samples were stored at either room temperature (process temperature) or 2-8 °C
(sample storage temperature). The drug substance samples were stored at 2-8 °C (sample storage tem-
perature). Standard pyrogen-free test tubes were used for the aliquoted samples.
8.8.2.3 Controls
The diluted spike concentration in LAL reagent water (LRW) and the sample were tested when initiat-
ing spiking. This control ensured that the expected spike level was met and that time would not be lost
if a simple mistake was made. The test was used as a prerequisite for continuing the experiment and
the values were not included in the evaluation of LER.
A sample without spike was included to establish the background level of endotoxin. The result would
normally be below the limit of quantification, so that was used to ensure that the endotoxin measured
in the spiked samples only derived from the spike. Unspiked water was included to confirm that the
bacterial endotoxin test (BET) assay was run normally.
Spiked LRW controls were included for each time point to indicate that no issues other than the
sample matrix would impact the study.
Spiked LRW at time point 0 was prepared in triplicate, as this measurement was used as a reference
point. The mean value of the three repetitions was used for further calculations.
8.8.2.4 Endotoxin Measurements

Endotoxin was measured according to the compendial kinetic chromogenic BET method. The ana-
lytical platform with reagents from Lonza and WinKQCL software was used for all measurements.
Standard laboratory nonpyrogenic plastic consumables were used for the endotoxin measurements.
The samples were diluted in LRW to a noninhibitory concentration as is done in routine analysis. The
solid samples were resuspended prior to spiking and diluted to a noninhibitory concentration imme-
diately before analysis.
8.8.2.5 Recovery Calculation Method

The mean of three repetitions of spiked LRW controls at T0 was used as a reference point. Recovery
was calculated for each time point as a percentage of the reference point.
8.8.3 Results
8.8.3.1 Drug Product
A drug product consisting of no published LER-prone components was investigated. Because the
product did not contain any published LER triggers, such as polysorbate or citrate, LER was not ex-
pected. Initially, the study was set up as a standard hold time experiment: spiking all samples and then
measuring the endotoxin content at different time points. A control study of water spiked with the
same concentration of endotoxin was performed simultaneously. Study results reported that recovery
of the water control during the study varied from 37% to 111%, compared to the spiked LRW control
at T0 (results not shown). With such variation in the water control, it was not possible to interpret if
recovery of endotoxin in the drug product was due to general variation or due to LER. Instead, a re-
verse study design was developed, where aliquots of undiluted sample were spiked at the different time
points during the hold time study and all samples are analyzed on the last day of the study. In this way,
all samples were analyzed in the same assay, eliminating assay-to-assay variation.
After shifting to the reverse study design, the hold time was investigated at both 2 °C-8 °C and 20 °C-
25 °C for a period of 14 and 3-7 days respectively. Hold time at 2 °C-8 °C was investigated because
samples for analysis are stored at that temperature after they are received in the laboratory. Hold time
at 20 °C-25 °C was investigated because it is the process temperature before sampling is performed.
Hold time studies for both temperatures were performed for three different batches of drug product.

For each of the three spiked batches and the spiked LRW control, the recovery of endotoxin as a per-
centage of endotoxin found in the LRW control at time 0 (mean of three repetitions) was calculated
and plotted against time (Figure 8.8.3.1-1 and Figure 8.8.3.1-2). For all studies (2-8 °C and 20-25
°C), the water controls were stable for the duration of the studies within a range of 85-120%. To
measure the spiked LRW controls in the same assays as the spiked product samples, the spiked water
controls had to be prepared and measured for each batch of drug product.
Likewise, the data presented in Figure 8.8.3.1-1 show that at 2-8 °C, the recovery was stable for the
duration of the study (14 days). The results at 2-8 °C were within the range of 87-154%. Thus, the
overall variation in the drug product samples was comparable to the variation in the water samples,
when the general variation of the method is considered. This data demonstrates that, at the refriger-
ated temperature, no decline in endotoxin recovery was observed during the two weeks beyond the
expected variation.
The study of samples held at 20-25 °C (Figure 8.8.3.1-2) was initially performed for seven days, but
as the percent recovery had declined to approximately 50% after only 72 hours, the two subsequent
batches were only investigated for three days, with additional time points within this period. Again,
the water controls were stable throughout the duration at 85-116%. The declining trend of recovery
of endotoxin spiked into the three batches of drug product was consistent with the declining trend of
recovery for the drug substances that were tested. Some 24-28 hours after spiking, the percent recovery
was above the predetermined goal of 50% for all three batches; however, after 48-52 hours, the percent
recovery was 45-65%. Also, at the 72-hour time point, data showed recovery below 50% for two of
the three test batches.
Since the drug product did not contain any of the published LER-inducing components (i.e., poly-
sorbate, citrate), the origin of LER was investigated (data not shown). This investigation was done by
testing the buffer without drug substance and testing the drug substance without other matrix compo-
nents. This investigation revealed that the trigger of LER could be mapped to the drug substance, in
this case, a protein (data not shown).
Figure 8.8.3.1-1 14-Day Hold Time Study for Drug Product at 2-8 °C with RSE
Percent recovery of RSE in three batches of drug product compared to percent recovery in water control as a function of
time. Percent recovery was calculated using the mean of three replicates of spiked water control as reference point. Limits of
acceptance (50-200%) are indicated in the plot.

Figure 8.8.3.1-2 7/3-Day Hold Time Study for Drug Product at 20-25 °C with RSE
Percent recovery of RSE in three batches of drug product compared to percent recovery in water control as a function of time.
Limits of acceptance (50–200%) are indicated in the plot. Percent recovery was calculated using the mean of three replicates
of spiked water control as reference point.
8.8.3.2 Drug Substance

This case study presents two examples of investigations conducted for drug substances with different
outcomes. In both cases, the drug substance was a solid material consisting of dry protein. In the nor-
mal quality control setting, these samples are received in the laboratories as dry material. An aliquot
of the material was weighed and resuspended to a noninhibitory concentration before conducting the
BET analysis. In initial experiments (data not shown), the spike solution was added directly to the
solid material; however, this procedure presented a challenge. The spiked solution was not distributed
uniformly in the dry drug substance, which resulted in variation in the recovery of endotoxin that
depended on the spike volume and degree of resolving the sample material. To overcome this chal-
lenge, the protein samples were dissolved in LRW before adding the spike solution. Because the drug
substances require a high degree of dilution in order to obtain a non-inhibitory concentration in the
BET assay, it was necessary to spike the solution with a high volume of endotoxin. To keep the protein
concentration relevant, the spike volume was taken into account when the sample was resuspended,
so the protein concentration after adding the spike was equal to the protein concentration in the pro-
duction. This way, a production-relevant investigation and spike in a liquid solution could be made.
The next consideration for the studies was temperature. The endotoxin sample of the drug substance
was taken immediately after the final temperature steps (of more than 30) and stored at 2 °C-8 °C until
analysis; consequently, that temperature was chosen to investigate hold time for the final drug substance.
The final experimental design for solid drug substances was as follows:
1. Resuspend sample in water.
2. Spike with endotoxin solution.
3. After spike production-relevant protein concentration is reached, store sample at 2-8 °C.
4. Perform reverse hold time study.
5. Further dilute sample immediately before BET to obtain noninhibitory concentration.

For one drug substance, no recoveries below 50% were observed in the given time frame (Figure
8.8.3.2-1). For this drug substance, the three batches were stable with no predominant trend. The
recovery remained within 54-85% over the eight days. The measurements of spiked LRW controls
showed no predominant trend over time.
For another drug substance, LER was present immediately upon spiking (data not shown). After nu-
merous unsuccessful attempts to recover the masked endotoxin, the upstream production process was
examined in more detail. Hold time was investigated for several intermediates and the probability of
inducing LER could be traced back to shortly before the final isolation step. In intermediates before
this step, the endotoxin could be fully recovered (Figure 8.8.3.2-2). For the intermediate investigated
in Figure 8.8.3.2-2, the three batches behaved similarly, and recovery was >50% until time point 250
hours. For the last time point, however, one batch displayed recovery <50%. Whether this was an
outlying result or a consequence of LER could not be concluded from this data.
Figure 8.8.3.2-1 8-Day Hold Time Study for a Drug Product at 2-8 °C with High-Potency Endotoxin, O55:B5 (Lonza)
Percent recovery of high-potency endotoxin, O55:B5 in three batches of drug substance compared to percent recovery in water control
as a function of time. Percent recovery was calculated using the mean of three replicates of spiked water control as reference point.
Figure 8.8.3.2-2 14-Day Hold Time Study for Intermediate of a Drug Substance at 2-8 °C with High-Potency Endotoxin,
O55:B5 (Lonza)
Percent recovery of high-potency endotoxin, O55:B5 in three batches of drug substance compared to percent recovery in water control
as a function of time. Percent recovery was calculated using the mean of three replicates of spiked water control as reference point.

8.8.4 Discussion
8.8.4.1 Variation and Study Design
In initial studies of endotoxin hold time, widely variable results were observed for both the spiked
samples and LRW controls, yet a conclusion regarding LER could not be drawn from these data. To
overcome the variability, the experimental design was changed to a reverse experimental design, where
samples were spiked at different time points and all samples were analyzed in the same assay. With
this reverse study design, the assay-to-assay variation could be removed, thus the results of the value of
spiked LRW control varied considerably less than in the initial study design. In general, results were
more variable when testing drug substance than when testing drug products even when no declin-
ing trend was observed (Figure 8.8.3.1-1, Figure 8.8.3.2-1, and Figure 8.8.3.2-2). Whether this
discrepancy in variation derived from the form (solid versus aqueous), the source of endotoxin used
in the experiments (RSE versus high-potency endotoxin O55:B5), the nature of the actual sample
in question, or other factors cannot be elucidated from the data available. For this experiment, the
reverse study design provided more consistent reproducible results for both drug substances and drug
products. Supplementing these studies, spiked LRW samples were prepared in parallel to the spiked
samples to serve as a control for proper design and execution of the study. The spiked water control
series enabled discrimination between random variation and actual LER. In the four studies presented
in this case, the water controls showed a variation of 10-32% which, compared to the variable nature
of the BET, was considered acceptable and provided sufficient resolution in the data to differentiate
LER from random variation. Percent recovery was calculated as the measured value in the sample rela-
tive to the measured value of endotoxin in spiked water at T0. As much emphasis was put on the T0
spiked water value, the mean value of three independent repetitions was used and the three measure-
ments were evaluated for outlying results. In this way, an outlying water control value did not lead to
false conclusions regarding LER. With this hold time study design, obtaining reproducible results for
all investigated products was possible.
8.8.4.2 Study Temperature

For the drug product, hold time studies were performed at both 2 °C-8 °C and 20 °C-25 °C; the first
temperature reflects the storage of the sample and the latter temperature reflects the process tempera-
ture before sampling. LER did not occur within during the study at 2 °C-8 °C and no apparent decline
was observed during the 14 days. At 20 °C-25 °C, the recovery of endotoxins in the drug product
drops below 50% after 48-72 hours, indicating LER under these conditions. From the hold time stud-
ies of the drug product at the two different temperatures, it became evident that masking of endotoxin
is a temperature-dependent phenomenon in agreement with published observations (3).
8.8.4.3 Solid Samples

The investigation of drug substances revealed several challenges: How can a liquid spike solution be
added to a dry sample and still have a relevant sample? How can enough endotoxin spike be added to
the sample, so it can still be detected after final sample preparation? What is the relevant temperature
for a complicated production process with multiple temperature steps over several weeks? In answer,
the samples were resuspended to a process-relevant concentration, including the volume of the spiked
solution, so the spike could be added without diluting the sample beyond the process-relevant con-
centration. A temperature of 2 °C-8 °C was chosen because it reflects sample storage temperature after
the production process is completed. With this setup, the laboratory was able to perform reliable,
reproducible studies of solid materials with focus on product relevance.
Based on the successful studies, several recommendations were made. Storage time reflecting the ob-
tained results should be implemented for the samples described in this case. To mitigate the risk of un-
derestimating endotoxin in samples, implementing additional in-process sampling is also recommended.

8.8.4.4 Origin of LER

In the literature, LER is associated with formulation components, such as polysorbate and citrate. The
drug product described in this case study does not contain any of the reported triggers of LER, there-
fore, it was unexpected when LER was observed (4). These studies revealed that LER could be mapped
directly to a specific protein. Although proteins are not generally reported as triggers of LER, all other
observations (e.g., time and temperature-dependency) indicate that the underlying mechanism can be
the same or similar. Furthermore, investigation of drug products with similar general composition, but
with different drug substances, did not reveal LER. Also, a hold time study of a different drug product
with a drug substance formulated in a buffer matrix containing known LER triggers showed no mask-
ing of endotoxin, which makes it very difficult to predict if a drug product is prone to LER. This case il-
lustrates that predicting LER may not be obvious solely by looking for known triggers in a formulation.
8.8.4.5 Mitigating Actions

When LER is observed, it may be addressed in different ways. Where the drug product at 20 °C-25 °C
(process temperature) displayed LER, an in-process control sample could be implemented. This in-
process control sample would be taken within 24 hours after the formulation of the products had been
started and, thereafter, placed at 2 °C-8 °C before being analyzed, ensuring that any endotoxin present
in the sample would be assayable. In addition, for this specific sample, a maximum storage time in the
laboratory at 2 °C-8 °C should be implemented. With this strategy for endotoxin sampling, the risk of
undetected endotoxin in the production process can be mitigated.
For the drug substance, LER could be mapped to a process step prior to the final drug substance. In
this case, an in-process sample implemented at the last process step ensured that potential endotoxin
can be fully recovered. Hold time for this sample was investigated and storage time was implemented
according to the experimental results. Since the production process following this step did not intro-
duce potential sources of endotoxin, such as watery solutions or the like, the risk of underestimating
endotoxin in the production was considered minimal.
8.8.5 Conclusion
In this case, a successful evaluation strategy has been presented. With a reverse study design, variability
could be reduced and reproducible results obtained. Also, an experimental design for solid samples was
developed and tested. Consequently, the study design will be applicable for any new products.
The products investigated in this case revealed protein-dependent LER. The LER was not driven by any
formulation components but it could be mapped to a protein. Two examples of investigation of drug
substances showed that one protein drug substance triggered LER, while the other did not. The case
illustrates that LER is not easily predictable. For the products in question, the recommendation is to
address LER by introducing sample-specific hold times based on data and by implementing in-process
controls both for the drug product and the drug substances. With implementation of the suggested mit-
igations, results obtained by the BET are considered valid and the process is considered to be in control.
8.8.6 References
1. Low Endotoxin Recovery: An FDA Perspective. Hughes, P, et al. April 10, 2015, BioPharma Asia. https://biophar-
ma-asia.com/magazine-articles/low-endotoxin-recovery-an-fda-perspective/ (accessed November 19, 2018).
(BPOG) member companies. Bolden, J, et al. 2017, Biologicals, Vol. 48, pp. 74-81.
3. Masking of endotoxin in surfactant samples: Effects on Limulus-based detection systems. Reich, J, et al.
2016, Biologicals, Vol. 44, pp. 417-422.
4. Biological Activity of Masked Endotoxin. Schwarz, H; et al. 2017, Scientific Reports, Vol. 7, doi:10.1038/srep44750.

8.9 Case Study 9: LER Case Study for a Monoclonal Antibody

A highly concentrated monoclonal antibody (mAb) drug product candidate containing both a chela-
tor (EDTA) and surfactant (PS80) was shown to exhibit low recovery when spiked with two highly
purified endotoxin preparations and tested in Lonza’s kinetic turbidimetric assay (KTA). LER was not
observed with four less-purified preparations spiked under the same conditions. Recovery losses using
the highly purified endotoxin preparations (RSE, CSE) in multiple batches of the fully formulated
product were observed within three days at ambient temperature at lower (3 EU/mL) and higher
(1700 EU/mL) spike levels. The effect of each formulation component on recovery of the more highly
purified endotoxins (RSE, CSE) was shown to be due to histidine/trehalose (buffer/cryoprotectant)
and EDTA at pH 5.5. Protein did not influence recovery, and the presence of polysorbate improved
recovery. PS80 at pH 5.5 was shown to improve recovery of CSE across a broad range (0.0001%
to 0.02%, v/v) and only concentrations as high as 0.5% were slightly inhibitory. Studies to resolve
LER in the product spiked with RSE or CSE, using various pretreatments or additives (commercial
dispersants, excess divalent cations, solvents, heat, acid/base), were all unsuccessful. The method for
calculating recovery (theoretical spike level versus data from spiked water controls) did not change the
recovery conclusions; however, Lonza’s Kinetic Chromogenic Assay (KCA) showed better recovery
than its KTA. In a more thorough evaluation of all the major suppliers of kinetic and chromogenic
limulus amebocyte lysate (LAL) kits, only the Charles River KTA prevented LER (>50%) for up to 14
days when recovery was calculated versus theoretical or when calculated versus results in spiked water.
In summary, the more purified/processed endotoxins showed the poorest recovery, and components in
this highly concentrated mAb (pH 5.5) product contributing to LER were the histidine and EDTA.
The presence of PS80 prevented some of the losses caused by the buffer and chelator. While LER was
not resolved by a number of pretreatments and additives, it was avoided by using one of the LAL
commercial assay kits (Charles River KTA). Based on these findings, resolution of LER appears to
be a highly empirical process requiring exploration of all the many tools available to practitioners of
LAL. Expectations from health authorities should be carefully explored, as some solutions (e.g., type
of endotoxin) may be more desirable than others.
8.9.1 Introduction
The product for this case was selected because it has been in clinical development for many years, is
well characterized, and contains a monoclonal antibody (mAb) with a common amino acid framework
(IgG2DA). Most importantly, its formulation has both a chelator (EDTA) and a surfactant (PS80),
components implicated in other studies to be causative for the LER phenomenon. Better recovery
has been reported with less purified forms of lipopolysaccharide (LPS) (1), so a number of endotoxin
preparations (more and less purified) were studied. Limulus amebocyte lysate (LAL), as well as recom-
binant sources of reagents, are commercially available in a variety of options for use with the bacte-
rial endotoxin test (BET). Each combination of assay standard and reagent is calibrated to reference
standard endotoxin (RSE), yet differences in results from various kits have been observed in a variety
of samples (2).
The primary purpose of this study was to produce data using a sample expected to have LER to de-
termine: 1) if endotoxin purity is a factor in recovery of endotoxin; 2) which components, or combi-
nation of components, are the main causes of LER; 3) if sample pretreatment can resolve LER; and
4) if any commercial endotoxin kits and reagents are able to resolve LER. Data analysis (recovery of
theoretical spike levels versus data from spiked water) was also of interest since both methods are cur-
rently in practice.

8.9.2.1 Protein, Matrix, and Characteristics
• Protein active pharmaceutical ingredient (monoclonal antibody produced in Chinese hamster
ovary cells): IgG2DA, 158.3 mg/mL, pI = 8.8

• Buffer: 20 mM histidine
• Cyroprotectant: 84 mg/mL trehalose
• Chelator: 0.05 mg/mL EDTA
• Surfactant: 0.02% (v/v) PS80
• pH: 5.5
8.9.2.2 LPS (Endotoxin) Suppliers (All Stored at 2 ºC-8 ºC Once Prepared or Reconstituted)
High potency control standard endotoxin (CSE), lot cat# L4391, was purchased from Sigma-Aldrich
and reported as >500,000 EU/mg, 1 mg/vial. This was purified from E. coli (O111:B4). LPS was re-
constituted with 1 mL LAL reagent water (LRW) and potency was measured by testing a number of
dilutions over several days at room temperature (ambient) using the Lonza KTA. The average mean
value was used to calculate spike levels to achieve target values of EU/mL.
CSE from the Lonza KTA kit, 410 EU/vial (see below) was reconstituted with 4100 μL of LRW to
achieve a nominal concentration of 100 EU/mL.
USP Reference Standard Endotoxin (RSE), 10,000 EU/vial (H0K354), was reconstituted with ap-
proximately 5 mL LRW in the initial study and with 1 mL LRW in the final study.
One of the naturally occurring endotoxins (NOEs) isolated from Serratia marcescens was prepared according
to Bowers et al. and determined to be 54,174 EU/mL by testing multiple dilutions in the Lonza KTA (3).
Three NOEs (S. marcescens, Escherichia coli, Enterobacter cloaca) were provided by John Dubczak of
Charles River Laboratories. Briefly, these minimally processed NOEs were prepared using depyro-
genated nutrient broth (dNB) to reconstitute lyophilized ATCC microbes. Following an overnight
incubation at room temperature, the organisms were subcultured into media containing 50% dNB
and 50% sterile water for injection (SWFI). The diluted dNB culture was again incubated overnight
at room temperature and then subcultured into medium containing 25% dNB and 75% SWFI. The
culturing of serial diluted depyrogenated nutrient broth was repeated until a final culture of 1% dNB
and 99% SWFI was obtained. Therefore, the endotoxin preparation generated from these relatively
starved microbes was representative of those found in water pretreatment systems. The preparation was
subjected to a final filtration (0.22 μm).
8.9.2.3 Bacterial Endotoxin Tests Conducted

For the six LPS preparations spiked into the protein sample and, for the individual matrix spiking
studies, Lonza’s KTA was used, with standard concentrations of 4, 1, 0.25, 0.063 and 0.016 EU/mL
fitted to a quadratic (power) equation.
Other endotoxin tests were as follows:

• Lonza Kinetic Turbidimetric Assay (KTA), Pyrogent 500, lot 0000619216
• Lonza Kinetic Chromogenic Assay (KCA), KQCL, lot 0000598261
• Charles River Kinetic Turbidimetric Assay (CR KTA), KTA2, lot H3642L
• Associates of Cape Kinetic Chromogenic Assay (ACC KCA), PyroChrom, lot CK0020
• Lonza recombinant factor C (PyroGene), lot 0000591563
To further explore the kinetics and influence of sample matrix components, a high potency CSE was
used to evaluate the causes of recovery issues. The high potency of CSE provided the maximum flex-
ibility for diluting LPS into “neat” samples (i.e., the smallest possible volume that can be accurately
measured into undiluted protein sample). High potency CSE (4,952,000 EU/mL) was diluted 10-fold
in LRW (20 μL into 180 μL). Thirty-two μL of this stock CSE was added to 980 μL of 158.6 mg/
mL protein sample. This was mixed and 450 μL of spiked protein added to 910 μL unspiked protein.

After mixing, 620 μL of this spiked protein was added to 600 μL unspiked protein. Finally, 335 μL of
this spiked protein was added to 500 μL of unspiked sample. This resulted in a concentration of 1056
EU/mL CSE in 158.3 mg/L protein (6.7 EU/mg). Finally, this was diluted to 3.4 EU/mL by addition
of 32 μL to a total of 10 mL of LRW. In parallel, a dilution in LRW, instead of protein sample, was
prepared at the same time. Combinations of buffers (histidine, trehalose) and PS80 were spiked with
the same CSE by adjusting volumes to achieve a nominal 3.4 EU/mL spike concentration. Samples
were diluted 16-fold for assay, but it was learned later that 10-fold was inhibitory in unspiked samples
tested by the traditional BET.
Three independently manufactured batches of the product were tested using Lonza’s KTA kit and
Lonza’s KCA kit. For these studies, a spiking solution was prepared by reconstituting a vial of RSE with
1 mL LRW to achieve a concentration of 10,000 EU/mL and diluted into neat protein solution for a
final concentration of 1700 EU/mL. Samples were diluted 1:1700 in two serial dilutions to achieve a
target assay concentration of 1.0 EU/mL. Three separate aliquots of each batch were spiked; an LRW
control was spiked and handled identically. One aliquot each of spiked sample and spiked LRW was
tested with both KTA and KCA to produce time zero (T0) results. The others were sealed and stored
at ambient temperature prior to testing three and seven days later (T3, T7).
The final aspect of this study was to determine if KTA, KCA, or Factor C (PyroGene) methodologies
are more or less prone to LER. A single lot of protein was spiked and diluted just prior to the BET
to achieve a target assay concentration of 1.0 EU/mL, a concentration where losses could be quanti-
fied. In this final study, RSE was used both as the spiking solution and to make the standard curve.
In this way, potential variances in CSE potency were avoided and the results between assay vendors
and modalities could be more directly compared. These kits included Lonza’s KTA and KCA, ACC
KTA, CR KTA, and Lonza’s PyroGene kit. The spiking solution was prepared by reconstituting a vial
of RSE with 1 mL LRW to achieve a concentration of 10,000 EU/mL and 438 μL spiked into 2500
μL of LRW and neat protein solution for a final concentration of 1491 EU/mL. Samples were tested
with all five endotoxin kits immediately after preparation for T0 results. (Compared to the prior study,
the BET test was conducted as soon as possible after preparation for a more exact T0 result.). Aliquots
from the same preparations were stored at ambient temperature and tested at 3, 7 and 14 days (except
PyroGene which was tested at T0 and T14).
Other attempts to resolve the LER included the following additives and/or pretreatments tested by
Lonza’s KTA, as described above, using the fully formulated drug product described in Section 8.9.2.3:
• Pyrosperse™ (Lonza, cat#N188) at 1:20 and 1:200 (v/v)
• MgCl2 was added at a final concentration of 10 mM
• MgSO4 was added at a final concentration of 50 mM
• TrisMgSO4 was added at a final concentration of 100 mM
• NaCL was added at final concentrations ranging from 1.7 to 34.2 mM
• Methanol was added at final concentrations ranging from 0.05 to 20% (v/v)
• Samples were heated at 70 ºC for 30 minutes, diluted, and then tested
• Samples were diluted, heated at 70 ºC for 30 minutes, and then tested
• Samples were heated at 95 ºC for 5 minutes, diluted, and then tested
• Samples were pretreated with Endo-Prep (BioDTech Cat#EDP-4001) by performing digestion at
1/10, 1/50 and 1/100 (v/v); the dilutions of each were tested at 1/10, 1/50 and 1/100 (v/v)
• Samples were hydrolyzed with 10 mM of 1.5M KOH and quenched with 0.3M H3PO4 in ace-
tonitrile; hydrolysis was conducted at 2-, 5- and 10-fold (v/v); these were tested with hydrolysis
alone and in combination with Pyrosperse™, MgCl2, and TrisMgSO4


The sources of LPS used for the following studies were calibrated for potency in multiple tests and the
mean value in EU/mL was used to calculate the dilutions required to achieve target levels in spiked
samples or LRW as follows:
• High-potency CSE purchased from Sigma-Aldrich was 4,952,000 EU/mL
• S. marscescens NOE prepared by our company was 54,174 EU/mL
• S. marscescens NOE prepared by Charles River was 729 EU/mL
• E. coli NOE prepared by Charles River was 742 EU/mL
• E. cloacae NOE prepared by Charles River was 2,860 EU/mL
The only exception was the CSE provided in Lonza’s KTA kit, which was used in the first study com-
paring six sources of LPS spiked neat into product. Figure 8.9.3-1 shows that when these preparations
were spiked neat into product at 100 EU/mL, by T3, recovery was below 50% for both of the highly
purified LPS materials (CSE, RSE); whereas, the four less-purified NOEs were all recovered within
50-200%. This result is similar to that obtained by others.
In the fully formulated product (mAb, buffer, cryoprotectant, chelator, surfactant, pH 5.5), there was
a loss in recovery of CSE even at T0 (43%) as shown in Figure 8.9.3-2. This loss may be due, in part,
to ordinary matrix inhibition in the traditional BET test, since the sample was diluted 16-fold and the
minimum sample dilution is 20-fold (which was 70% (data not shown)). This hold time study, there-
fore, may have exaggerated LER, which led to the subsequent, more controlled studies using small
volumes of higher spike levels and concomitant dilution factors. Figure 8.9.3-2, however, shows that
recovery in the fully formulated sample dropped from 43% to 21% after three days at ambient tem-
perature. Recovery was lower when PS80 was not present (sample with mAb, buffer, cryoprotectant,
chelator, pH 5.5) at both T0 (24%) and after three days (8%). This finding suggests that PS80 may
prevent recovery losses. When EDTA was not present in the formulation (sample with mAb, buffer,
cryoprotectant, PS80, pH 5.5), the recovery improved (T0=57%, T3=54%) to slightly above the LER
threshold (defined as recovery ≥50%). In the absence of both PS80 and EDTA, PS80 had no effect
on recovery (coincidentally identical results of T0=57%, T3=53%). However, recovery in protein and
buffers only showed low recovery (57%), possibly due, in part, to the fact that samples were not suf-
ficiently diluted for LAL, which warranted further investigation of buffer and surfactant effects.
To study the low recovery in protein and buffers and to avoid the potential sample dilution issue,
spikes of 100 EU/mL were made into histidine alone and in combination of trehalose, PS80, or both
trehalose and PS80, all at pH 5.5. These data (Table 8.9.3-1) demonstrate losses in one day (ambient)
in histidine alone or with trehalose, yet no losses when PS80 was present. The prior study showed that
in the presence of high concentrations of protein (mAb, 158 mg/mL) and the absence of EDTA, PS80
did not improve recovery. This suggests that either PS80 or protein provided a beneficial effect on the
recovery losses caused by histidine or histidine plus trehalose, whereas the inhibition caused by EDTA
in the presence of protein is improved by PS80. Although there are no obvious recovery losses caused
by 158 mg/mL mAb, evaluating the protein in the absence of any buffer was not possible because a
buffer system of some kind is required for its solubilization.
Table 8.9.3-1 Recovery of CSE Spiked into 20 mM Histidine +/- 83 mg/mL Trehalose, pH 5.5
EU/mL, T0 EU/mL, T1 day

20 mM Histidine 92 67
20 mM Histidine + 83 mg/mL Trehalose 124 48
Histidine + PS80 160 165
Histidine + Trehalose +PS80 138 134
LRW 116 119

To further elucidate the role of PS80, a broad range of concentrations were tested in histidine buffer
at pH 5.5. These samples were held at ambient temperature for six days in order to increase the sen-
sitivity for recovery losses. CSE was spiked at 100 EU/mL to allow high dilutions and avoid recovery
losses due to direct matrix effects on the BET. Table 8.9.3-2 illustrates that even the lowest PS80
concentration (0.0001%) showed better recovery (79%) after five days than in buffer alone after one
day (67%) as shown in Table 8.9.3-1. Table 8.9.3-2 shows no losses in recovery as a function of PS80
concentration; in fact, it shows a modest increase. The only exception was the highest concentration
(0.5% PS80), which showed a very modest loss (76% at T0). The test product contains 0.02% PS80,
which is in the middle of this distribution.
Figure 8.9.3-1 Percent Recovery of Six Preparations of LPS in LRW or Protein Sample at Room Temperature
Percent recovery of LPS spiked into mAb sample: Sample spiked at 50 EU/mL with the Lonza CSE supplied in the KTA assay kit
into LRW (top panel) or protein (bottom panel) and held for 3 days at ambient temperature. Values represent changes in recovery
from T0 (82–102% in LRW and 103-140% in protein).

Figure 8.9.3-2 Recovery of CSE Spiked into Protein vs. LRW Held for 3 days at Ambient Temperature
High-potency LPS (Sigma-Aldrich) spiked into neat sample and combinations of sample matrix components at time zero (T0) and
after three days (T3) at ambient temperature. LPS is spiked at a nominal concentration of 3.4 EU/mL (6.7 EU/kg) into product
formulation components (combinations of 158 mg/mL mAb, 20 mM histidine, 84 mg/mL trehalose, 0.05 mg/mL EDTA, 0.02%
[v/v] PS80, pH 5.5). Testing was with Lonza’s KTA BET method.
Table 8.9.3-2 Recovery of CSE in Various Levels of PS80 in 20 mM Histidine buffer, pH 5.5
EU/mL, T0 EU/mL, T6 days

LRW 100 96
0.0001% PS80 103 79
0.0004% PS80 110 107
0.002% PS80 115 111
0.01% PS80 130 126
0.02% PS80 (formulation control) 142 146
0.05% PS80 137 141
0.5% PS80 76 85
Because two sources of CSE (Sigma-Aldridge high-potency, Lonza LAL assay standard) had shown
losses in recovery, learning if any of the LAL reagents or methodologies were capable of improving re-
covery such that the LER phenomenon could be avoided was of interest. To study this in more depth,
three independent batches of the product were spiked with high-potency CSE at 1700 EU/mL. This
level was tested because it represented the level that, when diluted for administration in rabbits compa-
rable to human doses (mg/kg), would be expected to be pyrogenic. The same samples, plus identically
spiked LRW controls, were tested at three time points using both the KTA and the KCA kits according
to manufacturer instructions. Table 8.9.3-3 shows the recovery in LRW at T0, compared to the calcu-
lated theoretical spike level of 1700 EU/mL, was slightly lower by the KTA (1400 EU/mL) and slightly
higher by the KCA (2400 EU/mL). These recoveries (percent measured of theoretical based on dilutions
of RSE) at 82% and 141%, respectively, are well within the range of acceptable recoveries (50-200%)
and likely reflect either slight errors in pipetting or imperfect assignment of CSE standard concentration
to the RSE. Spiked product (versus LRW) at T0 also shows small losses in recovery, suggesting the LER
effect has begun even in the short period of time from spiking to testing. By T3, all three lots showed
further recovery losses, which are only lower incrementally by KCA, and no further by KTA.

Table 8.9.3-3 CSE Spiked at 1700 EU/mL in ~150 mg/mL Product
Kinetic Turbidimetric Assay (KTA) Kinetic Chromogenic Assay (KCA)

EU/mL,T0 EU/mL,T3 EU/mL,T7 EU/mL,T0 EU/mL,T3 EU/mL,T7
LRW 1400 1330 1400 2400 2070 2160
Lot 1 1030 201 247 2070 1140 962
Lot 2 1070 258 235 2160 1310 1050
Lot 3 1130 238 277 2360 1310 1050
To compare the methods of recovery analysis, the data in Table 8.9.3-3 was calculated as a percent-
age of theoretical (1700 EU/mL) and as a percent of an identical preparation spiked into LRW (one
for each time point to control for day-to-day variability). These data are presented in Figure 8.9.3-3,
where LER is observed for all three lots tested by KTA, regardless of calculation method (Panels A and
B); whereas, in the KCA method, LER is not present when results are compared to theoretical levels
and are questionable when calculated versus LRW (T7 has LER, T3 does not). Although this latter
case fails to meet the conditions for concluding that LER has occurred, another time point would be
needed to confirm whether the low recovery at T7 is repeated. Regardless, it is clear that 1) recovery
by KTA with the Lonza kit is improved versus the Lonza KCA kit for multiple lots of this product,
and 2) there are no significant numerical differences in one recovery calculation method over another,
regardless of whether the assay was conducted with KTA or KCA.
The prior study (Table 8.9.3-3) had suggested that the Lonza KCA kit might allow better recovery in
product in LER hold time studies than the Lonza KTA. However, this study was confounded by the
possible differences in spiked CSE from the standard CSE in the kits, which may or may not be cali-
brated exactly to the RSE. In addition, the passing LER at T7 was recovered only to 54-62% and only
when calculated against the theoretical, whereas recovery versus LRW failed (42-49%). To confirm
the Lonza KCA finding, as well as evaluate a number of other commercial endotoxin testing kits, the
product was spiked with RSE and subjected to an LER hold time experiment at ambient temperature.
The data in Figure 8.9.3-4A showed that at T0 in LRW the recoveries were very closely matched to
the theoretical (79-104%). This showed an improvement over the result in Table 8.9.3-3, although
a trend to higher recoveries at T14 is not completely explained (possibly evaporation in the screw top
containers). Compared to theoretical, recovery was above 50% for all except Lonza KTA and Lonza
KCA (which failed at both T3 and T7) as seen in Figure 8.9.3-4B. Results using ACC KCA were much
more variable and random than the other reagents. Finally, Figure 8.9.3-4C shows where recovery is
calculated versus LRW, only the CR KTA fails to exhibit LER. Note: Thorough sealing of samples
held at room temperature is important to avoid the confounding factor of evaporation over long time
periods.
Finally, a number of additives and pretreatments were studied with samples spiked with CSE in order
to see if the recovery issues could be resolved with conventional (or reported) solutions to LER for
samples containing this combination of PS80 and EDTA (the fully formulated product). None of
these alone or in combination was able to prevent LER (data not shown), including high concentra-
tions of divalent cations, solvents, and dispersants.
8.9.4 Conclusion
The more purified preparations of LPS are arguably less representative than endotoxin found during actual
contamination events. Counterarguments are that NOEs are not commercially available to all users, are
less characterized (i.e., “known”), and have not yet been mass-produced with consistent quality. Although
the purified forms of LPS may represent the worst-case scenario and may not be representative of LPS
actually present during production, they were used for these studies because they lend themselves to repro-
ducible evaluation. Likewise, temperatures studied were ambient, also “worst case” since recovery is im-
proved at 2 ºC-8 ºC, and this product is not actually processed or stored this long at ambient temperature.

Figure 8.9.3-3 LER Hold Time Spiked Protein Sample Study Calculated vs LRW and vs Theoretical
Data from Table 8.9.3-3 used results (EU/mL) for spiked protein samples as a percentage of theoretical RSE (1700 EU/mL) or
that as a percentage of results (EU/mL) in spiked LRW at each time point.
Figure 8.9.3-4 Comparison of Endotoxin Assay Kits in a Hold Time LER Study at Ambient Temperature
RSE Spiked into LRW (4A), Protein Sample (4B), and percent recovery of spiked protein vs spiked LRW (4C)
Perhaps the most surprising aspect of this work was finding that not only does PS80 not increase
propensity to LER, but it is preventative, even when a chelator is present. This argues against the
hypothesis that the particular combination of chelator and surfactant is deleterious to LPS recovery
in LER studies. Also surprising is that the protein appears to have no direct effect on recovery, even
though its high pI (versus pH) would suggest it is electropositive and, in theory, could bind and mask
the negatively charged LPS and cause lower recovery. As stated, the endotoxin used for these studies
is highly purified, so conclusions should not necessarily be extrapolated to actual cases of endotoxin
contamination. In fact, NOEs behave differently. For the strict application of LER studies, however, if
purified endotoxins (e.g., RSE) are used for spiking, remediation appears highly empirical. In this case,
after many attempts at LER resolution, only switching to a particular LAL vendor solved the problem.
The root cause is unknown, as the formulation of the lysate is proprietary to the manufacturer.
Because the majority of the endotoxin preparations studied do not show LER and, as is believed by
most microbiologists, NOEs are more representative of true adventitious contamination by facility
organisms, it would be ideal to have a different LPS preparation for LER evaluations. However, this
would require challenging decisions by an appropriate group of industry experts regarding how this
material should be produced, characterized, vialed, tested, and distributed globally.
8.9.5 References
2. Comparison of Limulus Amebocyte Lysate Test Methods for Endotoxin in Protein Solutions. Chen, L and
Mozier N. 2013, J Pharm Biomed Anal, Vol. Jun; 80, pp. 180-185.
3. Creation of an In-house Naturally Occurring Endotoxin Preparation for Use in Endotoxin Spiking Studies
and LAL Sample Hold Time Analysis. Bowers, K and Johnson, L. 2013, Am Pharm Review, Vol. Oct. 14
(Endotoxin Suppl).

8.10 Case Study 10: Endotoxin Recovery Studies with a Biologic Formulated
with Citrate and Polysorbate 20
Endotoxin recovery studies of Product X samples using control standard endotoxin (CSE) as the spike
analyte showed LER for drug substance and formulation buffer (both containing citrate and polysor-
bate 20). No LER was detected for the process step “Unconditioned Bulk,” which does not contain ci-
trate or polysorbate 20. LER for the drug substance was not fully confirmed when endotoxin, purified
from a plant isolate (Ralstonia pickettii, strain BMTU 3653), was used as an alternative spike analyte.
LER risk for Product X was mitigated by implementing a risk-based approach for batch release.
8.10.1 Introduction
The following document describes the endotoxin recovery studies performed to detect potential LER
for Product X, which is a humanized monoclonal antibody formulated with citrate and polysorbate.
It has recently been shown that these formulation ingredients cause LER when undiluted product
samples are spiked with a known amount of endotoxin and tested using the Bacterial Endotoxins Test
(BET), according to the pharmacopeias of the United States, Europe, and Japan (1).
8.10.2 Materials
8.10.2.1 Product and Buffer Samples
• Drug substance: One batch of Product X drug substance was tested (10 g/L API in 20 mM so-
dium citrate, 20 mM L-arginine HCl, 190 mM D(+) sucrose, 0.02% (w/v) polysorbate 20, pH
5.5). The API of Product X is a humanized monoclonal antibody with a pI of approximately 8.6.
Product X drug substance is manufactured from the process step “Unconditioned Bulk” by buffer
exchange using citrate and polysorbate 20-containing formulation buffer (Figure 8.10.2.1-1).
• Formulation buffer: One batch of formulation buffer was tested (20 mM sodium citrate, 60 mM
Arginine-HCl, 570 mM, sucros, 0.06% polysorbate 20, pH 5.5).
• Unconditioned bulk: One batch of process step “Unconditioned Bulk,” which did not contain
citrate or polysorbate, was tested.
Unconditioned Bulk
Containing API Formulation Buffer
20 mM sodium citrate,
60 mM Arginine-HCI,
570 mM sucrose,
0.06% polysorbate 20,
pH 5.5
Drug Substance
10 g/L API
20 mM sodium citrate,
20 mM L-arginine HCI,
190 mM D(+) sucrose,
0.02% (w/v) polysorbate 20,
pH 5.5
Figure 8.10.2.1-1 Manufacturing Steps for Biotech Product X Drug Substance
8.10.2.2 Endotoxins Used for Spiking Studies

• Control Standard Endotoxin: Charles River Laboratories (CRL), catalogue #E120, 10 ng/vial,
LPS type E. coli O55:B5)
• Ralstonia pickettii, strain BMTU 3653 endotoxin: Strain BMTU 3653 was isolated from the
same site where Product X was manufactured. Strain BMTU 3653 was identified as R. picket-
tii by ribosomal DNA sequencing. For endotoxin preparation, strain BMTU 3653 was grown
in tryptic soy bouillon at 32-35 °C. Cells were harvested by centrifugation, and endotoxin was
extracted from the resulting cell pellet by LPS-BioSciences, using the isobutyric acid-M ammo-
nium hydroxide patented method (2). Crude LPS were further washed with organic solvents,
i.e., CHCl3:MeOH (1:2 v/v) and CHCl3: MeOH: H2O (3:2:0.25, v/v/v), in order to remove
phospholipids and lipopeptides, then treated with enzymes DNAse, RNAse, and Proteinase K to
remove DNA, RNA, and protein contaminants (3).
8.10.3 Methods
8.10.3.1 Endotoxin Spiking and Sample Storage
Endotoxin spiking procedures and containers used for spiking and storage are listed in Table 8.10.3.1-1.
During spiking and prior to sampling from the spiked sample container, the sample container was
vortexed at full speed for at least 10 seconds. Drug substance (DS) and LAL reagent water (LRW)
control were independently spiked with CSE and R. pickettii, strain BMTU 3653 endotoxin whereas
other sample types and water controls were spiked with CSE only.
Table 8.10.3.1-1 Endotoxin Spiking and Container Types
Endotoxin Spiking
Sample Category Storage Container Type
(theoretical concentration 5.0 EU/mL)
4.5 mL DS and LRW were spiked with 0.5 mL endotoxin
Drug Substance stock solution (50 EU/mL, CSE or R. pickettii, BMTU 3653
Pyrokontrol borosilicate glass
endotoxin) in a glass container (10 mL, Lonza, Cat. No:
vials (ACILA, Product # 1958360)
80-507L) and 0.5 mL aliquots of the spiked solution were
transferred to storage containers.
55.8 mL formulation buffer was spiked with 6.2 mL
endotoxin stock solution (50 EU/mL CSE) in a glass container
(endotoxin-free Duran Flask) and 2 mL aliquots of the spiked
Formulation Buffer Glass container (10 mL, Lonza, solution were transferred to storage containers.
Cat. No: 80-507L)
1.8 mL LRW was spiked with 0.2 mL endotoxin stock
solution (50 EU/mL CSE) in a glass container (10 mL, Lonza,
Cat. No: 80-507L). Spiked LRW was not further aliquoted.
9 mL Unconditioned Bulk and LRW were spiked with 1 mL
Unconditioned Bulk Pyrokontrol borosilicate glass endotoxin stock solution (50 EU/mL CSE) in a glass container
vials (ACILA, Product # 1958360) (10 mL, Lonza, Cat. No: 80-507L) and 1 mL aliquots of the
spiked solution were transferred to storage containers.
8.10.3.2 Storage of Spiked Samples and Time Intervals

Storage containers were placed at process-relevant temperature (2-8 °C) within one hour of endotoxin spiking.
At each designated time point (Tmp = 0 h, Tmp = 1d, Tmp = 2d [only DS], Tmp = 3d, Tmp = 7d), spiked
product samples and spiked water controls were taken out of storage and assayed within one hour.
8.10.3.3 BET Method

The endotoxin assay was performed using the kinetic turbidimetric assay (KTA) with CRL reagents
(100 mM Tris buffer, LRW) and software (EndoScan). The CRL CSE was used to generate a four-
point standard curve (0.005, 0.05, 0.5, 5 EU/mL).

• Drug Substance: Drug substance sample was pH-adjusted with 100 mM Tris buffer (1:2 dilution,
pH of adjusted sample = 6.0) and further diluted with LRW (1:4, 1:8, 1:16). Dilutions 1:8 and
1:16 were tested at all time points. The water control (LRW spiked with 5 EU/mL endotoxin) was
diluted 1:8 with LRW prior testing and tested in parallel to the test samples.
• Unconditioned Bulk: Unconditioned Bulk was pH-adjusted with 100 mM Tris buffer and 1M
endotoxin-free NaOH (1:2 dilution, pH of adjusted sample = 6.5) and further diluted with LRW
(1:4, 1:8, 1:16). Dilution 1:8 was tested at all time points. The water control (LRW spiked with
5 EU/mL endotoxin) was diluted 1:10 with LRW prior to testing and tested in parallel to the test
samples.
• Formulation Buffer: Formulation buffer was pH-adjusted with 100 mM Tris buffer (1:2 dilution,
pH of adjusted sample = 6.0) and further diluted with LRW (1:4, 1:8, 1:16). Dilutions 1:8 and
1:16 were tested at all time points. The water control (LRW spiked with 5 EU/mL endotoxin) was
diluted 1:10 with LRW prior to testing and tested at T0 in parallel to the test samples.
8.10.3.4 Data Analysis and Interpretation

Water Control: The percentage of endotoxin recovery at each time point was calculated using the fol-
lowing formula:
% Endotoxin Recovery Endotoxin Concentration at Time Tmp (EU/mL)
= x 100
at Time Tmp 5 EU/mL (theoretical concentration)
Water Control: The percentage of endotoxin recovery at each time point was calculated using each the
following formulas:

= x 100
at Time Tmp 5 EU/mL (theoretical concentration)

= x 100
at Time Tmp Water Control at Time Tmp (EU/mL)

= x 100
at Time Tmp Water Control at Time T0 (EU/mL)
For each calculation approach, LER was defined as: No LER occurred if the endotoxin recovery in the
spiked product samples and respective water control was in the range of 50% to 200% throughout all
time points. If the endotoxin recovery from a spiked product sample at one time point was less than
50% and the low recovery maintained at all later time points, the test sample was considered to have
LER effect. If the calculated endotoxin recovery was less than 50% and only detected in any of the
middle time points, but not the end time points, the test sample was not considered to exhibit a LER/
masking effect.

The results of the endotoxin recovery studies are summarized in Tables 8.10.4.1-1 – Table 8.10.4.5-1.
8.10.4.1 Drug Substance

Table 8.10.4.1-1 Product X DS Spiked with 5 EU/mL CSE, BET with Dilution 1:8*
Endotoxin: CRL CSE, Catalogue #E120

Product #1, dilution 1:8 with 100 mM Tris buffer + LRW Water control, dilution: 1:8 with LRW
Recovery in Relation to Recovery in
Endotoxin Endotoxin
Theoretical Water Water Relation to
Time Point content PPC (%) Time Point content PPC (%)
Spike Control Control Theoretical
(EU/mL) (EU/mL)
(5.0 EU/mL) (T0) (Tmp) Spike
T0 0,9662 96 19% 23% 23% T0 4,1557 102 83%

4h 0,2962 94 6% 7% 5% 4h 5,4698 78 109%
1d 0,7590 103 15% 18% 15% 1d 5,0026 77 100%
2d 0,2016 102 4% 5% 4% 2d 4,7598 80 95%
3d 0,2836 100 6% 7% 6% 3d 4,8207 84 96%
7d 0,3655 88 7% 9% 8% 7d 4,6962 77 94%
* Recoveries <50% are highlighted in bold with a red background.
Table 8.10.4.1-2 Product X DS Spiked with 5 EU/mL CSE, BET with Dilution 1:16*

Endotoxin Endotoxin
(EU/mL) (EU/mL)
T0 1,5627 90 31% 38% 38% T0 4,1557 102 83%

4h 0,4649 104 9% 11% 8% 4h 5,4698 78 109%
1d 1,1031 107 22% 27% 22% 1d 5,0026 77 100%
2d 0,2694 112 5% 6% 6% 2d 4,7598 80 95%
3d 0,3971 100 8% 10% 8% 3d 4,8207 84 96%
7d 0,4866 87 10% 12% 10% 7d 4,6962 77 94%
Table 8.10.4.1-3 Product X DS Spiked with 5 EU/mL R. pickettii, BET with Dilution 1:8*
Endotoxin: Ralstonia pickettii, BMTU 3653

Endotoxin Endotoxin
(EU/mL) (EU/mL)
T0 5,8526 83 117% 226% 226% T0 2,5925 90 52%

4h 4,2672 73 85% 165% 78% 4h 5,4873 82 110%
1d 5,3617 78 107% 207% 146% 1d 3,6828 82 74%
2d 1,7484 83 35% 67% 48% 2d 3,6166 81 72%
3d 1,5451 82 31% 60% 46% 3d 3,3912 88 68%
7d 2,1925 74 44% 40% 69% 7d 3,1560 85 63%

Table 8.10.4.1-4 Product X DS spiked with 5 EU/mL R. pickettii, BET with Dilution 1:16*
Endotoxin: Ralstonia pickettii, BMTU 3653

Endotoxin Endotoxin
(EU/mL) (EU/mL)
T0 7,6338 84 153% 294% 294% T0 2,5925 90 52%

4h 5,3699 92 107% 207% 98% 4h 5,4873 82 110%
1d 7,0593 93 141% 272% 192% 1d 3,6828 82 74%
2d 2,0311 97 41% 78% 56% 2d 3,6166 81 72%
3d 1,7447 92 35% 67% 51% 3d 3,3912 88 68%
7d 2,7092 86 54% 105% 86% 7d 3,1560 85 63%
* Recoveries < 50% are highlighted in bold with a red background.
8.10.4.2 Drug Substance Summary

LER was detected from the first measured time point (T0) when CSE was spiked into undiluted Prod-
uct X DS samples and held at 2-8 °C for 7 days followed by BET testing of 1:8 and 1:16 dilutions.
Undiluted Product X DS samples spiked with R. pickettii BMTU 3653 endotoxin revealed LER when
recoveries were calculated in relation to the theoretical spike level (5 EU/mL) from test results of the
1:8 dilution. No LER was seen when BET data from 1:16 dilution were used or calculations were done
against water controls. These findings led to the assumption that R. pickettii BMTU 3653 endotoxin
was less affected by LER compared to the CSE. Reduced recoveries (<50%) may result from greater
assay variability of spiked R. pickettii BMTU 3653 endotoxin compared to CSE. This hypothesis is
supported by the higher degree of variance seen in the water control data for CSE (83-109%), com-
pared to the R. pickettii water control (52-110%).
8.10.4.3 Formulation Buffer

Table 8.10.4.3-1 Formulation Buffer Spiked with 5 EU/mL CSE, BET with Dilution 1:8*

Formulation buffer, dilution 1:8 with 100 mM Tris buffer + LRW Water control, dilution: 1:10 with LRW
Endotoxin Endotoxin Relation to
Time Point content PPC (%) Theoretical Water Water Time Point content PPC (%) Theoretical
(EU/mL) Spike Control Control (EU/mL) Spike
(5.0 EU/mL) (T0) (Tmp) (5.0 EU/mL)
T0 0,4575 88 9% 10% 10% T0 4,5641 134 91%
4h 0,0685 44* 1% 2% N/A 4h ND N/A N/A
1d 0,0400 76 <1% <1% N/A 1d ND N/A N/A
3d <0,0400 70 <1% <1% N/A 3d ND N/A N/A
7d <0,0400 75 <1% <1% N/A 7d ND N/A N/A
Table 8.10.4.3-2 Formulation Buffer Spiked with 5 EU/mL CSE, BET with Dilution 1:16*

Formulation buffer, dilution 1:8 with 100 mM Tris buffer + LRW Water control, dilution: 1:10 with LRW
Endotoxin Endotoxin Relation to
Time Point content PPC (%) Theoretical Water Water Time Point content PPC (%) Theoretical
(EU/mL) Spike Control Control (EU/mL) Spike
(5.0 EU/mL) (T0) (Tmp) (5.0 EU/mL)
T0 0,7928 102 16% 17% 17% T0 4,5641 134 91%
4h 0,1319 70 3% 3% N/A 4h ND N/A N/A
1d 0,1266 89 3% 3% N/A 1d ND N/A N/A
3d <0,0800 87 <2% <2% N/A 3d ND N/A N/A
7d <0,0800 87 <2% <2% N/A 7d ND N/A N/A
8.10.4.4 Formulation Buffer Summary

LER was detected from the first measured time point (T0) when CSE was spiked into undiluted
formulation buffer samples and held at 2-8 °C for 7 days followed by BET testing of 1:8 and 1:16
dilutions.
8.10.4.5 Unconditioned Bulk

Table 8.10.4.5-1 Unconditioned Bulk Spiked with 5 EU/mL CSE, BET with Dilution 1:8

Endotoxin Endotoxin
(EU/mL) (EU/mL)
T0 5,7528 76 115% 120% 120% T0 4,7754 109 96%

4h 4,8550 73 97% 102% 122% 4h 3,9896 88 80%
1d 4,4406 74 89% 111% 121% 1d 3,6644 115 73%
2d 4,8412 88 97% 132% 86% 2d 5,6221 114 112%
3d 4,9393 73 99% 88% 90% 3d 5,5092 115 110%
7d 4,1121 72 82% 75% 111% 7d 3,7034 112 74%
8.10.4.6 Unconditioned Bulk Summary

No LER was detected when CSE was spiked into undiluted Unconditioned Bulk samples and held at
2-8 °C for 7 days followed by BET testing of 1:8.
8.10.5 Conclusion
Endotoxin recovery studies of Product X samples using CSE as spike analyte showed LER for drug
substance and formulation buffer, both containing citrate and polysorbate 20. No LER was detected for
process step Unconditioned Bulk, which does not contain citrate or polysorbate 20. These findings agree
with previous studies, which showed that citrate and polysorbate 20-containing sample matrices can
cause LER (1,4,5). Endotoxin purified from a plant isolate (R. pickettii, strain BMTU 3653) was used
as an additional spike analyte. Different masking susceptibilities of CSE and R. pickettii, BMTU 3653
might be explained by different LPS structures. Further studies are needed to confirm this hypothesis.
LER risk for Product X was mitigated by implementing a risk-based, in-process control throughout
the drug substance and drug product manufacturing processes. The risk-based approach assessed en-
dotoxin contents of the furthest downstream step that does not exhibit LER (Unconditioned Bulk)

and raw materials, including process water used for preparation of formulation buffer. The risk-based
approach further took into account bioburden load of Unconditioned Bulk and process water used
for batch preparation since gram-negative bacteria might introduce endotoxins. Concerning process-
ing primary packaging components, the endotoxin content was specified for the incoming material
with further treatment, i.e., washing, sterilization, and depyrogenization, further reducing potential
contamination with endotoxins.
8.10.6 References
2. Novel method for isolating endotoxins. Caroff, M. 2002, International Patent. WO 2004062690 A1.
3. Simple method for repurification of endotoxins for biological use. Tirsoaga, A; et al. 2007, Appl Environ Micro-
biol, Vol. 73, pp. 1803-8.
4. Biological activity of masked endotoxins. Schwarz, H; et al. 2017, Sci Rep. Vol. Mar; 7(7), pp. 44750.
5. Masking of endotoxin in surfactant samples: Effects on Limulus-based detection systems. Reich, J, et al. 2016,
8.11 Case Study 11: No LER Effect in a Sodium Phosphate, Polysorbate 80

Formulation Matrix
An endotoxin hold time study was performed for a formulation matrix containing 25 mM sodium
phosphate, 0.0325% (v/w) polysorbate 80, 13% (w/w) mannitol buffer, and a final protein concentra-
tion of 5 mg/mL. The undiluted matrix was spiked with a known amount (1.20 EU/mL) of lipopoly-
saccharide (LPS) as the reference standard endotoxin (RSE), spiked sample containers were stored at
2-8 °C, and the recovery was evaluated over time (T0, T1, T4, and T8). More than 50% recovery of the
spiked product was demonstrated on average for all individual replicates over the included time points.
Though a decrease in endotoxin concentration over time was established, no LER effect was observed
for the sodium phosphate, polysorbate 80 matrix, as the recovery on T8 was still 59.08%. Although
typical LER-inducing matrix components were present in the buffer matrix, the data generated in this
case study demonstrate that the LER phenomenon is not always observed. Furthermore, this study
confirms that the LER effect might be less severe in sodium phosphate, polysorbate 80 matrices com-
pared to other surfactant chelator combinations in formulation matrices.
8.11.1 Introduction
For this case study, a formulation matrix containing 25 mM sodium phosphate, 0.0325% (v/w) poly-
sorbate 80, 13% (w/w) mannitol buffer, and a final protein concentration of 5 mg/mL was used for an
eight-day endotoxin (LPS) hold study. The LER effect is often associated with an excipient combina-
tion of polysorbate and phosphate. The surfactant and chelating agents cooperate in changing the LPS
configuration from a reactive to a nonreactive form (1). More specifically, the phosphate/polysorbate
formulation could chelate the divalent cations and destabilize these aggregates, which would result in
smaller aggregates or unreactive LPS monomers (2). In this case study, the LER effect was evaluated
for a typical phosphate/polysorbate formulation matrix.

A bulk volume of the formulation matrix (25 mM sodium phosphate, 13% mannitol; 0.0325% poly-
sorbate 80; pH 6.20) with a final product/protein concentration of 5 mg/mL (pI 3.8–5.4) was spiked
with a known amount of RSE (E150, Charles River Endosafe) corresponding to 2X the action limit
of the product, i.e., 1.20 EU/mL. In parallel, an identical bulk volume of LAL Reagent Water (LRW)
(W130 Charles River Endosafe) was spiked with an identical RSE concentration as a positive con-
trol. A multi-aliquot design was used for the endotoxin hold study experiment. On T0, the spiked
formulation matrix and LRW bulk volume were aliquoted into pyrogen-free 15 mL sample storage
tubes (352099, Corning) for each time point (T0, T1, T4, and T8). Five replicates per time point were
included. Spiked formulation matrix and spiked LRW containers were stored at 2-8 °C. Note: The
process took place at 6-10 °C.
The endotoxin content was determined using the Kinetic Turbidimetric Assay (KTA) with a sensitivity
of 0.01 EU/mL using CSE (E110, Charles River Endosafe) and a compatible KTA lysate (R15015;
Charles River Endosafe). Spiked samples were diluted 1/40 in LRW using Pyrotube D dilution tubes
(TB240; Associates of Cape Cod). Diluted formulation matrix and LRW samples were loaded onto a
96-well microplate ((35)3072; Falcon/ Becton Dickinson) and absorbance was read with a Biotek Elx
808TM Absorbance Microplate reader. Data analysis and calculations were performed by WinKQCL
software V4.0.4 (Lonza).
Endotoxin recovery results of the spiked formulation matrix were compared to the corresponding
spiked LRW control for each time point. A ≥50% recovery was set as the acceptance criterion on each
time point during the endotoxin hold study in order to refute LER. When two consecutive time points
showed an average endotoxin recovery <50%, the LER effect was confirmed and further troubleshoot-
ing initiated. For one product batch, the complete endotoxin hold study experiment was completed.
Two additional product batches were used to confirm the endotoxin recovery at the latest time point
for which no LER was detected.
8.11.2.1 List of Material and Methods

• Buffer Matrix, Concentrations, and pH: 25 mM sodium phosphate, 13% mannitol; 0.0325%
polysorbate 80; pH 6.20
• Product Concentrations and pI: Product concentration = 5 mg/mL; pI = 3.8 – 5.4 (heterogenic
pool: based on glycosylation, clipped forms and aggregates)
• LPS Supplier: Reference Standard Endotoxin (E150; Charles River Endosafe)
• BET Method and Reagents Vendor/Supplier: Kinetic Turbidimetric Assay (0.01 EU/mL sensitivity;
Charles River Endosafe); KTA lysate (R15015; Charles River Endosafe); CSE (E110; Charles River
Endosafe)
• Spiking/Hold Recovery Study Design: Multi-aliquot approach; Recovery at T0, T1, T4, T8 (5 rep-
licates per time point); 3 batches; complete hold study for one batch + confirmation of recovery
on latest time point without LER
• Hold Temperature and Endotoxin Spiking Concentrations (e.g., 20 EU/mL): Hold temperature:
2-8 °C; Process temperature: 6-10 °C; Spike concentration: 1.20 EU/mL
• Diluent Used for Sample Dilution and Dilution Factor: LAL Reagent Water (W110; W120;
W130; Charles River Endosafe); 1/40 dilution in LRW
• Controls: Spiked LRW as control across all time points
• Recovery Calculation Method: Recovery is calculated against the LRW control at the respective
time point; 50-200% acceptance criterion
• BET Equipment:
– Software: WinKQCL software V4.0.4 (Lonza)
– 96 well plate ((35)3072; Falcon/ Becton Dickinson)
– Biotek Elx 808TM Absorbance Microplate reader
– Dilution tube: Pyrotube D (TB240; Associates of Cape Cod)
– Pyrogen-free 15 mL sample storage tube (352099, Corning)

The average LPS (RSE) recovery on T0, T1, and T4 of the spiked sodium phosphate, polysorbate 80
matrix varied around the target spike of 1.20 EU/mL. On T8, the average LPS (RSE) recovery de-
creased to a value of 0.75 EU/mL (Figure 8.11.3-1). The average recovery of the positive control on
every time point was approximately the target spike of 1.20 EU/mL. These data indicate that the LPS
(RSE) spike could be adequately recovered from the sodium phosphate, polysorbate 80 matrix above
50% for at least eight days with regard to the time point’s respective positive control value (Figure
8.11.3-2).
LER is often linked to the presence of excipients, such as phosphate combined with polysorbate, in
formulation matrices (1). Reich, et al., established that the loss of LPS activity for different kinds of
chelators such as sodium phosphate, sodium citrate, and EDTA was lowest under phosphoric condi-
tions (2). With regard to this case study, no severe adverse effect of the sodium phosphate/polysorbate
80 formulation matrix on the endotoxin recovery was observed, as all recoveries remained above the
acceptance criterion of 50% recovery. Nonetheless, the results indicate that the sodium phosphate
buffer containing polysorbate 80 did exhibit a loss of LPS (RSE) reactivity overtime (T8=59.08%).
Moreover, the results generated during this case study confirm the results acquired by Platco (3).
During a 15-day endotoxin hold study under similar conditions, LPS (CSE) was spiked in a 10 mM
sodium phosphate, 0.01-0.02% Polysorbate 80 matrix (pH=6.5). A decrease in recovery overtime was
observed as well, though relative recoveries at T15 still remained above 50%.
Figure 8.11.3-1 Endotoxin Recovery

Endotoxin Recovery (EU/mL) of LPS (RSE spike of 1.20 EU/mL) in 25 mM sodium phosphate, 13% mannitol; 0.0325% polysorbate
80 (protein concentration: 5 mg/mL; pI 3.8 – 5.4; pH 6.20) and LAL reagent water control over time points T0, T1, T4 and T8 (n =
5; bars represent average of 5 replicates; error flags = standard deviation).
Figure 8.11.3-2 Average Relative Endotoxin Recovery Results

Average relative endotoxin recovery results for 25 mM sodium phosphate, 13% mannitol; 0.0325% polysorbate 80 (protein
concentration: 5 mg/mL; pI 3.8–5.4; pH 6.20) over time (T0, T1, T4 and T8). (n = 5; bars represent average of 5 replicates; error
flags = standard deviation)
8.11.4 Conclusion
Although typical LER-inducing matrix components were present in the buffer matrix, the data gener-
ated in this case study demonstrate that the LER phenomenon is not always observed in these situa-
tions. Furthermore, this study confirms that the LER effect might be less severe in sodium phosphate,
polysorbate 80 matrices compared to other surfactant, chelator combinations in formulation matrices.
8.11.5 References
2. Masking of endotoxin in surfactant samples: Effects on Limulus-based detection systems. Reich, J, et al. 2016,

8.12 Case Study 12: Lack of LER Phenomenon in A Therapeutic Protein

Biologic with Known LER-Causing Ingredients
The LER effect has been observed in biologic formulations containing a combination of polysorbate
with sodium citrate or sodium dihydrogen phosphate. This case study presents a new finding that L-
histidine in combination with polysorbate 20 and sodium dihydrogen phosphate in the product for-
mulation mitigates the LER phenomenon of a therapeutic protein product. Although the mechanism
of L-histidine in demasking the LER effect of polysorbate and phosphate is unknown, this finding
nevertheless provides additional choices in selecting non-LER-causing ingredients for formulating a
biological product in the presence of common surfactants such as polysorbate.
8.12.1 Introduction
LER is an endotoxin masking effect first described by Chen and Vinther whereby the spiked endo-
toxin recovery from undiluted biologic product is less than 50% of its original spike within a seven-
day incubation period using the compendial Bacterial Endotoxins Test (BET) methods (1). LER is
caused by formulation ingredients that contain a combination of polysorbate and sodium citrate or
sodium dihydrogen phosphate. These ingredients are commonly found in biologic products, including
monoclonal antibody and recombinant protein products. As a consequence of LER, endotoxin values
from a potential contamination event may be underreported. Chen and Vinther also reported that the
USP rabbit pyrogen testing (RPT) with a biologic product showed that spiked endotoxin undetected
by LAL methods still elicited a pyrogenic response; thus, endotoxin is masked but not inactivated.
Screening of LER has since become a requirement in U.S. FDA Biologics License Application (BLA)
for products containing polysorbate 20 or polysorbate 80 in the product formulations (1–3,5).
This case study describes a therapeutic protein product formulated with L-histidine, polysorbate 20,
and sodium dihydrogen phosphate but no demonstrated LER phenomenon.
8.12.2 Materials
• Biologic Drug Product: A recombinant human glycoprotein produced in a genetically engineered
Chinese hamster ovary (CHO) cell line in the dosage form of sterile solution for infusion at the
strength of 2 mg/mL. The drug product is formulated in sodium dihydrogen phosphate buffer
with 20 mM L-histidine and 0.01% polysorbate 20 at pH 6.0. The therapeutic protein has a neu-
tral pI between 6.5 and 7.4. Three drug product lots were analyzed for the impact of LER.
• Equipment and Reagents:
– Depyrogenated scissors and forceps – Temperature recorder
– Depyrogenated glass extraction vehicles – Pipetters and qualified dispensing tips
– Depyrogenated test tubes for dilutions – Parafilm M
of standards and samples – Balance
– 5 °C ± 3 °C Refrigerator – Charles River E. coli O55:B5 Control
– -20 °C ± 3 °C Freezer Standard Endotoxin (CSE)
– 37 °C ± 1 °C Incubator – Lysate (Charles River KTA2™)
– Vortexer – LAL Reagent Water (LRW)
– Pyrogen-free Syringes – Qualified Gloves
– Pyrogen-free Pipettes – Source of Spike Endotoxin: Charles River E.
– Pyrogen-free 96 well plates coli O55:B5 Control Standard Endotoxin
(CSE) in the concentration of 10 ng/vial
– 96 well plate reader
– Timer
8.12.3 Methods
8.12.3.1 LER Experimental Design
The first step in assessing the LER impact on the drug product is to qualify the compendial BET
method to demonstrate its suitability for the test article. LRW is the diluent to prepare the test ar-
ticle (drug product) at a 1:10 dilution for the kinetic turbidimetric Limulus amebocyte lysate (LAL)
method evaluation for sample inhibition and enhancement. Baseline testing was performed on the
test article at 1:10 dilution at time zero to determine if endogenous endotoxin was present. The LER
study was then conducted by spiking the undiluted test article at a concentration of 5.0 EU/mL CSE.
The storage of the spiked test article was performed in the original drug product glass vial. The spiked
sample was further aliquoted to individual glass vials to ensure sufficient sample volume for all incuba-
tion time points (0 h, 24 h, 48 h, 72 h, and 168 h). Spiked sample vials were then incubated at 5 °C ±
3 °C. Testing was then performed on spiked sample at a 1:10 dilution using LRW as the diluent from
each incubation time point to determine the recovery value of spiked endotoxin.
The LRW control was prepared by spiking at a concentration of 5.0 EU/mL with the same volume
and aliquot as the test article. The storage of the LRW control was performed in pyrogen-free glass
tubes and incubated at 5 °C ± 3 °C for all time points.
8.12.3.2 Calculation of Endotoxin Recovery

The arithmetic mean of the LRW control and each test article at the reported time point determinations
(Tmp: 0 h, 24 h, 48 h, 72 h, and 168 h) was calculated based on the formula of theoretical spike value.
X [Tmp]
% Endotoxin Recovery
= x 100
of Spiked Sample 5.0 EU/mL ÷ dilution factor
X [Tmp] = Arithmetic mean of the repeat determination at time point in EU/mL

8.12.4.1 LAL Assay Validity
The LAL assay used in determination of spiked endotoxin recovery must be qualified per compendia
and lysate manufacture instructions. Table 8.12.4.1-1 illustrates that the LAL assay validity criteria
were met for the test article and LRW control from all incubation time points.

Table 8.12.4.1-1 LAL Assay Validity Criteria Summary
Acceptance Criteria Met AC

Parameter Time Points Results
(AC) (Yes or No)
0h 36.9 : 37.0 °C Yes
24 h 36.9 : 37.0 °C Yes
Assay Incubation
37 ± 1°C 48 h 37.0 : 37.0 °C Yes
Temperature
72 h 36.9 : 36.9 °C Yes
168 h 36.9 : 37.0 °C Yes
0h -0.9994 Yes
24 h -0.9995 Yes
Linearity of Standard Absolute value of the
48 h -0.9998 Yes
Curve linearity is ≥ 0.980
72 h -0.9991 Yes
168 h -0.9990 Yes
0h -0.2680 Yes
24 h -0.2686 Yes
Slope of Standard
-1.0 < slope < -0.1 48 h -0.3263 Yes
Curve
72 h -0.2656 Yes
168 h -0.2759 Yes
0h <0.01 Yes
Must react slower 24 h <0.01 Yes
than the lowest point
Negative Control 48 h <0.01 Yes
in the standard curve
(<0.01) 72 h <0.01 Yes
168 h <0.01 Yes
0.33%, 0.19%
0h Yes
0.00%, 0.74%
Coefficient of 0.50%, 0.71%
24 h Yes
Variation of Reaction 0.00%, 1.95%
Time for CSE 0.29%, 0.22%
Standards <10% 48 h Yes
0.41%, 0.00%
(5.0, 0.5, 0.05, 0.01 0.19%, 0.47%
72 h Yes
EU/mL) 0.56%, 0.05%
0.92%, 0.51%
168 h Yes
0.27%, 0.28%
8.12.4.2 LER Assessment Results

The assessment of spiked endotoxin recovery from the undiluted test article (drug product lot 1025264)
and LRW control is summarized in Table 8.12.4.2-1 and Figure 8.12.4.2-1. The endotoxin recovery
values from the LRW control are consistently higher than the test article in all time points. This varia-
tion can be attributed by the spiking technique and intrinsic assay variability. However, all recovery
values from the test article and LRW control in all time points meet the 50-200% acceptance criteria.
Table 8.12.4.2-1 Endotoxin Recovery Results from the Drug Product and LRW Control
Calculated LER Effect

Acceptance Criteria Met AC
Parameter Time Points Endotoxin Exhibited
(AC) (Yes or No)
Recovery Results (Yes or No)
0h 86% Yes No
50%-200% Recovery 24 h 79% Yes No
Recovery in Spiked
Compared to 5.0 EU/mL 48 h 64% Yes No
Drug Product
Theoretical Spike 72 h 74% Yes No
168 h 79% Yes No
0h 117% Yes No
50%-200% Recovery 24 h 138% Yes No
Recovery in Spiked
LRW Control
168 h 135% Yes No
Figure 8.12.4.2-1 Endotoxin Recovery Curves from the Test Article and LRW Control
To ensure the repeatability of LER assessment results, two additional drug product lots (1032014
and 1032015) were selected for the same study. The test results show that no LER occurred (Table
8.12.4.2-2 and Figure 8.12.4.2-2).
Table 8.12.4.2-2 Endotoxin Recovery Results from Two Additional Drug Product Lots
Calculated LER Effect

Acceptance Met AC
Parameter Time Points Endotoxin Exhibited
Criteria (AC) (Yes or No)
Recovery Results (Yes or No)
0h 94% Yes No
Recovery in Spiked 50% – 200% Recovery 24 h 133% Yes No
Drug Product Lot Compared to 5.0 EU/mL 48 h 108% Yes No
1032014 Theoretical Spike 72 h 128% Yes No
168 h 109% Yes No
0h 83% Yes No
Recovery in Spiked 50% – 200% Recovery 24 h 92% Yes No
Drug Product Lot Compared to 5.0 EU/mL 48 h 81% Yes No
1032015 Theoretical Spike 72 h 89% Yes No
168 h 73% Yes No
0h 88% Yes No
50% – 200% Recovery 24 h 107% Yes No
Recovery in Spiked
LRW Control
168 h 112% Yes No
Figure 8.12.4.2-2 Endotoxin Recovery Curves from Two Test Articles and LRW Control

The lack of LER phenomenon demonstrated with the drug product containing the LER causing ingredients
of polysorbate 20 and sodium dihydrogen phosphate was unexpected based on existing literature (1,3). Fur-
ther analysis of all the formulation ingredients in the drug product reveals that the difference between this
biologic formulation and other published formulations is the unique presence of L-histidine in the formula-
tion, along with polysorbate and phosphate, suggesting that L-histidine may play a role in mitigating the LER
effect from the combination of polysorbate and phosphate. Confirming that L-histidine being non-LER im-
pacting, a study was conducted by Chen whereby 20 clinical biologics (proteins and monoclonal antibodies)
containing polysorbate and L-histidine in the drug product formulations were screened (4).
8.12.5 Conclusion
LER, which causes the failure to detect endotoxins in biological products by LAL detection methods,
has undesirable implications from the regulatory compliance perspective for medicinal product release.
This case study has demonstrated a new finding: L-histidine in combination with polysorbate 20 and
sodium dihydrogen phosphate has no LER impact on a therapeutic protein product. Although the
mechanism of L-histidine in demasking the LER effect of polysorbate and phosphate is unknown,
formulators canadd it to their choices for when selecting non-LER-causing ingredients to formulate
their unique biological products.
Acknowledgements
The author thanks Wu-Xi Apptec for conducting the entire LER studies.
8.12.6 References
2. Endotoxin challenge: A regulatory perspective. Hughes, P. Presented at the PDA 9th Annual Global Confer-
ence on Pharmaceutical Microbiology, Bethesda, MD, October 2014.
3. Risk Management on Common Biologics Affected by Low Endotoxin Recovery. Chen, J. Presented at the Phar-
maceutical Microbiology Forum Bacterial Endotoxin Summit, Philadelphia, PA, 2014.
4. Analytical Solutions of Low Endotoxin Recovery (LER). Chen, J and Von Wintzingerode, F. Presented at the
PDA 10th Annual Global Conference on Pharmaceutical Microbiology, Bethesda, MD, 2015.
5. General Chapter <85> Bacterial Endotoxins Test. USP 41–NF 36. U.S. Pharmacopeia. 2018. www.usp.org.
Terms of Usage for PDA Electronic Publications (Technical Reports, Books, etc.)
An Authorized User of this electronic publication is the purchaser.
Authorized Users are permitted to do the following:

• Search and view the content of the PDA Electronic Publication
• Download the PDA Electronic Publication for the individual use of an Authorized User
• Print the PDA Electronic Publication for the individual use of an Authorized User
• Make a reasonable number of photocopies of a printed PDA Electronic Publication for the individual use of an
Authorized User
Authorized Users are not permitted to do the following:

• Allow anyone other than an Authorized User to use or access the PDA Electronic Publication
• Display or otherwise make any information from the PDA Electronic Publication available to anyone other than
an Authorized User
• Post the PDA Electronic Publication or portions of the PDA Electronic Publication on websites, either available
on the Internet or an Intranet, or in any form of online publications without a license from PDA, Inc.
• Transmit electronically, via e-mail or any other file transfer protocols, any portion of the PDA Electronic Publica-
tion
• Create a searchable archive of any portion of the PDA Electronic Publication
• Use robots or intelligent agents to access, search and/or systematically download any portion of the PDA Elec-
tronic Publication
• Sell, re-sell, rent, lease, license, sublicense, assign or otherwise transfer the use of the PDA Electronic Publication
or its content
• Use or copy the PDA Electronic Publication for document delivery, fee-for-service use, or bulk reproduction or
distribution of materials in any form, or any substantially similar commercial purpose
• Alter, modify, repackage or adapt any portion of the PDA Electronic Publication
• Make any edits or derivative works with respect to any portion of the PDA Electronic Publication including any
text or graphics
• Delete or remove in any form or format, including on a printed article or photocopy, any copyright information
or notice contained in the PDA Electronic Publication
• Combine any portion of the PDA Electronic Publication with any other material
To License or purchase Reprints

• Licensing: Site licenses and licenses to distribute PDA Electronic Publication can be obtained for a fee.
To learn more about licensing options and rates, please contact:
Janny Chua, Publications Manager, +1 (301) 656-5900,
ext. 133. Written correspondence can be sent to: PDA, Inc. Attn: Janny Chua, 4350 East West Highway,
Suite 150, Bethesda, MD 20814.
• Reprints: Reprints of PDA Electronic Publication can be purchased for a fee.
To order reprints, please contact:
Janny Chua, Publications Manager, +1 (301) 656-5900, ext. 133.
Written correspondence can be sent to: PDA, Inc. Attn: Janny Chua, 4350 East West Highway, Suite 150,
Bethesda, MD 20814.


About PDA Technical Reports
PDA Technical Reports are global consensus documents, prepared by member-driven Task Forces (listed on inside
front cover) comprised of content experts, including scientists and engineers working in the pharmaceutical/bio-
pharmaceutical industry, regulatory authorities and academia. While in development, PDA Technical Reports are
subjected to a global review of PDA members and other topic-specific experts, often including regulatory officials.
Comments from the global review are then considered by the authoring Task Force during the preparation of the
final working draft. The level of expertise of the Task Force and those participating in the global review ensure a
broad perspective reflecting best thinking and practices currently available.
The final working draft is next reviewed by the PDA Advisory Board or Boards that sanctioned the project. PDA’s
Advisory Boards are: Science Advisory Board, Biotechnology Advisory Board, and Regulatory Affairs and Quality
Committee. Following this stage of review, the PDA Board of Directors conducts the final review and determines
whether to publish or not publish the document as an official PDA Technical Report.
While PDA goes to great lengths to ensure each Technical Report is of the highest quality, all readers are encouraged
to contact PDA about any scientific, technical, or regulatory inaccuracies, discrepancies, or mistakes that might be
found in any of the documents. Readers can email PDA at: [email protected].
Sanctioning Advisory Board: BioAB PDA Officers and Directors

Michael DeFelippis PhD, Eli Lilly and Company Officers
(Vice-Chair)
Chair: Rebecca Devine, PhD, Regulatory Consultant
John Geigert, PhD, BioPharmaceutical Quality Chair-Elect: Jette Christensen, PhD, Novo Nordisk
Solutions (Chair) Treasurer: Michael Sadowski, Baxter Healthcare
Michael Blackton, Adaptimmune, LLC Secretary: Steven Lynn, Lynn Consulting
Vijay Chiruvolu, PhD, Kite Pharma Imm. Past Chair: Martin VanTrieste
President: Richard Johnson
Olivia Henderson, PhD, Amgen
Christopher Hwang, PhD, Just Biotherapeutics China Directors
Margit Jeschke, PhD, Novartis Masahiro Akimoto, Otsuka Pharmaceutical Factory, Inc.
Stephan Krause, PhD, AstraZeneca Biologics Barbara Allen, Eli Lilly & Company
Morten Munk, NNE Pharmaplan Joyce Bloomfield
Veronique Davoust, Pfizer
Nadine Ritter, PhD, Global Biotech Experts, LLC
Ghada Haddad, Merck & Co./Merck Sharp & Dohme
Jens Schletter, PhD, Sandoz
Kerry Ingalls, Amgen
Michael VanDerWerf, Halozyme Therapeutics, Inc
Mary Oates, Pfizer
Karen Walker, Seattle Genetics Emma Ramnarine, Roche Pharma
Keith Wonnacott, PhD, Pfizer Stephan Rönninger, Amgen
Earl Zablackis, PhD, Sanofi-Pasteur Anil Sawant, Merck & Co./Merck Sharp & Dohme
Susan Schniepp, Regulatory Compliance Associates, Inc.
Melissa Seymour, Biogen
PDA Publication Production and Editing

Joshua Eaton, PDA Project Manager
Tina Morris, VP, Scientific & Regulatory Affairs
Walter Morris, Senior Director of Publishing
Katja Yount, Publication Designer
Marilyn Foster, Technical Editor/Writer
Bethesda
Bethesda Towers
Towers
4350
4350 East West
West Highway
Highway
Suite
Suite600
200
Bethesda,
Bethesda, MD MD 20814 USA
20814 USA
Tel:
Tel: +1 1(301)
(301)656-5900
656-5900
Fax: +1 1(301)
(301)986-0296
986-0296
E-mail:
E-mail: [email protected]
[email protected]
Web
Web site:
site:www.pda.org
www.pda.org

TR82 Low Endotoxin Recovery

Uploaded by

Copyright:

Available Formats

TR82 Low Endotoxin Recovery

Uploaded by

Document Information

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

TR82 Low Endotoxin Recovery

Uploaded by

Copyright:

Available Formats

Licensed to Goyal, Prateek/Lupin Ltd: Copying and Distribution Prohibited.

Technical Report No. 82

PDA Low Endotoxin Recovery Technical Report Team

Dayue Chen, PhD, Eli Lilly and Company, Co-leader

Low Endotoxin Recovery

Figures and Tables Index

2.0 Glossary of Terms

2 © 2019 Parenteral Drug Association, Inc. Technical Report No. 82

Nominal Value Potency

3.0 LER Hold-Time Studies

Technical Report No. 82 © 2019 Parenteral Drug Association, Inc. 3

3.1 Test Method

3.1.1 Reagents and Materials

3.1.2 Spiking of Undiluted Sample

4 © 2019 Parenteral Drug Association, Inc. Technical Report No. 82

A multiple-aliquot approach is recommended to avoid sampling repeatedly from a single container

3.1.3 Storage Containers

3.1.4 Batch Requirements

3.1.5 Storage Time and Temperature

Technical Report No. 82 © 2019 Parenteral Drug Association, Inc. 5

Table 3.1.5.1-1 Example for Identification of LER Process-Relevant Steps

3.1.9 Data Analysis and Interpretation

6 © 2019 Parenteral Drug Association, Inc. Technical Report No. 82

3.1.9.1 2–8 °C Studies

3.1.9.2 20–25 °C Studies

4.0 Proposed Mechanisms of LER

Technical Report No. 82 © 2019 Parenteral Drug Association, Inc. 7

4.1 Differentiation of LER and Test Interference

4.2 Investigation of LER Causes

8 © 2019 Parenteral Drug Association, Inc. Technical Report No. 82

Table 4.2-1 Investigation of LER Driving Forces in Biopharmaceutical Drug Products

Scenario Composition LER Root Cause

4.3 Proposed Two-Step Reaction Model of LER

Technical Report No. 82 © 2019 Parenteral Drug Association, Inc. 9

=Endotoxin =Divalent cation =Chelator =Surfactant

Figure 4.3-1 Schematic of Masking/Demasking Phenomenon

4.4 Factors Influencing Reaction Kinetics

4.4.1 Energy Input (Extrinsic)

10 © 2019 Parenteral Drug Association, Inc. Technical Report No. 82

4.4.2 Masking Capability of a Sample

4.4.2.1 Formulation Components

Technical Report No. 82 © 2019 Parenteral Drug Association, Inc. 11

4.4.3 Masking Susceptibility of Endotoxins

12 © 2019 Parenteral Drug Association, Inc. Technical Report No. 82

Figure 4.4.3-1 Covalent Modifications of Kdo2-lipid A in E. coli K-12 and Salmonella

4.5 Potential Aggregation States of Detectable and Non-detectable

Technical Report No. 82 © 2019 Parenteral Drug Association, Inc. 13

4.5.1 LER and Limited LPS Interaction with Factor C

4.5.2 Activity of Monomeric and Aggregated LPS

14 © 2019 Parenteral Drug Association, Inc. Technical Report No. 82

5.0 Mitigation of LER

5.1 Mitigation through Sample Treatment

Technical Report No. 82 © 2019 Parenteral Drug Association, Inc. 15

5.1.1 Evaluation of the Product

5.1.2 Addition of Dispersants to Samples

5.1.3 Addition of Excess Divalent Cations

5.1.4 Other Sample Treatments

5.1.5 Evaluation of LAL Assay Reagents

16 © 2019 Parenteral Drug Association, Inc. Technical Report No. 82

2 © 2019 Parenteral Drug Association, Inc. Technical Report No. 82

Technical Report No. 82 © 2019 Parenteral Drug Association, Inc. 3

4 © 2019 Parenteral Drug Association, Inc. Technical Report No. 82

Technical Report No. 82 © 2019 Parenteral Drug Association, Inc. 5

6 © 2019 Parenteral Drug Association, Inc. Technical Report No. 82

Technical Report No. 82 © 2019 Parenteral Drug Association, Inc. 7

8 © 2019 Parenteral Drug Association, Inc. Technical Report No. 82

Technical Report No. 82 © 2019 Parenteral Drug Association, Inc. 9

10 © 2019 Parenteral Drug Association, Inc. Technical Report No. 82

Technical Report No. 82 © 2019 Parenteral Drug Association, Inc. 11

12 © 2019 Parenteral Drug Association, Inc. Technical Report No. 82

Technical Report No. 82 © 2019 Parenteral Drug Association, Inc. 13

14 © 2019 Parenteral Drug Association, Inc. Technical Report No. 82

Technical Report No. 82 © 2019 Parenteral Drug Association, Inc. 15

16 © 2019 Parenteral Drug Association, Inc. Technical Report No. 82

Technical Report No. 82 © 2019 Parenteral Drug Association, Inc. 17

18 © 2019 Parenteral Drug Association, Inc. Technical Report No. 82

Technical Report No. 82 © 2019 Parenteral Drug Association, Inc. 19

20 © 2019 Parenteral Drug Association, Inc. Technical Report No. 82

Technical Report No. 82 © 2019 Parenteral Drug Association, Inc. 21

22 © 2019 Parenteral Drug Association, Inc. Technical Report No. 82

Technical Report No. 82 © 2019 Parenteral Drug Association, Inc. 23

24 © 2019 Parenteral Drug Association, Inc. Technical Report No. 82

25 © 2019 Parenteral Drug Association, Inc. Technical Report No. 82

26 © 2019 Parenteral Drug Association, Inc. Technical Report No. 82

28 © 2019 Parenteral Drug Association, Inc. Technical Report No. 82

Technical Report No. 82 © 2019 Parenteral Drug Association, Inc. 29

30 © 2019 Parenteral Drug Association, Inc. Technical Report No. 82

Technical Report No. 82 © 2019 Parenteral Drug Association, Inc. 31

32 © 2019 Parenteral Drug Association, Inc. Technical Report No. 82

Technical Report No. 82 © 2019 Parenteral Drug Association, Inc. 33

34 © 2019 Parenteral Drug Association, Inc. Technical Report No. 82

Technical Report No. 82 © 2019 Parenteral Drug Association, Inc. 35

36 © 2019 Parenteral Drug Association, Inc. Technical Report No. 82

Technical Report No. 82 © 2019 Parenteral Drug Association, Inc. 37

38 © 2019 Parenteral Drug Association, Inc. Technical Report No. 82

Technical Report No. 82 © 2019 Parenteral Drug Association, Inc. 39

40 © 2019 Parenteral Drug Association, Inc. Technical Report No. 82

Technical Report No. 82 © 2019 Parenteral Drug Association, Inc. 41

42 © 2019 Parenteral Drug Association, Inc. Technical Report No. 82

Technical Report No. 82 © 2019 Parenteral Drug Association, Inc. 43

44 © 2019 Parenteral Drug Association, Inc. Technical Report No. 82

Technical Report No. 82 © 2019 Parenteral Drug Association, Inc. 45