Establishing endotoxin limits to enhance the reliability of in vitro immunogenicity risk assessments

ABSTRACT

Immunogenic responses to biotherapeutics often lead to termination of their development because the resulting anti-drug-antibodies (ADA) can negatively impact pharmacology, safety, and efficacy. To mitigate ADA risks, in vitro risk assessment assays in non-clinical settings are essential to enhance safety and efficacy of protein-based therapeutics. This study aimed to develop and validate a human in vitro immunogenicity T cell proliferation assay. However, there is a lack of comprehensive guidelines for managing product-related factors such as endotoxin contamination, which can significantly influence assay sensitivity and accuracy. Our investigation of the impact of endotoxins revealed that levels above 0.1 EU/mg significantly induce T cell proliferation and CD14⁺ myeloid cell expansion, leading to potential false-positive outcomes in immunogenicity assessments. These findings suggest the importance of developing standardized protocols to enhance the predictive capability of in vitro methods, ensuring the assessment of therapeutic proteins accurately reflects their immunogenic potential without interference from contaminants.

Introduction

Immunogenicity, which refers to the ability of therapeutic antibodies to produce an anti-drug antibody (ADA) response, is a key concern in the development of protein therapeutics, as they can adversely affect drug efficacy and safety and complicate the interpretation of toxicity, pharmacokinetic, and pharmacodynamic data.^1–4 To address these challenges, assessing nonclinical immunogenicity risks during the early stages of drug development has become essential,⁵ prompting the European Medicines Agency and US Food and Drug Administration to publish a “Guideline on Immunogenicity Assessment of Biotechnology-Derived Therapeutic Proteins”⁶ and “Immunogenicity Assessment for Therapeutic Protein Products”⁷, respectively. In vitro assays serve as a critical tool, offering the advantage of fast, high throughput assessments, especially in the initial phase of drug development, which can potentially reduce development costs and improve drug approval success rates. Over the past decade, substantial advancements have been made in developing robust in vitro methods to predict immunogenic responses.^8–13

T cell-dependent responses are the key drivers of ADAs responses, involving interactions among antigen-presenting cells (APCs), T cells, cytokines, and B cells. Consequently, detection of ADA IgG responses often indicate T cell involvement in the immune response to a protein. In vitro peripheral blood mononuclear cell (PBMC) assays, particularly T cell proliferation (TCP) assays, provide straightforward insights into immunogenicity and are thus utilized across the biopharmaceutical industry to support studies evaluating the immunogenic potential of therapeutic proteins.^8,9,11–14

The immunogenicity of therapeutic proteins is influenced by a host of patient- and disease-related factors and product-related factors. Patient-specific variables, such as genetic predispositions, play a vital role; allelic polymorphisms in the major histocompatibility complex, for example, can alter immune responses to therapeutic proteins. Age, underlying health conditions, and concurrent treatments further impact these responses, potentially suppressing or activating the immune system. Past exposure to similar proteins can pre-sensitize patients, increasing the risk of cross-reactive immune responses. Additionally, the route and duration of administration, whether intravenous or subcutaneous, short-term or long-term, significantly affect immunogenic potential. Factors relating to the product, such as the protein’s origin and structure, post-translational modifications, and the presence of impurities like aggregates or residual host cell proteins, can influence the immunogenic profile. These considerations, alongside the formulation and storage conditions of the therapeutic protein, are integral to accurate immunogenicity assessments, which ultimately aim to enhance the safety and efficacy of biotherapeutic developments.^{3,6,7,15–20}

Among product-related factors, endotoxins, particularly lipopolysaccharides (LPS) from the outer membrane of Gram-negative bacteria, are impurities that may persist during the production process and trigger immune responses via Toll-like receptor 4 (TLR4). Upon interaction with pathogens, TLR4 recruits the adaptor protein myeloid differentiation factor 88 (MyD88), initiating a cascade of immune signaling pathways. This recruitment is crucial during the early phase of nuclear factor-κB (NF-κB) activation, which leads to the production of pro-inflammatory cytokines and other immune mediators.^20–23 These processes not only induce, but can also amplify immune reactions. Such potent immune activation underscores the importance of considering endotoxin contamination in the drug assessments of biotherapeutics, but currently there are no clearly defined acceptable levels of endotoxin contamination that would not lead to a false-positive signal in human TCP assays.

Recognizing these challenges, our study systematically examines the impact of endotoxin contamination on the in vitro immunogenicity assay. By investigating how endotoxins influence TCP and immune activation, we aimed to create specific recommendations for limiting endotoxin levels in test molecules. These recommendations are designed to ensure that in vitro immunogenicity assessments accurately reflect the true immunogenic potential of biotherapeutics, rather than artifacts induced by contaminant endotoxins.

Materials and methods

Tested material

Keyhole Limpet Hemocyanin (KLH) (Sigma, Cat. H7017) was used as positive control. Trastuzumab Lot 1, Lot 2, Lot 3, Evolocumab, Bococizumab, ATR-107, and Briakinumab, were produced internally.

Endotoxin testing

Endotoxin levels in all protein samples were measured using the Endosafe-PTS cartridge system, which provides a detection sensitivity of less than 0.05 EU/mg. The assay used reference endotoxin standards obtained from the United States Pharmacopeia (USP).

In vitro PBMC assay

Healthy donors were recruited from AbbVie ABC blood donor program by Sanguine. PBMCs were isolated from fresh blood. CD8⁺ T cells were depleted using CD8⁺ RosetteSep (StemCell Technologies Inc, London, UK). CD8⁺ depleted PBMCs were resuspended in phosphate-buffered saline (PBS) at a concentration of 1 X 10⁷ cells/mL and labeled with 0.5 uM of CellTrace according to the manufacturer’s protocol. CellTrace-labeled PBMCs were resuspended in CTS^TM OpTimizer culture media (Life Technologies, Carlsbad, CA) supplemented with serum-free replacement, GlutaMax (Gibco), penicillin/streptomycin (Gibco). 5 X 10⁵ PBMCs were plated in a 96-well flatbottom plate and cultured for 7 days with or without 40 µg/ml of test articles. Assays were performed in triplicate for each independent experiment.

Flow cytometry

Cells were washed and stained in PBS with Live/Dead fixable dead cell stain kit (Invitrogen, Cat. L34962) and then stained with antibodies to CD4 (OKT4, Cat. 317438), CD3 (UCHT1, Cat. 300424), CD19 (SJ25C1, Cat. 363012), CD14 (MɸP9, Cat. 566465), and CD45 (HI30, Cat. 304029). CD4⁺ proliferated cells were gated as CD4^hiCellTrace⁻ from non-CD14, non-CD19 and CD45⁺CD3⁺ gating. All samples were acquired using FACS LSRFortessa (BD Biosciences) and analyzed using FlowJo software (FlowJo LLC). All antibodies were purchased from BioLegend except anti-CD14 (BD Biosciences).

Cytokines

Interleukin (IL)-12p40, interferon (IFN)-γ, IL-6, IL-1β and tumor necrosis factor (TNF) protein levels were measured by MSD (MSD) according to the manufacturer’s protocols. Assays were performed in triplicate for each independent experiment.

Data analysis

For data analysis, the percentage of CellTrace negative cells was used to calculate the stimulation index (SI). The SI was calculated by dividing the percentage of proliferating cells of the test well by the percentage of proliferating cells of the media-only treated well for each donor. A response was considered positive if the SI was greater than 2.0 (SI > 2.0). The percentage of donors that responded (% donors responding) was calculated by taking the number of donors that had a positive response (SI > 2.0) divided by the total number of donors that were tested. Response Index (RI) was calculated by multiplying average of SI and % Responding Donors.

Statistical analysis

Statistical analyses were performed using GraphPad Prism 6 program (GraphPad Software. Statistical significance was defined as *p < 0.05, **p < 0.01, ***p < 0.001 and ****p < 0.0001.

Results

In-house in vitro immunogenicity risk assessment tool correlates with the rate of clinical immunogenicity

In vitro T cell activation/proliferation assays have been widely used to help predict the immunogenicity risk of biotherapeutics and can have several different readouts. We established an in-house TCP assay to mitigate risk in our candidate selection and validated its effectiveness in assessing the relative risk of immunogenicity in the clinic. Five mAbs with known rates of clinical immunogenicity were tested for their ability to elicit CD4⁺ T cell proliferative responses (Figure 1). Since KLH should be recognized by 100% of humans irrespective of HLA-DR types, KLH was used as a control to determine if the donor responds or not. Donors that did not respond to KLH were excluded from the assay. The SI was calculated by dividing the proliferative response of the test well (%) by the proliferative response of the medium-only treated well (%) for each donor, and each individual SI value was plotted in Figure 1a. A response was considered positive if the SI was greater than 2.0 (SI ≥ 2.0). The percentage of donors that responded (% donors responding) was calculated (Figure 1b) and these numbers were compared to % reported ADA in the clinic (Table 1). Remarkably, the percentage of responding donors correlates with the rate of clinical immunogenicity, suggesting that our TCP assay is well-designed to assess the T cell dependent immunogenicity and evaluate immunogenicity risk potential.

Figure 1.

The response of CD4 T cells in the CD8-depleted PBMC assay can assess T cell dependent immunogenicity with good correlation to clinical ADA. Biotherapeutic mAbs were tested in the immunogenicity PBMC assay at 40 ug/mL. Incubation with KLH at 20 ug/ml was used as a positive control. (a) the scatter plot presents the SI for each donor tested with each treatment. The horizontal line represents the threshold for a positive response (SI ≥ 2). S (b) the bar graph presents the percentage of positive donors for each of the treatments based on SI ≥ 2. (c) Response Index (RI) was calculated by multiplying the average SI and % responding donors.

Table 1.

Comparison of clinical ADA frequency and percentage of positive responding donors as determined by CD8-depleted PBMC assay.

	Trastuzumab	Evolocumab	ATR-107	Bococizumab	Briakinumab
% Reported ADA in the clinic	<1%	<1%	76%	48%	77%
% Responding donors in-house assay	16.7%	0%	50%	47.4%	76.7%

Endotoxin-induced TCP in the PBMCs via myeloid cell activation

Many factors can influence the immunogenicity of biotherapeutics in humans. Product- and process-related factors, such as post-translational modifications, host cell proteins, residual DNA, and chromatography ligands, are known triggers of immune responses. While these impurities are recognized as potential immunogenic factors in in vivo studies, their impact on in vitro immunogenicity assays has been less explored. Consequently, we evaluated the effects of endotoxin contamination on TCP to establish an acceptable endotoxin limit for in vitro immunogenicity risk assessment.

In vitro immunogenicity assays were conducted with samples spiked with endotoxin using PBMCs from 12 healthy donors (Figure 2). The average SI for all donors was plotted as shown (Figure 2a). Notably, 4 of 12 donors exhibited responses at 0.1 EU/mg of endotoxin, with a dose-dependent increase observed; ultimately, all donors responded at 3 EU/mg (Figure 2b). The RI, calculated by multiplying the average SI value by the percentage of responding donors, similarly increased in a dose-dependent manner (Figure 2c). These findings indicate that endotoxins, acting as contaminants, can indeed induce TCP. More importantly, such responses may result in false-positive outcomes during data interpretation. Given the clear evidence of endotoxin impact on TCP above 0.1 EU/mg in some donors and considering donor variability, we propose that the cutoff value for endotoxin levels in therapeutic antibodies and other biologics for in vitro immunogenicity assays should be set below 0.1 EU/mg.

Figure 2.

Endotoxin can induce T cell proliferation in the PBMC in vitro immunogenicity assay. (a) Stimulation Index dependent on endotoxin concentration. Each dot represents an average of 12 donors with error bars. (b) % Responding donors. (c) Response Index (RI).

Endotoxins, often synonymous with lipopolysaccharides (LPS), bind the CD14/TLR4/MD2 receptor complex in many cell types.^22–24 We specifically investigated the response of CD14⁺ myeloid cells to endotoxin exposure. As expected, we observed a significant dose-dependent expansion of CD14⁺ myeloid cells (Figure 3a,b), affirming the stimulatory effects of endotoxins on myeloid cell activity.

Figure 3.

Endotoxin can expand/activate CD14⁺ myeloid cells in the PBMC in vitro immunogenicity assay. (a) CD14⁺ myeloid cells were expanded upon endotoxin treatment in a dose-dependent manner. On an average of 12 donors plotted, the error bars indicate the various responses from different donors. (b) Representative FACS plot of CD14⁺ cells in response to each concentration of endotoxins (c) Proinflammatory cytokines in response to endotoxins at Day 1. (d) INF-γ was measured at Day 7.

Furthermore, we confirmed that endotoxin exposure triggers inflammatory responses in a similar dose-dependent manner (Figure 3c). These activated innate immune responses have a substantial impact on T cell activity, as demonstrated by the increased production of IFN-γ in samples exposed to endotoxins over a seven-day period (Figure 3d). Our findings in their entirely indicate that the presence of endotoxins profoundly influences both innate and adaptive immune responses. The dose-dependent expansion of CD14⁺ myeloid cells and subsequent increase in IFN-γ production highlight a significant inflammatory pathway activation that is not reflective of a typical immunogenicity-driven, T cell-dependent response.

Impact of endotoxin variability on TCP assay accuracy

To investigate the impact of endotoxin levels on T cell activation, we performed in vitro immunogenicity assays using three distinct lots of trastuzumab, each with varying endotoxin concentrations, across 10 healthy donors (Figure 4). Lot 1 contained less than 0.023 EU/mg of endotoxin, Lot 2 had 0.335 EU/mg, and Lot 3 was a remediated version of Lot 2, with endotoxin levels reduced to below 0.02 EU/mg. As anticipated, Lot 2, which had the highest endotoxin level, resulted in 50% of donors showing a response, the highest RI among the lots, consistent with our endotoxin titration experiments. In contrast, Lots 1 and 3 exhibited comparably low immunogenic responses. The results are summarized in Table 2. Notably, the comparison between Lot 2 and Lot 3 highlights that simply reducing endotoxin levels significantly decreases the incidence of T cell responses, strongly suggesting that residual endotoxins can indeed induce T cell proliferation.

Figure 4.

Different mAb lots with different levels of endotoxin have different capacity to induce T cell responses. (a) Stimulation Index. (b) % Responding donors. (c) Response Index (RI).

Table 2.

Summary of responses to endotoxin in the PBMC in vitro immunogenicity assay.

Sample	Lot	EU/mg	% Responding Donors (10)	RI
Trastuzumab	1	<0.023	10% (1/10)	21
	2	0.335	50% (5/10)	195
	3	<0.02	20% (2/10)	72

Discussion

Our study demonstrated the effectiveness of our in-house TCP assay in predicting the immunogenicity risk of biotherapeutics, showing a good correlation with clinical ADA outcomes. Given our limited donor pool, the percentage of responding donors may not always reflect accurate trends. For example, the low immunogenic mAb trastuzumab elicited responses in 3 of 18 donors, resulting in a response of 16.7%, though the intensity of these responses was low. While individual SI values and the percentage of responding donors provided useful information, we observed that these metrics alone did not always reflect the true strength of the response. To address this limitation, we introduced the RI, calculated by multiplying the average SI value by the percentage of responding donors. This metric better captures both the magnitude and frequency of the immune response, providing a more representative measure of antigenicity. While further validation is needed with an increased number of donors and therapeutic antibodies with known ADAs, our findings suggest that the in-house TCP assay, augmented with the RI, is well-suited for predicting immunogenicity risk and can be effectively incorporated into the discovery screening funnel. Future studies should explore a broader panel of therapeutics, as the evolving biologics landscape now includes immune-oncology molecules, TNF superfamily antibodies, bispecifics, antibody-drug conjugates, as well as other non-canonical molecules. These efforts will further validate the assay’s utility and extend its applicability across diverse therapeutic modalities.

Our study systematically examined the impact of endotoxin contamination on in vitro immunogenicity assays, focusing on TCP assays. We demonstrated that endotoxin levels can significantly influence assay outcomes by inducing nonspecific T cell proliferation. This was evident through the dose-dependent expansion of CD14⁺ myeloid cells and increased IFN-γ production. Of note, T cell responses from KLH and briakinumab, used as positive controls, indicated that immunogenic T cell responses did not consistently correlate with their CD14⁺ cell expansion (data not shown). These controls confirmed that endotoxins specifically stimulate CD14⁺ cells, leading to T cell responses. Notably, experiments with several trastuzumab lots further confirmed that higher endotoxin levels correlate with heightened T cell responses, underscoring the risk of false-positive results due to endotoxin contamination. Our findings demonstrated the critical importance of stringent endotoxin control in in vitro immunogenicity assessments. For both economical and labor efficiency reasons, not every step in the development process sufficiently purifies endotoxins. However, the presence of residual endotoxins can significantly affect the sensitivity and accuracy of TCP assays. In our study, we determined that endotoxin concentrations as low as 0.1 EU/mg could induce significant TCP and lead to false-positive results. Therefore, we recommend an endotoxin specification of less than 0.1 EU/mg for in vitro immunogenicity assays to ensure accurate and reliable data.

Since endotoxin has been recognized as an important factor engaging key inflammatory mechanisms in vivo, its levels are stringently limited in preclinical research to ensure safety.^25–28 The World Health Organization sets limits for standard endotoxin assays used in the pharmaceutical industry,²⁹ with the USP also standardizing the processes for endotoxin purification.³⁰ Some regulatory authorities require that the endotoxin used in human challenge studies be good manufacturing practice (GMP) grade endotoxin. However, comparable restrictions and recommendations are notably absent in the non-clinical in vitro setting, particularly for TCP assays, which have recently become widely used. It is important to note that higher endotoxin concentrations are often permissible in vivo due to the body’s robust immune mechanisms, which naturally tolerate, regulate, and filter endotoxin levels through processes such as hepatic detoxification and systemic clearance mechanisms. These physiological filters significantly mitigate the inflammatory effects of endotoxin. In contrast, in vitro environments lack these dynamic immune regulation systems, making them more susceptible to endotoxin-induced artifacts. This highlights the need for stricter endotoxin control in in vitro assays to ensure accurate immunogenicity assessments that truly reflect the biologics’ properties without confounding artifacts caused by residual endotoxin. Adhering to stringent endotoxin specifications during immunogenicity risk assessment of potential biotherapeutics helps prevent endotoxin-induced artifacts in in vitro immunogenicity assays and support informed decision-making.

In conclusion, this study first showed that residual endotoxins induce inflammatory responses and resulting in unwanted T cell responses at in vitro TCP assay. Our study provides valuable guidelines for researchers conducting in vitro PBMC TCP assays. Due to the complexity of immunogenicity risk factors, predicting the complete biological properties of a protein is challenging. However, these assays offer crucial insights into how product-related attributes, such as post-translational modifications, aggregates, residual host cell proteins affect immunogenicity. By assessing the impact of each attribute and validating their sensitivity in in vitro PBMC TCP assays, researchers can avoid artifacts in data analysis and enable better decision-making.

Funding Statement

AbbVie sponsored and funded the study.

Disclosure statement

No potential conflict of interest was reported by the author(s).

Abbreviations

ADA

Anti-drug-antibodies

APCs

Antigen-presenting cells

GMP

Good Manufacturing Practice

KLH

Keyhole Limpet Hemocyanin

LPS

Lipopolysaccharides

MyD88

Myeloid Differentiation factor 88

NF-κB

Nuclear factor κB

PBMC

Peripheral blood mononuclear cell

RI

Response Index

SI

Stimulation Index

TCP

T cell proliferation

TLR4

Toll-like receptor 4

USP

United States Pharmacopeia