Adaptation of an ELISA assay for detection of IgG2a responses against therapeutic monoclonal antibodies in a mouse immunization model

Abstract Biotherapeutic monoclonal antibodies (mAb) play important roles in clinical medicine but their potential to elicit immune responses in patients remains a major issue. In a study designed to investigate the effect of aggregation on immunogenic responses, mice were immunized with two monoclonal antibodies (mAb1 and mAb2). Serum levels of total IgG, IgG1, and IgG2a were measured by ELISA. An anti-mouse IgG2a monoclonal detection antibody cross-reacted with mAb2 but not mAb1, leading to high background when the ELISA plate was coated with mAb2. The problem was solved by use of a goat anti-mouse IgG2a polyclonal antibody that demonstrated the required specificity. IgG2a responses were similar for monomer- or aggregate-coated ELISA plates. The results demonstrate the importance of assessment of the specificity of individual reagents when measuring antibody responses against therapeutic antibodies by ELISA.


Introduction
In the last two decades biotherapeutics, and in particular monoclonal antibodies (mAb), have become increasingly important for treatment of a wide range of disorders (Geng et al. 2015). Currently, more than 250 approved biotherapeutics are available, and there are estimated to be in excess of 500 biotherapeutics at various stages of development (Shankar et al. 2007;Foltz et al. 2013). A variety of factors can influence the immunogenicity of biotherapeutics, including: product-related factors such as protein conformation or impurities, patient-related variables such as immune and genetic background, disease status, and treatmentrelated factors such as route and duration of exposure (Pendley et al. 2003;Tabrizi and Roskos 2007). These can impact the pharmacokinetics and efficacy of biologics; therefore, assessment of potential immunogenicity is an important element of the development and regulatory approval process (Anderson 2005;Chirmule et al. 2012;Maneiro et al. 2013;Sathish et al. 2013).
Products derived from non-human species are predicted to elicit a high incidence of antibody responses because the human immune system is not tolerated to non-human proteins. Consistent with this, there is a high incidence of Anti-Drug Antibody (ADA) development to most therapeutic proteins of bacterial origin, sometimes after a single therapeutic exposure. Aggregation is potentially an important factor which can influence immunogenicity: aggregated proteins are frequently more immunogenic than their non-aggregated counterparts (Ellis and Henney 1969;Moore and Leppert 1980;Rosenberg 2006). An experimental approach adopted in our laboratory is to use a murine intraperitoneal immunization model and humanized monoclonal antibodies or their fragments (Ratanji et al. 2017;Rane et al. 2019).
The purpose of these investigations was not to characterize the immunogenicity per se of a given biotherapeutic, but rather to understand the difference in behavior of the two mAb that are similar in structure, in the experimental settings. In the course of the experiments here, nonspecific binding and background issues were encountered in the ELISA experiments. Humanized biotherapeutic antibodies were used as ELISA substrates and the anti-murine IgG 2a antibodies were used for detection. Humanized mAb were stressed to generate aggregates which were used to immunize the mice and also as ELISA substrates. One of the mAb used performed well with experimental strategy used for ex vivo IgG 2a assessment; however, other mAb did not, generating high background and confounding ADA measurements. A strategy to mitigate this problem is proposed.

Material and methods
Animals BALB/c mice (female, 8-12 week-of-age) were obtained from Envigo (Bicester, UK) for use in these experiments. Mice were housed on sterilized wood bedding with materials provided for environmental enrichment. Food (Beekay Rat and Mouse Diet #1 pellets; B&K Universal, Hull, UK) and water were available ad libitum. The housing facilities were maintained at an ambient temperature of 21 ± 2 C with a relative humidity at 55 ± 10% and with a 12-hr light/dark cycle. All procedures were carried out in accordance with the Animals (Scientific Procedures) Act 1986, and approved by Home Office license.

Monoclonal antibodies
Two human IgG 1 monoclonal antibodies (designated mAb 1 and mAb 2 ) were used for the current study. mAb 1 has a theoretical molecular weight of $148 kDa, and mAb 2 (bi-specific antibody) has a theoretical weight of $204 kDa. Both mAb were provided by MedImmune (Cambridge, UK).

Aggregate formation
Both mAb were diluted to 1 mg/ml in Dulbecco's phosphate-buffered saline (DPBS) without Ca þ2 or Mg þ2 (Sigma, St Louis, MO). mAb 1 aggregates were generated using thermal stress at 60 C for 25 min. mAb2 aggregates were generated by shaking stress at 1,500 rpm in a bench top shaker (IKA V R MS 3 Digital, Oxford, UK) for 4 hr at 22 C.

Immunizations
Mice were immunized by intraperitoneal (IP) injection on Days 0, 7, and 14 with 250 mg of mAb 1 or 150 mg of mAb 2 in aggregated states. All mice were then exsanguinated on Day 21. Individual serum samples and samples pooled on group basis were prepared and stored at À80 C until analysis.

Elisa
Plastic Maxisorb V R plates (Nunc, Copenhagen, Denmark) were coated with 0.1 mg/ml monomeric or aggregated mAb 1 or mAb 2 (prepared in PBS) and incubated overnight at 4 C. Doubling dilutions of serum samples were then added and the plates incubated for 3 hr at 4 C. Sera from naïve mice (NMS) were used as controls. Serum blank wells (including all other reagents) were used to calculate plate background. Horseradish peroxidase (HRP)-conjugated sheep anti-mouse IgG antibody (1:4000 dilution, AAC10P, BioRad, Killington, UK) was used for the IgG ELISAs; HRP-goat anti mouse IgG 1 antibody (1:2000 dilution, STAR132P, BioRad) was used for the IgG 1 ELISAs. Bovine serum albumin (BSA, 2% in PBS) was added as a blocking reagent and the plates were then incubated at 37 C for 30 min. Serum dilutions were prepared in 1% BSA/PBS solution.
For the IgG 2a ELISA, plate wells were coated with 0.1 mg/ml monomeric or 0.05 mg/ml thermal stress-aggregated mAb 1 , or 0.1 mg/ml monomeric or shaking stress-aggregated mAb 2 (prepared in PBS), and then incubated overnight at 4 C. For IgG 2a detection in mice sera, various blocking reagents including 2% BSA in PBS, 5% HSA in PBS, 5% semi-skimmed milk in PBS, and 10% FCS in PBS were examined in ELISA pilot studies. On the basis of the results obtained, a 5% milk block was selected for use in all further experiments. After wells were washed with 0.05% Tween-20 in PBS, serum dilutions (prepared in 2.5% skimmed milk-PBS) were added to dedicated wells and the plates incubated for 2 hr at 4 C. After gentle washing, wells received HRP-conjugated monoclonal rat anti mouse IgG 2a antibody (clone: LO-MG2a-9, BioRad) or HRP-polyclonal goat anti-mouse IgG 2a antibody (BioRad), each diluted 1:1,000. For the IgG, IgG 1 , and both IgG 2a ELISAs, the plates were incubated for 2 hr at 4 C. The wells were then gently washed prior to addition of HRP substrate solution.
Plates then received substrate [1.6 mg/ml o-phenylenediamine and 0.4 mg/ml urea hydrogen peroxide in 0.5 M citrate phosphate buffer (pH 5)] and were incubated for 15 min in the dark at 22 C (RT). Reactions were then stopped by addition of 0.5 M citric acid. Absorbance in each well was then read at 450 nm using an ELx800 automated reader (BioTek Instruments, Winooski, VT), and evaluated using system-associated Gen 5 1.10 software. All data were reported out as OD 450 nm values (± SEM, where appropriate) and mean antibody titers. Titers were calculated as the maximum dilution of serum at which an OD 450 reading of ! 0.3 was recorded (i.e. three-times a reagent blank [all reagents except for serum] OD 450 reading of 0.1).

Statistical analyses
Statistical analyses were performed using the Prism 7 software (GraphPad, San Diego, CA). Analysis of variance (ANOVA) was used to determine the statistical significance of differences between groups. Experiments were analyzed by non-parametric one-way or two-way ANOVA followed by a Tukey's post-hoc test ( Ã p < 0.05, ÃÃ p < 0.01, ÃÃÃ p < 0.001).

Characterization of therapeutic mAb in pre-and post-stress
Two human IgG1 mAb, i.e. mAb 1 and mAb 2 , were used in the current study. Both mAb were prepared at 1 mg/ml in PBS and aggregates generated by application of thermal or shaking stress. mAb 2 showed no aggregation in response to thermal stress (at 45, 50 or 60 C) but aggregates of mAb1 were generated by both methods. Sizes of the generated mAb1 and mAb2 aggregates were analyzed by Dynamic Light Scattering (DLS) (Figure 1). Both mAb monomers showed a narrow size distribution at $10 nm, as anticipated. Application of thermal stress to mAb 1 generated an aggregate population within the sub-visible size range. Aggregate sizes were much larger ($1 mm) when formed by shaking stress mAb 2 . For aggregated mAb, DLS data shows single peak for aggregates with increased particle size, no monomer peak was observed in the same sample.

Assessment of anti-mAb IgG and IgG 1 responses in immunized mice sera
The aim of the study was to investigate effects of aggregation of two mAb on immuno-genic responses. Animals were immunized with 250 ll of 1 mg/mL of mAb 1 and 150 ll of 1 mg/ml of mAb 2 (monomeric or aggregated) in PBS. Following immunization of the mice with aggregated mAb 1 or mAb 2 , the vigor of the IgG response and the isotype distribution of the serum antibody response were analyzed by ELISA using monomeric or aggregate mAb substrates to coat the plates. Pooled serum samples from aggregate-immunized mice were analyzed and comparisons made with serum from naïve animals. Immunization with both mAb in aggregated form provoked vigorous IgG antibody responses compared with IgG levels in naive mice (Figure 2(A)). To further characterize the ADA, IgG 1 and IgG 2a were measured in a similar manner, but using appropriate detection antibodies for subclasses. There was little difference in IgG or IgG 1 antibody profiles for either mAb depending on whether monomeric or aggregated protein substrate was immobilized on the plate, although somewhat higher IgG 1 responses were observed for monomer mAb1 over that of aggregated substrate (Figure 2(B)).
IgG 2a ELISA assays were performed using similar experimental set-ups and anti-IgG 2a detection antibodies. This resulted in very low background (serum blank wells were < 0.3 OD 450 nm) and very low levels of binding of negative control (naïve) mouse serum samples. However, for mAb 2 monomer and aggregate substrate coated plates (Figure 3(A), left panel), the readings were similar in immunized and naïve mice serum samples, did not titrate out upon dilution of serum and serum blank wells also displayed high readings, indicating nonspecific binding of antimouse IgG 2a secondary antibody to the mAb 2 substrate (both monomeric and aggregated forms). The use of a number of different blocking reagents including bovine serum albumin (BSA) in PBS (2%, w/v), skimmed milk (5%, w/v), human serum albumin (10%, w/v), and various combinations thereof, were assessed to reduce background signals and increase the specific signal from immunized mice sera ( Supplementary Figures 1(A,B)). None of these blocking agents were effective at providing a specific signal with reduced background.
In an additional attempt to reduce the high background, analyses were repeated using an alternative detection antibody, a Figure 1. Characterization of mAb1 and mAb2 aggregates by DLS. mAb 1 and mAb 2 were diluted to 1 mg/ml in PBS. In both panels, the dashed line represents monomer (10 nm) for each mAb and the solid line represents thermal-($ 90 nm) and shaking ($1 mm)-stressed aggregated material from mAb 1 and mAb 2 , respectively. horseradish peroxidase (HRP)-conjugated polyclonal goat antimouse IgG 2a antibody. Somewhat surprisingly, despite being a polyclonal rather than a monoclonal antibody, use of this reagent markedly improved the signal to noise ratio, with reduced background for the analyses of the mAb 2 anti-sera (Figure 3(B)). Serum blank values were < 0.3 OD 450 nm, there was no signal above background for naïve mouse serum samples and a specific signal was detected for the serum samples from mAb 2 -immunized mice with the expected reciprocal dilution profile. The polyclonal detection antibody was also suitable for the analysis of the serum from mAb 1 -immunized mice (Figure 3(B)) with a vigorous signal detected regardless of whether the coating substrate was monomeric or aggregated protein. Comparison of blocking agents ( Supplementary Figure 1(A)) for the polyclonal detection agent in mAb 2 -immunized mice serum ELISAs indicated that 5% milk protein was the most appropriate block in terms of minimizing background and maximizing signal.
Analysis of the serum antibody titers revealed that there were no false positive IgG, IgG 1 or IgG 2a antibody readouts in the negative control samples, regardless of whether aggregated or monomeric protein was used as the substrate. Relatively high expression levels of mAb IgG, IgG 1 antibodies were found in sera isolated from monomer or aggregate immunized mice, with virtually identical titration curves irrespective of whether immunization was with the monomer or the aggregated form, or whether the substrate was monomeric or aggregated. In contrast, only immunization with the aggregated form of mAb resulted in a high level of IgG 2a antibody production. This was robust and reproducible finding observed in independently repeated experiments. Thus, in each experiment, equivalent IgG and IgG 1 antibody titers were recorded following immunization with either forms of the mAb, and regardless of the substrate used in the ELISA. A significantly higher titer IgG 2a antibody response was recorded in sera from mAb 1 aggregate compared with mAb 1 monomer immunized mice (Rane et al., 2019). But results were inconclusive for IgG 2a for mAb 2 immunization experiments. This paper gives the details on how this issue was resolved.
A summary of the results and IgG 2a proportion of total IgG generated in response to aggregated mAb 1 and mAb 2 immunizations (vs monomer or vs aggregate protein substrate) is shown in Figure 4. For mAb 1 ELISAs using monomeric and aggregate substrates, serum titers could be determined using both the anti-IgG 2a antibodies (polyclonal as well as monoclonal detection antibodies) and are shown as proportion of total IgG (Figure 4(A,B)). Pie charts illustrate the proportion of IgG 2a within total IgG for individual mice per group. Using monomer and aggregate substrates for ELISA experiments demonstrated subtle differences in individual mice responses; however, that did not have a profound effect in overall experimental outcome, and are demonstrated in a pie chart for individual mice (Figure 4(C)). Naïve mice sera were used for comparison and demonstrated very low IgG 2a titers compared to mAb 1 and mAb 2 immunization groups. Summary of serum titers obtained in mAb 1 -and mAb 2 -immunized groups using both the anti-IgG 2a detection antibodies and both monomeric and aggregate protein substrates is shown in Figure 4(D). Serum titers for IgG 2a were generated using both the detection antibodies for mAb 1 -immunized mice whereas for mAb 2 -immunized mice sera titers could be generated with polyclonal detection antibody only in ELISA experiments. Choice of monomeric or aggregated protein substrate used in the ELISA experiment did not have any significant effects on the results obtained. No significant difference between mAb 1 monomeric substrate with polyclonal or monoclonal detection antibody as well as aggregate substrate with polyclonal or monoclonal detection antibody was observed. With mAb 2 , no significant difference was observed between polyclonal detection on monomeric or aggregate substrate, though high background issues were encountered with monoclonal detection antibody.

Discussion
With the advent of novel analytical technologies, there are significant improvements in the ability to analyze and characterize . Total IgG antibody responses were measured in parallel. Individual serum titers were calculated (n ¼ 5/group) and are expressed as a ratio (A, B, and C) of IgG 2a titers to total IgG titers for the ELISAs using monomeric and aggregate substrates. IgG 2a : total IgG ratios for serum samples from mAb 1 -immunized animals using polyclonal or monoclonal detection antibody against monomeric substrate (A) or aggregated substrate (B). (C) IgG 2a : total IgG ratios for serum samples from mAb 2 -immunized animals using polyclonal detection antibody against monomeric substrate or aggregated substrate. Note: Due to high backgrounds of monoclonal detection antibody vs the mAb 2 substrate, it was not possible to compare between polyclonal and monoclonal detection antibodies for this substrate. (D) Individual IgG 2a titers for serum samples from mice immunized with mAb 1 or mAb 2 (closed symbols) or from naïve mice (open symbols) are displayed as titer (log 2 ), mean, and SEM. ELISA analyses were conducted with monomeric and aggregated substrates and with monoclonal (mAb 1 ) and polyclonal (mAb 1 and mAb 2 ) detection antibodies. Statistical significance of differences in antibody serum titers between all sera groups against each substrate was calculated using one way ANOVA. Ã p < 0.05, ÃÃ p < 0.01, ÃÃÃ p < 0.001.
biotherapeutics. Immunoassays provide cost-effective, highthroughput methods and could therefore represent an effective strategy for evaluation of potential immunogenicity. Humanized animal models are now being established to study these biological effects (Jiskoot et al. 2016). ELISA provides an effective method to monitor the emergence of antibody responses with time and treatment (Li et al. 2001;Tabrizi and Roskos 2007;Geng et al. 2015). In the current investigation, a BALB/c murine model was used and human mAb (at monomeric and aggregated states) were assessed to optimize the IgG and isotype ELISA assays. These observations were consistent with those reported previously (Ratanji et al. 2017;Rane et al. 2019).
When experiments were repeated using same experimental set up for mAb 2 -immunized group, nonspecific cross-reactivity was observed in IgG 2a ELISAs -with higher background readings in test wells as well as in negative controls. This was later resolved using different a detection antibody system. The ADA responses reflect impact on both the T-dependent and T-independent compartments of the immune system (Larocca et al. 1989;Amalfitano et al. 2001;Bertolotto et al. 2003), and underscores the need to characterize the immunogenic responses to biotherapeutic mAb individually. The current investigation highlighted the importance of optimization and evaluation of detection system used for individual biotherapeutic mAb before concluding negative observations, in this case absence of T H 1-polarized immunogenicity in response to mAb 2 immunizations.
Protein aggregation can be described as the self-association of monomers in their native or partially unfolded forms (Chi et al. 2003;Roberts 2007), and is a common phenomenon observed in biopharmaceutical preparations. There is evidence that aggregates can stimulate an anti-drug immune response which may impair drug efficacy. However, the mechanisms through which immunogenicity is enhanced or conferred on proteins are only poorly understood. Aggregates that may be present in protein products can range from dimers to subvisible or visible particles and can be formed during different stages of production, transport or delivery to the patient, in response to diverse stresses (Chi et al. 2003;Mahler et al. 2009) . Aggregates formed in biotherapeutic monoclonal antibodies under the influence of various stresses have been characterized by various techniques on the basis of their sizes, ranging from nm to micron dimensions (Fifis et al. 2004;Morefield et al. 2005;Filipe et al. 2010;Joubert et al. 2011). The mAb aggregates employed in this study fall within this range and can therefore be regarded as typical, at least in terms of size, compared with those studied previously.
The link between aggregation and enhanced immunogenic responses is well established in mouse models (including transgenic animals). For example, aggregate percentage and the extent of denaturation of interferon (IFN)b-1a have been shown to influence the ability of aggregates to break tolerance in transgenic mice (van Beers et al. 2010). Aggregates range in size and dimensions in the 0.1-10 lm range have been identified as being the most immunogenic (Cromwell et al. 2006). Characterization of the aggregated mAb using Raster Image Correlation Spectroscopy (RICS) has been reported previously (Rane et al. 2019). Percentages of aggregates to determine the level of aggregation that increases immunogenicity may not be relevant in the context of this study. Characteristics of aggregates which may contribute to immunogenic potential include the formation of neo-epitopes, multiple valency, post-translational modifications, concentration, and size (Braun et al. 1997;Ryff and Schellekens 2002;Schellekens 2002). The murine model used here helps us to understand that humanized mAb, though with similar backbone and structure, may behave differently under similar experimental conditions assessing mAb immunogenicity. Thus, it is important that each mAb is studied on a case-by-case basis.
In the current investigation, it was noted that the ELISA detection system used plays a crucial role in the experimental outcome and conclusions derived. The epitopes identified by the commercially-available detection antibodies have the potential for cross-reactivity with the ELISA substrate due to the exposed epitopes; the blocking reagents used also play important roles in preventing nonspecific binding of antigen and antibodies to microtiter plates as described previously (Chart et al. 1998). Assays utilized to assess and characterize immunogenicity should be designed to be sensitive and specific for the intended purpose. Appropriate experimental design, reagents used, and timepoints should be determined along with recommendations for an assay procedure, when needed, to detect meaningful antibody responses at pre-clinical and clinical stages in the development of ELISA assays using therapeutic mAb.

Conclusions
The present study confirmed that the binding epitope of antimouse secondary/detection antibodies used in the experiments play a critical role in study design as well as experimental outcome. It also proved beneficial in the current study to optimize and assess different blocking reagents used, commercially-available and commonly-used secondary antibodies that highlighted the differences in IgG 2a responses. It was also observed (with comparison to a previously-published study) that although both mAb have IgG1 subclass, subtle structural differences in the mAb can give rise to neo-epitopes inducing altered experimental outcome. This may play a critical role in characterization of ADA responses. Hence it is important to assess and optimize individual biotherapeutic monoclonal antibody proteins in the experimental systems.