An ELISA for detection of complement-bound circulating immune complexes in mice

Measurements of complement-bound circulating immune complexes (cCICs) in pre-clinical studies may provide important information about the etiology of certain pathology findings suggestive of being immune complex mediated. This article describes the development and qualification of a universal methodology to measure cCIC in mice after dosing with species foreign proteins. The assay is a sandwich enzyme-linked immunosorbent assay – exclusively based on commercially available reagents – that could detect mouse IgG bound to complement C3 independent of the test-substance present in the plasma sample. Heat-aggregated serum was used as positive control. The assay was qualified by assessment of acceptance criteria, stability of positive control, precision, and specificity. Finally, the performance of the assay was tested using plasma from mice administered either of three different proteins, i.e bovine serum albumin (BSA), a fully human monoclonal antibody, and a humanized monoclonal antibody.


Introduction
Anti-drug antibodies binding to administered test-substances may form circulating immune complexes (CICs) that can activate complement, resulting in complement fragments bound to CICs (cCIC) (Theofilopoulos and Dixon 1979;Brennan et al. 2010). Whereas complement C1q binding to CICs indicates activation early in the classical cascade, complement C3b fragments covalently attached to CICs indicate activation of later events in the complement cascade (Theofilopoulos and Dixon 1979;Murphy and Weaver 2016). Formation of cCICs may subsequently give rise to pathological changes due to cCIC deposition in the vasculature (Rojko et al. 2014), the inflammatory properties of complement fragments (Bourke et al. 1982), and recruitment of inflammatory cells (Mayadas et al. 2009;Krishna and Nadler 2016). Therefore, to explain any pathological findings when testing new therapeutics in animal models, it is very important to be able to detect various cCICs.
Several human-specific cCIC assays detecting endogenous IgG bound to C1q or C3 have been developed for clinical screening (Pereira et al. 1980;Soares et al. 2001;Stanilova and Slavov 2001). Similar assays are valuable for pre-clinical use. Compared to cCIC measurement in humans, few papers have described assays established for detection of cCIC in mice (Gabriel and Agnello 1977;Pereira et al. 1980;van Dam and Hack 1987;Hewicker et al. 1990;Machida et al. 2018). Qualified methods using a C3-IgG ELISA to measure cCIC in plasma from mice administered species-foreign proteins have not been described.
In this article, a simple sandwich ELISA was developed and qualified for detection of cCICs formed in vivo. The protocol employs a solid phase anti-C3 as a detecting antibody and a fluid phase horseradish peroxidase (HRP)-labeled anti-mouse (m)IgG antibody for capture; the entire assay is exclusively based on commercially available reagents. Assay performance/validity was tested using plasma isolated from mice that had been treated with human (h)IgG 1 , humanized (hz)IgG 4 , or bovine serum albumin (BSA).

Preparation of positive controls
Aggregated IgG is known to stimulate complement activation by the classical complement pathway (Lutz et al. 1996). Upon complement activation additional monomeric IgG is able to bind the cCICs (Van Dam and Hack 1987). To closely resemble complement bound hIgG-mIgG complexes formed in vivo heataggregated human c-globulin (HAGG) was used to stimulate complement activation in mouse serum. The material was used as a positive control (Figure 1).

cCIC ELISA
The assay principle is illustrated in Figure 2. Polyclonal rabbit anti-mouse C3 antibody (Bioss Inc., Boston, MA, cat. #BS-2934R) was diluted to 1 mg/ml in carbonate buffer (0.1 M, pH 9.6) and coated (at 100 ml/well) overnight at 4 C onto the walls of 96-well NUNC-Immuno MicroWell plates (Thermo Fisher Scientific, Roskilde, Denmark). Following washings (three wash cycles/300 ml dilution buffer/cycle) in an ELISA washer (Power Washer 384, Tecan, Gr€ odig, Austria), each plate was blocked (on shaker set at 300 rpm) for 1 h at 21 C using 200 ml dilution buffer/well. Blocking solution was then removed and 100 ml study samples/well (each diluted 1:80) and QCs were added to dedicated wells. The plates were then incubated 1 h at 21 C on a shaker. Each plate was washed with buffer and then 100 ml HRP-labeled goat anti-mIgG (Jackson ImmunoResearch, Newmarket Suffolk, UK, diluted 1:10 000 in dilution buffer, cat. #115-035-174) was added to each well. After a 1 h incubation at 21 C (on shaker), each plate was washed with buffer and the color reactions in each well initiated by addition of 100 ml trimethoprim (TMP One; Kem-En-Tec Diagnostics, Taastrup, Denmark) per well. Reactions were stopped after 6-7 min by addition of 100 ml 4 M o-phosphoric acid/well, and then spectrophotometric measurements of optical density (OD 450-620 ) were measured in a Sunrise ELISA plate reader (Tecan) using Magellan Software (Tecan).

Qualification of ELISA
The QCs were used for setting acceptance criteria; the calculation was based on QC high , QC medium , and QC low assayed in duplicate in seven independent assay runs. The lower limit acceptance criteria was calculated as: [Mean of OD] À [(one-sided t 0.01,df ¼ 3.365 for df:5)] Â [SD of OD], leading to a 1% rejection rate. If the QC were above the criteria and QC neg below the assay cut-point, the assay was accepted. To check the stability of the positive control, undiluted positive control was stored under different conditions: at À80 C for 5 d and 3 month, at 4 C for 1 and 5 d, after four repeated freeze/thaw cycles, and after eight repeated freeze/thaw cycles, followed by a 1:80-dilution and analysis in duplicate in the cCIC ELISA. An OD signal change < 20% was considered acceptable. Evaluation of intra-and inter-plate variations were based on QC analyzed in 11 replicates on the same plate and in six independent assay runs performed over two days. All samples were measured in duplicate, and mean OD values were reported as percentage of the coefficient of variation (% CV).
Protein G elution was used to separate a mouse plasma pool (previously shown to have a high cCIC level; K 2 EDTA C57BL/6 mouse plasma; BioIVT, Westbury, NY) into IgG and non-IgG fractions and to determine the signal from these. Eluation was performed in NAb Protein G Spin Columns (Thermo Fisher Scientific) according to manufacturer instructions, followed by freezing of the non-IgG and IgG fractions at À20 C until assayed. The two elution fractions, the positive controls and unfractioned plasma pool, were then diluted 1:20, 1:40, 1:80, or 1:160 in dilution buffer. A negative control consisting of monoclonal mIgG 2 (mouse anti-human IgG 1 , BD Pharmingen, San Jose, CA) was diluted to 10, 5, 2.5, or 1.25 lg/ml in dilution buffer. All samples were assayed in duplicate in the cCIC ELISA. All results are presented as mean OD values.
To assess potential signal disturbance between BSA and assay antibodies, BSA was diluted to 100 or 10 lg/ml in QC high , QC medium and QC low , or serially diluted from 2000 to 0.017 lg/ml in dilution buffer containing 1:80 plasma pool. The samples were assayed in duplicate and results given as mean OD values. Figure 1. Preparation of the positive control. The positive control was used as quality control in the enzyme-linked immunosorbent assay (ELISA). Mouse serum containing mouse immunoglobulin (Ig) and inactivated complement was incubated with heat-aggregated human IgG ("Complement Activator") to stimulate complement activation by the classical complement pathway. Monomeric mouse IgG was able to bind to aggregated human IgG, leading to formation of complement bound hIgG-mIgG complexes.

Assay cut-point determination
A panel of 36 drug-naïve murine plasma samples obtained from BALB/c and C57BL/6J mice (Charles River, Sulzfeld, Germany) were analyzed in duplicate in six independent assay runs. Data were analyzed using log-transformed S/N data and statistical information about distribution, means and variances were obtained using SAS.JMP version 13.0.0 (SAS, Cary, NC). A parametric cut-point approach allowing a false positive rate of 5% was based on the formula [Mean of S/N] þ [(t 0.05,df ¼ 1.645 for df:1) Â [SD of S/N], followed by logarithmic back-transformation, as described by Devanarayan et al. (2017).

Animals
C57BL/6J mice (48, female, 10-12 Week-of-age) mice were purchased from Charles River (Sulzfeld, Germany) and housed in pathogen-free facilities maintained at 21 C with a 60% relative humidity and 12-h light/dark cycle. All mice had ad libitum access to standard rodent chow and filtered tapwater. After at least 1 Week of acclimatizing, the mice were randomly allocated into groups (n ¼ 12/group) wherein they were injected subcutaneously twice weekly for 13 Weeks with vehicle, 10 mg/kg hIgG 1 (adalimumab, human anti-TNFa: Humira, AbbVie, North Chicago, IL), 10 mg/kg hzIgG 4 (antibody directed against trinitrophenyl [Novo Nordisk A/S, Måløv, Denmark]), or 50 mg/kg BSA (Sigma Aldrich, Schnelldorf, Germany). These animal experiments were approved by the Danish Animal Experiments Council under the Danish Ministry of Environment and Food, as well as the Novo Nordisk Ethical Research Committee.

Determination of cCIC in mouse plasma
One Day after the final injection, blood was collected from the periorbital plexus of each mouse and plasma obtained by centrifugation. The procedure was performed under isoflurane anesthesia and following sampling, the mice were euthanized by puncture of the abdominal aorta. Each isolated plasma sample was then stored at À80 C until analysis. At that time, the sample was thawed and then diluted 1:80 in dilution buffer just prior to the analysis. Samples were considered cCIC positive if above the assay cut-point.

Assay performance and cut-point evaluation
A lower-limit acceptance criterion for positive control quality samples was determined to an OD of 0.36, 0.19, and 0.08 for QC high , QC medium , and QC low , respectively; the OD mean of QC neg was 0.047 (Figure 3(A)). When maintained under different storage conditions, the positive control showed a stable OD signal, with changes of <20% CV (Figure 3(B)). The precision of the assay was considered acceptable with an intra-assay precision of QC high , QC medium , and QC low at 2.36, 3.48, and 4.48% CV, respectively, and an inter-assay precision of 20.36, 16.51, and 16.28% CV, respectively (Table 1). Protein G purification of a mouse plasma pool with high cCIC levels, showed that the majority of the signal in the cCIC assay derived from the IgG fraction (Figure 3(C)). In addition, the signal from the IgG-fraction was comparable to the positive control itself at 1:40-1:160 dilutions; at 1:20 dilution, the signal from the IgG fraction did not reach the same level as the positive control.
In the non-IgG-fraction (mouse plasma after protein G purification), the cCIC signal had clearly been decreased. Further, to test that antibody by itself did not induce a signal, a monomer mIgG (without C3) was tested. A low signal (between 0.28 and 0.12 at 10 and 1.25 lg/ml dilution, respectively) was observed, confirming specificity to CIC (data not shown). To ensure BSA from the test samples did not interfere in the assay, QC or plasma pool containing various concentrations of BSA were tested. No considerable changes of OD signals were observed (Figure 3(D,E)).
A cut-point was determined in order to assign test samples as positive or negative for cCIC. The Shapiro-Wilk W test showed normal distribution of log 10 -transformed (S/N) data in 5 of 6 runs (Shapiro-Wilks, p > 0.05) analyzing a panel of 36 drugnaïve murine plasma samples obtained from BALB/c and C57BL/ 6J mice. A Welch ANOVA and a Levene's test each showed homogeneity of means and variances (p > 0.05) and on pooled data from all six runs, a skewness coefficient at 0.43, suggesting a parametric cut-point approach was suitable after removal of two outliers. The cut-point was determined to 9.17 (10 0.936 ) S/N as calculated from the formula 0.455 þ 1.684 Â 0.301 ¼ 0.936.

Discussion
This article describes a generic method for measurement of in vivo-formed cCIC (consisting of C3 bound to CICs) in murine plasma samples. The assay was exclusively based on commercially available reagents, making it easy to set up in any department. Based on the performance qualifications, the assay was considered valid for detecting cCIC in mouse plasma. As previously described, heat-aggregated IgG was used as a positive control (van Dam and Hack 1987;Muratsugu 1996), and in order to mimic in vivo cCIC formation in mice administered with human proteins, the positive control was based on human HAGG in murine plasma.
The specificity of the assay was examined after protein G fractionation of plasma, showing that the signal was derived from the plasma IgG-fraction rather than the non-IgG fraction. Furthermore, limited signal was detected from monomer mIgG which confirms the specificity of the assay. The performance of the assay was tested using study samples from mice administered BSA or human(ized) antibodies. Different levels of cCIC in the different groups were observed, confirming the ability of the assay to detect cCIC in mouse plasma samples independently of the test-substance. However, cCIC was detected in one control mice, likely due to spontaneous hydrolysis of C3 independent of test substance, which may cause cCIC formation under normal circumstances (Pangburn and Muller-Eberhard 1983;Murphy and Weaver 2016). In addition, the cutpoint allowed for 5% to be false positive. All mice receiving BSA had considerable higher cCIC levels compared to the other test substances. The qualification showed no interference from BSA that could explain this and the higher level of cCIC observed was likely to be caused by higher immunogenic properties of BSA. The relatively high cutpoint was due to individual vehicle mouse responses measuring higher than the pool and high variability in the responses across the mice. However, this did not hinder cCIC detection in the current study.
The current assay only detected cCIC activated by C3; to ensure detection of all CICs that interact with complement, supplemental application using other CIC assays may be optimal (Lambert et al. 1978;Stanilova and Slavov 2001), e.g. using a C1q-IgG assay frequently used for detection of cCIC opsonized with C1q (Gabriel and Agnello 1977;Abrass et al. 1980;Devey et al. , 1982Hewicker et al. 1990;Hori and Abrass 1990;Chen et al. 2004) or CIC assays that are based on the ability of CIC to interact with complement receptors (Noble et al. 1987). control following storage at À80 C for 3 months or 5 d, at 4 C for 1 or 5 d, or after four or eight repeated freeze/thaw cycles. Samples were diluted 1:80 and analyzed (in duplicate) in cCIC assay. (C) Specificity of a mouse plasma pool previously analyzed with high cCIC levels, following separation into IgG and non-IgG fractions using protein G columns. The non-IgG fraction shows minimal signal, whereas the IgG fraction, shows larger signal, similar to the un-fractioned mouse plasma. The IgG-fraction signal was comparable to the positive control itself at dilution factor 1:40-1:160. (D,E) No considerable changes in assay readout from 1:80 plasma (D) or QC (E) containing various concentrations of BSA were observed, which suggested no interference between BSA and assay antibodies. (F) Positive cCIC results were detected in 1/12, 12/12, 7/12, and 1/12 mice dosed subcutaneously twice weekly for 13 Weeks with either vehicle, 50 mg/kg BSA, 10 mg/kg hIgG 1 , or 10 mg/kg hzIgG 4 . Dotted line represents cut-point. Signal expressed as optical density (OD) signal (A-E), signal to noise (S/N; OD signal/QC neg ) (F) and mean þ SD (F).

Conclusions
An assay for detection of cCICsregardless of antigenin murine plasma samples was successfully developed and qualified. The assay was shown to be specific, with high precision. Because it was based on commercially available reagents, this makes it easy to implement by other investigators. Inter-and intra-assay variation calculated based on six independent assay runs (inter-assay precision) and 11 replicates on same plate (intra-assay precision). Assay precision resulted in acceptable variation with coefficient of variation (CV) equal to or below 20%.