With the broad and increasing application of therapeutic monoclonal antibodies (mAbs) in clinical settings, IgG-induced allergic reactions, including passive systemic anaphylaxis (PSA), have attracted significant attention. However, it is not clear which types of IgG mAb–antigen complexes or IgG aggregates formed by antigen binding can trigger PSA, as not all immune complexes (ICs) are capable of triggering PSA. Here, we characterise mAb–antigen complexes capable of inducing murine PSA to evaluate and predict which ICs are able to induce PSA.
Thirty-six combinatory reactions with eight antigens and 27 corresponding mAbs were used to trigger PSA, which was defined by rectal temperature. Sandwich ELISA, passive cutaneous anaphylaxis (PCA) induction and flow cytometry analysis of CD16/32 (FcγRIII/II) expression were used to characterise the ICs. The dynamic concentrations of antigen in the peripheral blood were measured by ELISA.
Only 14 of the 36 ICs could trigger PSA and thus be termed anaphylaxis-inducing immune complexes (Ai-ICs). The Ai-ICs could be characterised by constructing sandwich ELISA, inducing PCA and down-regulating CD16/32 (FcγRIII/II) expression on blood neutrophils in vitro and in vivo. Additionally, the occurrence and severity of PSA was found to be associated with the instantaneous concentration of antigen in the peripheral blood in the presence of antibody.
Only Ai-ICs, not all ICs, could trigger IgG-mediated PSA, which could be characterised by the above simple methods. The occurrence and severity of PSA was associated with the instantaneous concentration of antigen in the peripheral blood in the presence of antibody.
Passive systemic anaphylaxis (PSA) is a life-threatening allergic reaction that can take place in hosts following antibody infusion [1-4]. Recently, IgG-induced PSA, a severe adverse effect that can take place following therapeutic treatment with monoclonal antibodies (mAbs) and immunoglobulins, has attracted significant attention [2, 3, 5, 6]. In clinical settings, antibody infusion-induced PSA is estimated to range in frequency from approximately 0.09 to 77% depending on the type of antibody used  and can occur rapidly, as quickly as several minutes after the first infusion of therapeutic mAbs [8-10]. To date, 30 therapeutic mAbs have been approved by the FDA, and more than 300 therapeutic antibodies have been studied in clinical trials .
In mouse models, IgE-induced anaphylaxis occurs through the classical pathway, while IgG-induced anaphylaxis uses an alternative pathway that can be mediated by basophils, macrophages and neutrophils [12-14]. All elements in the murine alternative pathway are thought to play a pivotal role in human IgG-induced anaphylaxis, suggesting that this alternative pathway may exist in both mice and human . Notably, not all ICs or IgG aggregates that are formed by binding with antigen can trigger PSA in mouse models, which is similar to what is observed in humans. Therefore, it is important to evaluate and predict what types of ICs are capable of triggering PSA.
In this study, we defined the characteristics of mAb–antigen complexes inducing PSA by observing 36 combinatory mAb–antigen reactions. We found that constructing sandwich ELISA, inducing PCA and down-regulating CD16/32 (FcγRIII/II) expression on blood neutrophils both in vitro and in vivo could be used to effectively identify ICs that are capable of inducing PSA. We also observed that the occurrence and severity of PSA was associated with the instantaneous concentration of antigen in the peripheral blood.
Female BALB/c mice (6–10-week-old) were purchased from the Laboratory Animal Centre of Sun Yat-Sen University (Guangzhou, Guangdong, China). All animal experiments were approved by the Ethics Committee for Experimental Animals at Southern Medical University and were performed according to the national guidelines for animal welfare.
Monoclonal antibodies for PSA
mAb No. 7 and No. 9 against ovomucoid (OVM) were provided by Li et al. ; mAbs against human serum albumin (HSA) and recombinant mannoprotein 1 of penicillium (MP1P) were provided by Ma et al.  and Wang et al. ; mAbs against recombinant human haemoglobin ζ (rHb zeta), human haemoglobin ζ (Hb zeta), human haemoglobin A2 (HbA2) and advanced oxidation protein products (AOPP) were prepared in our laboratory [19-22]. The specificity and isotype of mAbs are listed in Table S1.
Mice were sensitised with an intravenous (i.v.) injection of mAb in 200 μl PBS and challenged i.v. with antigen in 200 μl PBS 2 h later. Rectal temperature was measured using a digital thermometer (Model BAT-12; Physitemp Instruments, Inc., Clifton, NJ, USA) and recorded from 0 to 90 min postchallenge. Additionally, preformed ICs could be directly used for challenge without priming (Fig. 1C). To measure PSA-induced plasma exudation, the protocol was modified by challenging mice with antigen in 200 μl PBS containing 0.5% Evan's blue dye.
PCA in the paw was induced as previously described with slight modifications . Mice were sensitised with an intradermal (i.d.) injection of mAb in 20 μl PBS in the hind paws and i.v. challenged with antigen in 200 μl PBS containing 0.5% Evan's blue 2 h later. Both front paws, used as controls, and hind paws were photographed at 5 min postchallenge. The preformed ICs were also directly used for challenge (Fig. 2C and Table 2).
Determining mouse CD16/32 and human CD16 & CD32 expression on blood neutrophils
Both mouse CD16/32 and human CD16 & CD32 expression were analysed by flow cytometry as described in the supplement.
Detecting the instantaneous concentration of antigen in blood
The indicated doses of biotinylated OVM, biotinylated HSA and rHb zeta were injected i.v., and blood samples were taken at different time points postinjection. The serum biotinylated protein and rHb zeta were determined by either indirect or sandwich ELISA.
All statistical analyses were performed with SPSS 13.0 software (IBM Corporation, Armonk, NY, USA).
Further method details
The methods detailed in the supplement include the generation and identification of mAbs to OVM and ovalbumin (OVA), epitope mapping by the anti-OVM mAbs (Tables S2 and S3), the doses of mAb and antigen used for inducing PSA, PCA and blood neutrophil CD16/32 down-regulation (Table S4), and detailed protocols for ELISA and flow cytometry analysis.
Murine PSA can only be induced with a fraction of anti-OVM mAbs or mAb cocktails reacted with OVM
OVM and anti-OVM mAbs were used to establish a PSA mouse model (Fig. 1A). After priming with 100 μg mAbs, as little as 5 μg OVM could induce PSA. Increasing OVM doses could enhance the severity of PSA, but no further severity was observed when the antigen doses were increased to between 100 and 1000 μg OVM per mouse (Fig. 1B). Thus, we chose 10 μg OVM as the working dose, which was also verified by measuring plasma exudation of dye in the mouse ears (Fig. 1D). PSA of similar severity to that observed using 10 μg OVM could also be induced using 100 μg OVA containing trace OVM (Fig. S2). In addition, PSA could also be induced using ICs preformed in vitro by mixing OVM with mAb 2C2 (Fig. 1C). Moreover, we found that only 3 mAbs, 2C2, 5G12 and 6H4, could initiate PSA following OVM challenge, whereas the other 4 anti-OVM mAbs could not (Fig. 1E). Interestingly, using cocktails of mAbs, No. 7 plus No. 9, No. 7 plus 1A6 or No. 7 plus 5B8 could initiate PSA, but the other combinations could not (Fig. 1F). These results, further verified by testing the various mAb-antigen reactions shown in Table 1, indicated that only a fraction of mAbs or mAb cocktails bound with antigen could trigger PSA.
Table 1. PSA induction using 36 Antigen–mAb combinatory reactions
The antigen–mAb dose for PSA is summarised in Table S4.
Superscripts on mAb names indicate the IgG isotypes.
Maximum rectal temperature variation (Δtmax) during the 90 min following challenge was classified according to the standard created by Jonsson F et al.  with modifications. Grade 1: no shock, +1°C > Δtmax ≥ −1°C, no mortality; Grade 2: mild shock, −1°C > Δtmax ≥ −4°C, no mortality; Grade 3: severe shock, −4°C > Δtmax, possible mortality. The results are presented as mice in each grade vs total number of mice used.
OVM, ovomucoid, from chicken egg white.
HSA, human serum albumin.
MP1P, recombinant mannoprotein 1 of penicillium.
rHb zeta, recombinant human haemoglobin zeta chain; Hb zeta, human haemoglobin zeta chain from cord blood of Bart's thalassaemia.
To define the ICs capable or incapable of inducing PSA, we termed the ICs capable of inducing PSA as anaphylaxis-inducing ICs (Ai-ICs) to distinguish from non-anaphylaxis-inducing ICs (nAi-ICs).
An association between PSA induction in vivo and sandwich ELISA formation in vitro with the same ICs
Based on the requirement of non-monomeric IgG binding with multivalent antigen for both triggering PSA and forming sandwich ELISA, we set up a sandwich ELISA using OVM and the mAbs or mAb cocktails shown in Fig. 1E,F. The results indicated that the individual mAbs 2C2, 5G12 and 6H4, but not the other 4 mAbs, could function as both capture antibodies and detecting antibodies to form sandwich ELISA (Fig. 1G). Moreover, the mAb cocktails No. 7 plus No. 9, No. 7 plus 1A6 and No. 7 plus 5B8, but not the other three cocktails, could also form a sandwich ELISA (Fig. 1H). Notably, the PSA results in Fig. 1E,F show an association with the sandwich ELISA in Fig. 1G,H, as both used the same mAbs or mAb cocktails and OVM as antigen. In addition, the epitope mapping using competition ELISA showed that seven anti-OVM mAbs recognised three types of distinct epitope on OVM (Table S3), and the mAbs of mAb cocktails inducing PSA and forming sandwich ELISA recognised two different epitopes on OVM.
We further wondered whether this association would also exist for other antigen–antibody reactions. The results in Table 2 show that, of 19 ICs containing antigens and mAbs capable of forming sandwich ELISA (Table 2, reaction No. 1–19), 14 of these ICs could induce PSA (Table 2, reaction No. 1–14), and the IgG mAbs involved in the Ai-ICs could be IgG1, IgG2a or IgG2b. None of the 17 ICs incapable of forming sandwich ELISA could induce PSA in vivo (Table 2, reaction No. 20–36). Nevertheless, the sandwich ELISA was still useful as a preliminary criterion for determining Ai-ICs, as no ICs unable to form sandwich ELISA could induce PSA.
Table 2. Summary for characteristics of Ai-ICs inducing PSA and PCA
Ai-ICs induce PCA in parallel with the induction of PSA
Because IgG antibody–antigen complexes can also induce PCA, which is mediated by tissue mast cells expressing FcγRs [24, 25], PCA was used to evaluate the IgG ICs capable of triggering PSA. As shown in Fig. 2, PCA could be induced by i.d. injection of antibody followed by i.v. injection of antigen (Fig. 2B) or direct i.d. injection of preformed ICs (Fig. 2C). The results in Table 2 show full coincidence between PCA and PSA induction for all 14 ICs (Table 2, reaction No. 1–14, the primary photographic data are shown in Fig. S3. Importantly, inducing PCA with an i.d. injection of preformed ICs (approximately 2–20 μg antigens in ICs) might be a simple, sensitive and safe way to screen for Ai-ICs and predict PSA occurrence.
Ai-ICs induce down-regulation of blood neutrophil CD16/32 expression concurrent with PSA induction
Because the sandwich ELISA showed an incomplete coincidence with PSA induction, and PCA can only be evaluated in vivo, we tried to use an accurate and convenient in vitro test for Ai-IC prediction. The results in Fig. 3 show that 2C2–OVM complexes could induce significant down-regulation of mouse blood neutrophil CD16/32 expression in vivo (Fig. 3A) and in vitro (Fig. 3B) and could also induce down-regulation of CD16 and CD32 expression on human blood neutrophils (Fig. 3C). Furthermore, we conformed that Ai-ICs, but not nAi-ICs composed of other antibodies and antigens, could induce down-regulation of CD16/32 expression on murine blood neutrophils in vivo and in vitro (Table 2, reaction No. 1–14, the primary data are shown in Fig. S4).
Collectively, we summarised the characteristics of Ai-ICs in Table 2, which indicated that the occurrence of PSA exhibited a strong association with the sandwich ELISA and notably complete coincided with PCA induction and down-regulation of blood neutrophil CD16/32 expression.
The occurrence and severity of PSA is associated with the instantaneous antigen concentration in the peripheral blood in the presence of antibody
Generally, the dynamic serum concentration of a protein is associated with its clearance rate or half-life. The results presented in Fig. 4A, E and G show that the clearance rates of OVM and rHb zeta were faster than that of HSA in the absence of antibody. A postinjection drop in murine rectal temperature was also correlated with the instantaneous concentration of OVM (Fig. 4A, B, C, D), HSA (Fig 4E, F) and rHb zeta (Fig 4G, H) in peripheral blood in the presence of their corresponding antibodies, indicating that lower instantaneous concentrations of antigen were associated with milder drops in rectal temperature, with no drop observed in some instances. These results suggested that the dynamics of instantaneous antigen concentration change related to the clearance rate in peripheral blood could impact the occurrence and severity of IgG-induced PSA.
In this work, we proposed and demonstrated that only Ai-ICs, not all ICs, could trigger PSA. It should be emphasised that mAbs or mAb cocktails forming Ai-ICs must bind multiple epitopes to form non-monomer IgG aggregate to cross-link FcγRs on the effector cell membrane, and non-monomer IgG aggregate is also indispensable for the formation of sandwich ELISA. Therefore, sandwich ELISA as a simple method could be used for preliminary screening to predict the formation of Ai-ICs, as none ICs incapable of forming sandwich ELISA induce PSA, although several ICs forming sandwich ELISA are incapable of triggering PSA. This incomplete coincidence between triggering PSA and forming sandwich ELISA may be associated with the size and conformation of antigen molecule. The sandwich ELISA only needs the binding of two epitopes on antigen with capture and detecting mAb, while triggering PSA requires at least two IgG Fc portions to bind FcγRs on the effector cell membrane, in which the size and conformation of antigen as well as the distance between the two epitopes on antigen molecule may influence the ability of Fc portions to cross-link FcγRs. For instance, both recombinant Hb zeta and native Hb zeta could be detected by the same sandwich ELISA system, but only rHb zeta could trigger PSA, which may be attributed to the fact that rHb zeta is a monomer, but native Hb zeta from cord blood can exist as a monomer, dimer, trimer or even tetramer.
PCA can also be used to evaluate whether ICs can trigger PSA as the ICs capable of inducing PCA completely coincide with those able to induce PSA. This coincidence implies that PCA induced by direct i.d. injection with in vitro preformed ICs could be a simple and safe method to predict PSA before i.v. administering therapeutic antibodies.
As PCA may not be suitable for human use in vivo, we used a more convenient in vitro way to predict the formation of Ai-IC by consulting Khodoun's work that showed a down-regulation of CD16/32 expression on mouse and human neutrophils, but not on mouse monocytes, dendritic cells, NK cells or basophils during IgG-mediated PSA . In our work, noticeably, a decreased CD16/32 expression on mouse neutrophils and CD16 & CD32 on human blood neutrophils could only be induced by Ai-ICs containing murine IgG mAbs, but not by nAi-ICs. Furthermore, we found a correlation between severity of PSA and amplitude of decreased CD16/32 expression on mouse neutrophils (Fig. 1B and Fig. S5), in which a decreased CD16/32 expression could be detected by mAb 2.4G2 sensitively. As murine IgG1 cannot bind human CD16, but ICs containing murine IgGs can bind human CD32 that is an activating receptor for PSA [14, 27-29], it could be supposed that the decreased CD16 expression on human neutrophils induced by ICs containing murine IgG may be not a result directly from CD16 binding, but rather a result from the binding or action via CD32. Additionally, there was also a consideration that loss of CD16 during IgG-mediated anaphylaxis requires cell signalling involving receptor internalisation or shedding .
We also noticed that the clearance rate or instantaneous concentration of antigen in the peripheral blood in the presence of antibody could impact the occurrence and severity of PSA. These results suggest that it may be necessary to measure the target antigen concentration in the blood during treatment with therapeutic antibodies or IVIG. Particularly, cell components could act as antigens from dead tumour cells following antibody or chemical therapy, such as that observed in tumour lysis syndrome .
Moreover, some PSA models, using high-dose (approximately 0.5–4 mg) antigen or hapten conjugates [14, 31-33], led to a consideration that IgG-mediated PSA could not occur in human because it may be impossible for a person to have such high doses of antigen in vivo . However, Ishikawa's and our work using low doses antigen and mAbs suggest that it is more plausible that IgG-induced mouse PSA models can be used to investigate PSA resulting from the therapeutic antibodies infusion in humans .
In summary, we showed that only Ai-ICs could trigger IgG-induced PSA. Ai-ICs could be defined by construction of sandwich ELISA, induction of PCA and down-regulation of blood neutrophil CD16/32 expression. Our work is expected to provide new strategies for evaluating the safety in administration and development of therapeutic antibodies, including mAbs, polyclonal antibodies and immunoglobulins.
We thank Boquan Jin (Fourth Military Medical University, China) for his considerable advice and reagent support, Wenwei Tu (Hong Kong University) for reconfirming our animal models and Xiaoyan Che (Zhujiang Hospital, Southern Medical University) for reagent support and Lichang Liu for preparing experimental material.
DL Jiao and Y Liu performed most of experiment; X Lu and QJ Pan performed cytologic analysis and a part of preliminary experiment; J Zheng and YJ Liu established animal model and revised the manuscript; YF Wang, BY Liu performed the generation and purification of monoclonal antibodies and recombinant proteins. N Fu with help from DL Jiao and J Zheng designed all research and wrote the manuscript.
This work is supported by a research grant from the National Natural Science Foundation of China (30671970 to Ning Fu).
Conflict of interest statement
None of the authors has any conflict of interest to disclose regarding this manuscript.