Crucial Role for Immune Complexes but Not FcRn in Immunization against Anti–TNF-α Antibodies after a Single Injection in Mice

Anti–TNF-α Abs are widely used; some are even blockbuster biopharmaceuticals. Like adalimumab and infliximab, anti–TNF-α Abs are indicated in severe rheumatoid arthritis, in first-line treatment, or in mild cases when disease-modifying antirheumatic drugs or methotrexate alone have failed to control disease progression (1, 2). Yet immunization issues leading to interindividual variability tarnish their success, frequently forcing clinicians to switch to another treatment line.

The immunogenicity of infliximab and adalimumab is a major concern, because at least 14% and up to a third of the patients develop Abs also called antidrug Abs (ADA), directed against these anti–TNF-α Abs after just a few weeks of treatment (3, 4). These ADAs can lead to adverse effects during infusion or to a decrease in biologic concentration, which was associated with lower treatment efficacy (4).

Many factors have been suspected to be involved in the immunization against therapeutic Abs. Even though some have been refuted, like the degree of Ab humanization (5), or discussed, such as IgG allotypes (6–8), several other factors have been shown to be involved, like the dosing and administration schedule, underexposure to the anti–TNF-α drug (9), coadministration of other drugs (10), and formation of aggregates (11). Complexes can be formed between several molecules of the therapeutic Ab (aggregates) or with its target Ag as in classical immune complexes (ICs) to constitute target-related ICs (TRICs) (12). It has been shown that mAbs are immunogenic in vivo when they form TRICs (13). This phenomenon can be understood in the light of the recent literature concerning IC presentation by immune cells via the neonatal Fc receptor (FcRn) and with the work of Foss et al. (14), who recently demonstrated that TRICs are more efficiently transcytosed than monomeric mAb via FcRn. These mechanisms bring into play the well-described functions of FcRn, extension of IgG half-life, IgG biodistribution, and involvement in humoral and antitumoral immune response through IC presentation by immune cells (15, 16).

The present work focuses on the ability of TRICs to trigger a primary immune response (i.e., after a single injection) against anti–TNF-α Abs in the presence or absence of FcRn. To investigate this, anti–TNF-α Abs were administered to wild type and FcRn knockout (KO) mice and the formation of ADA was measured. The first Ab, infliximab, specifically recognizes human TNF-α (hTNF-α) whereas the other, adalimumab, cross-reacts with mouse TNF-α (mTNF-α), allowing for the evaluation of the role of TRICs and FcRn in ADA formation.

The following mAbs were used: infliximab and golimumab (anti–TNF-α IgG1; MSD), adalimumab (anti–TNF-α IgG1; Abbvie), and rituximab (anti-CD20 IgG1; Roche).

Six- to eight-week old C57BL/6 wild-type (WT) mice were obtained from Janvier (Saint-Berthevin, France). The B6.129 × 1-Fcgrttm1 Dcr/DcrJ (fcgrt^−/−) mice were generated by the Jackson Laboratory (Bar Harbor, ME). A targeting vector was designed to replace 1588 nucleotide fragments (encoding promoter sequence 5′ of the transcriptional start site, exon 1, intron 2, and most of exon 2) with a PGK-Neor cassette. The vector was electroporated into 129 × 1/SvJ-derived ESV/J-1182 embryonic stem cells. Correctly targeted embryonic stem cells were injected into recipient C57BL/6J blastocysts. The resulting chimeric animals were crossed to C57BL/6J mice. The mice were then backcrossed to C57BL/6J for 11 generations. All experiments were performed in accordance with national animal care guidelines (EC directive 86/609/CEE, French decree number 87–848), and were approved by the Val-de-Loire Animal Experiments Committee of Ethics (approval number 2012-04-09). Body weight was monitored throughout the study as an indicator of health status. After 1 wk of adaptation period in the same room, WT and FcRn KO mice were anesthetized with isoflurane 2.5% and intravenously injected by a retro-orbital route, with a single dose of 4 mg/kg adalimumab or infliximab. Blood samples were collected at 2 and 6 h, and on days 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 13, 14, and 21 after adalimumab or infliximab injection. Blood samples were centrifuged at 1500 × g for 15 min, and sera were stored at −20°C until measurement of infliximab, adalimumab, or ADA levels.

ICs were generated in vitro by mixing anti–TNF-α Ab and hTNF-α (Peprotech, Neuilly-sur-Seine, France), respectively in a 1:2 molar ratio, and left for 1 h at room temperature. The final concentrations in PBS were 500 μg/ml for anti–TNF-α Ab and 116 μg/ml for hTNF-α. Next, 200 μl were intravenously injected by a retro-orbital route. Blood samples were collected before mAb injection, and on days 5, 7, 10, and 21 for infliximab and adalimumab, or on days 3, 10, and 20 for golimumab.

The serum concentrations of infliximab and adalimumab were measured using validated ELISA techniques (17, 18). Briefly, recombinant hTNF-α was coated on the solid phase and recognized by infliximab or adalimumab, and the therapeutic Ab was detected by an anti-human IgG Fcγ-specific Ab conjugated to HRP.

Briefly, 96-well flat-bottom plates were coated with 1 μg per well of adalimumab, infliximab, or golimumab in 0.1 M NaCO₃/NaHCO₃ buffer (pH 9.6) in the case of ADA ELISA, or with 200 ng per well of human or murine TNF-α (Peprotech, Neuilly-sur-Seine, France) in TNF-α ELISA. The plates were incubated for 12 h at 4°C. Washing steps were performed with PBS, 0.1% tween 20, plus 5% casein in the blocking step or in mAb dilutions. Serum dilutions (1:100 unless otherwise specified) were incubated at room temperature for 1 h. Specific binding was detected by using either peroxidase-labeled donkey anti-mouse IgG (H+L) (Jackson Laboratory) for 1 h at room temperature in ADA ELISA, peroxidase-labeled goat anti-human IgG (H+L), peroxidase-labeled anti-IgG Ab (anti-Fc; Jackson Laboratory) or with a peroxidase-labeled anti-IgM Ab (anti-Fc; Jackson Laboratory). The reaction was developed by adding 100 μl of tetramethylbenzidine (Sigma-Aldrich, Saint Louis, MO) for 20 min and stopped with 50 μl H₂SO₄ 0.5 M. Absorbance was measured at 450 nm with a Mithras LB 940 Multimode Microplate Reader (Berthold Technologies, Thoiry, France).

Mice spleens were analyzed 2 or 6 d after Ab injection. Single-cell suspensions were isolated with Spleen Dissociation Kits (Miltenyi Biotec, Paris, France). Total leukocyte cell counts were performed and 1 × 10⁶ cells were analyzed by flow cytometry. FcγR were first saturated by cell incubation with an anti-FcγRII/FcγRIII mAb (2.4G2; BD Pharmingen, San Diego, CA) for 20 min. Then cells were incubated with fluorochrome-conjugated specific Abs, in PBS containing 2% SVF and 0.1% sodium azide at 4°C for 60 min. Cells were analyzed using a FACS Gallios equipped with Kaluza software (Beckman Coulter, Villepinte, France). Anti-mouse CD3-FITC, Ly-6c PE-CF594, CD11c PerCP-Cy5.5, CD45R/B220 PE-Cy7, CD11b APC-Cy7, and corresponding control isotypes were purchased from BD Pharmingen.

Flow cytometry and ELISA data are expressed as mean and SD of independent experiments. The Mann–Whitney U test was used to determine significant differences unless otherwise specified. Statistical analysis was performed using GraphPad Prism 5. The α risk was set at 5%. The level of significance is indicated on the figures as *p ≤ 0.05, **p ≤ 0.01, and ***p ≤ 0.001.

To measure the relation between anti–TNF-α mAb concentrations and the formation of ADA, adalimumab and infliximab concentrations were monitored for 21 d in the serum of WT and KO mice after a single 4 mg/kg mAb injection (Fig. 1). As expected, serum concentrations of both mAbs were undetectable after 7 d in FcRn KO mice. An acceleration of mAb elimination, with a nonlinear decrease, was observed especially for adalimumab in WT mice, suggesting the formation of ADA.

ADA formation against adalimumab or infliximab was monitored in both WT and FcRn KO mice after a single mAb injection. The results, presented in Fig. 2, show that ADA are detected earlier in FcRn KO than in WT mice after adalimumab injection (day 4 and day 6 respectively). Even though ADAs occurred earlier in FcRn KO mice, the titers obtained in WT mice seem to be higher (Fig. 2C). The two main mechanistic explanations for this phenomenon are a booster role of FcRn in the humoral response, leading to increased Ab production in WT mice, and the protection of these IgGs from lysosomal degradation by FcRn. ADA detection with an anti-mouse-IgG–specific or an anti-mouse-IgM–specific Ab revealed that the response is mainly an IgG response with very low IgM levels. By contrast, infliximab did not induce any ADA response up to 21 d after injection in both WT and FcRn KO mice (data not shown).

Our hypothesis was that infliximab did not induce ADA due to the lack of TRIC formation with the endogenous mTNF-α, whereas adalimumab cross-reacts with mTNF-α and therefore generates TRICs following injection. First, the species specificity of anti–TNF-α mAbs was assessed. As shown in Fig. 3, adalimumab binds both human and murine TNF-α whereas infliximab only binds hTNF-α. Rituximab, an anti-CD20 mAb, was used as a negative control. Then, infliximab-containing TRICs were constituted in vitro by mixing hTNF-α and infliximab prior to injection to WT and FcRn KO mice. No ADA could be detected by ELISA in WT or FcRn KO mice injected with infliximab alone. On the contrary, high amounts of infliximab ADA were measured in WT and FcRn KO mice injected with in vitro–formed infliximab-containing TRICs (Fig. 4). No anti–TNF-α Abs were detected in mice injected with recombinant hTNF-α alone. Altogether, these data showed that artificially formed TRICs are able to elicit an immune response to infliximab. In FcRn KO mice, ADA formation has different kinetics depending on the injected mAb, with anti-adalimumab Abs appearing more rapidly than anti-infliximab Abs, whereas ADA decreased more quickly in FcRn KO mice than in WT mice for both mAbs. We injected a 10-fold higher concentration of adalimumab (40 mg/kg), and showed that immunization occurred later in these mice, suggesting that dilution of the immune complexes in a tide of uncomplexed Abs delays ADA formation.

The lack of immunogenicity of infliximab alone could be because its variable domains are murine, whereas adalimumab has human variable domains. To address this, a new set of experiments was performed with golimumab, a fully human therapeutic anti–TNF-α mAb that is specific to hTNF-α, as shown in Fig. 5A. Golimumab was injected once, either alone or after in vitro IC generation by incubation of golimumab with hTNF-α. Golimumab alone did not elicit an ADA response in WT mice, in contrast to preformed golimumab-TRICs (Fig. 5B). These results were confirmed by repeating the experiment and performing ELISA with an anti-IgG HRPO-labeled murine-Fcγ-specific Ab (data not shown).

Spleens of WT and FcRn KO animals injected with adalimumab were harvested on day 2 or day 6 to examine the cellular population. Spleens from naive animals were used as control (day 0 in Fig. 6). As shown in Fig. 6A, the size of the spleen increased 2 d after adalimumab injection, with a largely more pronounced effect in FcRn KO mice. This phenomenon was described in all mice by total splenocyte count (Fig. 6A), and a return to a normal count was observed on day 6. Flow cytometry analysis revealed that the increase in cell count on day 2 after adalimumab injection was observed for B cells (B220⁺), T lymphocytes (CD3⁺), dendritic cells (Ly6c⁺/CD11b⁺), and neutrophils (CD11c⁺/CD11b⁺) with a 1.43–1.97 ratio (day 2/day 0) compiling all FcRn KO and WT mice data (Fig. 6B). As expected, ratios were higher in FcRn KO (1.73–2.24) as compared with WT mice (1.40–1.57) for all cell types. This increase in cell count was more pronounced for the B cell ratio (2.24 in FcRn KO versus 1.45 in WT mice), and dendritic cells (1.97 in FcRn KO versus 1.40 in WT mice), followed by a milder increase in neutrophils (1.84 in FcRn KO versus 1.43 in WT mice) and T cells (1.73 in FcRn KO versus 1.57 in WT mice).

The immunogenicity of therapeutic mAbs is a complex phenomenon that can occur during treatment with them, leading to the formation of ADA-related ICs and a profound alteration in mAb pharmacokinetics in most patients (9, 19, 20). These ADA-related ICs are defined as complexes formed by the binding of ADA to mAbs, particularly to the mAb idiotypes, because paratopes remain immunogenic whatever the degree of humanization of mAbs (5). This phenomenon has been seen in patients treated with infliximab or adalimumab, and measurement of ADA is often a warning sign of upcoming treatment failure after an initial response (9, 21, 22). To prevent immunization, it is necessary to understand the mechanisms leading to the formation of ADA. Our study shows that TRICs, but not FcRn, are essential to trigger an immune response (i.e., after a single injection) against anti–TNF-α Abs. Moreover, although the crucial role played by FcRn in the development of humoral immune responses has been demonstrated by several studies (23, 24), our data fit better with a booster role for FcRn (25). Most of these studies were carried out using OVA because the T epitopes are well characterized and can be followed up, but the specific case of anti–TNF-α Abs is different. In our study, the Ags are Abs that form TRICs that elicit an immune response after the first injection (or not in the case of infliximab). This point could modify the kinetics of the immune response. In our animal model, we first confirmed that the presence of ADA modifies the pharmacokinetics of therapeutic Abs. Adalimumab displayed a faster log-linear elimination than infliximab but, above all, an accelerated terminal elimination, characteristic of the development of ADA (26, 27). This acceleration was observed in all mice treated with adalimumab, independent of the FcRn status even if elimination was clearly faster in FcRn KO mice (28). Adalimumab binds mTNF-α as described by the European Medicines Agency (2) and confirmed by ELISA. We hypothesized that with adalimumab treatment TRICs would be formed and could in turn cause immunization. To address this, we incubated infliximab and TNF-α to form TRICs, which were able to induce anti-infliximab Abs after injection in both WT and FcRn KO mice.

The Abs used differ in several ways, and one could ask whether the results are biased due to confounding factors. Adalimumab is a fully human therapeutic mAb. To discard the possibility that immunization could be a result of the presence of human portions in the protein sequence, we performed similar experiments with golimumab, another fully human anti–TNF-α mAb very specific for hTNF-α. The absence of anti-golimumab Abs supports the idea that mAb humanization is not responsible for the development of ADA in our model. However, as observed with infliximab, golimumab-containing TRICs injected into mice were able to elicit an anti-golimumab response. Apart from the degree of humanization, allotypes have been suspected to participate in the immunization process in chronic inflammatory diseases, even though the results were negative (6). All the Abs used in this study shared the same allotype (G1m17; 1), which discards this hypothesis. Finally, the intrinsic characteristics of the TRICs themselves, such as their size (25), are involved in their trafficking and condition the type of immune response elicited (29). But FcRn^−/− mice also mount an immune response against these ICs, which discards this possibility. Moreover, infliximab incubated with hTNF-α also elicits an immune response. Altogether, these data confirmed that TRICs are necessary to elicit a primary immune response in the case of anti–TNF-α Abs.

FcRn is involved in the recycling and transcytosis of IgGs and monovalent immune complexes (30). But, surprisingly, ADAs were detected in both WT and KO FcRn mice, suggesting that FcRn is not necessary for immunization against anti–TNF-α Abs. In addition, anti-adalimumab Abs appeared more rapidly in FcRn KO mice than in WT animals. Several studies performed in FcRn KO models show that mAb pharmacokinetics are clearly impaired and that IgG biodistribution is affected differently depending on the organ tested (28, 31). We hypothesized that, in the absence of FcRn, adalimumab would not be recycled and that, its biodistribution outside the vascular compartment being altered, it would be fully available to form ICs with its target, especially in places such as mesenchymal lymph nodes, where ADA concentrations remain sufficient to form TRICs due to resident plasma cells as described by Baker et al. (16). TRICs would then be able to rapidly induce an immune response, characterized by IgG and low IgM ADAs, probably via FcγR binding and MHC class II presentation (30, 32, 33). Another nonexclusive hypothesis is that the concentrations of free adalimumab are lower in FcRn^−/− mice, and that their tolerogenic effect over TRICs is therefore shorter. Although dependence of Ab-mediated presentation of Ag on FcRn has been described in the case of secondary humoral response (15, 25), its absence could probably be overcome by other molecules for IC routing and MHC class II presentation as suggested in our model. The ADA response has a cellular component as observed by the increase in dendritic or monocyte counts as well as B or T cell counts in the spleen of immunized animals. This cellular immune response is necessary to lead to the humoral response and is proportional to the amount of ADA generated. Chen et al. (28) have shown that in the liver or the spleen, FcRn may exert its transcytosis function to transport IgGs out of the tissue. In the absence of FcRn, mAbs or TRICs might be retained in the spleen, thus participating in the increased immune response. Simultaneously, the absence of FcRn leading to accelerated clearance of adalimumab could also explain the faster clearance of ADA in FcRn KO mice as compared with WT. The involvement of this mechanism in ADA kinetic differences is sustained by the kinetics observed with preformed TRICs in our model. In this case, anti-infliximab Ab formation occurs without delay. The question remains as to which FcγRs participate in this immune response. The cellular counts performed on the harvested spleens show an increase in multiple cell populations 2 d after the injection of preformed TRICs, which include dendritic cells (CD11c⁺/CD11b⁺), neutrophils, and monocytes (LY6c⁺/CD11b⁺). These cells express the activating murine FcγRI (monocytes, dendritic cells), FcγRIII, and FcγRIV (dendritic cells, neutrophils, and monocytes), which could all play a role in this process. However, because of the numerous differences between humans and mice in terms of FcγR (34), caution must be used for extrapolation to the clinical situation.

In humans, ADAs have been detected in a large number of patients treated with anti–TNF-α mAbs. Underexposure to mAbs has been clearly demonstrated in patients as a risk factor of immunization (19, 20). This approach does not take into account the evaluation of specific TRICs, which are known to skew the dosage of ADAs in the serum (35). The nature of the target may also participate in the formation of ICs. In that context, TNF-α is a good candidate because it is composed of a trimer and exists as a soluble Ag that is more prone to form multimeric ICs. The three Abs used in this study target TNF-α in different ways. Adalimumab and infliximab stabilize TNF-α as a trimer, whereas golimumab binding does not prevent monomer exchange with unbound trimers (36), which could explain the rather important variability observed in the anti-golimumab responses. This could explain the higher levels of ADA found with anti–TNF-α than with biologics directed against other molecules, along with the inflammatory profile of the pathologies involving high titers of TNF-α.

It is important to bear in mind that this study was performed using FcRn KO mice, meaning that the bioavailability of the TRICs is altered compared with WT mice, and that full FcRn deficiency has not been seen so far in humans.

It has been documented that low concentrations of anti–TNF-α Abs are associated with increased ADA formation in patients (21), and hypothesized that maintaining high concentrations of monomeric Ab could prevent the immunization process (12). In the case of anti–TNF-α therapy and primary immune response in mice, we show that this is true.

We thank the animal facility of Université François Rabelais for kind help.

The authors have no financial conflicts of interest.