An Anti-EGFR IgA That Displays Improved Pharmacokinetics and Myeloid Effector Cell Engagement In Vivo

The EGFR constitutes an established tumor target antigen, which can be involved in tumorigenesis (1), and which is overexpressed or mutated in many common solid tumor types such as colorectal, head and neck, and lung cancer (2). Both tyrosine kinase inhibitors and monoclonal antibodies have been successfully developed and approved for clinical applications (3, 4). The two approved monoclonal antibodies cetuximab (chimeric IgG1) and panitumumab (human IgG2) bind to overlapping epitopes in domain III of EGFR and are similar in Fab-mediated effector functions such as inhibition of ligand binding, tumor growth inhibition, and receptor modulation (5). However, both antibodies differ in their efficacy to recruit effector cells for antibody-dependent cell-mediated cytotoxicity (ADCC): although the IgG1 antibody cetuximab is particularly efficient with NK cells, the IgG2 antibody panitumumab effectively recruits myeloid cells (6). Interestingly, both antibodies proved to be similarly effective in the treatment of patients with chemotherapy-refractory metastatic colorectal cancer (7). Furthermore, there is increasing evidence that myeloid cells are important effector cells in cancer and cancer immunotherapy (8–10). In vitro, myeloid effector cell engagement is as effective with human IgG2 as with human IgG1 antibodies (6), but is significantly improved with antibodies of IgA isotypes (11–13). Enhanced myeloid cell recruitment by human IgA compared with IgG1 antibodies has been observed against different tumor target antigens such as CEA, CD20, HER-2/neu, and EGFR. However, the immunotherapeutic potential of IgA antibodies has not been explored in humans.

Natural IgA antibodies constitute the dominant isotype in mucosal immunology (14). By concerting multiple mechanisms of action, IgA antibodies control commensals and prevent the invasion of pathogens, or neutralize those that passed through mucosal barriers (14). Important defense mechanisms, such as phagocytosis, oxidative burst, cytokine release, and antigen presentation, are mediated by the interaction of IgA antibodies with the myeloid IgA Fc receptor, FcαRI (CD89), which is expressed on monocytes, macrophages, granulocytes, and Kupffer cells (14–16). There are important differences in the antibody-FcR interaction between IgA and IgG antibodies (16): mutagenesis and X-ray crystal structure studies located the FcαRI interaction site to the interface of the Cα2 and Cα3 domains of the Ig α-chain and revealed a bivalent binding mode, although Fcγ receptors bind to IgG at the lower hinge in a 1:1 stoichiometry (16). Contribution of N-glycans to Fc receptor binding was shown for carbohydrates of FcαRI but was not observed for those of IgA antibodies (17–19). This is in contrast to IgG1 where glycosylation of both, FcγRIII and that of the antibody, does influence interactions with FcγRIII (20). Previous studies demonstrated the efficiency of IgA2 antibodies in activating myeloid effector cells for tumor cell lysis in ADCC assays in vitro (11–13). Thus, antibodies of IgA isotype may constitute interesting and potent molecules for targeted cancer therapy.

Humans have two genes for IgA heavy chains (HC) encoding two isotypes—named IgA1 and IgA2—with three allelic forms of IgA2, described as IgA2m(1), IgA2m(2), and IgA2n (21, 22). Each of these IgA molecules has distinct structural characteristics. For example, the elongated hinge of IgA1 increases its flexibility thereby possibly allowing binding of more distant epitopes (14), but the carbohydrates attached at the up to six O-glycosylation sites are difficult to control during recombinant expression (23) and are critically involved in the pathogenesis of IgA nephropathy (24). This elongated hinge region of IgA1 is not present in human IgA2 allotypes, resulting in a more compact structure (14, 25). IgA antibodies are stabilized by a variable degree of interchain disulfide bridges between both HCs in the Cα2 domain and between HC and light chains (LC; refs. 26, 27). However, LCs are not covalently connected to HCs in IgA2m(1), which is the most common allotype in Caucasians. The HCs of both IgA subclasses carry a C-terminal extension of 18 amino acids, called tail piece, which is critically involved in the formation and transcytosis of dimeric and secretory IgA by the polymeric immunoglobulin receptor (pIgR) onto mucosal surfaces (14, 28). Notably, cysteines 471 and 311 in monomeric IgA are required for covalent linkage to the J-chain and the extracellular domain of the pIgR to form dimeric or secretory IgA, respectively (14). In contrast to IgG1, IgA antibodies carry a more exposed and heterogeneous N-glycosylation (14, 29); there are two conserved N-glycans in both IgA isotypes with two and three additional N-glycans for IgA2m(1) and IgA2m(2), respectively. Increasing evidence underlines the impact of glycosylation on pharmacokinetic properties of IgA antibodies (30, 31). Naturally, IgA antibodies have a significantly shorter serum half-life compared with IgG in men and mice, which is partially explained by the lack of binding to the neonatal Fc receptor (FcRn; ref. 32). Additionally, recombinant IgA was previously reported to be rapidly cleared by the asialoglycoprotein receptor (ASGPR), which is predominantly expressed in the liver and clears proteins with exposed terminal galactose (30, 31, 33). Importantly, the serum half-life of IgG antibodies has been demonstrated to significantly impact on their therapeutic efficacy (34). Thus, glyco-engineering strategies to increase the serum half-life of IgA antibodies may improve the immunotherapeutic potential of this antibody isotype for clinical applications.

In the present study, we describe the rational design of an engineered IgA antibody directed against EGFR. We removed two N-glycosylation sites (N166 and N337), stabilized HC and LC linkage by introduction of a P221R mutation into the Igα HC and removed two free cysteins (C311 and C472). Next, we investigated the effects of this engineering on the productivity, assembly, glycosylation, stability, and functionality. Importantly, the engineered IgA antibody demonstrated increased productivity and thermal stability. This novel IgA antibody had a significantly improved serum half-life compared with wild-type IgA2 and was efficient in engaging myeloid effector cells in vitro and in vivo—thus constituting a promising antibody format for cancer immunotherapy.

Experiments with human material were approved by the Ethical Committee of the Christian-Albrechts-University (Kiel, Germany) in accordance with the Declaration of Helsinki. All experiments in mice were approved by the Animal Ethical Committees of the UMC Utrecht or Christian-Albrechts-Universität in Kiel.

Immunodeficient SCID mice (CB17/Icr-Prkdcscid/IcrIcoCrl) were used to evaluate the serum half-life in mice (four mice per group). Two hundred micrograms of the respective antibody was injected intravenously into the tail vein. Submandibular blood samples were collected daily during treatment, and antibody concentrations were evaluated using human IgA- or IgG-specific ELISA as described earlier (30). The tumor model with A431-luc2 in SCID mice transgenic and nontransgenic for human FcαRI was performed similarly as previously described (30). Experiments with B16F10-luc2-EGFR and BaF3-EGFR cells in C57BL/6 (BL6) and Balb/c mice transgenic (tg) or nontransgenic (ntg) for human FcαRI were performed as described earlier (30).

For pharmacokinetic imaging antibodies were labeled and separated from free dye using LICOR IRDye 800CW kit (LI-COR) according to the manufacturers' protocol. A total of 25 μg of labeled antibodies were injected i.v. into tail veins of 5- to 6-week-old female mice (Athymic Nude-Foxn1nu, purchased from Harlan Laboratories GmbH). For injection and imaging mice were anaesthetized with i.p. injections of ketamine (80 mg/kg, Aveco Pharmaceutical) and dorbene (0.5 mg/kg, Pfizer). Ventral views were imaged using a Peltier cooled charged-coupled device camera (NightOWL LB 983, Berthold Technologies). The excitation source is a ring light used for epi-illumination, mounted 12 cm above the mice. Filters of 740 and 790 nm were used to assess excitation and emission signals, respectively. The exposure time was set to 0.3 second. Fluorescent signal was quantified as counts per second using Indigo software (Berthold Technologies) with an automated peak search function.

The human epidermoid carcinoma cell line A431 [German Collection of Microorganisms and Cell Cultures (DSMZ), Braunschweig, Germany] and BHK-21 cells cotransfected with FcαRI (CD89) and the FcR γ-chain (35) as well as the luciferase-transduced A431 cells (30) were cultured in RPMI 1640 containing 10% heat-inactivated FBS, 100 U/mL penicillin, and 100 μg/mL streptomycin (media and additives from Life Technologies). Baby hamster kidney (BHK) transfectants were positively selected by addition of 1 mg/mL geneticin and 20 μmol/L methotrexate (Sigma). The human esophageal squamous carcinoma cell line Kyse30 (DSMZ) was kept in 45% RPMI 1640 with 45% Ham's F12, 10% FBS, and 1% antibiotics (100 U/mL penicillin, 100 μg/mL streptomycin). The human colon carcinoma cell line DiFi and the human glioblastoma cell line A1207 (European Collection of Cell Culture, ECACC, Salisbury, UK) were maintained in DMEM supplemented with 10% FBS, 100 U/mL penicillin, and 100 μg/mL streptomycin. A431-luc2, BaF3±EGFR, and B16F10-EGFR-luc2 [BaF3 (DSMZ) and B16F10 (ATCC)] cells were kept in RPMI 1640 containing 10% heat-inactivated FBS, 100 U/mL penicillin, and 100 μg/mL streptomycin (30). All cell lines were obtained between 2011 and 2013.

Purified human myeloma IgA2m(1) antibody was used as control IgA2 (Meridian Life Science). Monomeric wild-type 225-IgA2m(1), named 225-IgA2-wt, and a wild-type human IgG1, named 225-IgG1, were produced from the variable regions of the 225 antibody and human constant regions as previously described (13). The DNA-sequence encoding the mutated IgA2m(1) was generated by Entelechon. The translated protein contains the following mutations: asparagine to glycine at position 166, proline to arginine at position 221, cysteine to serine at position 311, asparagine to threonine at position 337, isoleucine to leucine at position 338, and threonine to serine at position 339. The nomenclature of the new IgA2m(1) is therefore 225-IgA2m(1)-N166G-P221R-C331S-N337T-I338L-T339S-dC471-dY472 (further named 225-IgA2.0). The coding sequences for the HCs of wild-type human IgA1, IgA2m(1), and the mutated IgA2.0 are aligned in Fig. 1. Further cloning, production, purification, as wells as the determination of antibody concentrations, specific production rates, gel electrophoresis, Western blotting, as well as the functional characterization were done as described earlier (13, 35).

The released N-glycans were analyzed as described in detail in a MIRAGE (36) compliant manner in the supplementary material (29, 37). Relative quantitation was performed using the Quant Analysis tool (Bruker), which determines the area under the curve obtained from the individual extracted ion chromatograms from multiple analyses (Supplementary Table S1).

Thermal stability was analyzed in thermal shift assay using Sypro Orange (Life Technologies). A volume of 7.5 μL 300× Sypro Orange were diluted with 12.5 μL 1× PBS (Life Technologies) and 5 μL of 2.5 mg/mL protein and transferred on a white 96-well thin-wall PCR plate and sealed with Optical-Quality Sealing Tape (both Roche). Plates were heated in a LightCycler 480 (Roche) from 37°C to 99°C with a heat-rate of 4°C/second and 1 minute incubation at each degree. Fluorescence was recorded simultaneously using 490 and 575 nm as excitation and emission wavelengths, respectively.

Surface plasmon resonance was investigated on a BIAcore T100 system equipped with BIAcore evaluation software V1.2 (both GE Healthcare). Recombinant human soluble FcγRIIIa and FcαRI (both R&D systems; 100 μg/mL) were coated using random amine coupling kit in sodium acetate buffer with pH 4.5 onto flow cell 1 (FcγRIIIa as reference) and flow cell 2 (FcαRI) of CM5 chips (kit and chips from GE Healthcare). Different concentrations (0–200 nmol/L) of 225-IgA2-wt and 225-IgA.0 dialyzed three times against HPS-EP⁺ running buffer (GE Healthcare) were injected with a constant flow rate of 30 μL/min. Binding to soluble FcγRIIIa and FcαRI was measured at 25°C. Sensograms were recorded and kinetics were calculated using a bivalent ligand model.

Growth inhibition of DiFi cells was analyzed using the 3-(3,5-dimethylthiazol-2-yl)-5-(3-carboxymethoxyphenyl)-2-(4-sulfophenyl)-2H-tetrazolium inner salt (MTS) assay and performed as described earlier (12). For the internalization assay, A431 cells were seeded (2.5 × 10⁴ cells) on 96-well plates (Sarstedt) and treated with 5 μg/mL of 225-IgG1 or respective IgA2 antibodies. A human kappa LC binding antibody fragment fused to a truncated version of Pseudomonas exotoxin A (α-kappa-ETA'; ref. 38) was applied in a dose-dependent manner. After 72 hours, MTS substrate was added, and absorption at 490 nm was measured after 24 hours to determine inhibition of cell growth as a measure for antibody internalization. All experimental points were set up in triplicates. Viable cell mass in the presence of control antibody served as reference (100% cell growth) to calculate growth inhibition by EGFR Abs according to the formula: absorption (EGFR Ab)/absorption (control Ab) × 100. Apoptosis of DiFi cells was induced by incubating cells with 20 μg of the respective antibodies using α-kappa-ETA'. Twenty-four hours later cells were yielded by trypsinization, stained with Annexin-FITC/PI kit (BD) and analyzed by immunofluorescence analyses. Preparation and engagement of effector cells was analyzed in ⁵¹chromium release assays as described earlier (13, 35). ASGPR-mediated uptake of antibodies was investigated using BHK cells, transfected with pCMV6-AC plasmid encoding ASGPR1 (OriGene) and Lipofectamine 2000 (Life Technologies). Positive single clones were selected by FACS using ASGPR1-directed antibody (Affinity Bioreagents) and FITC-labeled goat-anti-mouse Ig Fab₂ secondary antibody (Dianova). Cells (2.0 × 10⁴ cells) were seeded in 96-well plates and incubated for 96 hours with respective antibodies at indicated concentrations with or without the supplement of ahu-kappa-ETA. Viable cell mass in the presence of 1× PBS served as reference (100% cell growth) to calculate growth inhibition by EGFR Abs according to the formula: absorption (EGFR Ab)/absorption (1× PBS) × 100.

Data were generated from at least five independent experiments. Graphical and statistical analyses were performed using GraphPad Prism 6.0 (GraphPad Software). Group data are reported as mean ± SEM. Significance was determined by two-way ANOVA repeated measures test with Bonferroni's post hoc correction. EC₅₀ values were calculated from dose–response curves, reported as means ± SEM and compared by paired Student t test to calculate significant differences between data groups. Significance was accepted when P values were ≤0.05. HCs of IgA2 antibodies were modeled on the basis of the sequence available under the Genbank accession number AY647979 and aligned using I-TASSER (39).

Several strategies were combined for engineering an EGFR-directed IgA2m(1) antibody (Fig. 1A): first, the stability of the molecule was increased by exchanging proline to arginine at position 211 of the HC (P221R) to allow the covalent linkage of LCs and HCs. Second, the formation of dimeric aggregates and complex formation with serum proteins was reduced by deleting the C-terminal cysteine 471 and by exchanging cysteine to serine at position 311 (C331S). Additionally, the number of glycosylation sites was reduced by exchanging the amino acids at positions 166 and 337–339 to amino acids at corresponding positions in IgA1. The resulting monomeric molecule is therefore an Fc-engineered IgA2m(1) molecule with four instead of eight N-glycans. The three-dimensional structures of both IgA2 antibodies were predicted using I-TASSER (Fig. 1B). Highlighting red space fill spheres indicate surface exposed residues. Consequently, the exposed nature of the residues at the amino acid positions 166 and 337–339 indicates the accessibility of the respective N-glycosylation sites in wild-type IgA2m(1) antibodies. Next, HCs of wild-type and mutated IgA2 were modeled and aligned. Introducing the P221R mutation appears to allow a slight rotation of the HC, leading to a different orientation of the side chain of cysteine 241 in wild-type and mutated IgA2.0, as described previously (Fig. 1C; ref. 13). Furthermore, the tail piece was modeled differently for wild-type and mutated IgA2 (Fig. 1D). However, as IgA2.0 contains eight different mutations we may not conclude that the predicted structural changes are the result of one specific mutation.

Both IgA antibodies were produced in CHO-K1 cells growing serum-free under suspension culture conditions. Single clones stably expressing the antibodies were generated by limiting dilution cloning, and production rates of five 225-IgA2-wt and 225-IgA2.0 producing single clones were evaluated (Fig. 2A and B). Significantly higher production rates were detected for 225-IgA2.0 producing clones. Next, antibodies were affinity-purified and subjected to size exclusion chromatography to isolate monomeric IgA2 antibodies. Elution profiles of 225-IgA2-wt preparations displayed a significant amount of polymeric aggregates, whereas the preparations containing 225-IgA2.0 eluted mainly as a single peak (Fig. 2C). In both cases fractions containing monomeric IgA2 antibodies were pooled and purity of those preparations was assessed by size exclusion chromatography (Fig. 2D). Thus, higher production rates and lower formation of spontaneous aggregates suggested improved biopharmaceutical properties of IgA2.0 compared with wild-type IgA2.

Molecular mass and purity of monomeric IgA2 antibodies were analyzed by capillary electrophoresis (Fig. 2E). Under reducing conditions 225-IgA2-wt displayed bands ranging from approximately 75 to 110 kDa, whereas two distinct bands were detected for 225-IgA2.0 at approximately 75 and 80 kDa. On native PAGE, monomeric 225-IgA2-wt displayed a single band at the molecular mass of approximately 160 to 180 kDa, whereas monomeric 225-IgA2.0 was detected at a molecular mass of approximately 150 kDa (Fig. 2F). Next, affinity purified (using anti-human kappa LC directed beads) monomeric IgA2 antibodies were separated by gel electrophoresis under denaturing nonreducing and reducing conditions, and silver staining was used for protein detection (Fig. 2G). Again, the 225-IgA2.0 was detected at a lower molecular mass under reducing (∼50 and 60 kDa for HCs) and nonreducing conditions (∼150 kDa for monomeric antibody) than 225-IgA2-wt, which dissociates under nonreducing but denaturing conditions in homodimers of LCs (∼40 kDa) and HCs (∼100 kDa) as previously described (13). Interestingly, polymeric aggregates were detected under nonreducing conditions in the affinity-purified 225-IgA2-wt but not in 225-IgA2.0. Western Blot analyses using peroxidase (POX)-labeled anti-human α-chain–specific antibody proved the existence of polymeric aggregates in the affinity-purified wild-type but not in the IgA2.0 preparations (Fig. 2H). Western blot analyses using κ-chain–specific antibody confirmed the covalent linkage of LCs and HCs in 225-IgA2.0 but not in 225-IgA2-wt (Fig. 2I). Isoelectric focusing was performed using pH 3–10 isoelectric focusing gels under native conditions to determine microheterogeneity (Fig. 2J). Monomeric 225-IgA2.0 displayed a single band at pH 6, whereas 225-IgA2-wt displayed a remarkable heterogeneity with a pH from 4.5 to 6.5. Next, thermal stability of both wild-type and mutated IgA2 was analyzed in a thermal shift assay (Fig. 2K). Temperature (37°C–99°C) and fluorimetric shift of Sypro Orange due to protein unfolding was simultaneously measured in a RT-PCR cycler. The significantly increased melting temperature of the mutated compared with the wt IgA2 antibody, which is similar to that of 225-IgG1 (Table 1), indicated different kinetics in thermally induced protein unfolding between wild-type and engineered IgA2. Maintenance of functionality after incubation at denaturing temperatures was tested in ⁵¹chromium release assays using polymorphonuclear (PMN) as effector and A431 as targets cells, respectively (Fig. 2L). Although 225-IgA2-wt was fully denatured under the employed conditions, functionality of 225-IgA2.0 was at least partially maintained. Thus, after 5 minutes at 96°C, 1.7 ± 1.4% and 58.9 ± 19.5% specific lysis was detected for 225-IgA2-wt and 225-IgA2.0, respectively. Again, IgA2.0 displayed improved biopharmaceutical properties compared with wild-type IgA2.

Functional characterization of EGFR-directed IgA2 antibodies was assessed by investigating their capability to induce Fab-mediated effector mechanisms. First, both antibodies were compared for binding to EGFR-expressing A431 cells (Fig. 3A). In these experiments, 225-IgA2-wt and 225-IgA2.0 bound with similar avidity as confirmed by EC₅₀ values (Table 1). Next, their ability to block binding of the ligand EGF was investigated by incubating A431 cells with FITC-labeled EGF and increasing concentrations of both IgA2 antibodies (Fig. 3B). 225-IgA2.0 prevented binding of EGF with similar efficacy as 225-IgA2-wt (for EC₅₀ values see Table 1). Growth of EGFR-expressing DiFi colon carcinoma cells was inhibited similarly effective by 225-IgA2.0 and cetuximab, requiring lower concentrations than 225-IgA2-wt (Fig. 3C, EC₅₀ Table 1). Internalization of EGFR upon antibody binding was investigated using α-kappa-ETA' (Fig. 3D, Table 1). The toxic fusion protein induced growth inhibition of A431 cells, similarly when 5 μg/mL 225-IgA2-wt, 225-IgA2.0, or cetuximab were initially bound to target cells. Apoptosis induction was measured by incubating DiFi cells with EGFR-specific antibodies for 24 hours. Cells were then stained using Annexin V-FITC/PI and analyzed by flow cytometry. Results revealed all three antibodies to be similarly effective (Fig. 3E and F), indicating that Fab-mediated effector functions were not affected by the antibodies' Fc region.

The ability of EGFR-directed IgA2 antibodies to bind to the IgA Fc receptor, FcαRI, was analyzed first by surface plasmon resonance measurement (Fig. 4). Sensograms were derived from injection of different concentrations of 225-IgA2-wt (Fig. 4A) or 225-IgA2.0 (Fig. 4B) over randomly coupled FcαRI. Calculating association and dissociation constants and the steady-state affinity using a bivalent ligand model revealed both antibodies to have similar kinetics and apparent affinities in binding to FcαRI (Fig. 4C). These results are in line with previous reports describing FcαRI as a medium affinity receptor for monomeric IgA (17). Binding to human FcαRI was additionally measured by indirect immunofluorescence analyses using FcαRI/FcRγ-chain-cotransfected BHK cells. In these assays, 225-IgA2.0 demonstrated similar binding as wild-type 225-IgA2-wt as confirmed by similar EC₅₀ values (Table 1). Next, the potential of both IgA2 antibodies to mediate ADCC against EGFR-positive A431 cells was investigated in ⁵¹chromium release assays using increasing antibody concentrations and whole blood as a source of effector cells (Fig. 4D). In these assays, 225-IgA2.0 and 225-IgA2-wt were similarly effective in tumor cell killing, whereas cetuximab was not effective (Table 1). To address the potential of myeloid growth factors to enhance IgA2-mediated ADCC, whole blood of G-CSF-primed donors was used in ADCC assays (Fig. 4E). Both EGFR-directed IgA2 antibodies mediated significant lysis of A431 tumor cells and were similarly effective. The contribution of effector cell recruitment in whole blood ADCC assays was investigated by supplementing the FcαRI-specific murine antibody A77. ADCC mediated by both EGFR-specific IgA2 antibodies was significantly inhibited using increasing concentrations of A77, whereas IgG1-mediated lysis was not affected (Fig. 4F). Subsequently, we isolated FcαRI-expressing myeloid effector cells from the blood of healthy donors. In ADCC assays using A431 target cells and PMN (Table 1) and GM-CSF-primed PMN (Fig. 4G), monomeric 225-IgA2-wt and 225-IgA2.0 triggered similarly effective ADCC. In ADCC assays employing the colon carcinoma cell line DiFi, the glioblastoma cell line A1207, or the esophageal squamous carcinoma cell line Kyse30 both EGFR-specific IgA2 antibodies were similarly effective. In contrast, cetuximab did not engage granulocytes for ADCC against any of these target cells (Table 1). Both IgA2 antibodies were even more effective using GM-CSF primed compared with nonprimed PMN, as confirmed by EC₅₀ values (Table 1). Next, monocytes, which were isolated using CD14-specific magnetic beads (Fig. 4H), and human monocyte–derived macrophages (Fig. 4I) were analyzed as effector cells in overnight ADCC assays. In these experiments, both 225-IgA2 antibodies were similarly effective in mediating ADCC against A431 target cells by activating monocytes or macrophages for tumor cell killing (Table 1).

The efficacy of EGFR-directed antibodies was first evaluated in a syngeneic short-term i.p. model employing equal amounts of wild-type or engineered IgA2. One group of mice was pretreated with Ly6G-specific antibody to deplete granulocytes and monocytes. Briefly, BaF3 cells negative or positive for human EGFR (Supplementary Fig. S1) were labeled with 0.125 and 1 μmol/L carboxyfluorescein diacetate succinimidyl ester (CFSE; ref. 30), respectively, and were then injected i.p. into Balb/c mice ntg or tg for human FcαRI (Fig. 5A). After 16 hours, cells were recovered by peritoneal lavage and the relative amount of BaF3 cells was counted by calibrated flow cytometry, and the ratio of EGFR-negative to EGFR-positive cells was determined. Both IgA2 antibodies significantly decreased the amount of EGFR-expressing BaF3 cells in ntg mice with a significant increment in efficacy if human FcαRI was present in tg mice (Fig. 5B). Pretreatment with Ly6G-specific antibody 2 days before injection of tumor cells had no effect on the efficacy of 225-IgA2.0, although the treatment did efficiently deplete all neutrophils and most monocytes (data not shown). Importantly, this was in line with previous data demonstrating that peritoneal macrophages but neither granulocytes nor monocytes are responsible for the efficacy of IgA antibodies in this model (30). Next, a long-term peritoneal tumor model in SCID mice, tg or ntg for human FcαRI, was employed to further evaluate the in vivo efficacy of the novel IgA antibody (Fig. 5C). A431-luc2 tumor cells were administered at day 1 and allowed to grow until day 7 when EGFR-directed antibodies were applied (1 × 50 μg of 225-IgG1) or daily from day 7 to day 16 (10 × 50 μg of IgA antibodies) to compensate for the shorter serum half-life of IgA antibodies compared with IgG1. Growth of tumor cells was detected using BLI at indicated time points. Previous studies demonstrated that 225-IgG1 was similarly effective in FcαRI-tg and ntg mice. In ntg mice, 225-IgA2.0 but not the 225-IgA2-wt antibody blocked tumor outgrowth significantly, but the 225-IgG1 antibody was more effective than both IgA2 antibodies (Fig. 5D, left). However, in human FcαRI tg mice the efficacy of the engineered 225-IgA2.0 was significantly enhanced (Fig. 5D, right), indicating that FcαRI-positive myeloid effector cells contribute to the therapeutic efficacy of 225-IgA2.0 in transgenic animals. However, 225-IgA2-wt was not effective neither in tg nor in ntg mice (Fig. 5D). In Fig. 5E, representative BLI images are shown. Furthermore, ntg as well as tg mice treated with 225-IgA2.0 survived significantly longer than those treated with 225-IgA2-wt (Fig. 5F). In order to further investigate the immunotherapeutic activity of IgA2.0 in immune competent mice, we evaluated if the engineered IgA antibody was able to prevent tumor engraftment in a syngeneic long-term tumor model in BL6 mice (Fig. 5G). For this purpose, B16F10 cells, transfected to express luciferase and human EGFR (Supplementary Fig. S1), were injected i.v. into the tail veins of FcαRI-tg BL6 mice. Mice were then treated with PBS or 225-IgA2.0 (50 μg/day) from days 0 to 9. Growth of tumor cells was imaged using BLI at indicated time points (Fig. 5H). Engraftment of tumor cells was significantly delayed in FcαRI-tg mice treated with the engineered 225-IgA2.0, whereas in PBS-treated control mice tumors grew out more rapidly (Fig. 5H and I). Mice were sacrificed at day 26 to determine lung scores by visually scoring the number and size of lung metastases. In line with the BLI data, mice treated with 225-IgA2.0 displayed a significantly decreased number and size of lung metastases (Fig. 5J). Two representative lungs are displayed in Fig. 5K. Together, IgA2.0 demonstrated therapeutic activity in three in vivo models against tumor cells expressing different levels of EGFR. Importantly, the engineered IgA antibody was more effective than wt IgA in the long-term i.v. but not in the short-term i.p. model, suggesting that phamacokinetic differences may contribute to its enhanced activity.

Previous studies indicated that the serum half-life of recombinant IgA antibodies is affected by their glycosylation and by rapid ASGPR-dependent hepatic clearance. Thus, we investigated if the reduced number of N-glycosylation sites of the 225-IgA2.0 lead to an altered global glycosylation profile. For this purpose, wild-type and engineered IgA2 were separated on SDS-PAGE, transferred onto polyvinylidene difluoride (PVDF) membranes, and stained with direct blue 71 (Supplementary Fig. S2). The respective IgA-bands were cut out, N-glycans were enzymatically released using PNGase F, reduced, and subsequently analyzed in-depth by PGC nano-Liquid Chromatography-ESI MS/MS glycoprofiling (Fig. 6A and B). The systematic removal of two N-glycosylation sites in 225-IgA2.0 resulted in major glycosylation differences in terms of sialylation (and subsequently terminal galactoses) and oligomannosidic structures (Fig. 6A). In total, 86 different N-glycan structures (present in 50 different compositions) were identified (Supplementary Tables S2 and S3). 225-IgA2.0 displayed an increased proportion of terminally sialylated N-glycans (91%) compared with wild-type IgA2 (46%), simultaneously reducing the level of terminally galactosylated N-glycans significantly (Fig. 6B). Overall, the N-glycans of 225-IgA2.0 were almost exclusively of the complex type (>98%), whereas wild-type IgA2 contained just 73% complex type and 27% oligomannose/hybrid type N-glycans (Fig. 6B). The drastically different global glycosylation of the 225-IgA2.0 antibody also suggested an altered binding to the ASGPR. Therefore, ASGPR1-expressing BHK cells were used in an MTS assay to investigate the ASGPR-dependent internalization of EGFR-specific antibodies (Fig. 6C). Cells were incubated with dilutions of respective antibodies in the presence of α-kappa-ETA'. Significant growth inhibition was mediated by both wild-type and control IgA2 antibodies in a dose-dependent manner, whereas the 225-IgA2.0 and -IgG1 did not mediate target cell killing under these conditions. Subsequently, the serum half-life of IgA antibodies was investigated by injecting 200 μg of 225-IgA2-wt, 225-IgA2.0, and 225-IgG1 i.v. into tail veins of SCID mice. Blood was collected at different time points, and serum concentrations were measured by ELISA (Fig. 6D). These pharmacokinetic studies revealed a significantly increased serum half-life of 225-IgA2.0 compared with 225-IgA2-wt. Actually, the rapid elimination within the first 24 hours as well as the second elimination phase during the following days was significantly decelerated for 225-IgA2.0 compared with 225-IgA2-wt. Although the serum half-life of 225-IgA2.0 was significantly improved, there was still a difference to 225-IgG1, of which prolonged serum levels were likely maintained by FcRn-mediated recycling. Next, we analyzed the serum levels of EGFR-directed antibodies during treatment in A431-luc2 tumor-bearing SCID mice of Fig. 5. For this purpose, submandibular blood samples were collected daily during treatment, and serum levels of therapeutic antibodies were evaluated from day 8 to day 16 using human Ig-specific Sandwich ELISA (Fig. 6E). Results demonstrated a sustained higher concentration of 225-IgA2.0 during treatment compared with 225-IgA2-wt in transgenic and in nontransgenic mice (Fig. 6E, Table 1). Nevertheless, the serum levels of the respective 225-IgG1 were significantly higher than those of both 225-IgA2-wt and 225-IgA2.0 (Fig. 6E, Table 1). However, there was a significant decline in antibody concentrations detectable for 225-IgA2-wt and 225-IgG1, whereas the engineered 225-IgA2.0 demonstrated sustained high serum levels (Fig. 6E, Table 1). Next, fluorescently labeled IgA antibodies were injected into nude mice. Planar optical whole body and single organ fluorescence imaging indicated enhanced hepatic fluorescence for wild-type IgA2 compared with IgA2.0-treated mice (Fig. 6F and G; ref. 40). Thus, the significantly lower levels of terminal galactoyslation in IgA2.0 compared with wt IgA reduced hepatic up-take via the ASGPR, leading to the extended serum half-life of the engineered IgA antibody.

Human IgG1 is by far the most commonly selected antibody isotype in tumor immunotherapy (41). This decision is based on the well documented activity of human IgG1 to mediate complement-dependent cytotoxicity and to activate NK cells and monocytes/macrophages for ADCC (42) as well as its prolonged FcRn-mediated serum half-life (32). Furthermore, production and purification technologies are well established for human IgG1 antibodies (43). However, in vitro studies documented that myeloid effector cells—and PMN in particular—were more effective in ADCC by IgA than by IgG1 antibodies (11–13). Human IgA antibodies demonstrated significant antitumor activity in huFcαRI-transgenic mice, which are required to investigate Fc-mediated effector functions of this antibody isotype in vivo, because mice do not express a functional FcαRI orthologue (30, 44, 45). However, the serum half-life of recombinant human IgA antibodies in mice was unexpectedly short—probably due to rapid clearance by the ASGPR (30, 31) and the lack of binding to FcRn. This hepatically expressed receptor is known to mediate internalization and degradation of proteins with terminally exposed galactose (33), which was present on the majority of our previous IgA preparations (30). These results suggested engineering of an IgA antibody with the aim to improve its pharmacokinetic properties and to evaluate its therapeutic activity.

When we started to engineer IgA antibodies for tumor immunotherapy, we decided to improve an IgA2m(1) antibody, which is the most common Caucasian IgA2 allotype (22). The decision to use IgA2 rather than IgA1 is based on a number of reasons: IgA2 lacks the elongated hinge of IgA1, which is associated with potential disadvantages. For example, the IgA1 hinge region is targeted by bacterial proteases and contains difficult to control O-glycans, which are involved in the pathogenesis of IgA nephropathy (14). Furthermore, IgA2 proved significantly more effective in ADCC than IgA1 and has a prolonged serum half-life compared with IgA1 (11, 30). First, we introduced a P221R mutation into the Igα HC, which enabled a covalent linkage between LCs and HCs (13). Additionally, this mutation increased the thermal and long-term stability of the antibody. Next, the removal of the free and accessible thiol group in the tail piece further enhanced the stability, as it prevented the formation of dimeric aggregates and potential complexes with other serum proteins (35). The additional mutations described in this manuscript increased productivity and thermal stability as relevant pharmaceutical properties. Except for an yet unexplained difference in growth inhibition of DiFi cells (Fig. 3C), Fab- or Fc-mediated effector functions in vitro were not different between wild-type and engineered IgA2 (Figs. 3 and 4). Importantly, however, both molecules differed significantly in their glycosylation patterns.

The glycosylation of IgA antibodies is closely related to their pharmacokinetic properties (30), suggesting that glyco-engineering strategies may improve their therapeutic efficacy. By comparing sequences and structures of different IgA iso- and allotypes, deletion of the N-glycans at positions 166 and 337 appeared as a rational approach to reduce the overall level of glycosylation, because these sites are not conserved in IgA1. Interestingly, these mutations resulted in an altered global glycosylation profile. It will require further in-depth studies to elucidate whether this effect is caused by minor protein structure alterations making the sites of glycosylation better accessible to the glycan-modifying enzymes of the endoplasmatic reticulum and Golgi network, or whether the oligomannose and less sialylated N-glycans are specific for glycosylation sites Asn166 and Asn337 in the wild-type IgA2 protein expressed in CHO cells. Whether these alterations in glycosylation also affected antibody productivity or secretion would require additional studies. Potentially, the reduced number of N-glycans led to an accelerated posttranslational processing and thereby enhanced production. However, we cannot exclude that transfection or clone effects (higher gene copy number, gene amplification, and genomic loci of insertion) may have altered the production rates of the antibodies. Importantly, our data indicate that the increased terminal sialylation (which masks ASGPR recognized glyco-epitopes) of the engineered 225-IgA2.0 considerably delayed the clearance of IgA antibodies in mice—thereby increasing the otherwise short half-life of recombinant IgA compared with IgG antibodies (30). Hence, this report further confirms the important role of glycosylation and hepatic clearance by the ASGPR for the pharmacokinetics of IgA antibodies (30, 31). Although the engineering strategy described above resulted in altered glycosylation and enhanced pharmacokinetic properties of the IgA2.0 antibody, additional improvements appear reasonable. For example, C-terminal fusion of human albumin or FcRn-binding motifs to IgA antibodies, or the construction of IgG/A hybrid antibodies (46). Also, further glyco-engineering strategies or the combination of these novel approaches may further enhance the serum half-life and therapeutic efficacy of IgA antibodies (47).

Myeloid effector cells are considered as important effector cells in cancer and in cancer immunotherapy. Depending on the microenvironment, tumor-associated myeloid cells switch from tumor-promoting to tumor-preventing cells (8–10, 48, 49). Thus, myeloid cells constitute a numerous population of powerful effector cells, which are present at many tumor sites and which could possibly be recruited by tumor directed antibodies to kill malignant cells (13, 30, 35). Several studies have previously reported that FcαRI (CD89) is a potent trigger molecule to activate these effector cells in vitro and also in vivo using FcαRI-transgenic mice expressing FcαRI on myeloid cells (11–13, 30, 35). Furthermore, especially IgA2 antibodies were strong activators to induce myeloid effector cell-mediated tumor cell lysis, and co- or pre-stimulation with GM-CSF and G-CSF, respectively, could further enhance their efficacy (11). In this manuscript we describe the development of an engineered IgA2 antibody, which was as effective as wild-type IgA2 in engaging myeloid effector cells for ADCC in vitro against cell lines derived from different tumor entities. In a long-term xenogeneic model, the engineered 225-IgA2.0 was effective employing FcαRI-transgenic or nontransgenic mice, indicating recruitment of both Fc- and Fab-mediated effector functions (Fig. 5D–F). However, there was a clear increment in its efficacy if myeloid effector cell engagement via FcαRI-interaction was enabled in human FcαRI transgenic mice. In addition, the application of the engineered 225-IgA2.0 prevented engraftment and reduced the number of syngeneic tumor cells in FcαRI transgenic immune competent C57BL/6 and Balb/c mice, respectively. Thus, the engineered IgA antibody demonstrated significant in vivo efficacy against different target cells. Furthermore, the efficacy of IgA antibodies in wild-type mice lacking FcαRI in the short-term model supports their efficacy in Fab-mediated killing against low EGFR-expressing tumor cells.

Because there was no difference in Fab- or Fc-mediated effector mechanisms between wild-type and engineered IgA2 in vitro, the increased efficacy of the mutated IgA2 compared with wild-type IgA2 in the long-term tumor model in vivo could be explained by the sustained higher concentrations during treatment in transgenic as well as nontransgenic mice, whereas wild-type IgA2 was rapidly cleared. According to a previous report, where in vivo mechanisms of EGFR-directed antibodies were shown to depend on local antibody concentrations (50), we may argue that the higher serum levels of 225-IgA2.0 compared with wild-type IgA2 may constitute more appropriate conditions for the engineered antibody to induce growth inhibition or signaling abrogation in nontransgenic mice and ADCC in FcαRI transgenic mice, respectively. Thus, the prolonged serum half-life of IgA2.0 led to a therapeutic benefit in a long-term, but not in short-term treatment models, in which pharmacokinetic properties apparently did not contribute to therapeutic efficacy. Nevertheless, serum levels of IgG were multiple times higher than those of the respective engineered IgA2 antibody although higher amounts of the latter were applied, although both demonstrated similar in vivo activity under these conditions. Further studies are required to evaluate the relation between the serum half-life of recombinant IgA antibodies and local antibody concentrations, which are sufficient for preventing tumor growth in vivo. Noteworthy, antibody isotype comparisons in mice are difficult to translate into humans because, e.g., FcRn binding, Fc receptor biology, and effector cell functions are all very different between men and mice.

In conclusion, an engineered IgA antibody with improved pharmacokinetic properties demonstrated enhanced efficacy in vivo by employing both Fc- and Fab-mediated effector functions. Engagement of myeloid effector cells via FcαRI constitutes a promising approach for immunotherapy against EGFR-expressing tumors. Furthermore, our results demonstrate that the presented molecule is a novel and innovative antibody format that overcomes some of the limitations (production, stability, glycosylation, serum half-life, and in vivo efficacy) of previous IgA antibodies. Thus, these results further support the concept of introducing IgA antibodies into clinical development.

P. Sondermann is a CSO at SuppreMol GmbH. No potential conflicts of interest were disclosed by the other authors.

Conception and design: S. Lohse, L.A.P.M. Meulenbroek, S. Derer, T. Valerius

Development of methodology: S. Lohse, U. Möginger, D. Kolarich, T. Valerius

Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): S. Lohse, S. Meyer, L.A.P.M. Meulenbroek, J.H.M. Jansen, A. Kretschmer, U. Möginger, S. Tiwari, D. Kolarich

Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): S. Lohse, U. Möginger, D. Schewe, D. Kolarich, J.H.W. Leusen, T. Valerius

Writing, review, and/or revision of the manuscript: S. Lohse, J.H.M. Jansen, K. Klausz, U. Möginger, S. Derer, T. Rösner, C. Kellner, D. Schewe, P. Sondermann, D. Kolarich, M. Peipp, J.H.W. Leusen, T. Valerius

Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): S. Lohse, M. Nederend, T. Valerius

Study supervision: S. Lohse, S. Derer, T. Valerius

Other (support in engineering the IgA molecule): P. Sondermann

The authors thank the excellent technical assistance from Kathinka Tüxen, Christyn Wildgrube, Yasmin Brodtmann, the group of Prof. Dr. Axel Scheidig (Section of Structural Biology, Centre of Molecular Biosciences, Christian-Albrechts-University, Kiel), and the group of Prof. Dr. Susanne Sebens (Department of Internal Medicine I, Inflammatory Carcinogenesis, Christian-Albrechts-University, Kiel) for their help with BIAcore and Real Time PCR instruments, respectively.

This work was supported by the German Research Organization (Lo 1853/1-1, Va 124/7-2), the Wilhelm Sander-Foundation (2009.098.1 and 2), the Max Planck Society, European Union Seventh Framework Program (grant number PCIG09-GA-2011-293847), and by intramural funding from the Christian-Albrechts-University.

The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.