Mutation of the TGN1412 anti‐CD28 monoclonal antibody lower hinge confers specific FcγRIIb binding and retention of super‐agonist activity

Abstract The agonistic action of several immunomodulatory monoclonal antibodies (mAbs) requires both target antigen binding and clustering of this mAb:target complex by the Fcs interacting with Fcγ receptors (FcγRs), in particular FcγRIIb, on neighboring bystander cells. Fc mutations were made in the immunoglobulin G4 (IgG4)‐based TGN1412 anti‐CD28 mAb to define the role of FcγR interactions in its “super‐agonist” activity. The dual mutation, IgG4‐ED269,270AA, ablated interaction with all human FcγRs and agonistic action was consequentially lost, confirming the FcγR dependence on the action of TGN1412. The IgG4 lower hinge region (F234L235G236G237) was modified by L235E mutation (F234E235G236G237), a mutation commonly used to ablate FcγR binding, including in approved therapeutic mAbs. However, rather than ablating all FcγR binding, IgG4‐L235E conferred specific binding to FcγRIIb, the inhibitory Fc receptor. Furthermore, in combination with the core hinge‐stabilizing mutation (IgG4‐S228P, L235E), this mutation increased affinity for FcγRIIb compared with wild‐type IgG4. In addition to having FcγRIIb specificity, these engineered TGN1412 antibodies retained their super‐agonistic ability, demonstrating that CD28‐ and FcγRIIb‐specific binding are together sufficient for agonistic function. The FcγRIIb‐specific nature of IgG4‐L235E has utility for mAb‐mediated immune agonism therapies that are dependent on FcγRIIb interaction and of anti‐inflammatory mAbs in allergy and autoimmunity that harness FcγRIIb inhibitory signaling.


INTRODUCTION
Monoclonal antibodies (mAbs) are potent and effective biotherapeutics used in a broad range of diseases. 1 Most therapeutic mAbs are based on human immunoglobulin G1 (IgG1) and can interact with the immune system's effector molecules. However, for some therapeutic approaches, engaging effector molecules, such as Fcc receptors (FccRs), is redundant or even detrimental. A common strategy to minimize FccR interaction and associated effector functions is the use of IgG2 or IgG4 backbones which have naturally restricted FccR binding specificity. 2 This strategy has been successful for the development of many therapeutic mAbs, for example, the IgG4-based anti-interleukin-13 mAb tralokinumab 3 or the IgG2-based anti-interleukin-17Ra mAb brodalumab. 4 However, IgG4 binds to FccRIIb and is also a highaffinity ligand for the activating FccRI, and so is not an optimal backbone if a truly nonfunctional Fc is desired. Indeed, in its first human trial, the agonistic IgG4-formatted anti-CD28 TGN1412 mAb 5 induced a life-threatening cytokine storm and multi-organ failure involving FccRIIb acting as a scaffold for TGN1412 and crosslinking of CD28 on the T cell surface. 6 A more direct approach to ablate FccR interaction is mutation of the Fc domain creating "FccR-null" mAbs. 2,[7][8][9] One widely used modification for this purpose is the L 235 E mutation of the lower hinge, which was first used in OKT3, a mouse anti-human-CD3 mAb associated with severe cytokine storm. 10 This mutation eliminated T cell activation presumably by ablation of FccR binding. 10 The L 235 E mutation has since been incorporated into clinically approved mAbs, namely, sutimlimab, an anti-C1s mAb formatted on a hinge-stabilized IgG4 backbone; durvalumab 11 and anifrolumab, 12 which are anti-programmed death-ligand 1 and anti-interferon alpha receptor mAbs, respectively; and recently in the severe acute respiratory syndrome coronavirus 2 neutralizing mAb cocktail of tixagevimab/cilgavimab formatted on an IgG1 backbone also containing IgG4-like modifications of the lower hinge (L 234 F) and F/G loop (P 331 S). 13,14 In an initial evaluation of the mechanism of the TGN1412-induced cytokine storm, the TGN1412 IgG4-L 235 E FccR-null control was as potent as TGN1412. 15 However, inconsistencies are apparent in subsequent evaluations of the mechanistic basis of the TGN1412 cytokine storm which suggested that FccRIIb interaction was necessary for the severe response. 6 In this study, we investigated the effect of the L 235 E mutation on the super-agonist activity and FccR interactions of the IgG4 anti-CD28 mAb TGN1412. Notably, we found that both IgG4-L 235 E and IgG4-S 228 P, L 235 E were not Fc binding-inactive but exhibited specific binding to FccRIIb and thereby retained super-agonist function.

RESULTS AND DISCUSSION
The anti-CD28 specificity of super-agonist antibody TGN1412 was formatted as wild-type (WT) IgG4 or three IgG4 mutants as follows: the presumed "FccR-inactive" lower hinge mutant L 235 E (IgG4-L 235 E); this mutation together with the stabilizing core hinge mutation, S 228 P, (IgG4-L 235 E, S 228 P) and a novel FccR-null mutant IgG4-ED 269,270 AA. The WT and Fc-mutated mAbs bound equivalently to CD28 on peripheral T cells at saturating concentrations ( Figure 1a).
The binding to the individual cell surface-expressed human FccRs revealed surprising differences ( Figure 1b). As expected, the monomeric WT IgG4 bound to the high-affinity FccRI (Figure 1b Surprisingly, the lower hinge L 235 E mutation did not universally inactivate FccR binding. Notably, FccRIIb interaction was retained at levels equivalent to WT IgG4 and was further increased when combined with hinge stabilization (Figure 1b-iv). As expected, IgG4-L 235 E binding to FccRI was ablated, while FccRIIa and FccRIIIa binding was negligible (Figure 1b-i). Thus, the L 235 E mutation confers FccRIIb specificity.
The CD28 agonistic activity of the mAbs was evaluated in vitro. 16 The WT IgG4-and FccRIIb-selective mutants (IgG4-L 235 E and IgG4-S 228 P, L 235 E) induced tumor necrosis factor, interferon gamma and interleukin-10 release, whereas the novel FccR-null mutant, IgG4-ED 269,270 AA, did not ( Figure 2). This lack of FccR binding and lack of agonism by the IgG4-ED 269,270 AA mutant unequivocally demonstrate the Fc receptor dependence on its agonistic action. This also contrasts with the retention of agonistic potency by the FccRIIbspecific L 235 E mutants, which is consistent with the proposed critical role of FccRIIb in the TGN1412 cytokine storm. 6 Furthermore, despite similar or increased FccRIIb-binding avidity of IgG4-L 235 E and L 235 E, S 228 P mutants, respectively, over IgG4-WT, they induced less tumor necrosis factor and interferon gamma compared with WT IgG4 TGN1412, indicating that FccRIIb-binding avidity is not the only determinant for maximum agonistic activity. Interestingly, this small reduction in agonism of the IgG4-L 235 E mutant compared with WT IgG4 TGN1412 contrasts with a previous study that showed near equivalent levels of agonism of WT IgG4 and its IgG4 L 235 E mutant; however, FccRIIb was not investigated in that study. 15 The reason for the differences between the two studies is not clear. Altered intrinsic affinity for FccRIIb is ruled out as TGN1412 and its L 235 E mutant antibodies have equivalent binding avidity to cell surface-expressed FccRIIb (Figure 1b). More likely are cellular factors that are known to influence TGN1412 agonism, which include the nature of the culture systems such as co-incubation of peripheral blood mononuclear cell on human umbilical vein endothelial cell monolayers, 15 which were not used in our study. The levels of cell surface FccRIIb on the cells under different culture conditions could be another major factor. 6 Nonetheless, it is clear that L 235 E mutation of the IgG4 anti-CD28 mAb TGN1412 confers specific binding to the human inhibitory Fc receptor, FccRIIb, and that its super agonistic action is retained.
We observe FccRIIb specificity and retention of FccRIIb-dependent agonism with FEGG mutation on IgG4 backbone TGN1412. In comparison, the triple mutation L 234 F, L 235 E, P 331 S in IgG1 near-completely abrogated FccRI, FccRIIa, FccRIIIa and C1q binding, 17 while the same IgG1 triple mutation of an anti-tumor necrosis factor mAb did not bind FccRI, but retained, though reduced, binding to FccRIIa, FccRIIb and FccRIIIa in a high-avidity formatted binding assay. 18 Thus, it should be noted that normal human IgG4 already has intrinsically restricted FcR binding specificity interacting only with FccRIIb and FccRI. By contrast, human IgG1 is the universal ligand binding to all human FccR. 2 It is clear from mutagenesis and structural studies that the interaction between IgG subclasses and the different FccR is topologically similar; however, the binding of IgG subclasses is nuanced by differences in individual interacting residues in the receptors and IgG ligands. Thus, the effects attributable to the FEGG mutation may be influenced by the choice of IgG backbone and the attributes of FEGG on an IgG4 backbone may not necessarily be recapitulated in IgG1.
Although the L 235 E mutation is historically viewed as an FccR binding-ablating mutation, it clearly retains specific FccRIIb binding. This could play a role in the efficacy of some approved therapeutic mAbs using this mutation [11][12][13] by utilizing the physiologic action(s) of FccRIIb such as its inhibitory, scaffolding or immune complex removal functions. 19 Thus, for example, the anti-severe acute respiratory syndrome coronavirus-2 mAbs tixagevimab/ cilgavimab may harness the potent clearance of mAb-virus immune complexes via the liver sinusoidal endothelium 19 or interact with FccRIIb of airway smooth muscle. 20 (ii-vi), the binding of complexed CD28 mAbs to low-affinity receptors FccRIIa, FccRIIb and FccRIIIa was performed by preincubating antibodies (20 lg mL À1 ) with Alexa Fluor 647-conjugated F(ab 0 ) 2 fragments of anti-human IgG F(ab 0 ) 2 (10 lg mL À1 ), then titrating these complexes and incubating with the FcR cells. Analysis was performed with FlowJo. Data are the mean AE standard error of the mean for 5-15 experiments analyzed by two-way ANOVA with Dunnett's multiple comparisons test, comparing the main column effect with unmodified IgG4; ns, not significant, *P ≤ 0.05, **P ≤ 0.01, ***P ≤ 0.001, ****P ≤ 0.0001. APC, allophycocyanin; FccR, Fcc receptor; FITC, fluorescein isothiocyanate; Ig, immunoglobulin; mAb, monoclonal antibody; MFI, mean fluorescent intensity; PBMC, peripheral blood mononuclear cell; WT, wild-type.
Other Fc modifications reported to improve binding to FccRIIb nonetheless confer very high affinity interactions with some activating-type FccRs. 21 By contrast, FccRIIb specificity but low affinity of the IgG4-L 235 E modification makes this format well-suited to applications where agonistic activity is dependent on avidity conferred by IgG-Fc complexing, such as scaffolding specifically by FccRIIb. 22 Furthermore, there are other potential applications of this IgG4-L 235 E format in therapeutic mAbs for treating autoimmunity and allergy. Such mAbs could selectively exploit the physiological antiinflammatory action of FccRIIb, such as the suppression of immunoreceptor tyrosine-based activation motifdependent activation pathways of the B-cell receptor or the activating FccRs, particularly the high-affinity IgE receptor, FceRI. 19,23

Generation of recombinant mAbs
The TGN1412 VH, VL and kappa CL region sequences 24 were codon optimized and synthesized together with the IgG4 heavy (H) chain constant domains using gene synthesis services [Thermo Fisher Scientific, Waltham, MA (GeneArt) or Bioneer Corporation, Daejeon, South Korea]. Additionally, several mutants were generated by introducing the L 235 E (IgG4-L 235 E) mutation, the L 235 E and S 228 P (IgG4-L 235 E,S 228 P) mutations or the E 269 A and D 270 A (ED 269,270 AA) mutations in the Fc. The H or L chain sequences were subcloned into pCR3 or pcDNA3.4 (Thermo Fisher Scientific) and produced by transient transfection in Expi293 cells and purified by protein A and size exclusion chromatography as described. 25 FccR-binding assays FccR binding was analyzed by flow cytometry. 26,27 In brief, human FccRI, FccRIIa (H 131 or R 131 allotype), FccRIIb and FccRIIIa (F 158 or V 158 allotype) were expressed on FcR-deficient IIA1.6 cells and binding to the high-affinity FccRI or to the low-affinity FccRIIa, FccRIIb and FccRIIIa was determined using monomeric or complexed anti-CD28 IgG, respectively, and detected using Alexa Fluor 647-conjugated F(ab 0 ) 2 fragments of anti-human IgG F(ab 0 ) 2 (Jackson ImmunoResearch Inc, West Grove, PA). 26,27 Cytokine release assays Peripheral blood mononuclear cells were isolated from healthy human donors or buffy coats obtained from the Australian Red Cross approved by the Alfred Human Research Ethics Committee. All participants gave written informed consent. Peripheral blood mononuclear cells were obtained using Ficoll density gradient centrifugation (Sigma-Aldrich, St Louis, MI, USA). Any remaining red blood cells were lysed with 155 mM NH 4 Cl, 10 mM KHCO 3 and 0.1 mM EDTA 2 Na. Prior to CD28-mAb stimulation, peripheral blood mononuclear cells were cultured at high density (1 9 10 7 cells mL À1 ) in 24-well plates 16 for 48 h in Roswell Park Memorial Institute-1640 supplemented with L-glutamine, nonessential amino acids, N-2-hydroxyethylpiperazine-N-2-ethane sulfonic acid (HEPES), b-mercaptoethanol, sodium pyruvate, penicillin/streptomycin and 10% heat-inactivated AB-positive human serum. Cells (1 9 10 5 /well) were then transferred to roundbottomed 96-well plates with monomeric anti-CD28 antibodies (0.1-10 lg mL À1 ). After 48 h, supernatants were harvested and cytokine levels were measured by ELISA (ELISAkit.com; Jomar Life Research, Scoresby, Australia).
Dunnett's multiple comparisons test; P ≤ 0.05 was considered significant.

ACKNOWLEDGMENTS
This work was supported by the NHMRC Project grant GNT1147303, Margaret Walkom Charitable Trust and Nancy Pendergast Trust and a Sponsored Research Agreement between Burnet Institute and Genmab A/S. The authors gratefully acknowledge the support of the AMREP flow cytometry core facility. Open access publishing facilitated by Monash University, as part of the Wiley -Monash University agreement via the Council of Australian University Librarians.