Specific location of galactosylation in an afucosylated antiviral monoclonal antibody affects its FcγRIIIA binding affinity

Abstract

Monoclonal antibodies (mAbs) comprise an essential type of biologic therapeutics and are used to treat diseases because of their anti-cancer and anti-inflammatory properties, and their ability to protect against respiratory infections. Its production involves post-translational glycosylation, a biosynthetic process that conjugates glycans to proteins, which plays crucial roles in mAb bioactivities including effector functions and pharmacokinetics. These glycans are heterogeneous and have diverse chemical structures whose composition is sensitive to manufacturing conditions, rendering the understanding of how specific glycan structures affect mAb bioactivity challenging. There is a need to delineate the effects of specific glycans on mAb bioactivity to determine whether changes in certain glycosylation profiles (that can occur during manufacturing) will significantly affect product quality. Using enzymatic transglycosylation with chemically-defined N-glycans, we show that galactosylation at a specific location of N-glycans in an afucosylated anti-viral mAb is responsible for FcγRIIIA binding and antibody-dependent cell-mediated cytotoxicity (ADCC) activity. We report a facile method to obtain purified asymmetric mono-galactosylated biantennary complex N-glycans, and their influence on bioactivity upon incorporation into an afucosylated mAb. Using ELISA, surface plasmon resonance and flow cytometry, we show that galactosylation of the α6 antenna, but not the α3 antenna, consistently increases FcγRIIIA binding affinity. We confirm its relevance in an anti-viral model of respiratory syncytial virus (RSV) using an adapted ADCC reporter assay. We further correlate this structure-function relationship to the interaction of the galactose residue of the α6 antenna with the protein backbone using 2D-¹H-¹⁵N-NMR, which showed that galactosylation of at this location exhibited chemical shift perturbations compared to glycoforms lacking this galactose residue. Our results highlight the importance of identifying and quantifying specific glycan isomers to ensure adequate quality control in batch-to-batch and biosimilar comparisons.

Keywords: 2D NMR; effector activity; glycosylation; monoclonal antibody; structure-function activity.

Figure 1

Schematic representation of transglycosylation of palivizumab to generate homogeneous glycoforms. (A) Deglycosylation of commercially-available palivizumab using EndoS D233Q generates the truncated GlcNAcFuc disaccharide (blue square = GlcNAc; red triangle = fucose) remaining on palivizumab. Defucosylation is achieved using fucosidase GH29, and subsequent glycan ligation of purified glycan oxazoline is achieved using EndoS 233Q. (B) Enzymatic reaction and purification of truncated glycans (mono-galactosylated at the α3 antenna (G1 (3)T) and digalactosylated (G2T)) using EndoS and α2-3,6,8 neuraminidase to cleave truncated glycans between the reducing end chitobiose core, followed by purification by porous graphitic carbon (PGC)-HPLC. (C) Synthesis of truncated agalactosylated (G0T) and monogalactosylated at the α6 antenna (G1 (6)T) glycans using LacZ β1-4 galactosidase (* indicates reaction byproduct peaks), followed by purification by PGC-HPLC. HPAEC-PAD trace of overlaid spectra of purified glycans are shown on the right in each of (B, C).

Figure 2

Fab binding of glycan-remodeled palivizumab to RSV-A2 F-protein is not affected by remodelled Fc-glycans. Comparable dose-response curves are obtained in the various glycoform analogues in an ELISA assay. RSV-A2 F-protein was coated at 0.25 μg/mL, followed by mAb binding. An anti-human Fc-HRP and Ultra-TMB were used for detection at 450 nm. Error bars show mean ± standard deviation (n=3 replicates); non-linear regression (4-PL) was used for curve-fitting in GraphPad. EC50 values generated and statistically analyzed (One-way Anova, Tukey post-hoc analysis) by GraphPad showed no significant differences between the various groups (p = 0.3184).

Figure 3

FcγRIIIa (CD16A) binding shows galactose on the α6 antenna significantly affects activity – 2D binding assays. (A,B) Dose response curves are obtained in the various glycoform analogues in an ELISA assay. (A) CD16A V 176 (high affinity) or (B) CD16A F176 (low affinity) were coated, followed by mAb binding. An anti-human Fab-HRP antibody, and Ultra-TMB were used for detection at 450 nm. Error bars show mean ± standard deviation (n=4 replicates); non-linear regression (variable slope, four parameters) was used for curve-fitting in GraphPad. (C-F) Surface Plasmon Resonance (SPR) of each palivizumab glycoform for either (C, E) CD16A V176 or (D, F) CD16A F176. (C, D) Overlaid sensograms and (E,F) dissociation constant (KD, M) are shown. For (E, F), error bars show mean KD ± standard deviation from 3 replicate binding experiments; **p < 0.01; ****p <0.0001, one-way ANOVA, Tukey post-hoc analysis.

Figure 4

FcγRIIIa (CD16A) binding shows galactose on the α6 antenna significantly affects activity - Cell based assays. (A-C) NK92-CD16A expressing cells (V Variant) and (D) Jurkat T-cell luciferase reporter cell line (CD16A V Variant). (A) A representative flow cytometry histogram is shown of NK92-CD16A (V176) cells bound to 0.41 µg/mL of various palivizumab glycoforms that were detected using an anti-human IgG (F(ab’) ₂ –specific) conjugated to AlexaFluor(AF)-647. (B) Dose-response curve of % GFP+APC+ cells for each palivizumab glycoform analogue (C) EC 50 (µg/mL) shows that galactosylation in the α6 antenna result in approximately twice as low EC50 compared to glycoforms lacking galactosylation in this position. EC50s are calculated using GraphPad. *, p< 0.05, One-way ANOVA, Tukey post-hoc analysis. (D) Dose-response curve of % GFP+APC+ cells for each palivizumab glycoform analogue to NK92-GFP cells expressing the low-affinity CD16A F176 variant. (E) Dose-response curve and (F) % relative activity of palivizumab glycoforms in the ADCC reporter assay, where % relative activity is the ratio of the EC50 of G2-palivizumab to the EC50 of each sample in each biological replicate. For (B-F), error bars show mean ± standard deviation (n=3 biological replicates); for (B, D, E) non-linear regression (4-PL) was used for curve-fitting in GraphPad; for (C, F) EC50 values were calculated and statistical analyses were performed using GraphPad. *, p< 0.05, One-way ANOVA, Tukey post-hoc analysis.

Figure 5

Synthetic scheme and NMR characterization of ¹⁵N-isotopically labelled Fc subunits with homogenous monogalactosylated glycoforms. (A) Synthetic scheme showing the remodeling of ¹⁵N-labelled Fc proteins expressed in P. pastoris and remodelled with sequential treatment with EndoH and EndoS D184M and purified glycan oxazolines. (B, C) Overlay of 2D-¹H-¹⁵N NMR spectra of agalactosylated ¹⁵N-G0-NISTmAb-Fc (red) with (B) ¹⁵N-G1(6)-NISTmAb-Fc (bright green) and (C) ¹⁵N-G1(3)-NISTmAb-Fc (dark green). Peptide backbone amino acid residues are labelled, and Combined Chemical Shift Differences (CCSDs) are calculated for those labeled in red and are shown in Table S1 .

Figure 6

Mapping chemical shifts perturbations of Fc glycoproteins bearing a galactose residue at either the α3- or α6-antenna. (A) In the α6-antenna, the galactose is in close proximity with the protein backbone, whereas (B) in the α3-antenna, the galactose residue is exposed to the outer solvent. Residues that experiencing chemical shift perturbations are depicted in green. β-sheets are depicted in cyan, α-helix in red, and loops/turns are shown in gray. X-ray structure of glycosylated Fc (PDB ID: 4byh).