Fc Sialylation Prolongs Serum Half-Life of Therapeutic Antibodies

The term “antibody” was coined by Paul Ehrlich to predict the existence of protective receptors able to neutralize toxins (1). This medicinal prospect became reality with the advent of immortalization of Ab-producing clonal B cells following their somatic cell fusion with myeloma cells (2). The resulting mAbs have revolutionized scientific research and medicine. Therapeutic mAbs have proven successful in the treatment of a wide range of diseases, including cancer and autoimmune and infectious diseases (3). Most therapeutic mAbs are human or humanized molecules of the IgG isotype, composed of the Fab and the Fc, which contains a single N-glycan structure attached to asparagine 297 (N297).

Among the four subclasses of human IgG, IgG1 are broadly used as anti-cancer mAbs. IgG1 exert strong effector functions through binding of their Fc to proteins of the lytic complement pathway and cross-linking of FcγRs widely expressed on many innate immune cells, such as NK cells, monocytes, macrophages, and neutrophils (4). In humans, the FcγRs include the activating receptors FcγRI, FcγRIIA, FcγRIIC, FcγRIIIA, and FcγRIIIB and the inhibitory receptor FcγRIIB (5). IgG interaction with FcγRIIIA is essential for the immune effector cells to engage mechanisms of Ab-dependent cell-mediated cytotoxicity (ADCC) against tumor cells. Because of a natural polymorphism in its extracellular domain (V/F 158), FcγRIIIA displays modulated affinities for IgG, influencing mAb therapeutic responses (6). Patients homozygous for the higher-affinity allele of FcγRIIIA (V158) show more favorable anti-tumor clinical responses with mAbs (7–9). Consequently, Fc protein-engineered mAbs with increased FcγRIIIA affinity, resulting in higher ADCC, have been developed to enhance efficacy by the introduction of Fc amino acid substitutions (10–12). In addition, Fc glycoengineered mAbs were also designed because modulation of the Fc N-glycan composition exerts profound influence over IgG functions (13). Indeed, aglycosylated mAbs, produced in Escherichia coli or mutated at N297 (N297A or N297G), lack effector functions (ADCC and complement-dependent cytotoxicity [CDC]) (14), whereas the absence of fucose residues on the Fc glycan moiety is associated with enhanced affinity for FcγRIIIA (15–17). Moreover, the anti-inflammatory activity of sialylated IgG has been broadly described (18–22) and is proposed to play an important role in the immunomodulatory activity of IVIg (23, 24).

The long catabolic t_1/2 (∼21 d) of IgG (IgG1, 2, and 4) augments their functional efficacy. This property is mainly attributed to IgG molecular size above the renal filtration threshold and interaction with the neonatal Fc receptor (FcRn). This receptor acts as a salvage receptor that prevents IgG degradation through the lysosomal pathway. FcRn binds IgG in a strictly pH-dependent manner. After internalization in endothelial cells through pinocytosis, IgG bind to FcRn with high affinity, owing to the low pH within the endosomes (pH 5–6). Upon binding, the IgG-FcRn complexes are recycled back to the cell surface where IgG are released at neutral pH into the blood circulation (pH 7–7.5) (25). Apart from endothelial cells, FcRn is also expressed by immune cells such as macrophages, monocytes, and dendritic cells (26). FcRn exerts numerous functions in immunological (27, 28) and nonimmunological (29, 30) mechanisms of catabolism and transport. Because longevity of therapeutic mAbs augments their efficacy, Fc engineering strategies have been developed to optimize the pharmacokinetic (PK) properties of IgG. Fc protein-engineered IgG with increased FcRn affinity and conserved pH dependency have been designed (31–33). These variants generally include two to three substitutions in or close to the FcRn binding site localized at the CH2/CH3 interdomain region (34, 35). The obtained Fc-engineered mAbs demonstrated longer serum half-lives in vivo in humanized FcRn (hFcRn) transgenic mice and cynomolgus monkeys (33, 36–38), which has recently been extended to humans (39).

Using random mutagenesis and a pH-dependent phage display selection process (40, 41), we previously identified several FcRn-optimized (FcRn⁺) variants, of which, one variant (C6A-66) raised our particular interest. Relative to the other FcRn⁺ variants, C6A-66 only modestly improved FcRn binding, yet it demonstrated the longest serum t_1/2 in hFcRn mice (40). The C6A-66 FcRn⁺ mutations included two substitutions located within the FcRn binding site as well as a glutamate residue deletion at position 294 (E294Del; “Del” hereupon). Assuming that the improved FcRn binding is attributed to the substitutions in the FcRn binding site, this could imply that the Del mutation might augment the IgG serum t_1/2 independently of FcRn binding. This hypothesis is explored in this study. To augment the sensitivity of detecting serum persistence, we combined the Del mutation with either of two optimal FcRn⁺ variants to salvage the IgG mutants from lysosomal degradation. We demonstrate that the Del-FcRn⁺ variants do not modify the affinity for the FcRn relative to the native FcRn⁺ variants but significantly augment their persistence in the blood circulation. The improved serum persistence of the Del-FcRn⁺ mutants could be directly attributed to the hypersialylation of the N297 sugar moieties. This study, therefore, demonstrates what is to our knowledge a hitherto unknown contribution of Fc sialylation on the serum t_1/2 of mAbs in vivo.

Directed mutagenesis was performed to construct the Fc variants by overlap PCR using standard protocols. The Fc variants were produced as full IgG variants with an anti-CD20 specificity, based on the variable domains of CAT 13.6E12 (42), using the YB2/0 cell line (CRL-1662; ATCC) in stable pools and the HEK293 Freestyle cell line (Invitrogen) by transient transfection, as previously described (41). After production, cell supernatants were clarified by low-speed centrifugation, concentrated by ultra filtration on a Pellicon XL Filter (Millipore), and subjected to affinity chromatography using HiTrap Protein A Sepharose Fast Flow column (GE Healthcare). Bound Abs were eluted with 0.1 M sodium citrate (pH 3) and neutralized with 1 M Tris-HCl (pH 9). Fractions containing the IgG were pooled and dialyzed against PBS (pH 7.4), sterile filtered (0.22 μm) and stored at 4°C. The purified IgG were analyzed using SDS-PAGE, under nonreducing and reducing conditions, and Coomassie Brilliant Blue staining. Low aggregate levels and absence of contaminants were verified by analytical gel filtration on Superdex HR/200 with an AKTA Prime system (GE Healthcare). None of the mutations tested induced IgG aggregation. The limulus amebocyte lysate endotoxin test with the gel-clot method was used to measure endotoxin levels.

Western blot analyses were performed with Sambucus nigra agglutinin (SNA), a lectin which preferentially binds to sialic acid attached to terminal galactose in α(2,6)- and, to a lesser degree, α(2,3)-linkage. Purified IgG (0.3 μg) were analyzed by SDS-PAGE in reduced conditions using a 10% Nu-PAGE gel and transferred on a PVDF membrane (Invitrogen). After saturation (TBS [pH 7.35] + 1% BSA + 0.05% Tween 20) overnight, the membrane was incubated, first with 2 μg/ml Biotinylated SNA (Vector Laboratories) for 2 h and then with 0.5 μg/ml Immunopure Streptavidin HRP Conjugate (Thermo Scientific) for 1 h. After a final incubation with F(ab')₂-Goat anti-Human κ L Chain Secondary Ab HRP Conjugate (Thermo Scientific; dilution 1/8000) for 1.5 h, the membrane was revealed by chemiluminescence using SuperSignal West Pico PLUS Chemiluminescent Substrate (Thermo Scientific). All incubation phases were performed at room temperature (RT) under gentle agitation. Between each phase, the membrane was washed five times for 5 min in 0.9% NaCl + 0.05% Tween 20.

A total of 500 μg of each sample was desalted with a Zeba Spin Desalting Column 7K MWCO (Thermo Fischer) and vacuum dried. Samples were then redissolved in PNGase F digestion buffer (20 mM sodium phosphate [pH 7.5], containing 0.02% sodium azide). N-Glycans were enzymatically released from the glycoprotein with PNGase F (GKE-5006B; Prozyme). Deglycosylated proteins were cold-ethanol precipitated, and glycans were collected in the supernatant. Glycans were then separated into four fractions. After a step of vacuum drying, the first tube was kept for native condition. The second tube was treated with sialidase A (GK-80040; Prozyme). The third tube was treated with sialidase A and fucosidase (GKX-5006; Prozyme). The last tube was treated with sialidase A, galactosidase (GKX-5014; Prozyme), and hexosaminidase (GKX-5023; Prozyme). Alternatively, one sample was treated with sialidase A and α-galactosidase (Glyko α(1–4,6)-Galactosidase [GKX-5007]) to determine the presence of Gal-Gal epitopes. Glycans were labeled with a fluorophore (8-aminopyrene-1,3,6, trisulfonate) and analyzed by capillary electrophoresis (CE; CESI 8000 plus; SCIEX). A glucose ladder standard (G20) was analyzed at the same time. Once the migration times of the samples are normalized by the standards, the corresponding glucose unit values are calculated and used for database search for possible matching structures. For that search, an internal database obtained with standard glycan structures released from mAbs and the Glycostore database (http://www.glycostore.org) were used.

Purified IgG (300 μg) were desalted, vacuum dried, and redissolved in 60 μl of digestion buffer (50 mM NaH₂PO₂ and 150 mM NaCl [pH 6.3]). Samples were separated into three aliquots. The first was digested with 100 U of IdeS (FabRICATOR Kit; Genovis) to produce the F(ab′)₂ and Fc fragments. The second was digested with 100 U of IdeS and 100 U of EndoS (IgGZERO Enzyme; Genovis) to determine the level of core afucosylation. Aliquots 1 and 2 were incubated at 37°C for 1 h. The third aliquot was first digested with 50 milliunits of PNGase F (GKE-5006B; Prozyme) for 24 h at 37°C and then digested with 100 U of IdeS. This sample was subsequently used to determine the mass values of the deglycosylated Fc/2 fragment. Following digestion, these fragments were denatured with 8 M urea and 0.4 M NH₄HCO₃ and reduced with 5 μl of a 250-mM DTT solution for 30 min at 50°C to produce the Fc/2, Lc, and Fd IgG subunits (∼25 kDa). Chromatographic separation of IgG subunits was performed using a Waters HPLC System equipped with a Waters XBridge BEH 300 C4 column heated to 70°C. Mobile phase A consisted of 0.02% trifluoroacetic acid in water, and mobile phase B consisted of 0.016% trifluoroacetic acid in acetonitrile. A total of 25 μg of IgG subunits mixture was injected on the column and separated with a linear ramp from 20 to 45% B over 6 min at a flow rate of 600 μl/min. Eluted species were then analyzed by a time-of-flight analyzer (LCT Premier XE; Waters). Operating in W mode with electrospray ionization from m/z 500 to 4000, the spectrometer was calibrated with NaICs (sodium iodide and cesium iodide mixture), and a lock mass solution of leucine enkephalin (500 pg/ml at 5 ml/min) was used. Electronic parameters were optimized with purified Ab as described before (43, 44). Spectra were analyzed with MassLynx software (Waters). Structures of glycans were determined theoretically with the differences of masses between the nonglycosylated species (deglycosylated fragment obtained after PNGase F treatment) and the experimental results by using an internal database and GlycoWorkbench software (EUROCarbDB).

To eliminate sialylated residues from glycans, the IgG variant T5A-74Del was treated with sialidase A (GK80040; Prozyme) to release α(2,3)-, α(2,6)-, α(2,8)-, and α(2,9)-linked N-acetylneuraminic acid. Four milligrams of IgG in PBS was incubated with 0.8 U of sialidase A at 37°C for 24 h. A control without the enzyme, T5A-74Del (mock), was treated in the same conditions. Samples were then purified with a HiTrap Protein A column (GE Healthcare) to eliminate the enzyme, concentrated, dialyzed against PBS, sterile filtered (0.22 μm), and stored at 4°C.

The IgG variants were tested for their binding to human FcγRs by ELISA. Recombinant His-tagged extracellular domains of FcγRs (FcγRI, FcγRIIA R131 and H131, FcγRIIB, FcγRIIIA F158 and V158; R&D systems) were coated at concentrations between 25 and 100 ng/well in PBS overnight at 4°C on MaxiSorp immunoplates (Nunc) or for 1 h at 37°C on Ni-NTA Plates (Pierce). Plates were then washed two times with PBS/0.05% Tween-20 and saturated with PBS/4% BSA for 2 h at 37°C. In parallel, purified IgG variants were diluted in PBS to a final concentration of 0.5 or 1 μg/ml, mixed with HRP F(ab′)₂ goat anti-human F(ab′)₂ at the same concentration, and incubated for 2 h at RT. F(ab′)₂-aggregated IgGs were then incubated under gentle agitation for 1 h at 30°C on the saturated ELISA plates (at the exception of FcγRI-coated plates) at serial dilutions (from 1 μg/ml to 31.25 ng/ml). For FcγRI-coated plates, classical ELISA was performed without complexes. IgG variants were first incubated on saturated plates for 1 h (at different concentrations from 1 μg/ml to 31.25 ng/ml), and the HRP F(ab′)₂ goat anti-human F(ab′)₂ was then incubated for 1 h at 37°C. For all plates, after washing, bound complexes were revealed with TMB (Pierce) and OD absorbance was read at 450 nm using a plate reader (Tecan). Ratios of OD IgG variant/wild type (WT) were calculated at an IgG concentration of 0.25 μg/ml, a concentration determined below the IgG-binding saturation plateau. Two to four independent experiments were performed, and mean and SD were calculated using PRISM 5.01 (GraphPad Software).

CD20-expressing Raji cells (1 × 10⁵ cells) were incubated at 4°C for 20 min with 100 μl of Ab (anti-CD20 variants and isotype control) at different final concentrations (0, 0.1, 0.5, 1, 2; 10 μg/ml) in PBS with 1% FCS. After washing, Abs were visualized with a goat F(ab′)₂ anti-human IgG Fc fragment coupled to PE (100 μl of a dilution of 1:100 in diluent; Jackson Immunoresearch) at 4°C for 20 min. The cells were washed, and mean fluorescence intensity (MFI) was studied with flow cytometer (FC500; Beckman Coulter).

CD20-expressing Raji cells (0.5 × 10⁵ cells) were incubated with increasing concentrations of anti-CD20 Abs (0–500 ng/ml) in the presence of 2.5 × 10⁵ Jurkat cells transfected with human FcγRIIIA_F158 and 10 ng/ml of PMA (Sigma-Aldrich). After 16 h of incubation at 37°C, the quantity of IL-2 released by Jurkat cells was measured by colorimetry (DuoSet Kit IL-2; R&D Systems). EC₅₀ (quantity of Ab that induces 50% of maximum IL-2 release) was calculated using PRISM 5.01 (GraphPad Software).

1.5 × 10⁵ CD20-expressing Raji cells were incubated with different concentrations of anti-CD20 Abs (0–5000 ng/ml) in the presence of baby rabbit serum as a source of complement (dilution to 1/10; Cedarlane). After 1 h of incubation at 37°C, the level of lactate dehydrogenase released in the supernatant by the lysed target cells was measured chromogenically (Roche Applied Sciences Cytotoxicity Detection Kit) to quantify CDC mediated by the Abs. Maximum lysis (1% Triton X-100) and spontaneous lysis levels (without Ab) served as control. Results were expressed as a percentage of lysis. EC₅₀ (quantity of Ab that induces 50% of maximum lysis) and E_max (percentage of maximum lysis) were calculated using PRISM 5.01 (GraphPad Software).

Two assays were set up to characterize the interaction of the IgG variants with hFcRn: 1) biolayer interferometry (BLI) and 2) competitive binding assay on FcRn-expressing Jurkat cells (1). FcRn affinities were measured by BLI using a RED96 OCTET system (Pall ForteBio). Recombinant hFcRn (α- and β2-microglobulin chains) was produced using the baculovirus system, as described before (40), and biotinylated with the EZ-link NHS-PEO kit (Pierce). After binding of 0.7 μg/ml biotinylated FcRn to streptavidin biosensors (300 s), IgG variants were associated at pH 6 (60 s; concentration of 200, 100, 50, 25, 12.5, 6.25, 3.125, and 0 nM) and then dissociated (30 s) from the receptor before regeneration (120 s) of the biosensors at pH 7.8. All dilutions were done in 0.1 M phosphate buffer, 150 mM NaCl, and 0.05% Tween 20. For data analysis, the 0-nM wells were used as a reference, the y-axis was aligned on the last second of the baseline, the interstep correction was done based on dissociation, and the fitting was done following a 1/1 model during the first 10 s of dissociation (2). For the competitive immunofluorescence assay, IgG variants were diluted in PBS (pH 6) at a final concentration ranging from 0 to 500 μg/ml and incubated with 1.5 × 10⁵ Jurkat cells expressing a truncated form of FcRn in the presence of Alexa-conjugated Rituximab (50 μg/ml) on the cell membrane. After 20 min, the cells were analyzed by flow cytometry to quantify the fluorescence signal. The results were expressed as a percentage of MFI; 100% refers to the MFI obtained with Alexa-conjugated Rituximab alone (i.e., without competitor), and 0% refers to the MFI value measured when Jurkat FcRn cells were not incubated with Alexa-conjugated Rituximab. Each experiment was done in duplicate. For each IgG, the MFI was plotted versus the log of IgG concentration. IC₅₀ (the Ab IC₅₀ of Rituxan-A488 binding) was calculated using PRISM 5.01 (GraphPad Software).

A temperature-controlled DynaPro Nanostar dynamic light scattering (DLS) instrument (Wyatt Technology) was used to record the thermal stability of the mAbs. Samples at a concentration of 1 mg/ml in PBS were centrifuged for 10 min at 10,000 × g. Thermal stability was recorded by ramping the temperature from 25 to 80°C at 1°C/min (5 measures per degree). Melting temperatures (Tm), defined as the onset temperature of unfolding/aggregation, were determined using thermal transition curves (i.e., hydrodynamic radius and scattering intensity, according to increasing temperature) in agreement with the manufacturer’s instructions (V6 Software). These curves indicate the structural modifications of the molecule under increasing temperature, permitting us to obtain the temperature when radius starts to increase, indicating the start of protein unfolding/aggregation.

Animals were housed in the animal center of Commissariat à l'Energie Atomique Saclay, and all studies were approved by the Responsible Animal Care and Use Committee (authorization number: 2012_121, CEEA 26). hFcRn mice (mFcRn^−/− hFcRnTg 276 heterozygote on a B6 background) for PK studies were obtained by breeding homozygous Tg276 hFcRn mice with FcRn^−/− mice obtained from The Jackson Laboratory. These mice are deficient in the mouse FcRn α-chain and carry an hFcRn α-chain gene, instead. Each animal (male and female, 8–12 wk old) received a single dose of 5 mg/kg of Ab i.v. via the retro-orbital sinus, after isoflurane 3–3.5% anesthesia. A total of 30 to 40 μl of blood samples was collected from the retro-orbital sinus of each mouse at the following time points: 1, 24, 48, 72, 168, 216, 336, and 432 h. The samples were then transferred into tubes. Blood samples were left at RT for at least 1 h, processed to serum, and stored at −80°C until analysis. A qualified anti-human IgG immunoassay was used to determine human IgG levels in the mouse serum samples and was performed by C. RIS Pharma (Saint Malo, France). Briefly, an affiniPure donkey anti-human IgG (H⁺L) polyclonal Ab was used for capture, and an affiniPure donkey anti-human IgG (H⁺L) peroxydase polyclonal Ab was used for detection (both from Jackson ImmunoResearch Laboratories). IgG concentrations were extrapolated from standard curves using the same purified IgG variant used for injection. The assay range was established with spiked quality control samples. The lower limit of quantification was 5 ng/ml. Sample results were corrected for dilution; values within the standard curve range were averaged. The serum concentrations were used to perform a noncompartmental PK analysis using Phoenix WinNonlin (Enterprise version 6.4 and 8.0) to estimate the PK parameters of each tested IgG for the six different PK studies. The apparent elimination t_1/2 was determined using data points from the terminal phase. The other PK parameters determined include the following: maximum plasma concentration extrapolated at time of injection, area under the serum concentration-time curve extrapolated from zero to infinity, volume of distribution at steady-state, clearance or the volume of plasma from which a substance is completely removed per unit time, and mean residence time (MRT) extrapolated from zero to infinity.

Statistical significance was calculated using an unpaired Student t test (t_1/2, MRT, and K_D) using PRISM 6 (GraphPad Software). The p values were as follows: ****p < 0.0001, ***p < 0.001, **p < 0.01, and *p < 0.05.

This study aims to understand whether the Del mutation (in yellow in Fig. 1A) can impact IgG serum t_1/2. As mentioned, we previously identified the Del mutation within a triple Fc mutant, C6A-66, containing two additional substitutions located in the FcRn binding site (T307P and N434Y, residues in orange in Fig. 1A) (40, 41). We tested the Del mutation alone and in combination with either of two optimal FcRn⁺ variants, the C6A-74 and T5A-74 variants (variants C6A-74Del and T5A-74Del; Fig. 1A). The FcRn⁺ C6A-74 and T5A-74 variants both contain one mutation directly within the FcRn binding site (N434Y) known to strongly increase FcRn binding, combined with additional mutations within the CH2 and CH3 domains outside the FcRn binding site (residues in purple in Fig. 1A). These Fc variants were produced as full-length human IgG1, using the variable regions of the anti-CD20 CAT 13.6E12 (42) in the YB2/0 cell line (EMABling platform, LFB Biomanufacturing). These cells naturally express low levels of FUT8 fucosyltransferase (45), resulting in low fucosylation of the Fc glycan moiety known to strongly enhance ADCC (46, 47). The native WT IgG was produced in parallel in the same expression system. The IgG WT and variants were purified on protein A and characterized for their Fc glycosylation profiles. The anti-CD20 Fab fragment was confirmed not to be N- or O-glycosylated (data not shown). The human IgG1 Fc fragment contains a single N-glycosylation site located at N297 in the CH2 domain. This N-glycan structure is a well-defined complex biantennary structure composed of a core heptasaccharide that can vary by the addition of fucose, N-acetylglucosamine (GlcNAc), galactose, and sialic acid (Fig. 1B). Strikingly, the N-glycosylation profiles of the Fc domains revealed that all molecules containing the Del mutation endowed high sialylation. First, probing a Western blot with SNA, a sialic acid–specific lectin, demonstrated high binding to IgG variants Del, C6A-66, T5A-74Del, and C6A-74Del (Fig. 2A). By contrast, this SNA reactivity was absent for the WT IgG as well as the C6A-74 and T5A-74 variants, which do not include the Del mutation. The composition of the Fc N-glycans was assessed using two complementary approaches: first, high-performance CE laser-induced fluorescence (HPCE-LIF) (48, 49) and second, an orthogonal technique based on liquid chromatography (LC) and mass spectrometry (MS) (44). For HPCE-LIF analyses, N-glycans were released by PNGase F treatment (Fig. 1B), fluorophore labeled, and separated by high-performance CE. Classically (50), sequential top-down enzymatic digestions with fucosidase, sialidase, and galactosidase permitted a bottom-up identification of sugar residues, for example, sialic acid, fucose, galactose, and bisecting GlcNAc (Fig. 3A–C, Supplemental Fig. 1, Table I). For LC-MS analyses, Fc fragments were first released from IgG by IdeS treatment and, after reduction of disulfide bridges, monomeric Fc fragments containing N-glycans were separated by LC and analyzed by MS (Fig. 4A–C, Table II). Deglycosylated Fc fragments, after PNGase F treatment, were also analyzed by LC-MS. This established their mass and revealed their low level of oxidation (<25%) and high level of H chain C-terminal lysine clipping (>85%), which are classical modifications described for mAbs (51). HPCE-LIF and LC-MS analyses confirmed low fucose content (between 10 and 40%) for the WT and FcRn⁺ variants (C6A-74 and T5A-74) as well as a low galactosylation rate (<80/200) and very low bisecting and sialylated forms. On the contrary, Del and Del-FcRn⁺ variants (C6A-66, C6A-74Del, and T5A-74Del) were highly galactosylated (between 160 and 200/200) and sialylated (∼90% by HPCE-LIF and ∼150/200 by LC-MS), while conserving a low fucose content (<60%). Remarkably, among the Fc-sialylated forms, a majority (∼60%) were fully bisialylated, in correlation with the high sialylation rate observed. The Del-FcRn⁺ variants also exhibited increased bisecting GlcNAc (up to 50%). Further analyses demonstrated that Del variants could contain trace amounts of Gal-Gal epitopes (<2.8%) and monosialylated N-glycolyl neuraminic acid [NGNA; <10% (Fig. 4B)]. No immunogenic response is expected from these minor structures because it was shown that Gal-Gal epitopes on Fc fragments are not immunogenic (52) and that at least two Fc NGNA seem to be required to elicit immunogenicity in humans (53). Then, to address the relative contribution of the YB2/0 cell line to the glycosylation profile, we produced the IgG WT, T5A-74, and T5A-74Del variants in HEK293 cells and compared the N-glycosylation profiles. Probing for SNA reactivity by Western blot demonstrated positivity for the T5A-74Del (HEK) variant but clearly at a lower level than the T5A-74Del variant produced in YB2/0 (Fig. 2B). Indeed, LC-MS analyses (Table II) demonstrated a low level of galactosylation (<80/200) and sialylation (20/200) for the T5A-74Del (HEK), resulting in <5% of bisialylated forms. High fucosylation was confirmed for the three IgG variants produced in HEK293 cells (75–80%). The T5A-74Del (HEK) variant also exhibited increased bisecting GlcNAc (+29% compared with the T5A-74 variant). These analyses clearly showed that the level of Fc sialylation induced by the Del mutation highly depends on the cellular production system used.

The IgG variants produced in YB2/0 cells were tested for their capacity to bind to human FcγRs by ELISA (Table III). For low-affinity receptors (FcγRIIIA, FcγRIIA/B), the avidity of the IgG variants was augmented by preincubation with a secondary anti-F(ab′)₂ Ab to form aggregates mimicking immune complexes. For the high-affinity receptor FcγRI, a classical ELISA was performed with monomeric IgGs. Results for each receptor were expressed as a ratio of specific signal obtained for the IgG variant divided by the signal of the WT IgG (Table III, upper part). IgG variants containing the Del (C6A-74Del and T5A-74Del) exhibited significantly decreased binding to FcγRs (ratios variant/WT: 0.05–0.27). In addition, a fully desialylated version of the T5A-74Del variant was prepared for comparative studies by sialidase A treatment (T5A-74Del + sial. A). Specific enzymatic removal of sialic acids and integrity of a control sample processed in the same conditions without the enzyme (T5A-74Del [mock]) were monitored by SNA Western blot, HPCE-LIF, and LC-MS (Figs. 2C, 3C, 4C, Table II). Importantly, a low-binding profile was also observed for the desialylated variant (T5A-74Del + sial. A), demonstrating that the Del has a destabilizing effect on FcγRs binding independently of the sialylation. As previously described (41), the C6A-74 variant displayed a conserved binding profile to FcγRs, whereas the T5A-74 variant had increased binding to FcγRI (2-fold increase). The IgG variants produced in HEK293 cells were also tested by ELISA and compared with their WT counterpart (Table III, lower part). Similarly, the mildly sialylated T5A-74Del (HEK) variant showed decreased binding to all FcγRs. The T5A-74 (HEK) variant displayed a 2-fold increased binding to FcγRI and to FcgRIIIA_V158. This last binding increase was not observed for the T5A-74 variant produced in YB2/0 cells, probably because the IgG low-fucose content already resulted in high binding to this receptor, reducing the additional impact of the mutations (15).

To evaluate the effect of the mutations on IgG effector functions, Ag-dependent functional activities of the IgG variants were evaluated using a human B cell line expressing CD20 (Raji cells) as target. First, we verified that neither the mutations nor the desialylation procedure affected the Ag recognition on these cells (Fig. 5A, Supplemental Fig. 2A, 2B). Second, Jurkat cells expressing human FcγRIIIA_F158 were used in presence of Raji cells to detect the ability of the IgG variants to engage this receptor upon Ag recognition, resulting in IL-2 secretion. In corroboration with the binding results, the WT, T5A-74, and C6A-74 variants induced equivalent IL-2 secretion (EC₅₀: ∼0.8 ng/ml), whereas the Del, C6A-74Del, and T5A-74Del (HEK) variants had no detectable activity (Fig. 5B, Supplemental Fig. 2C–E). The T5A-74Del and T5A-74Del (mock) variants induced a very low level of IL-2 secretion, which was slightly increased after desialylation (EC₅₀ not measurable). This demonstrates that the Del has a negative impact on FcγRIIIA_F158-induced activity. Finally, CD20-expressing Raji cells were used to test the ability of the IgG variants to induce CDC activity in presence of baby rabbit serum (Fig. 5C). The T5A-74 (HEK) variant displayed no CDC activity, as previously documented (41). As expected, the WT (HEK), WT, and C6A-74 variants induced robust CDC in an equivalent manner with an EC₅₀ of 116, 101, and 82 ng/ml, respectively. In contrast, the Del and C6A-74Del variants had no detectable CDC activity, showing the strong negative impact of the Del on this IgG functional activity, as already observed for the C6A-66 variant (41).

The affinity of the IgG variants for hFcRn was evaluated using BLI with biotinylated recombinant hFcRn (Fig. 6A, 6B). Association and dissociation phases were performed at pH 6, and regeneration of the biosensors was performed at pH 7.8, confirming the pH dependence of the IgG variants. To confirm the ability of the IgG variants to recognize FcRn, a competitive assay was performed using Jurkat cells expressing hFcRn at their cell surface (Fig. 6C, 6D, Supplemental Fig. 2F). The IgG WT molecules, produced in YB2/0 and HEK cells, displayed the same affinity for FcRn (K_D = 28.8 and 29.4 nM, respectively), confirming that changes in fucose content have no impact on FcRn binding. As expected, the FcRn⁺ variants C6A-74 and T5A-74 exhibited an equivalent increased affinity for FcRn (K_D = 5.0 and 3.9 nM, respectively), independent from the production cell line used (K_D = 3.6 nM for the T5A-74 [HEK]). A modest decreased affinity was measured for the Del variant by BLI (K_D = 38.2 nM) compared with the WT (K_D = 28.8 nM), which was confirmed with the cell-based assay (Fig. 6C). When cumulated with the FcRn⁺ variants, the Del had very low impact on FcRn binding, as observed on cells (Fig. 6C, 6D). This decrease was not significantly detectable by BLI. Moreover, both techniques showed that desialylation of the T5A-74Del variant had no impact on FcRn affinity (K_D between 3.9 and 4.6 nM; Fig. 6C). Finally, we performed comparative thermal stability experiments using DLS, a spectroscopic technique that measures the hydrodynamic radius of molecules in suspension. At 25°C, monodispersed and monomodal profiles were obtained for all the IgG molecules, indicating an absence of aggregation induced by the Fc mutations. Then, the temperature was ramped from 25 to 80°C, and Tm were measured for all the IgG molecules (Table IV). All IgG variants have conserved high thermal stability (Tm > 60°C), comparable to the IgG WT molecules (∼63°C). These data demonstrate that the Del does not degrade the integrity of the mAbs as measured by thermal stability.

Because the IgG variants have improved FcRn binding properties in vitro, we tested whether FcRn affinities correlated with longer serum persistence in vivo. PK studies were performed in hFcRn mice that are homozygous for a knockout allele of murine FcRn and heterozygous for an hFcRn-α transgene. These hFcRn mice, unlike WT mice, have been shown to be a reliable surrogate for studying human IgG serum t_1/2 (36). Studies using marketed mAbs, for which PK data in humans and/or primates were available, convincingly demonstrate that the hFcRn mice reliably predicted their serum t_1/2 in humans (54, 55). The absence of Ag recognition in mice by the anti-CD20 IgG variants allowed us to study target-independent mechanisms of clearance. Each animal received a single i.v. injection of IgG at 5 mg/kg via the retro-orbital sinus. Blood samples were collected at multiple time points from 1 to 432 h and titrated by ELISA. For each IgG variant, one to three independent PK studies were performed in comparison with the IgG WT molecule and PK parameters of t_1/2, and MRT were determined. MRT represents the average time a molecule stays in the body, therefore estimating its serum persistence. As previously described (40), the C6A-74 and T5A-74 variants displayed increased t_1/2 (ratio variant/WT = ∼1.5) but also increased MRT compared with the IgG WT molecule (ratio variant/WT = 2.8 for both variants). The Del variant, only containing the Del and highly sialylated, had increased t_1/2 (ratio variant/WT = 1.8) but unchanged MRT compared with the IgG WT molecule (Fig. 7A, 7B). Finally, the variants combining FcRn⁺ mutations and the Del, C6A-74Del, and T5A-74Del, resulting in high sialylation, displayed further increased t_1/2 (ratio variant/WT of 2.5 and 3.9, respectively) and MRT (ratio variant/WT = 5.9 and 8.8, respectively) compared with their nondeleted counterpart (C6A-74 and T5A-74; ratios t_1/2/WT = ∼1.5 and ratios MRT/WT = ∼3). Although the deletion of E294 alone has no effect on MRT (ratio variant/WT = ∼1), it has a synergistic effect on IgG persistence (MRT parameter) when combined with both C6A-74 and T5A-74 variants. This synergistic effect on MRT is not dependent on FcRn affinity because the combined IgG variants are not FcRn improved compared with their nondeleted counterpart (Fig. 6). This implied a possible role for the N297-associated sugar moieties in the in vivo mAb persistence. To directly test this hypothesis, we focused on the T5A-74Del variant for which the most dramatic increase in serum persistence was observed (Fig. 7C). First, we desialylated the variant T5A-74Del and compared its serum PK profile with the control IgG (T5A-74Del [mock]), and a statistically significant decrease in both t_1/2 and MRT was observed (Fig. 7D, 7E, Supplemental Fig. 3). The magnitude was striking as the t_1/2 was reduced to that of the native T5A-74 variant, and 70% of the impact on the MRT was abolished. This indicates that the synergistic impact of the Del on the serum persistence is mediated by the sialylation of the N-glycan at the position N297. This conclusion is strengthened by the results obtained for the T5A-74 and T5A-74Del variants produced in HEK293 cells, resulting in low Fc sialylation. Indeed, these two variants showed identical PK profiles and parameters (Fig. 7F), confirming that the unusual N-glycan profiles of the IgG variants produced in YB2/0 cells are certainly responsible for the effects observed on PK parameters.

We next assessed whether this phenomenon was unique for the Del. To this end, we identified a number of Fc substitutions (F241A, F243A, V262A, and V264E) that cause hypersialylation in the literature (56–58). We choose to introduce the substitution V264E on the T5A-74 variant because it demonstrated the highest level of sialylated structures among the four mutants tested. Second, similar to the Del, the V264E variant allowed a 30% sialylation in HEK293 cells (57). When produced in YB2/0 cells, the T5A-74/V264E variant displayed similar in vitro properties as the T5A-74Del variant (Supplemental Fig. 2A–F), with an equally high affinity to FcRn measured by BLI (K_D = 4.9 and 4.6 nM, respectively). Most importantly, HPCE-LIF analysis of the T5A-74/V264E variant revealed that close to 100% of the N-glycan chains were sialylated (Supplemental Fig. 4), a sialic acid content slightly increased compared with the T5A-74Del variant. When tested in hFcRn mice in a PK study as described before, the T5A-74/V264E variant clearly demonstrated equivalent PK profile as the T5A-74Del variant (Fig. 7C). Calculation of PK parameters confirmed the extended serum persistence of the T5A-74/V264E variant with a t_1/2 of 178 h (ratio variant/WT = 4.8) and an MRT of 227.6 h (ratio variant/WT = 11.3).

Fc engineering has proven highly efficient in tuning the effector functions of mAbs. In this article, we describe the impact of a single amino acid deletion (E294) in the CH2 domain on Fc glycosylation, IgG effector functions, and IgG serum t_1/2. This mutation was identified in a phage display–derived Fc domain containing two mutations that sit within the FcRn binding site in addition to the Del. This C6A-66 variant demonstrated a long t_1/2 even when compared with IgG variants with a higher affinity for FcRn. This is unexpected because IgG serum t_1/2 is dominantly regulated by the FcRn (13). To study whether the Del could explain this discrepancy, we combined the Del with each of two well-characterized FcRn⁺ variants to limit the degradation of the variants in lysosomes. Critically, the combined Del-FcRn⁺ variants conserved high affinity for the FcRn relative to the native FcRn⁺ variants. When injecting the Del-FcRn⁺ variants into hFcRn mice, we observed a striking augmentation in their serum t_1/2 and persistence (MRT) relative to the FcRn⁺ variants. This improved serum persistence of the Del-FcRn⁺ mutants could be directly attributed to the hypersialylation of the N297 sugar moieties. To date, and to our knowledge, not only has such a contribution of Fc sialylation on the serum t_1/2 of mAbs never been reported, it is also widely accepted that the presence or absence of terminal galactose and/or sialic acid residues has little influence on IgG t_1/2 (13, 59, 60). However, sialylation has long been implicated in the salvaging of glycoproteins (non-IgG) from their degradation in the liver (61). This has been attributed to the asialoglycoprotein receptor that binds asialylated galactose residues present on glycoproteins. The addition of sialic acid on galactose residues masks the epitope, thereby inhibiting their removal from circulation (62).

A 6- to 9-fold augmentation in serum persistence could be observed for the Del-FcRn⁺ variants. This striking impact was caused by a synergistic contribution of the FcRn⁺ mutants (3-fold increase in MRT) and the Del (1-fold increase in MRT). This synergy was mediated by the Del-induced sialylation of the N297 glycan moiety. Treatment with sialidase A permitted desialylation of the Del-FcRn⁺ variants, which resulted in the PK profile to return to that observed for the FcRn⁺ variants. This critical contribution reflects the density of N297 sialylation (150/200) and the dominance of bisialylated moieties. This efficacy was only achieved when producing the Del variants in the FUT8-deficient YB2/0 cells. Classically, recombinant mAbs are produced in HEK293 and CHO cells. The sialylation obtained under these conditions is usually below 5%, and even the Del variants reached 30% sialylation at best. Few other studies have reported an impact of N297 glycosylation on PK changes. N-Glycans with high mannose (Man5-9) or terminated GlcNAc could bind to the mannose receptor (ManR) expressed on macrophages/dendritic cells, leading to the accelerated clearance of IgGs (63, 64). Fucosylation has been formally excluded to impact serum t_1/2 (65). Similarly, binding to FcRn is not impaired by glycan removal (66).

In addition, the striking phenotype of the Del was fully reproduced with a second mutation, V264E. Both mutations allowed for a prolonged IgG serum t_1/2, Fc hypersialylation, and reduced FcγRs binding. They even matched for 30% sialylation when the variants were produced in HEK293 cells. By extrapolation, the mutations F241A, F243A, V262A, V264E, and V265A might all prove to be phenocopies because they all permitted 30% sialylation when produced in HEK293 cells (56–58, 67, 68). Nuclear magnetic resonance (69) and crystallographic (66) analyses showed that these residues, in particular F241 and F243, are in close contact with the N-glycan chains (68). They contribute to the N-glycan motion restriction that impacts the structure of the C’E loop (aa Y296 to R300, in red in Fig. 1A), where sits the N297 glycosylation site. This restricted Fc structure resembles an “open” conformation that is favorable to FcγRIIIA binding (70). Changes in Fc glycosylation can shift this C’E loop, thereby influencing Fc conformation and consequently binding to FcγRs (13, 71). Fc bearing oligomannose-type glycans, associated with increased FcγRIIIA binding, adopts an asymmetrically open conformation with an expansion of the interdomain space (72), whereas low-fucosylated Fc seems, instead, to accommodate a glycan of FcγRIIIA by a subtle local conformation alteration around residue Y296 (17, 73). In contrast, the sialylated F241A and F243A variants were shown to adopt a “closed” conformation, unfavorable to FcγRIIIA binding, characterized by a structural change inducing the two C’E loops to come closer. This closed conformation was also described for the glycoengineered bisialylated Fc fragment and was associated with increased conformational flexibility of the CH2 domain (57, 70, 74). Altogether, Fc sialylation seems to adversely impact FcγRIIIA binding and ADCC activity in the context of core fucosylation but not in its absence (75–77). In addition, the anti-inflammatory effect endowed to naturally sialylated Fc was fully recapitulated by the F241A variant, confirming structural and functional mimicry (78).

The position we studied, E294, is structurally localized close to the N297 glycosylation site, at the C’E loop base in the CH2 domain (Fig. 1A). It is therefore conceivable that the Del impacts the structure of the C’E loop favoring a closed conformation, in correlation with our observations of increased sialylation and decreased binding to FcγRIIIA. The Del variant also induces a loss of binding to all human FcγRs and abolishes CDC activity. Desialylation had little effect on these properties, demonstrating the dominant effect of the Del in imposing a conformational change. Likewise, desialylation of the F243A variant had no effect on its ADCC activity (58), and desialylation of the F241A variant had no influence on its anti-inflammatory properties (78), demonstrating the dominant impact of these mutations on the structural changes induced. On the contrary, mutations directly in the FcγRs binding site (L234F/L235E/P331S) also resulted in a complete lack of effector functions but had no major impact on Fc glycosylation nor Fc conformation (79). Such a direct effect on FcγR binding is highly unlikely for our variant because of the structural localization of E294 and the strong impact on Fc glycosylation.

Sialylated Abs are natural constituents of the humoral response that are induced over the course of viral and bacterial infections (80, 81). Experimentally enforcing N297 glycan sialylation endows a potent anti-inflammatory capacity in models of inflammatory disease (21, 22, 82). Whether N297 sialylated moieties impact the longevity of IgG in the blood circulation is not known. The closest comparison we provide is the impact on Ab t_1/2 by the Del single variant when compared with WT IgG. This impact is modest and composite. The FcRn binding affinity of the Del variant (K_D = 38.2 nM) versus the WT Ab (K_D = 28.8 nM) indicates weaker binding of the Del variant to FcRn. This was corroborated in the competitive FcRn binding experiments. By contrast, the in vivo persistence of the Del variant was similar in terms of MRT and doubled in terms of t_1/2. By extrapolation, this suggests that the serum persistence of sialylated Abs might be requiring both FcRn-dependent and -independent mechanisms.

What could be the mechanism by which N297 sialylation prolongs IgG serum longevity? The IgG preservation is independent of FcRn or FcγR binding. The DLS analysis has demonstrated that neither the FcRn⁺ mutations nor the Del impaired the thermal stability of the Fc-engineered mAbs. This implies that the mechanism of action does not result from an aberrant structural alteration. By contrast, the desialylation experiments directly implicate N297 terminal galactose and/or sialic acid in IgG serum longevity. It is conceivable that the N297 sugar moiety is directly accessible for receptor binding. During the Fc–FcγRIIIA binding, the sugar residues of Asn162 of the Fc receptor interact with the sugar moiety of the Fc Asn297 via carbohydrate–carbohydrate interactions. Fucosylation of Fc-N297 reduces the affinity of this interaction by steric hindrance within the Fc cavity, obstructing the carbohydrate–carbohydrate interaction (17). No notable conformational change is observed for fucosylated Abs (73). In addition, the Del is a phenocopy of the F241A mutation for which the crystal structure has been elucidated (74). This mutant demonstrated an increased conformational flexibility in the CH2 domain that appears to facilitate access to the cavity near the C’E loop. This modest conformational modification might further facilitate access to the N297 sugar moieties by receptors that influence the serum t_1/2 of circulating proteins. Consequently, we expect that the Asn297 sugar moieties are directly accessible for binding by receptors that influence the serum t_1/2 of circulating proteins. The asialoglycoprotein receptor is a strong candidate; sialic acid masks its binding to asialylated galactose residues, thereby salvaging the glycoproteins from degradation (62).

In conclusion, our study reveals a hitherto unrecognized contribution of Fc sialylation to the persistence of Abs in vivo. On therapeutic Abs, the introduction of Fc sialylation provoked a synergistic increase in serum persistence when combined with FcRn⁺ mutations. Structurally, the Del mutation is well tolerated, as demonstrated by the conserved thermal stability. In this aspect, the Del mutation is an improvement over the well-known FcRn⁺ YTE variant, which increases mAb serum t_1/2 in humans at the cost of highly decreasing the thermal stability of the mAb (83). In addition, as assessed by two in silico algorithms that predict epitope binding to human HLA alleles (Antitope [iTope and TCED] and Lonza [Epibase]), the Del mutation failed to generate neo-epitopes with consensus HLA-binding residues. This is reassuring with regards to the potential immunogenicity of Del-modified therapeutic mAb. These long-lived mAbs could hold therapeutic promise, notably in clinical applications where Ag binding alone suffices for the therapeutic activity (neutralization, agonism, or antagonism) or when anti-inflammatory effects are sought.

We thank Alain Longue for expertise in cell culture, Ouarda Messaoudi for complementary cellular experiments, and Bianca Boussier and Sylvie Le Ver for mAb injections in hFcRn mice.

All authors, with the exception of L.T.M., are employees of LFB Biotechnologies, which patented the Fc mutations described.