Engineered aglycosylated full-length IgG Fc variants exhibiting improved FcγRIIIa binding and tumor cell clearance

ABSTRACT

FcγRIIIa, which is predominantly expressed on the surface of natural killer cells, plays a key role in antibody-dependent cell-mediated cytotoxicity (ADCC), a major effector function of therapeutic IgG antibodies that results in the death of aberrant cells. Despite the potential uses of aglycosylated IgG antibodies, which can be easily produced in bacteria and do not have complicated glycan heterogeneity issues, they show negligible binding to FcγRIIIa and abolish the activation of immune leukocytes for tumor cell clearance, in sharp contrast to most glycosylated IgG antibodies used in the clinical setting. For directed evolution of aglycosylated Fc variants that bind to FcγRIIIa and, in turn, exert potent ADCC effector function, we randomized the aglycosylated Fc region of full-length IgG expressed on the inner membrane of Escherichia coli. Multiple rounds of high-throughput screening using flow cytometry facilitated the isolation of aglycosylated IgG Fc variants that exhibited higher binding affinity to FcγRIIIa-158V and FcγRIIIa-158F compared with clinical-grade trastuzumab (Herceptin®). The resulting aglycosylated trastuzumab IgG antibody Fc variants could elicit strong peripheral blood mononuclear cell-mediated ADCC without glycosylation in the Fc region.

Introduction

In the past 20 years, substantial advancements have been made in the research, development, and clinical application of therapeutic monoclonal antibodies. Monoclonal antibodies have become one of the fastest growing sectors of human therapeutics for treating various human diseases such as cancer, inflammation, infections, and autoimmune diseases. As of May 25, 2017, 61 monoclonal antibodies have been approved by the US Food and Drug Administration (FDA),¹“ and the market size of monoclonal antibodies has been projected to increase to over $125 billion by 2020.²

In addition to the intrinsic antigen recognition capability, which is achieved by specific interactions between the hypervariable regions of an antibody and a target antigen, monoclonal IgG antibodies exert various therapeutically critical effector functions via the interaction between the Fc region of IgG and Fc-binding ligands such as Fc gamma receptors (FcγRs) and serum complement molecules. In addition, the circulating serum half-life of IgG antibodies can be prolonged by pH-dependent binding of the Fc region to the human neonatal Fc receptor (FcRn), resulting in intracellular trafficking and recycling of serum IgG.³

Human immune leukocytes express transmembrane-activating FcγRs (FcγRI, FcγRIIa, and FcγRIIIa) and inhibitory FcγRIIb on their surface.^4,5 IgG antibodies bound to a target antigen on tumor cells engage with FcγRs expressed on the surface of various immune cells and trigger intracellular signaling in immune effector cells to induce antibody-dependent cell-mediated cytotoxicity (ADCC) by releasing cytotoxic reagents such as perforin or granzyme.⁶ In antitumor immunotherapy using therapeutic monoclonal antibodies such as trastuzumab (Herceptin®) and cetuximab (Erbitux®), numerous research and clinical results indicate that ADCC is one of the most critical tumor cell-killing mechanisms.⁷ In addition to the ADCC effector function, antibodies recruit phagocytic cells such as macrophages to induce antibody-dependent cell-mediated phagocytosis (ADCP). They also recruit serum complement molecules via association of the antibody Fc with C1q, which generates a membrane attack complex that results in complement-dependent cytotoxicity (CDC) for the clearance of aberrant cells.³

In a natural IgG antibody molecule, a pair of N-linked glycans is appended at Asn297 of homodimeric Fc regions, and the presence of oligosaccharide chains is critical for the antibody structure and therapeutic functions.^3,8 In silico modeling and experimental studies report that N-linked glycans stabilize dynamic CH2 region,^9-11 resulting in resistance to the thermal denaturation of the region¹² and low pH-induced aggregation.¹³ In addition, effector functions, such as ADCC, ADCP, and CDC, elicited by binding to Fc-binding ligands, such as FcγRs and C1q, are largely affected by the glycan composition.^14,15

From a bioprocessing viewpoint, aglycosylated antibodies, which can be produced in prokaryotic hosts with reduced manufacturing costs and shorter manufacturing time without the necessity of time-consuming cell line development and without glycan heterogeneity issues, are advantageous compared with glycosylated antibodies produced in mammalian cells. Aglycosylated IgG antibodies show almost identical antigen-binding affinity, stability at a physiological temperature, and in vivo serum persistence compared with glycosylated counterparts.^14,16,17 Moreover, recent cellular and genetic optimization has enabled the generation of full-length aglycosylated antibodies with improved productivity, and several aglycosylated antibodies are being assessed in clinical trials without any reported immunogenicity issues.^14,17

However, the absence of N-linked glycans abrogates nearly all FcγR-binding affinity and immune effector functions that are essential for clearing antibody-opsonized tumor cells. Therefore, all US FDA-approved monoclonal antibodies that demand effector functions are glycosylated antibodies and are produced in mammalian cells such as Chinese hamster ovary (CHO), NS0, and Sp2/0.¹⁸ To explore the issue and isolate effector functional aglycosylated IgG antibodies that might have therapeutic use, Jung et al. performed bacterial display and high-throughput screening of a randomized Fc library and generated aglycosylated Fc variants that exhibited highly selective binding to FcγRI, without significant binding to other FcγRs such as FcγRIIa, FcγRIIb, and FcγRIIIa.¹⁹ The trastuzumab Fc variant that comprises the resulting engineered Fc contained two mutations (E382V/M428I) that stabilized the upper CH2 region of IgG Fc⁹ and potentiated dendritic cell-mediated ADCC,¹⁹ which was not observed using conventional glycosylated IgG antibodies. Sazinsky et al. isolated an aglycosylated Fc mutant (S298G/T299A) that exhibited restored FcγRIIa-binding affinity using a yeast surface display technique, and the resulting full-length IgG Fc variant could trigger effector function that was mediated by the Fc–FcγRIIa interaction.²⁰ More selective binding to the activating FcγRIIa vs. inhibitory FcγRIIb was also achieved by directed evolution of aglycosylated Fc. An aglycosylated trastuzumab with identified mutations (S298G/T299A/N390D/E382V/M428I) significantly enhanced macrophage-mediated ADCP compared with glycosylated antibodies.²¹

Natural killer (NK) cells are one of the most potent cytotoxic effector cell types for eliminating malignant cells. The antibody–tumor cell immune complexes associate with FcγRIIIa (CD16) on NK cells and induce the activation of effector cells to release cytotoxic granules that contain perforins or granzymes.⁶ Therefore, the significance of high FcγRIIIa binding and antitumor effects mediated by ADCC has been well established, and ADCC activity is critical for the success of antitumor therapies using monoclonal antibodies.^7,22,23 Here, we report the generation of engineered aglycosylated Fc variants for FcγRIIIa binding and effector function. By capitalizing on a comprehensive directed evolution, we isolated aglycosylated Fc variants that exhibited higher binding affinity to FcγRIIIa-158V and FcγRIIIa-158F than the clinical-grade trastuzumab (Herceptin®) and exhibited strong binding affinity to other FcγRs. The isolated trastuzumab Fc variants could induce potent peripheral blood mononuclear cell (PBMC)-mediated ADCC activity against human epidermal growth factor receptor (HER)2-expressing cancer cells. The results showed that an aglycosylated full-length IgG antibody was successfully evolved for FcγRIIIa binding, which would be useful for activating immune effector cells in antitumor therapies.

Results

Isolation of AglycoT-Fc1004-IYG exhibiting improved FcγRIIIa binding

Human FcγRIIIa expressed on immune effector cells binds to the glycosylated Fc region of human serum IgG with low affinity.⁴ Conversely, an aglycosylated Fc region does not display significant detectable binding to low-affinity FcγRs.^14,19 For efficient analysis and screening of FcγRIIIa–Fc interactions using flow cytometry or enzyme-linked immunosorbent assay (ELISA), we fused a streptavidin core domain to the C-terminus of FcγRIIIa to generate a tetrameric form of FcγRIIIa-158V, which enabled us to increase the apparent affinity by the avidity effect. The tetrameric FcγRIIIa‐158V, prepared by transient transfection into human embryonic kidney 293F (HEK293F) cells, followed by Ni-NTA affinity chromatography, showed strong noncovalent tetramerization while maintaining some tetrameric forms even after boiling for 10 min (SI Fig. 1A). The tetrameric FcγRIIIa-158V was fluorescently conjugated with Alexa Fluor 488 and retained good IgG-binding affinity (SI Fig. 1B).

Figure 1.

Isolation of aglycosylated Fc variants that exhibited high affinity to FcγRIIIa binding. (A) Schematic diagram showing the flow cytometric sorting of spheroplasts displaying aglycosylated trastuzumab Fc variants with high FcγRIIIa binding. (B) FACS analysis of isolated aglycosylated Fc variants. Values in the parenthesis of the 3D-overlay histogram indicate mean fluorescence intensity (MFI) on binding of spheroplasts displaying aglycosylated trastuzumab Fc variants to 5 nM tetrameric FcγRIIIa-158V-Alexa Fluor 488.

To engineer an aglycosylated Fc domain with significantly improved FcγRIIIa binding, we displayed full-length trastuzumab Fc variants on the inner membrane of E. coli. We used the covalently anchored full-length IgG display system, which comprises the coexpression of IgG light chains with NlpA leader peptides, as well as soluble forms of IgG heavy and light chains, in the periplasmic region of E. coli.²¹ In a previous study, an aglycosylated full-length IgG antibody Fc variant (Fc1004: S298G, T299A, E382V, N390D, and M428L) exhibited selective high binding affinity to FcγRIIa-131R compared with FcγRIIb.²¹ In another study, IgG Fc-AIYG (T299A, K326I, A327Y, and L328G), containing mutations in the C′/E loop and F/G loop of the CH2 region, showed improved binding to FcγRIIIa.²⁴ First, the binding affinity of anchored trastuzumab Fc variants that contained previously identified mutations, aglycosylated trastuzumab-Fc1004 (AglycoT-Fc1004) and aglycosylated trastuzumab-AIYG (AglycoT-AIYG), to fluorescently labeled tetrameric FcγRIIIa-158V-Alexa Fluor 488 were compared with that of wild-type trastuzumab using the full-length IgG display system. Spheroplasts displaying AglycoT-Fc1004, which was previously isolated for selective high binding affinity to FcγRIIa-131R,²¹ did not show significant binding to FcγRIIIa, as expected (SI Fig. 2A). Conversely, spheroplasts displaying aglycosylated trastuzumab-AIYG (AglycoT-AIYG) showed a slightly increased fluorescence-activated cell sorting (FACS) signal on binding to tetrameric FcγRIIIa-158V conjugated with Alexa Fluor 488. When we combined Fc1004 and AIYG mutations, the resulting AglycoT-Fc1004-IYG (S298G, T299A, K326I, A327Y, L328G E382V, N390D, and M428L) showed dramatically improved binding to FcγRIIIa-158V, which is in sharp contrast to AglycoT (wild-type aglycosylated trastuzumab), AglycoT-Fc1004, and AglycoT-AIYG (SI Fig. 2B).

Figure 2.

Mutations of aglycosylated Fc variants evolved for high FcγRIIIa binding. (A) Alignment of sequences for isolated aglycosylated Fc variants with new identified mutations is presented in red. (B) The mutations of Fc1004-IYG and new identified mutations are annotated in the Fc crystal structure (PDB:1FC1) by sticks and spheres, respectively.

Engineering AglycoT-Fc1004-IYG for improved FcγRIIIa binding

To isolate aglycosylated Fc variants that exhibited further improved FcγRIIIa-binding affinity compared with AglycoT-Fc1004-IYG, we randomized the Fc region of AglycoT-Fc1004-IYG by error prone PCR (target error rate, 0.5%). The PCR fragments were subcloned into a plasmid for IgG heavy chain expression and transformed into E. coli Jude1 cells that harbored the plasmid for anchoring IgG light chains, leading to the generation of a large aglycosylated trastuzumab Fc variant library (library size, 1.14 × 10⁹). The error rate of the library was 0.457% on the basis of the sequences of 20 randomly picked clones.

For associating FcγRIIIa with aglycosylated full-length IgG Fc variants displayed on the inner membrane of E. coli, the outer membrane and peptidoglycan layer of the library cells were eliminated as previously described,²⁵ and the resulting spheroplasts were labeled with tetrameric FcγRIIIa conjugated with Alexa Fluor 488. The region in the fluorescence histogram corresponding to the highest 3% of fluorescent spheroplasts following tetrameric FcγRIIIa-Alexa Fluor 488 binding was gated and sorted using flow cytometry (Fig. 1A). To enrich spheroplasts that exhibited high FcγRIIIa-binding affinity more efficiently, sorted spheroplasts were re-sorted again using the same gating strategy. After five successive sorting rounds, 90 individual clones were analyzed for FcγRIIIa binding using FACS, and finally, four aglycosylated IgG Fc mutants that exhibited significantly high affinity to FcγRIIIa were isolated (Fig. 1B). Among them, the AglycoT-MG48 mutant showed the highest FACS-binding signal with approximately a 6-fold and >17-fold improved signal compared with AglycoT-1004-IYG and AglycoT, respectively (Fig. 1B).

In the sequence analysis of the four isolated Fc variants, all the clones contained V264E mutation, and the selected clones contained either T350A or N421S mutation, indicating that these mutations were critical for increased FcγRIIIa-binding affinity (Fig. 2A and B). Moreover, the two mutations (T350A and N421S), which were located very far from the putative FcγRIIIa-binding region and covered the range between the lower hinge and upper CH2 regions, might induce conformational changes in the FcγRIIIa-binding epitope for enhanced FcγRIIIa binding (Fig. 2B).

Characterization of AglycoT-Fc variants for human FcγRIIIa-158V and FcγRIIIa-158F binding

Because all engineered Fc variants have previously identified mutations in the canonical N-linked glycosylation motif (Asn297-X-Ser/Thr), we produced aglycosylated trastuzumab Fc variants in mammalian cells without glycosylation at Asn297. For transient expressions, we subcloned the isolated Fc regions into a plasmid encoding the trastuzumab heavy chain and cotransfected the resulting plasmid into HEK293F cells along with a plasmid encoding the trastuzumab light chain. After transient expression in HEK293F cells and purification using protein A affinity chromatography (Fig. 3A), the kinetic and equilibrium binding constants of AglycoT-Fc variants toward FcγRIIIa prepared from HEK293F cells (Fig. 3B) were measured using a Biacore instrument (SI Fig. 3 and Table 1). The AglycoT-MG59 engineered for higher FcγRIIIa-158V binding from the parent clone (AglycoT-Fc1004-IYG) showed binding affinity similar to wild-type glycosylated trastuzumab (GlycoT). Moreover, the AglycoT-MG48 mutant exhibited over 16-fold and 24-fold improved binding affinity to FcγRIIIa-158V and FcγRIIIa–158F, respectively, relative to AglycoT-AIYG. This indicated an even higher affinity to both FcγRIIIa-158V and FcγRIIIa–158F than to the clinical-grade Herceptin prepared from CHO cells or wild-type glycosylated trastuzumab (GlycoT) expressed in HEK293F cells. The AglycoT-MG48 mutant, which was screened by high FcγRIIIa–158V binding, displayed a much higher affinity improvement ratio in FcγRIIIa–158F binding, with an approximately 7.5-fold higher affinity to FcγRIIIa–158F than that to clinical-grade Herceptin (SI Fig. 3 and Table 1).

Figure 3.

Production of IgG Fc variants and dimeric forms of FcγRIIIa-158V/FcγRIIIa-158F. (A) SDS-PAGE analysis for the purified trastuzumab Fc variants GlycoT, AglycoT-AIYG, AglycoT-MG59, AglycoT-MG87, and AglycoT-MG48. Each trastuzumab Fc variant was expressed in HEK293F cells and purified using protein A affinity chromatography. (B) SDS-PAGE analysis of the dimeric form of FcγRIIIa-158V and FcγRIIIa-158F. Each receptor, expressed in HEK293F cells, was purified using anti-GST affinity chromatography.

Table 1.

Quantitative affinity constants and equilibrium dissociation constants for trastuzumab Fc variants.

	FcγRIIIa-158V			FcγRIIIa-158F
	kon (M−1S−1)	koff (S−1)	KD (M)	kon (M−1S−1)	koff (S−1)	KD (M)	Fold increase (V/F)
AglycoT	N.D	N.D	N.D	N.D	N.D	N.D	0
AglycoT-AIYG	3.988 × 104	0.04426	1.110 × 10−6	8.141 × 103	0.1414	1.737 × 10−5	1
GlycoT	9.019 × 104	0.01082	1.200 × 10−7	1.778 × 104	0.1028	5.782 × 10−6	9.25/3.00
Herceptin	6.030 × 104	0.005076	8.418 × 10−8	1.771 × 104	0.09107	5.142 × 10−6	13.19/3.38
AglycoT-MG48	9.128 × 104	0.006201	6.793 × 10−8	2.212 × 104	0.01496	6.763 × 10−7	16.34/25.68
AglycoT-MG59	6.945 × 104	0.007424	1.070 × 10−7	2.251 × 104	0.02362	1.049 × 10−6	10.37/16.56

Binding affinities of AglycoT-Fc variants to other human FcγRs, FcRn, and C1q

To examine how mutations in isolated Fc variants affect binding to other human Fc receptors that are crucial for the activation status of various immune leukocytes and serum persistence of IgG antibodies, we prepared low-affinity Fc receptors, FcγRIIa-131H, FcγRIIa-131R, FcγRIIb, FcγRIIIa-158V, FcγRIIIa-158F, and FcRn, as dimeric forms with a C-terminal glutathione S-transferase (GST) fusion, which enabled more efficient analysis of binding using the avidity effect of the dimeric Fc receptors. FcγRIIa-131H-GST, FcγRIIa-131R-GST, FcγRIIb-GST, FcγRIIIa-158V-GST, FcγRIIIa-158F-GST, and FcRn-GST were expressed in HEK293F cells, purified using anti-GST affinity chromatography (SI Fig. 4), and then employed for binding analyses of trastuzumab Fc variants using ELISA.

Figure 4.

ELISA for binding of engineered aglycosylated IgG Fc to Fc receptors and C1q. (A) Schematic representation illustrating ELISA strategies for analyzing the binding of trastuzumab Fc variants to Fc receptors and human C1q. (B – G) Binding of trastuzumab Fc variants (GlycoT, AglycoT-AIYG, AglycoT-MG48, AglycoT-MG59, or AglycoT-MG87) to monomeric FcγRI (B), dimeric FcγRIIb (C), dimeric FcγRIIa-131H (D), dimeric FcγRIIa-131R (E), dimeric FcRn at neutral pH 7.4 (F), and dimeric FcRn at weak acidic pH 5.9 (G), respectively. (H) Binding of trastuzumab Fc variants (GlycoT, AglycoT, AglycoT-AIYG, AglycoT-MG48, AglycoT-MG59, or AglycoT-MG87) to human C1q.

In contrast to previously identified aglycosylated Fc variants, AglycoT-Fc5¹⁹ and AglycoT-Fc1004,²¹ trastuzumab Fc variants isolated in this study (AglycoT-MG48, AglycoT-MG59, and AglycoT-MG87) exhibited significant binding to all human FcγRs while displaying higher binding affinity to all human FcγRs compared with AglycoT-AIYG. Moreover, they showed lower and higher binding affinity to FcγRI and FcγRIIb, respectively, relative to GlycoT (Fig. 4A–E).

The pH-dependent FcRn binding is critical for the prolonged serum half-life of IgG antibodies. Serum IgG antibodies that are transported to the acidified endosome by pinocytosis are bound to FcRn, and then recycled back to the serum via dissociation from FcRn at the neutral pH in serum.³ AglycoT-MG48, AglycoT-MG59, and AglycoT-MG87 showed excellent binding at an endosomal pH (pH 5.5–6.0) and dissociation at a neutral serum pH, suggesting that IgG Fc variants display good FcRn-mediated recycling and prolong the serum half-life of IgG antibodies (Fig. 4F and G).

Serum complement C1q binds to the CH2 region of IgG²⁶ and initiates CDC for clearing target cells. Unlike GlycoT, which is the wild-type glycosylated IgG antibody, the lack of glycosylation in AglycoT-MG48, AglycoT-MG59, and AglycoT-MG87 significantly ablated the C1q binding activity, with a slightly increased binding compared with AglycoT-AIYG (Fig. 4H).

Evaluation of ADCC activity of trastuzumab Fc variants elicited by human PBMCs as effector cells

FcγRIIIa expressed on the surface of immune effector cells, including NK cells, interacts with the Fc region of IgG immune complexes and activates effector leukocytes for the clearance of antigen-sensitized target cells. To explore the antitumor activity of aglycosylated trastuzumab Fc variants by ADCC activity in a model more relevant to human physiological conditions, we used human PBMCs prepared from five anonymous donors (SI Table 1) as effector cells instead of using either NK cells purified from human leukocytes or NK92 cells, an immortalized cell line derived from an NK cell lymphoma patient. As target cells, two different breast cancer cell lines, MCF-7 and SKBR-3, were tested, and the expression of HER2 on the cancer cell lines was analyzed using FACS and clinical-grade Herceptin conjugated with Alexa Fluor 488. SKBR-3 cells showed significantly higher HER2 expression levels than MCF-7, which is in good agreement with previously reported results^27,28 (Fig. 5A). ADCC activity was measured by monitoring lactate dehydrogenase (LDH) release. Using clinical-grade Herceptin, a typical ADCC result was obtained, which displayed E:T ratio-dependent ADCC activity against both cancer cell lines (Fig. 5B), and we analyzed ADCC activities of AglycoT-Fc variants at a 25:1 effector:target cell ratio. No detectable cytotoxicity was observed with normal human serum IgG and AglycoT sensitized effector cells. In contrast, Fc-engineered AglycoT-MG48 and AglycoT-MG59 displayed potent ADCC activities when using PBMCs as effector cells. As expected, SKBR-3 cells displaying higher HER2 antigens on their surface (Fig. 5A) showed higher ADCC activities (Fig. 5C). Despite higher FcγRIIIa-binding affinity of AglycoT-MG48 compared with either GlycoT or clinical-grade Herceptin (prepared in HEK293F cells and CHO cells, respectively), the ADCC activity of AglycoT-MG48, which also exhibited higher binding affinity to the inhibitory FcγRIIb, was lower compared with that of GlycoT and Herceptin, indicating the high relevance of FcγRIIb-binding affinity in the ADCC potency of therapeutic IgG antibodies (Fig. 5C).

Figure 5.

Analysis of ADCC activities for trastuzumab Fc variants using human PBMCs as effector cells. (A) HER2 expression level of target SKBR-3 and MCF-7 cell lines. (B) Lysis of SKBR-3 or MCF-7 target cells using clinical-grade Herceptin depending on E (effector cell):T (target cell) ratio. (C) ADCC activity of trastuzumab Fc variants. IgG antibodies were added at a 25:1 (E:T) ratio and ADCC activity was then measured by LDH release after incubation for 4 h. Results of cytotoxicity were calculated on the basis of three experimental repeats performed in duplicate.

Prediction of immunogenicity of Fc variants

To evaluate a potential T-cell response triggered by an exogenous peptide-bound MHC II, we analyzed MHC II binding to the peptides that can be derived from Fc variants through the Immune Epitope Database (IEDB).^29,30 From the Fc sequences of trastuzumab (wild-type human IgG1), AglycoT-MG48, and AglycoT-MG59, a series of 15-amino acid peptides was generated and the percentile rank (CPR) of each peptide for binding to the HLA-DRB1*0401, an MHC II allele, which covers nearly 95% of human populations, was scored. A lower CPR score indicates a potential higher MHC II affinity binder, suggesting a high probability of eliciting immunogenicity. In the evaluation of all 103 peptides containing the identified Fc mutations, no peptides showed significantly reduced CPR score under CPR < 2 (SI Table 2). Moreover, the assessment of the 9-amino acid core peptide regions derived from the 15-amino acid peptides showed that there is no predicted peptide in the category of high MHC II affinity binder (IC₅₀ value < 50 nM) and that the isolated new mutations did not change any peptide from low affinities (IC₅₀ value < 500 nM) to intermediate affinities (IC₅₀ value < 5000 nM) (SI Table 3). Therefore, these results clearly suggest that the possibility of immunogenicity resulting from introducing the new identified mutations is very low.

Table 2.

Plasmids used in the current study.

Plasmid	Relevant characteristics	Reference or source
pMAZ-FcγRIIIa-158V-FLAG-Streptavidin-His	FcγRIIa158V‐FLAG-streptavidin gene in pMAZ‐IgH‐GlycoT	Current study
pPelBFLAG	Cmr, lac promoter, tetA gene, skp gene, C-terminal FLAG tag	32
pPelB-AglycoT(H)-Fc-FLAG	Trastuzumab H chain gene in pPelBFLAG	26
pPelB-AglycoT(H)-Fc1004-FLAG	Trastuzumab-Fc1004 mutant H chain gene in pPelBFLAG	26
pPelB-AglycoT(H)-AIYG-FLAG	Trastuzumab AIYG mutant H chain gene in pPelBFLAG	Current study
pPelB-AglycoT(H)-Fc1004-IYG-FLAG	Trastuzumab Fc1004-IYG mutant H chain gene in pPelBFLAG	Current study
pBAD30-KmR	Kmr, BAD promoter	20
pBAD-AglycoT(L)-His	Trastuzumab L chain gene in pBAD	26
pMAZ-IgH-GlycoT	Trastuzumab H chain gene in pMAZ-IgH-H23	32
pMAZ-IgL-GlycoT	Trastuzumab L chain gene in pMAZ-IgL-H23	32
pMAZ-IgH-GlycoT-AIYG	Trastuzumab AIYG mutant H chain gene in pMAZ-IgH-GlycoT	Current study
pMAZ-IgH-GlycoT-MG48	Trastuzumab MG48 mutant H chain gene in pMAZ-IgH- GlycoT	Current study
pMAZ-IgH-GlycoT-MG59	Trastuzumab MG59 mutant H chain gene in pMAZ-IgH- GlycoT	Current study
pMAZ-IgH-GlycoT-MG87	Trastuzumab MG87 mutant H chain gene in pMAZ-IgH- GlycoT	Current study
pMAZ-IgH-GlycoT-FcN297D	Trastuzumab N297D mutant H chain gene in pMAZ-IgH- GlycoT	26
pMAZ-FcγRIIa-131H-GST	FcγRIIa131H‐GST gene in pMAZ‐IgH‐GlycoT	26
pMAZ-FcγRIIa-131R-GST	FcγRIIa131R‐GST gene in pMAZ‐IgH‐GlycoT	26
pMAZ-FcγRIIIa-158V -GST	FcγRIIIa158V‐GST gene in pMAZ‐IgH‐GlycoT	26
pMAZ-FcγRIIIa-158F-GST	FcγRIIIa158F‐GST gene in pMAZ‐IgH‐GlycoT	26
pMAZ-FcγRIIb -GST	FcγRIIb‐GST gene in pMAZ‐IgH‐GlycoT	26
pcDNA3-FcRn-GST	Ampr CMV promoter, FcRn-GST gene in pcDNA3	43

Table 3.

Primers used in this study (underlining indicates restriction enzyme sites).

PRIMER NAME	PRIMER NUCLEOTIDE SEQUENCE (5′→3′)
MJ#2	CTGCCCATGTTGACGATTG
MJ#36	CGCAGCGAGGCCCAGCCGGCCATGGCGGAGGTTCAATTAGTGGAATCTG
MJ#37	CGCAATTCGGCCCCCGAGGCCCCTTTACCCGGGGACAGGGAG
MJ#38	CAAGGAGTACAAATGCAAGGTCTCCAACATTTATGGCCCAGCCCCCATCGAGAAAACC
MJ#39	GGTTTTCTCGATGGGGGCTGGGCCATAAATGTTGGAGACCTTGCATTTGTACTCCTTG
MJ#42	CGGGAGGAGCAGTACAACAGCGCGTACCGTGTGGTCAGCGTCC
MJ#43	GGACGCTGACCACACGGTACGCGCTGTTGTACTGCTCCTCCCG
MJ#44	ACAAGATTTGGGCTCAACTTTCTTGTCG
MJ#45	CGACAAGAAAGTTGAGCCCAAATCTTGT
MJ#46	CGCAATTCCGGCCCCCGAGGCCCC
MJ#49	CGCAGCGAGCGCGCACTCCATGGCGGAGGTTCAATTAGTGGAATCTG
MJ#50	CCCTAAAATCTAGACCTTTACCCGGGGACAGGGAG

Discussion

FcγRIIIa (CD16a) expressed on the surface of various types of immune leukocytes, including monocytes, macrophages, and NK cells,^5,31 is a key molecule that links innate and adaptive immunities. NK cells that express only the activating FcγRIIIa account for approximately 15% of peripheral blood lymphocytes, and they are the main effector cells for eliciting an ADCC effector function, which is the most critical tumor cell-killing mechanism of therapeutic IgG antibodies. Considering that the engagement of IgG antibody Fc with FcγRIIIa (CD16a) determines the activation status of NK cells and that >90% of NK cells in the human blood are of the CD16a^high phenotype,^32-34 tremendous efforts have been made in the biopharmaceutical industry to engineer IgG Fc regions with improved FcγRIIIa binding and NK cell activation for tumor cell clearance.

Currently, all marketed therapeutic IgG antibodies that require effector functions are glycosylated. They contain heterogeneous N-linked glycans that are appended at Asn297, and their presence and composition are largely affected by the type of cell line and the downstream processes for manufacturing therapeutic IgG antibodies. Thus, a high capital investment for controlling the complicated glycan status to maintain product consistency is needed. As an alternative to glycosylated antibodies, aglycosylated IgG antibodies have been developed, and some of them are currently under clinical trials.¹⁷ Because aglycosylated IgGs have no glycan heterogeneity issues, development of these antibodies allows cheap production and high-throughput screening in various prokaryotic hosts. However, an aglycosylated IgG Fc, which has a highly dynamic upper CH2 region,⁹ cannot engage with FcγRs expressed on the surface of immune leukocytes.

To generate aglycosylated Fc variants that confer high binding affinity to FcγRIIIa expressed on immune effector cells and, in turn, elicit potent ADCC effects, we employed directed evolution strategies. By combining mutations previously identified for improved binding to FcγRIIa and FcγRIIIa, we generated Fc variants with significantly improved FcγRIIIa binding. Following this, creation of a randomized Fc library and high-throughput screening of IgG Fc variants displayed on the inner membrane of E. coli allowed the selection of Fc variants with significantly improved FcγRIIIa-binding capability. The resulting aglycosylated Fc variants exhibited much higher binding to FcγRIIIa-158V and FcγRIIIa-158F by over 16-fold and 25-fold, respectively, than that to a previously identified aglycosylated Fc variant for FcγRIIIa binding, AglycoT-AIYG. AglycoT-MG48 and AglycoT-MG59 displayed increased binding to FcγRIIIa-158F, with a 7.6-fold and 4.9-fold higher affinity than Herceptin, which would potentially make them useful as therapies for homozygous FcγRIIIa-158F patients.

Despite the higher FcγRIIIa affinity of AlycoT-MG48 and AglycoT-MG59 than clinical-grade Herceptin, ADCC activity was lower than glycosylated IgG antibodies. The activation-to-inhibition ratio of an IgG antibody, which is determined by the discriminative binding affinity to activating FcγR relative to inhibitory FcγRIIb, is crucial for the activation status of immune effector cells.^35,36 It is likely that PBMCs comprise multiple types of immune effector cells, and the lower ADCC activity of aglycosylated Fc variants compared with Herceptin might be derived from the increased binding affinity to FcγRIIb, an inhibitory receptor for downregulating immune cell activation for antitumor activity.^37,38 Further engineering to reduce binding to inhibitory FcγRIIb will generate Fc variants that exhibit improved ADCC and tumor cell-killing activities.

Wild-type aglycosylated trastuzumab and aglycosylated trastuzumab Fc variants showed nearly similar production yields compared with wild-type glycosylated trastuzumab in HEK293F cells without any change in aggregation propensity. Also, mutations identified in this study did not significantly affect the expression level of aglycosylated trastuzumab when we analyzed it using a bacterial display system, suggesting that introducing the identified mutations into the aglycosylated Fc region of IgG may not significantly destabilize it (SI Fig. 5). The clinical analysis of several aglycosylated antibodies has not indicated issues associated with immunogenicity.³⁹ The potential immunogenicity resulting from newly identified mutations were evaluated using in silico MHC II binding prediction, and our results indicated that all the peptide epitopes derived from Fc variants were predicted to have almost equivalent MHC II binding affinity compared with those originating from wild-type human Fc sequence. The properties of the engineered aglycosylated IgG Fc variants, which are capable of eliciting PBMCs-mediated tumor cell-killing ADCC activity, will be assessed in developability, non-clinical and clinical studies in the future.

Materials and methods

Reagents

All restriction endonucleases and Vent polymerase were purchased from New England Biolabs (Ipswich, MA). Taq polymerase and oligonucleotide primers were purchased from Biosesang (Seongnam, Korea) and Cosmogenetech (Seoul, Korea), respectively. The Alexa Fluor 488 Labeling kit and ultra-TMB substrate solution were obtained from Thermo Fisher Scientific Inc. (Waltham, MA). Difco™ Terrific Broth (TB) was from Becton Dickinson Diagnostic Systems (Sparks, MD). Ni-NTA agarose and protein A agarose were obtained from Qiagen (Hilden, Germany) and Genscript (Scotch Plains, NJ), respectively. Human FcγRI was from R&D Systems (Minneapolis, MN; Cat. No. 1257-FC). Horseradish peroxidase (HRP)-conjugated goat anti-GST and anti-His antibodies were from GE Healthcare (Piscataway, NJ; Cat. No. RPN1236) and Sigma-Aldrich (St. Louis, MO; Cat. No. A7058-1VL), respectively. PBMCs were purchased from Cellular Technology Limited (Cleveland, OH). All other analytical and molecular biological grade biochemical reagents were purchased from Sigma-Aldrich (St. Louis, MO), unless stated otherwise.

Construction of plasmids

All plasmids and primers used in this study are given in Tables 2 and 3. Two DNA fragments, which were amplified by PCR using the template pPelB-AgylcoT(H)-Fc1004-FLAG²¹ with two primer sets, MJ#36/MJ#43 and MJ#42/MJ#37, respectively, were assembled to generate full-length heavy chains with three additional mutations (K326I, A327Y, and L328G). These were SfiI digested and then ligated into pPelBFLAG digested with the same restriction endonuclease to generate pPelB-AglycoT(H)-Fc1004-IYG-FLAG. To construct plasmids for expressing trastuzumab IgG Fc variants in mammalian cells, genes encoding each Fc variant, MG48, MG59, MG87, or AIYG, were PCR amplified using the two primers MJ#49 and MJ#50, and the templates derived from the plasmids recovered after library screening or from the pPelB-AglycoT(H)-AIYG-FLAG. Each Fc variant gene was assembled with the gene encoding the VH-CH1 region of trastuzumab and then ligated into the pMAZ-IgH-GlycoT vector using BssHII and XbaI restriction enzyme sites to generate pMAZ-IgH-GlycoT(H)-MG48, pMAZ-IgH-GlycoT(H)-MG59, pMAZ-IgH-GlycoT(H)-MG87, and pMAZ-IgH-GlycoT(H)-AIYG.

Preparation of FcγRs

pMAZ-FcγRIIIa-158V‐FLAG-Streptavidin-His, pMAZ-FcγRIIa-131H/131R-GST, pMAZ-FcγRIIIa-158V/158F-GST, pMAZ-FcγRIIb-GST, and pcDNA3-FcRn-GST, which are the plasmids for the expression of extracellular regions of Fc receptors encoding tetrameric FcγRIIIa‐158V, dimeric FcγRIIa-131H/131R, dimeric FcγRIIIa-158V/158F, dimeric FcγRIIb, and dimeric FcRn, respectively, were prepared using the eCube plasmid DNA midi kit (PhileKorea, Korea) and were transfected into HEK293F cells using polyethyleneimine (PEI) as previously described.^40,41 The transfected cells, cultured in GIBCO FreeStyle™ medium (Invitrogen) for 6 days, were harvested by centrifugation at 2,000 rpm for 10 min. The supernatants were mixed with 40 ml of 25 × phosphate-buffered saline (PBS) per liter of culture and were incubated with 1 ml of Ni-NTA agarose resin slurry at 4°C for 16 h for binding. FcγRs bound to the resin were purified by passing through a polypropylene column (Thermo Fisher Scientific), with subsequent washing with 100 ml of 1 × PBS, 25 ml of 10 mM imidazole buffer, and 25 ml of 20 mM imidazole buffer, followed by elution with 2.5 ml of 250 mM imidazole buffer. The buffer of the purified FcγR was then exchanged using an Amicon Ultra‐4 spin column (Merck Millipore; 3-kDa cutoff). The concentration of purified proteins was obtained by measuring absorbance at 280 nm using a spectrophotometer (BioTek), and the purity was assessed by running the samples on 4%–15% SDS‐PAGE gels (Bio-Rad).

E. coli culture conditions and spheroplast generation

For displaying full-length IgG Fc variants, Jude1 cells that harbored two plasmids, pPelB-AglycoT(H)-Fc variant-FLAG and pBADAglycoT(L)-His,²¹ were cultured at 37°C for 16 h with shaking at 250 rpm in 25 ml of TB containing 2% (wt/vol) glucose, chloramphenicol (40 µg/ml), and kanamycin (50 µg/ml) in a 250-ml Erlenmeyer flask. The overnight cultured cells were diluted 1:100 in 110 ml of fresh TB media supplemented with chloramphenicol (40 µg/ml) and kanamycin (50 µg/ml) in a 2-L Erlenmeyer flask. After incubation at 37°C until the optical density at 600 nm reached 0.6, the cells were cooled at 25°C for 20 min and induced with 1 mM of isopropyl-1-thio-D-galactopyranoside and 0.2% arabinose for protein expression. After incubating the cells at 25°C for 20 h, the cells were pelleted by centrifugation at 14,000 rpm and washed twice with 1 ml of cold 10 mM Tris‐HCl (pH 8.0). The outer membrane was removed by resuspending in 1 ml of cold STE solution (0.5 M sucrose, 10 mM Tris‐HCl, 10 mM EDTA; pH 8.0) and incubating at 37°C for 30 min on a rotator. The cells, pelleted by centrifugation at 14,000 rpm for 1 min, were washed in 1 ml of cold Solution A (0.5 M sucrose, 20 mM MgCl₂, 10 mM MOPS; pH 6.8) and incubated in 1 ml of Solution A with 1 mg/ml of chicken egg lysozyme at 37°C for 15 min. After centrifugation at 14,000 rpm for 1 min, spheroplasts were recovered in the pellet.

Library construction

The pMopac12-PelB-VH-CH1-CH2-CH3 (Fc1004-IYG)-FLAG plasmid was used as a template for constructing an aglycosylated Fc randomized library. The VH-CH1 region was amplified by PCR with two primers, MJ#36 and MJ#44, using Vent polymerase. Random mutations were then introduced in the hinge-CH2-CH3 region with a 0.5% target error rate using primers MJ#45 and MJ#46 based on the standard error-prone PCR technique.⁴² Heavy chains of trastuzumab Fc variants were generated through assembly PCR with two fragments (VH-CH1 and hinge-CH2-CH3) and two primers (MJ#36, MJ#46). Randomized heavy chain genes digested with SfiI were ligated into the pPelBFLAG vector, and the resulting plasmids were transformed into Jude1 cells that harbored the light chain plasmid pBADAglycoT(L)-His.

Library screening using flow cytometry

One mg of tetrameric FcγRIIIa‐158V was labeled with Alexa Fluor 488 fluorescent dye using an Alexa Fluor 488 Protein Labeling Kit (Thermo Fisher Scientific). Diluted spheroplasts (3:7) in cold PBS were labeled with 30 nM of tetrameric FcγRIIIa-158V-Alexa Fluor 488 by incubating at room temperature for 1 h, pelleting by centrifugation at 14,000 rpm for 1 min, and washing with 1 ml of cold PBS. Then, 1 ml of fluorescent probe-labeled spheroplasts was diluted in 20 ml of PBS and passed through a cell strainer to remove aggregates (BD Bioscience). Spheroplasts that exhibited the highest 3% of fluorescence intensity were sorted using an S3 cell sorter (Bio-Rad) and immediately re-sorted to enhance sorting efficiency. From sorted spheroplasts, the genes encoding IgG heavy chain Fc variants were rescued by PCR amplification using two primers (MJ#36 and MJ#2). The amplified IgG heavy chain Fc variant genes, digested with SfiI restriction endonuclease, were ligated into the pPelBFLAG vector¹⁹ and were transformed into Jude1 cells that harbored the pBADAglycoT(L)-His plasmid. In the next round of screening, spheroplasts were labeled with 30 nM (for second and third rounds) or 10 nM (fourth and fifth rounds) of tetrameric FcγRIIIa-158V-Alexa Fluor 488 and were sorted in the same way as the first round of flow cytometry screening.

Expression and purification of aglycosylated trastuzumab Fc variants

For transient expression of aglycosylated trastuzumab, Fc variants (AIYG, MG59, MG87, and MG48) in HEK293F cells and two plasmids, pMAZ‐IgL and pMAZ‐IgH, were transfected into HEK293F cells using PEI as a transfection agent, and then cultured in GIBCO FreeStyle™ medium. After culture for 6 days, the cells were harvested by centrifugation at 2,000 rpm for 10 min, and the recovered supernatants were mixed with 40 ml of 25 × PBS per liter of culture, filtered through a 0.2-μm bottle-top filter, and bound to 1 ml of Protein A agarose (Genscript) by incubating at 4°C for 16 h. After loading into the polypropylene column and washing twice with 10 ml of 1 × PBS, purified trastuzumab Fc variants were eluted using 3 ml of 100 mM glycine‐HCl (pH 2.7) and were immediately neutralized by collecting the eluate in a tube containing 1 ml of 1 M Tris (pH 8.0). After exchanging the buffer with 1 × PBS and concentrating using Amicon Ultra‐4 spin columns with a 3-kDa cutoff, the purity of trastuzumab Fc variants was evaluated by analyzing on 4%–15% resolving SDS‐PAGE gels.

ELISA

For ELISA, 50 μl of 4 μg/ml IgG antibodies, GlycoT, aglycosylated trastuzumab Fc variants (AglycoT-AIYG, AglycoT-MG48, AglycoT-MG59, and AglycoT-MG87), or clinical-grade Herceptin purchased from Roche were diluted in 0.05 M Na₂CO₃ (pH 9.6) and were coated onto flat-bottomed polystyrene high-bind 96-well microplates (Costar) by incubating at 4°C for 16 h. Each well was blocked by adding 100 μl of a blocking solution, 5% bovine serum albumin, PBST (1 × PBS, 0.05% Tween20) and was incubated at room temperature for 2 h. The plates were washed four times with 180 μl of PBST and then 50 μl of the following FcγRs serially diluted in the blocking solution were added: monomeric FcγRI-His (R&D Systems), dimeric FcγRs (FcγRIIa-131H-GST, FcγRIIa-131R-GST, FcγRIIIa-158V-GST, FcγRIIIa-158F-GST, FcγRIIb-GST, and FcRn-GST), or tetrameric FcγRIIIa-158V Alexa Fluor 488 conjugate. After 1-h incubation at room temperature and four washes, 50 μl of the following antibodies were added: mouse anti-polyhistidine IgG HRP conjugate (1:10,000; Sigma-Aldrich) for detecting FcγRI-His and tetrameric FcγRIIIa-158V Alexa Fluor 488 conjugate or goat anti-GST IgG HRP conjugate (1:5,000; GE Healthcare) for detecting FcγRIIa-131H-GST, FcγRIIa-131R-GST, FcγRIIIa-158V-GST, FcγRIIIa-158F-GST, FcγRIIb-GST, and FcRn-GST. After washing the plate with PBST and adding 50 μl of 1-Step ultra-TMB-ELISA substrate solution (Thermo Fisher Scientific), the ELISA signals were analyzed using an Epoch microplate spectrophotometer (BioTek) by measuring the absorbance at 450 nm after adding 50 μl of 2 M H₂SO₄ to each well for quenching the colorimetric reaction.

Biacore analysis

Surface plasmon resonance was performed using a Biacore T200 instrument (GE Healthcare). Herceptin, Glyco-T, Aglyco-T, AglycoT-AIYG, AglycoT-MG48, and AglycoT-MG59 were directly immobilized onto individual cells of the CM5 sensor chips (GE Healthcare) using an amine coupling kit (GE Healthcare). Dimeric FcγRIIIa-158V-GST or dimeric FcγRIIIa-158F-GST were injected over the variant IgG-coated sensor chips at a flow rate of 30 μl/min for 1 min and dissociated for 2 min with HBS-EP running buffer (GE Healthcare). The kinetic assays were performed by 2-fold serial dilution starting from 0.5 μM and 2 μM for dimeric FcγRIIIa-158V-GST and dimeric FcγRIIIa-158F-GST, respectively. Regeneration was achieved with a sequential injection of 50 mM glycine (pH 3.9), 50 mM glycine (pH 9.5), and 3 M NaCl for 2 min each. Dissociation constants (K_D) for monovalent receptor binding were calculated by fitting a 2:1 bivalent analyte model using the Biacore T200 evaluation software 3.0 (GE Healthcare), as previously described.²¹

ADCC assay using human PBMCs

The V-bottom 96-well plates (Corning) were coated with 1 × 10⁴ SKBR-3 or MCF-7 target cells/50 µl. A 10 µl aliquot of each IgG antibody, namely AglycoT-MG48, AglycoT-MG59, AglycoT-N297D, GlycoT, clinical-grade Herceptin, or normal human serum IgG (Sigma-Aldrich, Cat. No. 12511), was serially diluted and added at 0, 0.032, 0.16, 0.8, 4, and 20 µg/ml to the target cell-immobilized plate. Then, 50 µl of human PBMCs (2.5 × 10⁵ cells) derived from five individual donors (Cellular Technology Limited) (SI Table 1) were added, centrifuged at 100x g for 1 min, and incubated in a CO₂ incubator at 37°C for 4 h. After centrifuging the plate at 300 × g for 3 min, 50 µl of supernatant from each well was transferred to a Spectraplate 96-well plate (PerkinElmer) and then 50 µl of Cytotox 96® Reagent (Promega) was added. The plate was further incubated at room temperature for 30 min, and cytotoxicity was then measured by reading the absorbance at 490 nm after adding 50 µl of the stop solution. The experiments were repeated three times in duplicate, and the average percentage of cytotoxicity was calculated.

In silico prediction of MHC II epitopes of isolated Fc variants

For the prediction of potential T-cell epitopes in Fc sequences of trastuzumab (wild-type human IgG1), AglycoT-MG48, and AglycoT-MG59, binding to the HLA-DRB1*0401 as MHC II allele, which covers nearly 95% of human populations worldwide, was assessed using the consensus prediction method through the Immune Epitope Database (IEDB).^29,30 Fc sequences consisting of 227 amino acids were fragmented into a series of 15-amino acid peptides that can be presented as an antigenic peptide by MHC II, and the predicted binding to the MHC II was scored as a CPR, with a lower CPR score indicating a higher binding affinity to MHC II. In addition, MHC II binding for core peptide regions comprising 9 amino acids, which were derived from the 15 amino acid peptides, were investigated. Each core peptide had IC₅₀ value for the average relative binding to MHC II, in which IC₅₀ values of <50 nM, <500 nM, and < 500 nM were considered high affinity, intermediate affinity, and low affinity, respectively.

Abbreviations

ADCC

antibody-dependent cell-mediated cytotoxicity

ADCP

antibody-dependent cell-mediated phagocytosis

AglycoT

aglycosylated trastuzumab

CDC

complement-dependent cytotoxicity

ELISA

enzyme-linked immunosorbent assay

FACS

fluorescence-activated cell sorting

Fc

fragment crystallizable

FcγR

Fc gamma receptor

FcRn

neonatal Fc receptor

FDA

Food and Drug Administration

GlycoT

glycosylated trastuzumab

HEK293F

human embryonic kidney 293F

LDH

lactate dehydrogenase

MHC

Major histocompatibility complex

NK cells

natural killer cells

PBMC

peripheral blood mononuclear cell

Disclosure of potential conflicts of interest

No potential conflicts of interest were disclosed.

Acknowledgments

This work was supported by a grant from the Basic Science Research Program (2016R1C1B2007434), Bio & Medical Technology Development Program (2017M3A9C8060552), and the Pioneer Research Center Program (2014M3C1A3051460) through the National Research Foundation of Korea (NRF) funded by the Ministry of Science, ICT and Future Planning, and the National R&D Program for Cancer Control, Ministry of Health and Welfare, Republic of Korea (1420160).

Funding

Ministry of Science, ICT and Future Planning, 2016R1C1B2007434, Ministry of Science, ICT and Future Planning, 2017M3A9C8060552, Ministry of Science, ICT and Future Planning, 2014M3C1A3051460, Ministry of Health and Welfare, 1420160.