Cell-based glycoengineering for production of homogeneous and specific glycoform-enriched antibodies with improved effector functions

Significance

The efficacy of therapeutic antibodies is determined by two activities, one is to capture the target through specific antigen recognition and the other is to recruit immune cells to the target site and kill the target through effector functions. The type of immune cells recruited by the antibody to exhibit optimal effector functions is determined by the interaction of the Fc moiety with the Fc receptors on immune cells, and this interaction is influenced by the glycosylation of the Fc moiety. In this study, we established a cell-based method through glycosylation pathway engineering to produce antibodies with enriched Fc-α2,6-sialyl complex type (Fc-SCT) glycan that exhibit optimal binding to the Fc receptors associated with the effector functions in target killing.

Abstract

Glycosylation of humanized antibody at Fc-Asn297 significantly affects the Fc-mediated killing of target cells through effector functions, especially antibody-dependent cellular cytotoxicity (ADCC), antibody-dependent cell-mediated phagocytosis (ADCP), and antibody-dependent vaccinal effect (ADVE). Previous studies showed that therapeutic immunoglobulin G (IgG) antibodies with α2,6-sialyl complex type (SCT) glycan attached to Fc-Asn297 exhibited optimal binding to the Fc receptors on effector cells associated with ADCC, ADCP, and ADVE. However, the production of antibodies with homogeneous Fc-SCT glycan requires multiple in vitro enzymatic and purification steps. In this study, we report two cell-based methods to produce Fc-GlcNAc antibody and Fc-SCT-enriched antibodies with improved effector functions. First, we expressed endoglycosidase S2 in Expi293F GnT1- cells to trim all N-glycans to Fc-GlcNAc antibody for in vitro transglycosylation to generate homogeneous antibodies with well-defined Fc glycan. Second, we engineered the glycosylation pathway of HEK293T cells through knock-out of undesired glycosyltransferases and knock-in of desired glycosyltransferases to produce Fc-SCT-enriched antibodies and evaluated their binding to Fc receptors, and we found that the Fc-SCT-enriched antibody is like or better than the homogeneous Fc-SCT antibody in binding to the Fc receptors associated with ADCC, ADCP, and ADVE.

Antibodies play a significant role in humoral immune response and have diverse plasticity and biological functions. The prototype of antibody is a major component of B cell receptor (BCR) on naive B cells. When the BCRs are activated through pattern recognition (PR), the alternative splicing of messenger RNA (mRNA) helps produce antibodies as secreted forms of immunoglobulin M or immunoglobulin D. With the assistance of dendritic cells and T helper cells, the activated B cells could further undergo clonal expansion, somatic hypermutation, and class switch and are differentiated into plasma cells to produce secreted immunoglobulin G (IgG), immunoglobulin A, or immunoglobulin E antibodies with high antigen specificity. Once a target cell or pathogen is opsonized by an antibody, it triggers different downstream effector functions, including phagocytosis by macrophages, cellular cytotoxicity by nature killer (NK) cells, vaccinal effect by dendritic cells, etc. The activation of these effector functions required the binding of the antibody Fc domain to specific Fc receptors on immune cells. Each subtype or glycoform of antibody can be recognized by several Fc receptors rendering activating or inhibitory signaling to generate different effector functions. For example, the binding of IgG antibody to the FcγIIIA receptor on NK cells will activate NK cells to release cytokines to kill target cells, a process called antibody-dependent cellular cytotoxicity (ADCC) (1–3), and binding to the FcγIIA receptor on macrophages will activate macrophage to kill the target cell through phagocytosis, also called antibody-dependent cell-mediated phagocytosis (ADCP). In addition, binding of antibody–antigen complex to the FcγIIA receptor on dendritic cells can facilitate dendritic cells to engulf the complex, and through processing and maturation to induce memory CD8 T cell response (4).

Glycosylation is a common post- or cotranslational modification found on most cell surface proteins. During N-glycoprotein synthesis, the dolichol-linked high mannose glycan is first synthesized in the ER and then transferred to the growing polypeptide chain followed by chaperone-mediated protein folding and quality control to form glycoproteins. The high-mannose glycoproteins are trimmed by mannosidases and translocated to the Golgi apparatus for sequential addition of more monosaccharides by different glycosyltransferases before moving to the cell surface. Unlike protein and DNA, glycosylation is template independent, and the glycan sequence is determined by the repertoire of glycosyltransferase and glycosidases expressed in different cell types. Moreover, the cellular environment and protein accessibility also affect the activity and specificity of glycosyltransferases and glycosidases. Thus, the glycans generated by glycosylation and deglycosylation usually exhibit more diversity than other modifications (5–7).

Antibodies of each subtype are also glycosylated at different places within the constant region of the heavy chain (HC), and it is known that the glycosylation would regulate antibody functions (1). For example, the preBCR assembly is important for B cell development and is regulated by the N-glycan at N46 on μHC (8). The N-glycan at N402 on μHC has been linked to antibody oligomerization and complement activation (9, 10). Besides, IgG N-glycosylation at Fc-N297 on μHC plays a critical role in complement activation and Fcγ receptor activation, leading to various effector functions (11–13). Therefore, it is important to identify optimized glycosylation to improve the efficacy of therapeutic antibodies.

Generally, the most frequently used mammalian cell lines for recombinant antibody production are CHO, HEK293T, and NSO cells. However, antibody production using these cell lines often gives mixtures of heterogeneous glycoforms. The glycoforms with core-fucosylated biantennary complex type glycans (14) are not optimal for binding to the FcγIIIA receptor because of the inhibitory function of core-fucosylation (11), and those with terminal galactosylation may cause off-target delivery (15). Therefore, modification of the Fc glycan on antibody along with protein sequence optimization to improve specific receptor binding and the corresponding effector function is important for development of therapeutic antibodies. One strategy used previously is to block the core fucosylation pathway using the cell line deficient in GDP-fucose synthesis to produce the antibody with an increase in binding to FcγIIIA receptor thereby enhancing the ADCC activity (11, 16). Other cell-based approaches include use of precise gene editing tools such as Zinc Finger Nuclease or CRISPR/CRISPR-associated protein 9 (Cas9) to knock-out the core fucosylation (17) and alter the glycosylation pathway in cells to produce therapeutic antibodies with improved efficacy (18, 19). Another strategy to alter protein glycosylation is through in vitro glycoengineering by enzymes or chemical ligation. A common approach was endoglycosidase-mediated transglycosylation (7). Usually, the antibody with high mannose type glycan derived from insect cells (20), yeast (21), or GNT1 deficient cell line (Expi293F-GnT1-) (22–24) was treated with endoglycosidase H or S2 to generate mono-GlcNAc- or GlcNAc-Fuc-bearing glycoform. The antibody with either one of these two glycoforms can be a good acceptor substrate for glycan remodeling (7) to make homogeneous antibody glycoforms via transglycosylation using a mutated endoglycosidase as synthase and a glycan oxazoline as donor substrate to generate a glycan chain. So far, many glycans have been transferred to antibodies to study the functional impact on receptor binding and effector functions. It is known that the fully galactosylated complex type glycan without core fucose contributes significantly to improve antibody binding to the FcγIIIA compared to the wild-type (WT) antibody (22, 23). However, it reduces binding to FcγIIA and facilitates the antibody clearance by liver cells through asialoglycoprotein receptor binding. Our further studies revealed that the α2,6 linked sialyl complex type (SCT) biantennary glycan provides enhanced binding to FcγRIIIA and FcγRIIA receptors which are associated with ADCC, ADCP, and antibody-dependent vaccinal effect (ADVE) (22–24). Although this strategy could produce homogeneous antibody glycoforms, the process may not be cost-effective as it requires a separate synthesis of glycans, multiple enzymatic reactions, and purification steps. In addition, it is not clear whether the homogeneous Fc-SCT antibody exhibits optimal binding to all FcRs and activity in vivo. In this report (Fig. 1), we used humanized antibody chMC813-70 (with human sequence in the heavy chain to minimize immunogenicity) that targets specific stage embryonic antigen 4 (SSEA4) on various cancer cells as a model to produce and evaluate the homogeneous Fc-SCT, Fc-SCT-enriched, and WT antibody glycoforms in binding to activating and inhibitory Fc receptors and their activity of cellular cytotoxicity.

Fig. 1.

Illustration of CRISPR/Cas9 KO/KI strategies for glycosylation pathway engineering. The Fc-SCT-enriched antibody was generated by sequential knock-in of hB4GalT1 and hST6Gal1 into the target gene locus of HEK293T cells. The Fc-GlcNAc- antibody was generated by either knock-in of Endo H or Endo S2 into the target gene locus of Expi293F-GnT1- cells.

Results

Development of Fc-GlcNAc Antibody–Producing Cells by CRISPR-Cas9.

Antibodies are generally produced from mammalian cell lines which contain core-fucosylated complex type glycans (14). To generate the Fc-GlcNAc antibody for in vitro transglycosylation, antibody glycans need to be trimmed by endoglycosidase and fucosidase to generate a homogeneous antibody glycoform containing only the innermost GlcNAc in the Fc domain as an acceptor (7). We aimed to shorten these steps by CRISPR-Cas9 to generate Fc-GlcNAc antibody using the N-acetylglucosaminyltransferase 1 (GnT1) deficient cell line (Expi293F-GnT1-). The N-glycans expressed by this cell line are high mannose types which can be preferentially cleaved by Endo S2 or Endo H to generate Fc-GlcNAc antibody. Thus, we knocked in Endo S2 or Endo H in this cell line by CRISPR-Cas9 and found that the Endo S2 knock-in (KI) cells grew well and were more stable than the Endo H KI cells (SI Appendix, Fig. S1). The antibodies expressed by these two KI cells were analyzed by electrophoresis and the result showed that the antibody heavy chain (AbHC) expressed by these cells had faster mobility than AbHC from its parental cells (Expi293F-GnT1-) and most AbHCs were downshifted (Fig. 2A). We then analyzed the glycans on this antibody by intact protein mass analysis, and the results showed that the antibody contained only GlcNAc in the Fc domain (Fig. 2B), though a minor unknown product was observed in the antibody produced from Endo H KI cells (SI Appendix, Fig. S2). This antibody was confirmed to be Fc-GlcNAc antibody (22, 23) which was subjected to transglycosylation with high efficiency (Fig. 2C). These results indicate that CRISPR-Cas9-mediated Endo S2 KI Expi293F-GnT1- cells can produce homogeneous Fc-GlcNAc antibody and is an efficient platform for further application.

Fig. 2.

Expression of Endo S2 in Expi293F-GnT1- cells results in cleavage of antibody N-glycans to GlcNAc. (A) The chMC813-70 antibody glycoforms expressed by engineered and parental Expi293F-GNT1- cells were subjected to electrophoresis and a band shifting was observed. (B) The chMC813-70 antibody expressed by engineered Expi293F-GnT1- cells was subjected to intact protein mass analysis to examine the molecular weight and glycan changes. (C) The chMC813-70 antibody expressed by engineered Expi293F-GnT1- cells was subjected to transglycosylation, and the efficiency was evaluated by protein electrophoresis.

Glycosylation Pathway Engineering in Cells to Produce Fc-SCT-Enriched Antibodies by CRISPR-Cas9.

Another approach to generate antibody with enriched Fc-SCT glycoform was to directly edit the expression of glycosyltransferases in cells. As previously described, the antibody produced from commonly used cell lines, like CHO and HEK293T cells, EK 293KHusually contained biantennary complex type N-glycans with core-fucose and terminated mainly with N-acetylglucosamine and galactose (14, 18, 19, 25). Therefore, in order to produce Fc-SCT enriched antibody, we decided to increase the level of galactosylation and 2,6-sialylation on the antibody (19, 25). In addition, the core fucosylation, which reduced binding to FcγRIIIA and depleted ADCC (11), was eliminated by knocking out FUT8 (26). Furthermore, terminal sialylation with α2,6 linkage, not the α2,3 linkage, was introduced as it was found to enhance receptor binding (27) and improve ADCC, ADCP, and ADVE (23, 24, 28). Although the mRNA of human α2,3-sialyltransferases could be detected in HEK293T and CHO cells (18, 29), we did not observe any obvious sialylation on the antibody glycans from these cell lines (18, 19, 25) (Fig. 3, Upper). Therefore, we are not concerned about the α2,3-sialylation when we introduce the α2,6-sialyltransferases for sialylation. To establish the Fc-SCT enriched cell line, we sequentially knocked out hFUT8 gene and knocked in human β-1,4 galactosyltransferase (hB4GalT1) and human α-2,6 sialyltransferase (hST6Gal1) in HEK293T cells through CRISPR-Cas9 (Fig. 1) and this engineered cell line was evaluated for growth, stability, and production of desired products. The antibody expressed by this engineered cell line was purified by protein A and subjected to intact protein mass analysis. We observed a significant increase of galactosylated and sialylated antibody glycoforms (Fig. 3, Lower) compared to the antibody generated by WT cells (Fig. 3, Upper). The glycoform analysis also revealed a significant increase in antibody galactosylation and sialylation, with a lack of core-fucosylation. The content of α2,6-linked Fc-SCT antibody is approximately 60% in the mixture of glycoforms after purification by protein A (SI Appendix, Figs. S3 and S4) based on the intact protein mass analysis and lectin assay that recognized the α2,6-linked sialosides. This mixture of Fc-SCT-enriched glycoforms was used for assessment of Fc receptor binding and ADCC activity and the results were compared with that from the WT and the homogeneous Fc-SCT antibody.

Fig. 3.

The glycoform composition of chMC813-70 antibodies expressed from WT and glycoengineered cells. The antibodies expressed in WT (Upper) and glycosylation pathway–engineered HEK293T cells (Lower) with hFUT8 KO and hB4GalT1/hST6Gal1 KI were purified by protein A, and the N-glycans were analyzed by intact protein mass analysis.

The Sialylation on Antibody Is α2,6-Linkage.

To confirm the sialylation on the antibody was mediated by hST6Gal1, the antibody was treated with two different sialidases, Streptococcus pneumonia α2,3 neuraminidase and Clostridium perfringens neuraminidase respectively. After overnight reaction, these antibodies did not show any obvious difference in the distribution of sialylation between the untreated and α2,3 neuraminidase-treated groups (Fig. 4, Upper and Middle), indicating that the sialylation of antibody was not due to the endogenous α2,3 sialyltransferases. On the other hand, using neuraminidase which could remove all linkages of sialylation largely eliminated the terminal sialic acids. In addition, the antibody with increased galactosylation generated from cells with knock in of hB4GalT1 only, the terminal sialylation was still not observed (SI Appendix, Fig. S5). These results confirmed that the sialylation on the antibody was introduced by the hST6Gal1 and the increase of hB4Gal1 expression indeed elevated the galactosylation of antibody (Fig. 3). To further explore the capacity of sialylation or galactosylation in cells, we transiently overexpressed hST6Gal1 or hB4GalT1 in the engineered cells followed by antibody production to see whether the galactosylated or sialylated antibody was increased additionally. However, we did not observe further elevation of any galactosylated or sialylated antibody after transient overexpression of hST6Gal1 or hB4GalT1 (SI Appendix, Fig. S6).

Fig. 4.

The sialylation on SCT-enriched antibody is α2,6 linkage. The SCT-enriched chMC813-70 antibody was expressed in glycosylation pathway–engineered cells. After incubation with indicated enzyme overnight, the antibody was purified by protein A and analyzed by intact protein mass analysis.

Protein Glycosylation Varies with Expression Time.

In order to understand whether the amount and time of antibody overexpression in cells affected glycosylation, we extended the culture time to 5 d and then collected the medium at day 3 and day 5 to examine the antibody glycans by intact protein mass analysis. We unexpectedly found that the glycan pattern was different between these two time points (SI Appendix, Fig. S7A). In the Intact protein mass analysis, a relatively strong sialylation was observed in the first collection (first 3 d) (SI Appendix, Fig. S7A, D0-3), and then, sialylation was decreased in the second collection (last 2 d), accompanied by an increase of terminal galactosylation (SI Appendix, Fig. S7A, D4-5). This result indicated that the proteins and enzymes involved in glycosylation may not be sufficient to effectively produce Fc-SCT antibody at later stage. Because the hST6Gal1 and hB4GalT1 involved in Fc-SCT antibody production were nonendogenous and constitutively expressed in the cells, we examined whether endogenous glycosyltransferases also faced the same problem of reduced glycosylation. By expressing the antibody in FUT8 knock-out (KO) cells, the antibody was almost terminally glycosylated with at least one or more galactoses in the first collection (SI Appendix, Fig. S7B, D0-3), but the proportion of galactosylated antibody decreased and even the antibody glycans without any galactoses became the dominant glycoforms in the last 2 d (SI Appendix, Fig. S7B, D4-5). The result confirmed that short glycans are observed on endogenous glycosylation after protein overexpression. This finding promoted us to analyze the glycosylation pattern impacted by antibody production at various time points. We purified antibody every day after plasmid transfection and analyzed the dynamic change of antibody glycans from multiple engineered cells. To our surprise, the antibody with full sialylation was significantly detected by intact protein mass analysis at the first day (Fig. 5) and then gradually decreased daily (Fig. 5). Regarding galactosylation, the antibody glycans were capped by at least two galactoses (2 ~ 4 galactoses in total) on the first day, but started to decline on the second day, as evidenced by the appearance of one galactose on antibody glycans (Fig. 5). These observations suggest that antibody glycosylation was almost complete at the beginning of overexpression and gradually decreased over time. Thus, cell supernatants can be purified within 3 d after transfection to obtain Fc-SCT-enriched antibodies with optimized sialylated glycans.

Fig. 5.

Antibody expression and time-dependent changes of glycans on antibody. The chMC813-70 antibody expressed in glycosylation pathway–engineered cells was collected every day and purified by protein A for intact protein mass analysis.

Fc Receptors’ Binding of WT, Homogeneous Fc-GlcNAc Antibody, and Fc-SCT-Enriched Antibodies.

As we can obtain Fc-SCT-enriched antibody from glycosylation pathway–engineered cells, the next step is to compare the binding of this antibody to Fc receptors with homogeneous and WT antibodies (3). FcγIIA and the inhibitory FcγIIB are usually expressed simultaneously by the antigen-presenting cells, such as dendritic cells and macrophages, and these two receptors work together to modulate immune response, especially ADCP and ADVE (3). FcγIIIA and the inhibitory FcγIIIB are critical for immune cells, especially NK cells to mediate ADCC (3). Therefore, the binding of antibody to these receptors was evaluated by the ELISA to determine the impact of their glycan composition. We did not study the binding to FcγIA here because the glycoforms of complex type N-glycans on antibody would not affect the binding (23, 24). We observed increased binding of Fc-SCT-enriched antibodies to FcRs to the tested Fc receptors (Fig. 6) and the increased binding to FcγIIA and FcγIIIA was consistent with previous study (22–24) (Fig. 6 A and B). The WT antibody showed very weak binding to inhibitory FcγIIB while the Fc-SCT-enriched antibody showed a better binding than the homogeneous Fc-SCT antibody to FcγIIA which is associated with ADCP and ADVE. Although the Fc-SCT-enriched antibodies showed binding to the inhibitory FcγIIB and FcγIIIB at higher concentrations, the binding was much weaker than the binding to FcγIIA and FcγIIIA and thus the inhibitory effect is insignificant in the normal dose of antibody used in therapy (Fig. 6 C and D).

Fig. 6.

Improved receptor binding of Fc-SCT-enriched antibody compared to homogeneous and WT antibodies. The Fcγ receptor-coated plate (A), FcγIIA; (B), FcγIIIA; (C), FcγIIB; (D), FcγIIIB was incubated with humanized WT antibody (black), homogeneous SCT antibody (blue), or Fc-SCT-enriched antibody (red) at indicated concentrations. The binding of the antibody was measured by anti-human IgG Fc antibody conjugated with HRP. The binding of Fc-SCT-enriched antibody to receptors was greatly improved compared with WT and homogeneous Fc-SCT-humanized antibodies.

The Fc-SCT-Enriched Antibody and Homogeneous Fc-SCT Antibody Are Similar in Functions.

Since the binding of Fc-SCT-enriched antibody to FcγIIIA was significantly elevated, we compared the binding with that of the homogeneous Fc-SCT antibody. The result showed that Fc-SCT-enriched antibodies exhibited similar binding avidity to homogeneous Fc-SCT antibody toward the activating and inhibitory Fc receptors, though the Fc-SCT-enriched antibodies exhibited slightly better binding to FcγIIA (Fig. 6), perhaps due to the presence of other glycoforms. The increase in binding to FcγIIA may have important implication in cancer therapy as macrophages have better access to the tumor site than NK cells, i.e., ADCP is more important than ADCC in cancer therapy. Then we further evaluated whether the Fc-SCT-enriched antibody could evoke a similar effect as homogeneous Fc-SCT antibody in ADCC reporter assay. At the lowest antibody concentration used for induction, WT antibody did not induce any obvious cell activation, but the homogeneous Fc-SCT and Fc-SCT-enriched antibodies can trigger cell activation rapidly and reached the maximum at next concentration (Fig. 7). Though WT antibody can reach the maximum induction at higher concentrations, the activity was still much lower than that of homogeneous Fc-SCT or Fc-SCT-enriched antibody (Fig. 6).

Fig. 7.

Functional comparison of homogenous Fc-SCT-, Fc-SCT-enriched, and WT antibodies. Indicated antibody, target cells, and effector cells were coincubated for 6 h followed by detection of luciferase activity. Fc-SCT-enriched antibody and homogenous Fc-SCT antibody showed similar effects on induction of ADCC reporter response, and both elicited higher response than WT antibody.

Discussion

In this work, we simplified the process to produce Fc-SCT-enriched antibodies with desired effector functions using a glycoengineered cell line. First, we demonstrated that endoglycosidase can be expressed in GnT1-deficient Expi293F-GnTI- cells. The cells were transfected with plasmids bearing the Endo S2 gene and overexpressed Endo S2 in vivo to cleave the high mannose type N-glycans to Fc-GlcNAc antibody which can be applied for in vitro enzymatic transglycosylation. We also found that Endo S2 is more efficient and specific than Endo H in cleavage of high mannose antibody glycoforms expressed in Expi293F-GnT1- cells.

In a second approach, we established a glycoengineered cell line that can produce Fc-SCT-enriched antibodies directly and showed high binding activity to Fc receptors compared to WT antibody. The experiments were performed by knocking out the core-fucosylation and constitutively expressed the glycosytransferases required to produce antibodies with enriched Fc-SCT glycoforms. It is noted that a previous attempt to increase the sialylation by elevating the expression of glycosyltransferases in CHO cells showed incomplete sialylated glycans (25). Human hST6Gal1 is known as the only glycosyltransferase which mediates α2,6 sialylation and preferentially uses the α1,3 branch arm of the biantennary glycan as substrate (30). The in vitro studies showed that glycoengineering of antibody by hST6Gal1 is difficult to yield fully sialylated antibody glycans (27, 30) because if the galactose on G1 glycan was not on the α1,3 branch arm, it would not be sialylated by hST6Gal1. Therefore, the expression level of glycosyltransferases is not the bottleneck in this situation, rather the substrate preference and enzyme activity may be the key factors (5–7). Thus, it was suggested that using the α2,6 sialyltransferases from other organisms with a better activity than hST6Gal1 may improve the glycoengineering process (31). This observation was aligned with the fact that enhancement of α2,6-sialylation was the key factor in intravenous Ig (IVIg)-mediated inflammatory suppression (32–34). Although the increased binding to the two opposing functional receptors, FcγIIA and FcγIIB, appears to be conflicting, the eventual immune response was determined by the expression level and overall signaling strength of FcRs on effector cells. We found that the Fc-SCT enriched antibody exhibited significantly better binding to FcγIIIA and FcγIIA than to FcγIIIB and FcγIIB and also displayed higher binding activity than WT antibody toward the Fc receptors in the comparative study. This enhancement is mainly contributed by removal of core-fucose to enhance binding to FcγIIIA and addition of α2,6-linked sialic acid residues to enhance antibody stability and binding to FcγIIA (35). The increased binding of Fc-SCT-enriched antibody to FcγIIA compared to the homogeneous Fc-SCT antibody is interesting and remains to be further investigated. In addition, further comparative study on the binding of homogeneous and Fc-SCT-enriched antibodies to different Fc receptor–bearing immune cells and in vivo efficacy will be conducted to establish a correlation between the in vitro and in vivo activities and provide a better understanding of antibody action.

In conclusion, since the Fc-SCT-enriched antibodies exhibited similar or better FcR binding profiles and Fc-mediated killing compared to the homogeneous Fc-SCT antibody, the production of Fc-SCT-enriched antibodies from glycosylation pathway–engineered cells is apparently more cost-effective as the cell-based method would avoid the high-cost glycan synthesis and multiple enzymatic reactions as well as purification steps in vitro and is compatible with the current antibody manufacturing process. Endo S2 and Endo H can be introduced to cells by CRISPR-Cas9 to process the high mannose N-glycans to Fc-GlcNAc antibody for in vitro transglycosylation. Endo S2 is more efficient and has a broader substrate specificity than Endo H as it will also accept complex type, hybrid type, and bisecting N-glycans as substrates (36, 37). In addition, this cell-based method can be used to produce other mono-GlcNAc-glycoproteins, such as influenza hemagglutinin and SARS-CoV-2 spike protein as vaccines to elicit broadly protective immune responses against the highly conserved epitopes shielded by glycans (38, 39). We believe that the cellular manufacture methods demonstrated in this, and preliminary studies (40) should simplify the production of therapeutic antibodies with improved effector functions, especially ADCC, ADCP, and ADVE.

Materials and Methods

Cell Culture.

HEK293T cells (ATCC) were cultured in the DMEM (Thermo Fisher) supplied with 10% FBS (Thermo Fisher) at 37 °C with 5% CO₂ in the cell incubator. Expi293F-GnT1- cells (Thermo Fisher) were cultured in the Expi293 expression medium (Thermo Fisher) at 37 °C with 8% CO₂ in the cell incubator.

Plasmid.

The plasmids for CRISPR/Cas9 expression were constructed as manual (Thermo Fisher). Briefly, the validated sgRNA sequence targeting glycosyltransferase was synthesized (Integrated DNA Technologies) and cloned into the vector for Cas9 and sgRNA expression. The synthetic codon-optimized gene insert (Integrated DNA Technologies), flanked by the homologous arm of the target gene at the sgRNA target site, was cloned into empty vector as the donor plasmid.

Antibody Production.

The plasmids for antibody chMC813-70 were transfected into cells followed by incubation. Transfection of HEK293T cells and Expi293F-GnT1- cells were mediated by TransIT-293 (Mirus Bio) or expi293fectamine (Thermo Fisher) following the reagent manual. At the time of harvest, the culture medium was collected and then subjected to protein A sepharose beads (GE HealthCare) column to purify the antibody.

Protein Gel Analysis.

The proteins were heated at 95 °C for 5 min in the LDS sample buffer supplied with 2-mercaptoethanol and then subjected to gel electrophoresis with 12% SDS-PAGE. The protein on gel was stained by Coomassie blue buffer (ApexBio).

Enzymatic Transglycosylation Assay.

The antibody was incubated with SCT glycan-oxazoline and Endo S2 mutant in Tris buffer at 37 °C for the time indicated before harvest for purification or gel electrophoresis (22, 23).

Homogeneous Fc-SCT Antibody.

The procedure is similar to the procedure reported previously (22–24). The Fc-GlcNAc antibody (Fc-GlcNAc-chMC813-70) was prepared from Expi293F-GnT1- cell line with Endo S2 KI. After protein A purification, the antibody was subjected to transglycosylation, which was incubated with SCT glycan-oxazoline and an Endo S2 mutant in Tris buffer at 37 °C for 1 h, followed by protein A purification.

Intact Protein Mass Analysis.

The samples were diluted with LC/MS grade water at a concentration of 10 μM and analyzed by 6230 TOF LC/MS with a Dual AJS ESI ion source (Agilent Technologies) with PLRP-S 1,000 Å 5 µm column (Agilent Technologies). Solvent A was 0.1% Formic Acid in H₂O and solvent B was 0.1% Formic Acid in ACN.

Glycosidase Treatment.

The proteins were incubated with α2,3-neuraminidase, neuraminidase, or PNGase F (New England Biolabs) in the supplied buffer at 37 °C overnight followed by purification or gel electrophoresis.

ELISA Binding Assay.

Like the procedure described previously (22, 23), the recombinant soluble FcγR (R&D) was coated with 50 ng/well in bicarbonate/carbonate buffer (50 mM, pH 10) at 4 °C overnight before blocking the well by 5% BSA in TPBS (0.05% Tween 20 in PBS) at 4 °C overnight. The antibody was added into wells with final concentration started at 100 μg/mL with fivefold series dilution and incubated at room temperature for 1 h followed by incubation with HRP-conjugated goat anti-human IgG Fc antibody (Jackson ImmunoResearch) for another 1 h at room temperature. Finally, the TMB substrate (Bethyl Laboratories) was added to react with HRP at RT prior to stopping the reaction by H₂SO₄. The absorbance at 450 nm was detected by SpectraMax M5 spectrum reader (Molecular Device). The wells were washed by TPBS 3 to 5 times between each step.

ADCC Reporter Assay.

ADCC reporter assay (Promega) was performed as kit manual (22, 23). Briefly, the SSEA4 expressing target cells SKOV-3 was seeded into a 96-well plate followed by addition of antibody with final concentration started at 1 μg/mL with fivefold series dilution. Then, the FcγRIIIA expressing effector cells (effector: target cell ratio, 6:1) was added and incubated with target cells in the incubator at 37 °C for 6 h. The plate was placed at room temperature for 15 min prior to adding the luciferase substrate. After 5 min incubation, the luminescence was measured by SpectraMax M5 spectrum reader. The induction fold was calculated by RLU (induced–background)/RLU (no antibody control–background).

Glycoform Analysis.

The FASP (filter-aided sample preparation) method was used for in-solution protease digestion. The procedure was like what was developed previously (22–24). Briefly, 5 μg of protein samples were loaded onto filter units (Microcon YM-30) and mixed with 100 μL of 8 M urea in 0.1 M Tris-HCl pH 8.5 (UB) each for protein denaturation. The filter units were centrifugated with 14,000 × g for ~30 min until the solution was removed, followed by the addition of 100 μL of 25 mM DTT in UB each for protein reduction, and incubated at room temperature for 10 min. After centrifugation with 14,000 × g for ~30 min to remove the DTT solution, 100 μL of 50 mM iodoacetamide in UB was added to each sample for alkylation, and the filter units were incubated in the dark for 10 min, followed by centrifugation to remove iodoacetamide solution. To each filter unit, 100 μL of UB was added again. After another centrifugation to remove the solution, 100 μL of 50 mM ammonium bicarbonate in water (ABC) was added to change the buffer system to ABC for in-solution protease digestion. After centrifugation to remove the solution, 40 μL of ABC with 0.1 μg Trypsin/Lys-C Mix (Promega, sequencing grade) was added to each sample and incubated in a wet chamber at 37 °C overnight. The digested peptides in samples were spun to collection tubes, and an additional rinse with 40 μL ABC for each sample was performed. The filter units were centrifuged at 14,000 × g for 10 min to collect the flow-through, which contains digested peptides. Samples were acidified with FA to a final concentration of 0.1%. The peptide samples were dried in a SpeedVac evaporator and stored at −20 °C until analysis with LC-MS/MS.

Samples were analyzed using the LC-ESI-MS method on an Orbitrap Fusion mass spectrometer (Thermo Fisher Scientific, San Jose, CA) equipped with an EASY-nLC 1200 system (Thermo, San Jose, CA) and an EASY-spray source (Thermo, San Jose, CA). A 5 μL digestion solution was injected at a flow rate of 1 μL/min onto an easy column (C18, 0.075 mm × 150 mm, ID 3 μm; Thermo Scientific). Chromatographic separation utilized 0.1% formic acid in water as mobile phase A and 0.1% formic acid in 80% acetonitrile as mobile phase B, operated at a flow rate of 300 nL/min. The gradient employed was from 5% buffer B at 2 min to 60% buffer B at 55 min. For full-scan MS, the conditions included a mass range of m/z 375 to 1,800 (AGC target 5E5) with a lock mass, resolution set at 60,000 at m/z 200, and a maximum injection time of 50 ms. MS/MS was conducted in top speed mode with 3 s cycles using both CID and HCD, with a dynamic exclusion duration of 60 s and a 10-ppm tolerance around the selected precursor and its isotopes. The electrospray voltage was maintained at 1.8 kV, and the capillary temperature was set to 275 °C.

Byonic software version 4.2.10 (Protein Metric) was used for the identification of summary formulas of glycans associated with glycopeptides. The glycan database contained 132 entries, and the parameters used were as follows: precursor mass tolerance of 10 ppm, fragment mass tolerance of 0.5 Da, maximum missed cleavages set to 5, and cysteine carbamidomethylation and methionine oxidation considered.

Supplementary Material

Appendix 01 (PDF)

Data, Materials, and Software Availability

All study data are included in the article and/or SI Appendix.