Glycoengineering of pertuzumab and its impact on the pharmacokinetic/pharmacodynamic properties

Pertuzumab is an antihuman HER2 antibody developed for HER2 positive breast cancer. Glycosylation profiles are always the important issue for antibody based therapy. Previous findings have suggested the impact of glycosylation profiles on the function of antibodies, like pharmacodynamics, antibody-dependent cellular cytotoxicity (ADCC) and complement-dependent cytotoxicity (CDC). However, the roles of fucose and sialic acid in the function of therapeutic antibodies still need further investigation, especially the role of sialic acid in nonfucosylated antibodies. This study focused on the pharmacokinetic and pharmacodynamic properties of pertuzumab after glycoengineering. Herein, nonfucosylated pertuzumab was produced in CHOFUT8−/− cells, and desialylated pertuzumab was generated by enzymatic hydrolysis. Present data indicated that fucose was critical for ADCC activity by influencing the interaction between pertuzumab and FcγRIIIa, nevertheless removal of sialic acid increased the ADCC and CDC activity of pertuzumab. Meanwhile, regarding to sialic acid, sialidase hydrolysis directly resulted in asialoglycoprotein receptors (ASGPRs) dependent clearance in hepatic cells in vitro. The pharmacokinetic assay revealed that co-injection of asialofetuin can protect desialylated pertuzumab against ASGPRs-mediated clearance. Taken together, the present study elucidated the importance of fucose and sialic acid for pertuzumab, and also provided further understanding of the relationship of glycosylation/pharmacokinetics/pharmacodynamics of therapeutic antibody.

controversially discussed 22,[25][26][27][28][29][30] . Galactose has been reported to influence Fcγ RIIIa binding while not affect ADCC activity. Although, many studies indicated the correlation between the levels of sialic acid and the CDC activities, but there is still have no agreement on their interactions with Fcγ R 29,31,32 . Until now, whether the correlations between mAbs and glycosylation profiles are positive or negative is not determined and is likely dependent on the properties of different antibodies 21 . However, antibody glycoengineering, such as the deletion of FUT8 gene, could enhance the Fcγ RIIIa binding affinity and lead to an increased ADCC activity 33,34 . On the other hand, the impact of glycosylation on the PK behavior of mAbs was also investigated among these years, especially for terminal sialic acid. It is found that the sialylation degree of a glycoprotein can play a significant role in the serum half-life [35][36][37] . Glycoproteins can be removed from the circulation via a sugar-based interaction with ASGPRs on hepatic cells [37][38][39] . Several studies have reported that the ASGPRs induced endocytosis can lead to the clearance of glycoproteins in circulation 36,40,41 . Sialic acid content varies depending on several factors including cell types and fermentation process, all these factors can result in different terminal sugar moieties of mAbs. Whereas, whether sialic acid could influence the PK behavior of mAbs is not clear. To this end, we have engineered pertuzumab by CRISPR-Case9 and enzymatic hydrolysis to better understand the roles of the fucose and sialic acid. Herein, this study indicated that fucose and sialic acid descended the ADCC activity of pertuzumab, while the removal of sialic acid could result in an increased CDC activity, but induced a relatively poorer PK behavior. Additionally, we found that sialic acid affected the PK/PD relationship of pertuzumab, and would be critical quality attributes (CQA) for mAbs production.

Characterization of pertuzumab expressed in Glycoengineered CHO cells.
To generate nonfucosylated pertuzumab, FUT8 gene was knocked out by CRISPR/Cas9 using LCA-selection method. Followed, the humanized pertuzumab glycovariants were produced from wild-type CHO-K1 cells and CHO FUT8−/− cells using zeocin/blasticidin selection. Stably expressing cell pools were cultured for 5 days and the secreted antibodies were purified from the collected supernatants by Protein A affinity chromatography, and exchanged into PBS buffer using Desalting column. The heavy chain (~50 KD) and light chain (~25 KD) of concentrated antibodies were separated by SDS-PAGE (Fig. 1A). Western blot analysis showed that both pertuzumab Fuc+ and pertuzumab Fuc− binds strongly with 185 KD HER2 protein from SK-BR-3 Cell lines (Fig. 1B). Then FITC-LCA lectin blot assay was performed to detect the fucose in the antibodies (Fig. 1C). As expected, pertuzumab derived from CHO FUT8−/− cells displayed no visible fluorescence. For sialic acid determination, sialic acid was released and derivatized with DMB and analyzed by reversed-phase UPLC (Fig. 1D). The results showed the average sialic acid level for pertuzumab was 2.3~2.6% (mol/mol) (Table S1) and no significant difference were observed compared to nonfucosylated pertuzumab.
In vitro binding properties of pertuzumab associated with the level of sialylation and fucosylation. Both the pertuzumab and nonfucosylated pertuzumab were desialylated by sialidase A to generate four glycoforms. Microscale thermophoresis (MST) method was used to determine the binding properties of the four pertuzumab glycoforms to HER2 and Fc receptors, respectively. For this purpose, HER2, CD16a and FcRn were fluorescently labeled. Accordingly, binding of pertuzumab to fluorophore-labeled molecules change the thermophoretic mobility, and resulted in a saturation binding curve for gradient pertuzumab concentration ( Fig. 2A,B,C and D). All glycoforms of pertuzumab bound to HER2 and FcRn with a similar binding affinity ( Fig. 2A and D, Table 1), whereas nonfucosylated pertuzumab showed significantly higher binding affinity to CD16a. Moreover, the removal of sialic acid can also significantly enhance the binding to CD16a ( Fig. 2B and C, Table 1). All these demonstrated that the Fc fucose and sialic acid was critical for the binding to Fcγ RIIIa. Determination of ADCC activities of different glycoforms of pertuzumab. HER2 positive cell line SK-BR-3 was used as target cells (T) to test the ADCC reactivity and peripheral blood mononuclear cells (PBMCs) from healthy human donors were chosen as effecter cells (E). First, different effecter and target ratios were tested (E:T) (Fig. 3A), and the increased cytotoxicity was clearly depend on the E:T ratio. 1 μ g/mL Herceptin was also set as positive control and achieved equivalent ADCC activity with pertuzumab ( Fig. 3B and Table S2). Further, four types of glyco-modified pertuzumabs were applied for the ADCC measurements (Fig. 3C). The result showed that 5~6 fold higher cytotoxicity was observed for pertuzumab Fuc−SA+ compared with pertuzumab Fuc+SA+ , whereas pertuzumab Fuc−SA− achieved a significant ~20 fold higher ADCC activity (Table S3). The incensement of ADCC reactivity for glyco-modified pertuzumabs correlated with the binding properties observed in MST assays. These results are in accordance with increased Fcγ RIIIa binding affinity on human NK cells for nonfucosylated and desialylated pertuzumab, and which resulted in higher ADCC activity.
Correlation of CDC potency with different glycoforms of pertuzumab. The CDC activities of four glyco-modified pertuzumabs against SK-BR-3 cell lines were measured using fresh human serum from healthy donors as a complement source (Fig. 4A). The cell viability decreased with raised concentration of human serum compared with inactivated human serum. Followed, 10% human serum was used to assess the CDC potency of pertuzumab. There was no significant difference between the fucose (+ ) type and the fucose (− ) type of pertuzumab ( Fig. 4B and Table S4). Meanwhile, an approximate 5 fold higher CDC activity was found for the sialic acid (− ) type of pertuzumab, and pertuzumab Fuc−SA− exhibited the highest CDC activity ( Fig. 4B and Table S4).
In vitro cellular clearance of Pertuzumab and its glycoform variants. Previous findings suggested that ASGPRs of hepatocytes could mediate the endocytosis of desialylated proteins and undergo intracellular degradation progress. We investigated the endocytosis properties of glyco-modified pertuzumabs. FITC labeled asialofetuin (ASF) was proved to be the ligand of ASGPRs 42 , and the endocytosis of pertuzumab was determined as the amount of intracellular FITC-ASF that internalized into hepatocytes. The data presented in Fig. 5A and B showed that the sialic acid (− ) types of pertuzumab could significantly inhibit the endocytosis of FITC-ASF compared with sialic acid (+ ) types of pertuzumab, whereas the level of fucose was not seem to influence the endocytosis of FITC-ASF. Further, gradient concentrations of pertuzumab were used to antagonize the endocytosis of FITC-ASF ( Fig. 5C and D). Although the percentage decline of FITC-ASF endocytosis was about 10%, desialylated pertuzumabs displayed more ability to suppress the endocytosis of FITC-ASF (Table S5), and these suggested that the sialic acid dependent ASGPRs interactions could play a role in the intracellular clearance of pertuzumab.

Pharmacokinetic properties of the pertuzumab sialic acid variants in mice.
Given the connection between fucose/sialic acid content and the in vitro cellular clearance properties of pertuzumab, we further characterized the impact of fucose and sialic acid content of pertuzumabs on their PK profiles. The PK of four types of pertuzumab was evaluated in ICR mice, and the free pertuzumab in mice serum were measured based on HER2 binding ( Figure S2). 100 mg/kg Asialofetuin was intravenous injected 30 min before the experiment in sialic acid (− ) groups (Fig. 6). The plasma concentration-time curve displayed a biphasic clearance profile -a rapid distribution phase and a longer elimination phase for each glyco-modified petuzumabs (Fig. 6). The distribution phase of four glyco-modified pertuzumabs was less than 3 h, the terminal elimination half-life was between 10 and 15 days. Compared with the sialic acid (+ ) type, the sialic acid (− ) type pertuzumabs showed a significantly increased clearance values (~11% incensement, Table 2), which displayed an 11%~35% decreased half-life. Additionally, no significantly differences were observed for fucose (− ) types of pertuzumab. To confirm whether the clearance differences of pertuzumabs was due to the interaction with ASGPRs in vivo, we pre-injected asialofetuin and evaluated the PK of sialic acid (− ) types of pertuzumab. In asialofetuin pre-injected groups, the average clearance values decreased significantly, and the elimination half-life was prolonged (Table 2).

Discussion
Many studies have shown that Fc glycosylation of monoclonal antibodies play a critical role for their biological activities 18,21 . However, few studies have addressed the effects of sialic acid on the PK/PD relationship of nonfucosylated antibodies. In this study, we employed the glycoengineering technique in vitro to generate four glyco-modified HER2 antibodies. In the present study, the anti-HER2 IgG1 amino acid sequence was equivalent to that of pertuzumab approved for the treatment of HER2 positive breast cancer (See supplemental sequences).
FUT8 is known to be the only gene to encode fucosyltransferase, which catalyzes the transfer of fucose from GDP-fucose to N-linked type complex glycoprotein. FUT8 knockout cell lines would be theoretically to produce nonfucosylated glycoprotein. To achieve this goal, CRISPR/Cas9 was employed previously to knockout the FUT8 gene in CHO-K1 cell using LCA-selection method, and lectin based blot was applied to detect the fucosylation of glycoprotein. We therefore chose the FUT8 knockout CHO cell line and wild-type CHO-K1 as host cell lines to generate pertuzumab respectively. Furthermore, we used sialidase A to produce sialic acid (− ) type pertuzumab in vitro by enzymatic digestion for 48 h in 4 °C ( Figure S1). Consequently, we succeeded to get four combinations on the fucose and sialic acid levels of pertuzumab. Binding affinities of four glyco-modified pertuzumabs were quantified and both nonfucosylated pertuzumab and desialylated pertuzumab showed the same binding affinity with HER2 antigen in vitro. Fucosylated pertuzumab showed a weaker binding to CD16a, and desialylated pertuzumab could enhance the CD16a binding. And these were consistent with previous reports 43,44 . In the ADCC assay, the nonfucosylated pertuzumab showed stronger ADCC reactivity. On the other hand, the strongest CD16a binding activity of pertuzumab Fuc−SA− induced the strongest ADCC activity. Accordingly, the weakest binding activity of pertuzumab Fuc+SA+ resulted in a weakest ADCC activity. The differences of ADCC activity among these pertuzumab glycoforms suggested the negative role of fucose and sialic acid, for their roles in binding with Fcγ RIIIa in vitro, which subsequently influence the interaction with Fcγ RIIIa in NK cells in vivo. Followed, the CDC assay mediated by human serum was carried out. Desialylated pertuzumab induced a relatively higher   CDC activity, whereas the fucose did not seem to affect the CDC activity for pertuzumab, and we believe the non-reducing end of N-linked oligosaccharides would be critical for C1q interaction rather than core-fucose. Glycosylation impact on the PK properties of antibodies has not been well characterized in the previous literatures. Glycoproteins with exposed terminal sugar residues such as galactose are known to be internalized by ASGPRs on hepatocytes, which can be rapidly removed from circulation and led to a short half-life. So oligosaccharides of glycoprotein are considered to play a critical role in in vivo clearance of glycoprotein. In the case of glyco-antibodies, although human IgG1 antibodies can undergo an FcRn-mediated transcytosis, glycan-receptor  mediated clearance did exist, including ASGPRs mediated 37,45,46 and Mannose receptors mediated [47][48][49] clearance. Desialylation are common during antibody production, and we demonstrated desialylated pertuzumab was more effectively internalized into hepatocytes in vitro using an endocytosis competition assay. Actually, the desialylated type of pertuzumab showed a more rapid clearance in mice, and co-injection of asialofetuin as an antagonistic agent for ASGPRs decreased the clearance rate significantly. And we confirmed that the desialylated pertuzumab from CHO FUT8−/− had the shortest in vivo half-life in mice, and then desialylated pertuzumab from CHO-K1, nonfucosylated type and normal pertuzumab from CHO-K1, the clearance seemed to depend on the sialic acid levels. Meanwhile, a similar FcRn-binding affinity was observed among the four different glyco-modified pertuzumabs in our MST assay. This finding is consistent with previous reports that glycosylation may not influence the  interaction with FcRn 50-52 . And they all displayed a relative long serum half-life in mice, about 10 days. Therefore, our findings are also reflecting the clearance profiles of human IgGs in human in vivo.
In conclusion, as shown in Fig. 7, we used glycoengineering technique to generate four types of pertuzumabs by altering the fucosylation and sialylation. We found core fucose was important for the ADCC activity, and removal of terminal sialic acid could enhance both ADCC (2~4 fold) and CDC (~5 fold) activity of the pertuzumab, whereas the poorly sialylated pertuzumab also led a relatively higher clearance rate in mice and the co-injection of asialofetuin could protect the desialylated pertuzumab against ASGPRs induced endocytosis in hepatocytes.
Preparation of glycoengineered pertuzumab. Two antihuman HER2 (pertuzumab) IgG1-producing clones from CHO FUT8−/− or CHO-K1 were expanded into 1 L shake flask and cultured in serum-free medium CD-CHO at 37 °C, 5% CO 2 for 5 days. Human IgG1 was purified from the cultured medium using Protein A column (GE, USA). The purified antibodies were exchanged into PBS (pH = 7.2) by a desalting column (GE), and concentrated using an Amicon Ultra-15 centrifugal filter (Millipore, USA). The obtained mAb was aliquoted and stored at − 80 °C. Desialylated pertuzumab were generated by the removal of sialic acids using Arthrobacter ureafaciens Glyko Sialidase A (PROzyme, CA). Further, reversed-phase UPLC (detailed protocol see supplemental file) was used for the detection of sialic acid content in pertuzumab after hydrolyzed by acid and derived by DMB.

Western blot and Lectin blot.
Purified pertuzumabs were analyzed by 10% SDS-PAGE and transferred to a PVDF membrane. For Lectin blot, the membrane was blocked and incubated with 5 μ g/mL FITC-LCA for 2 h at room temperature in dark. The fluorescence was detected by Tanon 5200 Multi (Tanon, China). Further, SK-BR-3 cells were lysed, the concentration of cell lysate was determined by Bradford method. 40 μ g/well proteins were separated by 10% SDS-PAGE in non-reduced status, and transferred to a PVDF membrane. The membrane was blocked and incubated with purified anti-HER2 mAbs overnight at 4 °C. HRP-Mouse anti-Human IgG1 (Life    Technologies, USA) was used to incubate for 1 h at room temperature following four washes with TBST, and the specific bands were detected using ECL detection system (Miliproe).
MST assays. MicroScale Thermophoresis (MST, NanoTemper, Germany) was used to detect the affinity of pertuzumab with Human HER2, CD16a and FcRn. Before the MST assays, the ligands (HER2, Sino Biological, China; CD16a, R&D system, USA; FcRn, R&D system) were labeled according to the manufacturer's instructions (Nano Temper) using Monolith NT protein labeling kit RED-NHS respectively. 10 μ L labeled ligands were mixed with 10 μ L gradient mAbs, and filled into capillaries to measure the fluorescence intensity in a Monolith NT.115 system. In the measuring, 40% was chosen for IR-laser power. The binding data were fit using the Hill equation in Prism (Graphpad, USA).
Antibody-dependent cellular cytotoxicity (ADCC) assays. The collection of PBMCs and serum from healthy volunteers has been carried out in accordance with the Declaration of Helsinki, and was approved by the Institutional Review Board of Zhongda Hospital (Southeast University, Nanjing, China). A signed informed consent was obtained from each volunteer. ADCC activity was determined by the LDH release assay. SK-BR-3 was chosen as target cells (T). Human PBMCs purified from healthy donor using Lympholyte-H (Cedarlane, USA) were used as effector cells (E). E/T ratio was determined by mix SK-BR-3 with PBMCs in the presence of antihuman HER2 antibodies. Further, PBMCs mixed with SK-BR-3 at an E/T ration of 50/1, and incubated with different pertuzumab concentrations. After 4 h incubation at 37 °C., the plate was centrifuged at 250 g for 5 min and transferred 50 μ L of the supernatants to another new 96-well plate. According the manufacturer's instructions, 50 μ L/well substrate was added to each well and incubate for 30 min, the reaction was stopped and read the absorbance of 490 nm. Besides, effector cells self-release, target cells self-release, target cells maximum release, blank control group and positive control group were also set.
Complement-dependent cytotoxicity (CDC) assays. CDC activity was determined by ATP assay.
SK-BR-3 cells acted as target cells, normal human serum was selected as complement resource. Firstly, different human serum dilutions (normal serum and inactive serum) were applied to incubate with SK-BR-3 in the presence of antihuman HER2 antibody. Followed, 10% normal human serum was chosen to investigate the CDC activity of different concentrations of pertuzumab. After 1 h incubation at 37 °C., CellTiter-Glo 2.0 reagent (Promega, USA) was used to measure the intracellular ATP content according to the manufacturer's instructions.
Endocytosis assays. Endocytosis assays were carried out using FITC-asialofetuin in HepG2 and L-02 cells.
In vitro pharmacokinetics of pertuzumab. All experimental procedures with animals used in the present study were according to the Guide for the Care and Use of Laboratory Animals as adopted and promulgated by the United States National Institutes of Health, and had been given prior approval by the China Pharmaceutical Experimental Animal Manage Committee under Contract 2016 (su) − 0020. For each glyco-modified pertuzumab, five 6-week-old male ICR mice (n = 5, Comparative Medicine Centre of Yangzhou University, China) were injected into tail vein (intravenous bolus) with 1 mg/kg antibody. For asialofetuin co-injection groups, 100 mg/kg asialofetuin was pre-injected 30 min before. Mice blood was collected from tail vein at time points: 0.5, 1, 2, 4, 8 and 24 h; and 2, 3, 4, 5, 7, 9, 14, 20, and 26 days, and serums were separated by centrifugation at 4000 rpm for 30 min at room temperature. The collected serums were stored at − 80 °C for further analysis. The antibody concentration was measured by HER2 antibody specific ELISA, which coated with 100 ng human HER2 antigen per well, 200 ng/mL, 100 ng/mL, 50 ng/mL, 25 ng/mL, 12.5 ng/mL, 6.25 ng/mL, 3.125 ng/mL, 1.5625 ng/mL and 0 ng/mL pertuzumab were added as standards. The HER2-bound pertuzumab was detected by horseradish peroxidase (HRP) labeled mouse anti-human IgG Fc secondary antibody (Thermo Fisher). The concentrations of pertuzumab in the collected mouse serum were calculated according to the standard curve. The data were fit to a two-comparmental model, C(T) = A*EXP (− ALPHA*T) + B*EXP (− BETA*T), and PK parameters of AUC, CL, half-life, V1 and Vss were determined (Win Nonlin 5.2; Pharsight, USA).
Statistics. Student's t-test was used to calculate statistical significance between two groups. When more than two groups are compared One-Way-ANOVA analysis were applied. In all figures, error bars denote standard deviation.