Analytical Characterization for Similarity Assessment Between an Aflibercept Biosimilar SB15 and Reference Product (Eylea^®)

The commercially available aflibercept RP sourced from US and EU (US/EU-aflibercept) were purchased through local distributors. The purchased RPs were stored and handled according to the manufacturer’s instructions.

Liquid chromatography-electrospray ionization-tandem mass spectrometry (LC–ESI–MS) method was used for the analysis of primary structure and post-translation modification (PTM) analysis including N-glycan identification. An ACQUITY ultra performance liquid chromatography (UPLC, Waters, Milford, MA, USA) coupled to a Synapt G2 quadrupole time-of-flight mass spectrometer (Q-ToF, Waters, Milford, MA, USA) were used. The acquired data were processed using MassLynx v4.1 (Waters, Milford, MA, USA) and/or BiopharmaLynx v1.2 software (Waters, Milford, MA, USA).

Molecular weight was measured using LC–ESI–MS method in protein level. Since the aflibercept has high heterogeneity in N-glycan profile, SB15 and RP samples were deglycosylated and reduced by treating Dithiothreitol (DTT) and PNGase-F prior to LC–ESI–MS analysis. The BEH300 C4 column (Waters, Milford, MA, USA) was employed for the protein separation.

In the peptide level LC–ESI–MS analysis, samples were denatured and reduced by 8 M urea and 1 M DTT. After reduction, samples were alkylated using 1 M isodoacetamide (IAA) and buffer exchanged into the digestion buffer (1 M urea in 50 mM Tris–HCl, pH 7.5). Multiple enzymes including trypsin, Lys-C, and Asp-N were used for digestion. The BEH300 C18 column (Waters, Milford, MA, USA) was used for the peptide separation, while the disulfide linkage analysis reduction steps were ignored.

The chromatograms of SB15 and aflibercept RP were compared for the peptide mapping analysis. The amino acid sequence analysis, including N-terminal and C-terminal sequences, was performed using MS/MS-based peptide identification. Moreover, semi-quantitative analysis of PTM levels was conducted using the acquired data.

N-glycan identification was performed by LC–ESI–MS method. To release N-glycan, samples were treated with PNGage-F after denaturation. Released N-glycans were separated from protein using ethanol precipitation and dried using EZ-2.3 vacuum evaporator (Genevac, Ipswich, EN, USA). Dried N-glycan samples were reconstituted with distilled water (DW), labeled by procainamide, and loaded onto a BEH amide column (Waters, Milford, MA, USA). N-glycan annotation was conducted by using GlycoWorkBench software.

For the N-glycan quantitation, 2-aminobenzamide (2-AB) was used for labeling of N-glycan. 2-AB labeled samples were separated through a BEH amide column (Waters, Milford, MA, USA), which is connected to UPLC system with fluorescence detector (Waters, Milford, MA, USA). Detection was done by fluorescence at λ_ex = 330 nm and λ_em = 420 nm. Data were acquired and processed using Empower 3 software (Waters, Milford, MA, USA).

SB15 and RP samples were diluted with each formulation buffer (SB15: 7.78 mM sodium phosphate, 8% sucrose (w/v), 0.03% polysorbate 20 (w/v), pH 6.2; aflibercept RP: 10 mM sodium phosphate, 40 mM NaCl, 5% sucrose (w/v), 0.03% polysorbate 20 (w/v), pH 6.2) for the far-ultraviolet (UV) and the near-UV circular dichroism (CD) analysis. A Chirascan Q100 (Applied Photophysics, Leatherhead, UK) with a 0.1 mm path length cell (far-UV) and 10 mm path length cell (near-UV) was used. The far-UV CD scan and the near-UV CD scan were acquired within 200–260 nm and 250–350 nm range, respectively. The acquired CD spectra were blank-subtracted to the corresponding formulation buffer.

Fourier transform infrared (FT-IR) spectroscopy of liquid formulations was performed on a Tensor 27 FT-IR spectrometer (Bruker Optics, Coventry, UK) by using the BioATR II attenuated total reflectance (ATR) unit. All samples were analyzed without dilution. The spectra were recorded at a controlled temperature of 25 °C from wave numbers of 4000–850 cm⁻¹ with a resolution of 4 cm⁻¹. The SB15 and RP samples were analyzed against the corresponding formulation buffer as background. Each measurement was an average of 60 scans. Sample spectra were recorded by using atmospheric compensation (elimination of disturbing H₂O and/or CO₂ bands in the result spectra). The second derivative of sample spectra was calculated with nine smoothing points and normalized by vector normalization. Data evaluation of secondary structure was performed with OPUS 7.5 QUANT2 software (Bruker Optics, Ettlingen, Germany) by using the alpha-helix and beta-sheet databases.

A MicroCal Auto VP-Capillary differential scanning calorimetry (DSC) system (Malvern Instruments. Malvern, UK) was used to analyze the melting temperature (T_m) of the SB15 and aflibercept RP samples. The samples and the corresponding buffer were heated from 20 to 100 °C with a heating rate of 60 °C/h. The μDSC cell was pressurized to prevent boiling of the sample during heating. All samples were diluted to 1 mg/ml in SB15 formulation buffer.

A baseline run was performed by loading the formulation buffer. Baseline was subtracted from each sample measurement. Thermal data was normalized for the protein concentration. The T_m values were determined at the center of the peak or shoulder by using derivative analysis of the heating scan. Data analysis was performed by using Origin DSC software (OriginLab, Northampton, MA, USA).

SB15 and RP samples were diluted with D₂O then incubated over 10 s, 1 min, 10 min, 1 h, and 4 h. After incubation, samples were quenched and injected onto a hydrogen/deuterium exchange-MS system (H/DX-MS) consisting of nanoACQUITY UPLC system including H/DX manager (Waters, Milford, MA, USA) and Synapt G2-Si Q-ToF MS (Waters, Milford, MA, USA) for online digestion and peptide level LC–MS analysis. A immobilized BEH pepsin column (Waters, Milford, MA, USA), BEH C18 VanGuard Pre-column (Waters, Milford, MA, USA), and BEH C18 analytical column (Waters, Milford, MA, USA) were installed on the H/DX-MS system. The analysis of data was performed by using PLGS software (Waters, Milford, MA, USA) for peptide identification, and DynamX software (Waters, Milford, MA, USA).

SB15 and RP samples were injected onto a TSKgel G3000 SWXL column (Tosoh, Tokyo, Japan) attached to a 3.10 Size Exclusion-High-Performance Liquid Chromatography (SE-HPLC) system consisted of a Waters Alliance 2695 Separation Module (Waters, Milford, MA, USA) with a Waters 2487 Dual λ Absorbance Detector (Waters, Milford, MA, USA). The separated monomer and size impurities were detected at 280 nm. Data acquisition and processing were performed by Empower 3 software (Waters, Milford, MA, USA).

Capillary electrophoresis-sodium dodecyl sulfate (CE-SDS) analysis was performed under non-reduced and reduced condition. In non-reduced CE-SDS analysis, SB15 and aflibercept RP samples were mixed with 250 mM IAA, 10-kDa internal standard, and SDS-MW sample buffer and heated for 5 min at 70 °C. For reduced CE-SDS, 2-mercaptoethanol (BME) was used instead of IAA. After heating, samples were loaded onto a bare-fused silica capillary (Beckman Coulter, Fullerton, CA, USA) connected on PA 800 Plus CE system (Beckman Coulter, Fullerton, CA, USA). Electropherograms were obtained at 220 nm using 32 Karat software (SCIEX, Brea, CA, USA).

Before the imaged capillary isoelectric focusing (icIEF) analysis, sialic acid of SB15 and aflibercept RP samples were removed by treating of sialidase A. After desialylation, samples were mixed with pharmalyte 3–10, pharmalyte 8–10.5, pI 6.61 marker, and pI 9.50 marker. The mixture was injected onto an icIEF cartridge (Protein Simple, San Jose, CA, USA) installed on iCE3 instrument (Protein Simple, San Jose, CA, USA). Data were obtained and processed using CFR (Protein Simple, San Jose, CA, USA) and Chrom perfect software (Protein Simple, San Jose, CA, USA), respectively.

Enzyme-linked Immunosorbent Assay (ELISA) is employed to determine VEGF-A 165/121/189 binding activity of aflibercept. VEGF-A 165 (293-VE/CF, R&D Systems, Minneapolis, MN, USA), VEGF-A 121 (4644-VS/CF, R&D System, Minneapolis, MN, USA) or VEGF-A 189 (8147-VE/CF, R&D System, Minneapolis, MN, USA) was absorbed onto a 96-well plate and then, BSA-containing buffer was treated into the plate for blocking non-specific binding. Aflibercept dilution series were added for binding to coated VEGF-A 165, 121 or 189 and then, the sequential addition of HRP-conjugated anti-hIgG (Fc specific) antibody, tetramethylbenzidine (TMB, T0440, Sigma-Aldrich, St. Louis, MO, USA) substrate, and 1 N sulfuric acid is followed. A measure to absorbance at a 450-nm wavelength can determine a relative binding activity of aflibercept to VEGF-A 165, 121 or 189. The VEGF-A 165/121/189 binding activities of aflibercept were calculated relative to a reference standard via a parallel line analysis (PLA) software (Stegmann Systems, GmbH, Rodgau, Germany).

Human umbilical vein endothelial cells (HUVEC, C2519A, Lonza, Basel, Switzerland) was employed for the determination of a relative anti-proliferation potency of aflibercept. VEGF-A 165 (293-VE/CF, R&D Systems, Minneapolis, MN, USA) and aflibercept dilution were treated onto a microplate, and then, HUVEC in the endothelial cell medium (ECM, 1001, ScienCell) were sequentially treated. The proliferation of HUVEC was induced for 4 days at 37 °C, in 5% CO₂ incubator. The changes of HUVEC growth were measured through Cell Titer-Blue^® cell viability system (G8082, Promega, Madison, WI, USA) by quantifying the intensity of fluorescence dye activated by the metabolic process of live cells. The relative potency of HUVEC anti-proliferation by aflibercept was determined relative to a reference standard via a PLA software.

VEGFR-2-overexpressed engineered cell line (NFAT-RE-Luc2P/KDR HEK293 cells, E8510, Promega) in the assay media (DMEM, 11995, Gibco, Grand Island, NY, USA) was added onto a microplate and, VEGF-A 165 (293-VE/CF, R&D Systems) was treated in turn. Aflibercept dilution series were added onto a plate and incubated for about 6 h at 37 °C, in 5% CO₂ incubator to induce luciferase gene expression by VEGF signaling. A quantitation of a luminescence was measured in a dose-dependent manner through using a microplate reader (EnVision, E2104, Perkin Elmer, Waltham, MA, USA). A relative VEGF-A 165 neutralization potency was calculated and determined relative to a reference standard via a PLA software.

Surface Plasmon Resonance (SPR) is employed to evaluate neonatal fragment crystallizable receptor (FcRn) binding affinity of aflibercept. Recombinant human FcRn (8639-FC, R&D Systems, Minneapolis, MN, USA), PlGF-1 (264-PGB, R&D Systems, Minneapolis, MN, USA), PlGF-2 (6837-PL, R&D Systems, Minneapolis, MN, USA), or VEGF-B 167 (751-VE/CF, R&D Systems, Minneapolis, MN, USA) was immobilized on CM5 chip (BR-1005-30, Cytiva). Aflibercept dilution series was prepared using HBS-EP buffer. Diluted aflibercept was injected into a flow cell and dissociation was conducted. The kinetic constants were calculated by the sensorgrams regressed to 1:1 binding model of BIAevaluation™ software.

Aflibercept is a recombinant fusion glycoprotein composed of the second domain of VEGFR-1, the third domain of the human VEGFR-2, and the Fc domain of human immunoglobulin G (IgG). Aflibercept consists of polypeptide chain linked with disulfide bonds and has five N-glycosylation sites on each chain. Due to its heterogeneity of N-glycan species on five N-glycosylation sites, it is difficult to accurately measure the molecular weight of the intact protein. Therefore, the molecular weights of SB15 and US/EU-aflibercept were compared for the deglycosylated and reduced samples. All experimental values for both SB15 and US/EU-aflibercept were equal to the theoretical values within 0.01%. In addition, identified disulfide linked peptide of SB15 were identical to aflibercept RP (Table 2). Using site-specific N-glycosylation analysis, it was also confirmed that SB15 and aflibercept RP have similar predominant N-glycan for all five N-glycosylation sites (data not shown). Based on the results of single chain molecular weight, disulfide linkage, and site-specific N-glycosylation analysis, the intact molecular weight of SB15 and US/EU-aflibercept were considered comparable.

The peptide map by trypsin, Lys-C, and Asp-N were compared and amino acid sequence were confirmed using MS/MS data. As shown in Fig. 1, SB15 and US/EU-aflibercept showed similar profile without any missing or newly observed peaks. The amino acid sequences were confirmed as identical with 100% sequence coverage (data not shown). For all three products, the pre-dominant C-terminal variant was the lysine-removed form and only one N-terminal sequence was identified. The methionine oxidation level and asparagine deamidation level of SB15 showed slight differences compared to US/EU-aflibercept (data not shown). However, no meaningful differences were observed in biological activities (Table 3). Even in the oxidative and basic stressed samples, biological activities were not changed regardless of increased oxidation and deamidation levels, respectively (data not shown).

The higher-order structures of SB15 and US/EU-aflibercept were compared using various orthogonal methods. In the far-UV CD, minimum and maximum wavelength of spectrum and their residue molar ellipticity of SB15 were comparable to US/EU-aflibercept (Fig. 2A). Also, the resulting relative contents of secondary structural features were comparable (data not shown). The near-UV CD spectrum showed no meaningful difference in terms of profile and maximum wavelength between the three products which support the similarity in tertiary structure (Fig. 2B).

The secondary structure was additionally characterized by using FT-IR spectroscopy. The second derivative FT-IR spectra of SB15, and US/EU-aflibercept, showed similar profiles which overlap almost of amide I and amide II region, suggesting the similar structural contribution of the three products in secondary structure (Fig. 2C). The DSC thermograms of SB15 were comparable to US/EU-aflibercept and there were no meaningful differences in the measured melting points. Thus, it was demonstrated that the thermal transition and thermal stability of SB15 were similar to those of US/EU-aflibercept (Fig. 2D).

In addition, tertiary structure and conformational dynamics were characterized using H/DX-MS. The dynamics of deuterium uptake over time (10 s to 4 h) showed symmetry between compared samples (Fig. 3A, B) and no meaningful difference were observed in the individual difference and the sum of differences (Fig. 3C, D).

The purity and impurity in terms of size heterogeneity were analyzed. The SE-HPLC chromatograms of SB15 and US/EU-aflibercept are presented in Fig. 4A. The samples were separated into an intense main peak (denoted as monomer) and a high molecular weight impurity peak (denoted as HMW). The relative quantity of HMW (%HMW) of SB15 (0.9%) were lower than US/EU-aflibercept (2.4 and 2.5%, respectively). The HMW impurity was confirmed as dimeric aggregates of aflibercept by SEC-MALS analysis (data not shown) and the aggregates of aflibercept can be increased during storage. The observed difference in %HMW has no biological impact (Table 3).

In non-reduced CE-SDS, samples were separated into an intense main peak (denoted as main) and two impurity peaks related to low molecular weight (LMW) fragment (Fig. 4B). For all three products, %main were comparably high (99.6–99.7%). Under reduced condition, two main peaks (denoted as main 1 and main 2) and LMW impurity peaks were observed (Fig. 4C). The main 1 and main 2 peaks were related to partially N-glycosylated single chain and fully N-glycosylated single chain, respectively. The level of main 1 (%main 1), main 2 (%main 2), and LMW (%LMW) of SB15 were similar to those of US/EU-aflibercept.

The charge heterogeneity of SB15 and US/EU-aflibercept were compared using icIEF method. Since aflibercept (SB15 and RP) is a highly glycosylated protein with terminal sialic acid on VEGFR-1 and VEGFR-2 domain, the charge profile of aflibercept (SB15 and RP) is highly complex. Thus, sialic acids were removed prior to icIEF analysis to efficiently assess the individual charge variant level and the sialic acid contents were measured using the orthogonal method.

In the electropherogram of the samples (Fig. 4D), the most intense single peak was denoted as main. The multiple peaks with lower pI value than main were assigned as acidic and the higher pI peaks were assigned as basic. The acidic levels (%acidic) of SB15 (67.1%) were lower than US/EU-aflibercept (75.5 and 75.6%) and the main levels (%main) were opposite. These observed differences in acidic and main level were mainly related to the level of deamidation including isoaspartate formation (data not shown). Also, the basic levels (%basic) of SB15 were lower than that of US/EU-aflibercept connected to different C-terminal variant levels. Although, a difference in charge heterogeneity was observed, it was not linked to changes in biological activities (Table 3).

The N-glycan species on SB15 and US/EU-aflibercept were identified using LC–ESI–MS. From the three products, 34 N-glycan species were identified and there was no N-glycan species detected in only SB15 nor US/EU-aflibercept (data not shown).

In the quantitation analysis, N-glycan species were categorized into 4 N-glycan groups (galactosylated glycan, afucose, charged glycan, and high mannose). As presented in Table 4, slight differences were observed in galactosylated and afucose level between SB15 and US/EU-aflibercept. However, aflibercept has no Fc-effector functions such as antibody-dependent cellular cytotoxicity (ADCC) and complement-dependent cytotoxicity (CDC) in its MoA. In addition, pharmacokinetics (PK) profile of SB15 was comparable to that of its RP [16] and no meaningful difference was observed in biological activities (Table 3).

The analytical similarity in terms of biological activity between SB15 and US/EU-aflibercept RP was demonstrated by MoA-related bioassays, including VEGF-A 165 binding, VEGF-A 121 binding, HUVEC anti-proliferation, and VEGF-A 165 neutralization assay, and Fc-related bioassay such as FcRn binding assay (Table 3). For the similarity assessment of biological activity, similarity ranges were established as mean ± kSD. In addition, binding assays to VEGF family ligands and Fc gamma receptors between SB15 and US/EU-aflibercept were performed. The results are provided in Figs. 5 and 6.

MoA-related bioassays were conducted including VEGF-A 165 binding, VEGF-A 121 binding, HUVEC anti-proliferation potency, and VEGF-A 165 neutralization potency.

First, the evaluation of a relative VEGF-A 165 binding activity showed that the average of relative VEGF-A 165 binding activities of SB15 and US/EU-aflibercept were 104, 97, and 98%, respectively. There was no meaningful difference between SB15 and US/EU-aflibercept in VEGF-A 165 binding activities for all SB15 tested samples which were within the US/EU similarity range (Fig. 5A).

Second, the average of VEGF-A 121 binding activities of SB15 and US/EU-aflibercept were 100, 96, and 97%, respectively. There was no meaningful difference between SB15 and US/EU-aflibercept and VEGF-A 121 binding activities for all SB15 tested samples which were within the US/EU similarity range (Fig. 5B).

Third, the average of HUVEC anti-proliferation potency of SB15 and US/EU-aflibercept was 102, 97, and 99%, respectively. There was no meaningful difference between SB15 and US/EU-aflibercept and HUVEC anti-proliferation potencies for all SB15 tested samples which were within the US/EU similarity range (Fig. 5C).

Finally, the average of VEGF-A 165 neutralization potency of SB15 and US/EU-aflibercept was 101, 98, and 99%, respectively. There was no meaningful difference between SB15 and US/EU-aflibercept and VEGF-A 165 neutralization potencies for all SB15 tested samples which were within the US/EU similarity range (Fig. 5D).

Fc-related bioassay was conducted to evaluate FcRn binding affinity, using SPR. The average of FcRn binding affinity of SB15 and US/EU-aflibercept was 99, 100, and 100%, respectively.

Relative FcRn binding affinities for all SB15 tested samples which were within the US/EU similarity range (Fig. 5E).

Aflibercept is a fusion protein of human VEGF receptor 1 and 2 extracellular domains with an Fc fragment of IgG1 and it can bind to VEGF-A, VEGF-B, and PlGFs.

SPR and ELISA methods were employed in order to assess binding activities of aflibercept (SB15 and RP) to VEGF family ligands such as VEGF-A 189, VEGF-B 167, and PlGFs.

Consequently, there were no meaningful differences between SB15 and US/EU-aflibercept in terms of VEGF family ligands binding activities (VEGF-A 189, VEGF-B 167, PIGF-1, and PlGF-2), respectively (Fig. 6).