Characterization of Biosimilar Monoclonal Antibodies and Their Reference Products Approved in Japan to Reveal the Quality Characteristics in Post-approval Phase

Abstract

Background

Promoting the use of biosimilars is a global issue from the perspective of reducing medical costs. In Japan, the replacement of original biopharmaceuticals with biosimilars has not progressed as expected. To promote the use of biosimilar monoclonal antibodies, it is necessary to increase the public’s understanding of biosimilars by evaluating and confirming the quality of biosimilar products. However, there are limited data to compare among multiple biosimilars and/or the reference, and among their product lots.

Objective

In this study, we evaluated quality attributes of multiple lots of reference products and their biosimilars to determine the extent of distribution in quality attributes between products as well as the quality consistencies from lot to lot.

Methods

As the quality attributes, the glycosylation profile, charge heterogeneity, binding affinity for the antigen and Fcγ receptors, and high-molecular-weight species and subvisible particles were measured.

Results

The degree of similarity in quality attributes with a reference product was different for each biosimilar product. We confirmed the differences between reference and biosimilar product reported in previous articles or review reports, and that some quality attributes of biosimilars were out of the quality ranges of reference products. Although lot-to-lot variations and trends at some degrees were observed for some products, no clear drifts among lots were observed.

Conclusion

Analyzing several lots of these products enabled us to capture a profile of quality characteristics for each product. Overall, the extent of variability of each quality attribute among reference products and biosimilars was revealed by this study for the first time.

Graphical Abstract

Supplementary Information

The online version contains supplementary material available at 10.1007/s40259-025-00722-4.

Key Points

Key quality attributes of multiple lots of reference products and their biosimilars were measured to determine the extent of distribution of each attribute among products as well as the quality consistencies from lot to lot.

These comparative data among multiple biosimilar products are useful for facilitating healthcare providers’ understanding of the quality of biosimilars and will promote the replacement rate of biosimilars for reference products.

In terms of the global trend of reducing clinical trials in biosimilar development, accumulating the data of commercially available biosimilars regarding the range or distributions of key quality attributes will be of great importance, because these data may suggest the clinically acceptable ranges of quality attributes.

Introduction

Biopharmaceuticals, such as monoclonal antibodies (mAbs), are a very important category of drugs. Although the development of biopharmaceuticals is accelerating, the rising medical cost has become an issue. In Japan, the drug price of biosimilars is set at 70% of the reference (REF) product in an attempt to curb soaring medical costs and provide patients with better access to medicines [1]. However, biosimilars are not widely used in Japan. For example, as of 2020, the replacement rate of filgrastim biosimilars was approximately 90%, whereas for typical therapeutic mAbs, the replacement rate of rituximab (RTX) was approximately 70%, etanercept (ETN) was 40%, and infliximab (IFX) was 20%, which varied greatly depending on the product [2]. In the EU and the USA, regulatory agencies make efforts to promote healthcare professionals’ and patients’ understanding of biosimilars [3, 4], although decisions on how to implement the policy are managed by the individual country or state [5, 6]. As a result, for example, the replacement rate of biosimilars for tumor necrosis factor α (TNFα) inhibitors, such as IFX, is close to 80% in France, where substitution is permitted unless “no substitutions allowed” is noted in the prescription [7]. According to the results of a Japanese survey of physicians regarding biosimilar adoption, one of the reasons for the lower replacement rate is the lack of clinical trial data for minor indications. Moreover, compared to the situation where information on the quality of generic drugs is sufficiently available [8–10], the information regarding the quality of biosimilars is limited and not well understood by healthcare professionals and patients [11]. The lack of comparative data among multiple biosimilar products may also be a factor in the reluctance of hospitals to adopt them.

With respect to the quality characteristics of biosimilars, developers may have published data comparing biosimilars with a REF product; however, data comparing each quality attribute have not been published for all biosimilars distributed in Japan. Information regarding the quality characteristics of biosimilars may also be found in the regulatory agency’s published review reports, which describe whether the quality characteristics were considered comparable. The US Food and Drug Administration’s (FDA’s) review report includes detailed information on the range of measured values for each quality attribute and the method for evaluating their similarity. However, such information is not found in the review reports from Japanese regulatory agencies. Furthermore, most of the information obtained from these reports is based on comparisons of the REF product with a certain biosimilar product. Thus, there are limited reports that compare the quality characteristics of the REF product and multiple biosimilar products. In addition, biopharmaceuticals (REF and biosimilar products) may be subject to drift in some quality attributes over time [12–16]. Therefore, to promote the use of biosimilars, it is necessary to clarify the degree of differences in the quality characteristics between biosimilars and their REF products as well as among the lots of each product.

In this study, we measured the attributes of multiple lots of REF products and their biosimilars to determine the extent of distribution in quality attributes between products as well as the quality consistencies from lot to lot. For the quality attributes, the glycosylation profile, charge heterogeneity, aggregation (high-molecular-weight aggregates [HMW] and subvisible particles), and binding affinity for the antigen and Fcγ receptors were selected as potential variable quality attributes that depend on the manufacturing and storage conditions. In terms of the global trend of reducing clinical trials in biosimilar development, accumulating biosimilar data regarding the range or distributions in each attribute will be of great importance.

Materials and Methods

Materials

All original and biosimilar IFX, trastuzumab (TRA), RTX, bevacizumab (BEV), and ETN products were purchased from commercial suppliers from 2019 to 2022. The list of products and each lot number is shown in Tables S1–5 (see the electronic supplementary material). Peptide-N-glycosidase F (PNGase F) was purchased from Roche Diagnostics (Basel, Switzerland). 2-Aminobenzamide (2-AB), sodium dodecyl sulfate (SDS), and dimethyl sulfoxide (DMSO) were purchased from SigmaAldrich (St. Louis, MO). Dithiothreitol (DTT) was purchased from Thermo Fisher Scientific K.K. (Tokyo, Japan). One percent methylcellulose and pI markers were purchased from Protein Simple (Minneapolis, MN). Pharmalyte 3-10, CM5 sensor chip, Human Antibody Capture Kit, Biotin CAPture Kit, and other reagents related to surface plasmon resonance (SPR) were purchased from Cytiva (Tokyo, Japan). Carboxypeptidase B (CPB) was sourced from Sigma-Aldrich (C9584).

All of the products were measured before their expiration dates. Lyophilized formulations were dissolved in a designated solution or distilled water. For subvisible particle analysis using light obscuration (LO) and flow imaging (FI), measurements were conducted on the day the products were opened or dissolved. After subvisible particle analysis, each solution was used for the other analysis within 2 weeks or stored at − 80 °C until further analysis. The effect of freeze-thaw on measurement values was confirmed to be negligible.

Glycan Analysis

The glycosylation profile was measured by 2-AB labeling and hydrophilic interaction liquid chromatography-fluorometric detection (HILIC-FL) analysis. An aliquot of protein (100 µg) was desalted and dissolved in 50 µL of 20 mM sodium phosphate (pH 7.5) containing 0.25% SDS and 25 mM DTT. The sample was heated at 65 °C for 15 min. After cooling to room temperature, 5 µL of 10% Triton X-100 and 1 µL of 1 Roche Unit/µL PNGase F solution were added, and then, the sample was incubated at 37 °C for 1 h while mixing. After digestion, 50 µL of 1% acetate was added, and the sample was incubated for 1 h. The digest was applied to an OASIS HLB 1 cc (30-mg) solid phase extraction (SPE) column conditioned with 1 mL of methanol and 1 mL of water. Then, 0.9 mL of 1% acetate was applied to the column. The flowthrough was collected and dried using a centrifugal vacuum concentrator. Next, 5 µL of 2-AB derivatization solution, which consisted of 0.37 M 2-AB and 1 M sodium cyanoborohydride in DMSO and an acetic acid mixture (7:3), was added to the dried glycan, and the sample was incubated at 37 °C overnight [17]. After adding 1 mL of acetone, the sample was vortexed, centrifuged at 15,000 × g for 15 min, and the supernatant was removed by decantation. 2-AB-labeled N-glycan was dissolved in 25 µL of water, and 1 µL of the sample was analyzed by HILIC-FL. The 2-AB-labeled glycans were separated on an ACQUITY UPLC Glycan BEH Amide, 130-Å column (Waters, 2.1 × 150 mm, 1.7 µm) and detected using a fluorometer (excitation wavelength = 320 nm, emission wavelength = 420 nm). Mobile phase A was 100% acetonitrile, and mobile phase B was 100 mM ammonium formate, pH 4.5. The flow rate and column temperature were set at 0.25 mL/min and 45 °C, respectively. The gradient started at 75% B for 2 min and was followed by a linear gradient from 75% to 50% B over 45 min. The total area was recorded from 0 to 45 min.

Size-Exclusion Chromatography

Size-Exclusion chromatography (SEC) analysis was performed by ACQUITY UPLC H-Class Plus Bio (Waters) with a titanium flow cell (5-mm pathlength, 1500-nL volume), using a TSKgel G3000SWXL, 7.8 × 300 mm, 5 μm (Tosoh Bioscience) with a guard column. The mobile phase was 100 mM sodium phosphate buffer, pH 7.0, containing 300 mM NaCl. The flow rate was 1.0 mL/min, and the column temperature was set at 25 ℃. Each sample was diluted with water to 10 mg/mL as necessary, and a volume of 10 µL (100 μg) was injected. Protein was detected by UV at 280 nm.

Imaged Capillary Isoelectric Focusing

The charge profiles of the mAbs were obtained using Maurice (ProteinSimple) with the operational software Compass. The master mix was prepared by mixing 74 μL of deionized water, 70 μL of 1% methylcellulose, 8 µL of Pharmalyte 3–10, 4 µL of 500 mM arginine, and 2 µL each of pI marker 3.38 and 9.99 or 10.17. Each sample was diluted to 1.0 mg/mL with deionized water, and 40 μL was mixed with 160 μL of master mix. After vortexing, the sample was centrifuged at 13,000 × g for 5 min, and 160 μL was transferred to a 96-well plate. CPB treatment was performed by incubation with CPB at a ratio of 1:200 (CPB to mAbs) in 50 mM Tris-Acetate, pH 7.0 for 1 h at 25 ℃. The sample was loaded for 55 s, then focused at 1000 V for 1 min, followed by 3000 V for 7 min onto the instrument using a capillary cartridge. The separated proteins were detected by UV absorbance (0.005-s exposure).

Surface Plasmon Resonance (SPR), Affinity for Antigens

The interaction between antibodies and antigens was analyzed by SPR using a Biacore 8K instrument (Cytiva). Detailed measurement conditions including sensor chip, running buffer, antigen, captured level, and contact and dissociation time are listed in Table S6 (see the electronic supplementary material). To analyze the interaction of IFX, RTX, TRA, and BEV with antigens, the antibodies were first captured on the CM5 sensor chip immobilized with anti-human IgG (Fc) antibody (immobilization level, approximately 7000 RU), and serially diluted antigen solutions were injected. For the analysis of the interaction of ETN with TNFα, Biotin CAPture Reagent was first injected on a Senser Chip CAP based on the manufacturer’s instructions, and biotinylated human TNFα was subsequently captured on the chip. The serially diluted ETN was injected. The measurements were conducted at 25 ℃. For the measurement of the interaction of IFX with TNFα, the same REF product was used as a control to correct for differences in the measured values between runs. Data analysis was performed using the evaluation software (Cytiva). The dissociation constant (K_D), association rate constant (k_on), and dissociation rate constant (k_off) were calculated using a global fitting analysis assuming a Langmuir binding model and a stoichiometry of 1:1 or a bivalent fitting model.

SPR, Affinity for FcγRIIIa

The interaction between antibodies and human FcγRIIIa Val158/Phe158 was analyzed by SPR in a Biacore 8K instrument (Cytiva). Phosphate buffered saline (PBS) supplemented with 0.005% Tween-20 was used as a running buffer. The anti-His-tag antibody was first immobilized on the CM5 sensor chip. The His-tagged human FcγRIIIa (158V and 158F) reagent (Sino biologicals, Beijing, China) was subsequently injected and captured on the sensor chip. The aimed capture level of FcγRIIIa was 60 RU. As analytes, antibodies serially diluted from 360 to 22.5 nM for FcγRIIIa 158V and from 1440 to 90 nM for FcγRIIIa 158F were injected. The contact and dissociation times were 150 and 250 s, respectively. The measurements were conducted at 25 °C. Data analysis was performed using the evaluation software (Cytiva). The K_D, k_on, and k_off were calculated using a global fitting analysis assuming a Langmuir binding model and a stoichiometry of (1:1).

Light Obscuration

KL-04A (Rion Co., Ltd., Tokyo, Japan) was used for the LO experiments. The instrument was qualified according to the Japanese Pharmacopoeia <6.07> Insoluble Particulate Matter Test for Injections [18]. Based on our previous study on a reduced test volume for the LO method, the LO measurement was conducted four times with a sample volume of 0.2 mL [19]. The tare volume was set at 0.2 mL. Four results were collected, and three results were analyzed after discarding the first result. The particle counts were obtained from the particle numbers in the range of 2.0–100 μm.

Flow Imaging

FlowCam 8100 (Fluid Imaging Technologies, Inc., Scarborough, ME, USA) was used for the FI experiments. Before the measurements, the performance of the instrument was verified by confirming that the particle counts and size of the 10-μm polystyrene standard particles (COUNT-CAL Count Precision Size Standards CC10-PK, Thermo Scientific, CA) were within the certificated value. The flow rate, sample volume, autoimage rate, sampling efficiency, and segmentation threshold (dark/light) were set at 0.1 mL/min, 0.2 mL, 16 frames/s, approximately 70%, and 10/10, respectively. The instrument was flushed with distilled water and, if necessary, 1% Tergazyme (Alconox Inc., White Plains, NY, USA) solution at a flow rate of 5 mL/min to ensure that there were no major particles in the flow cell. After setting the instrument parameters, to assess the qualification of each instrument, 700–1000 μL of distilled water were tested to confirm that fewer than 100 particles > 2 μm in size were present in each milliliter of fluid and to ensure the absence of any particles > 10 μm in size. For a single assay, 200 μL of the sample was applied to the instrument and measured three times. The results were collected and analyzed in triplicate. The particle counts were obtained from the particle numbers in the ranges of 2.0–100 μm.

Circular Dichroism

J-1100 CD spectrometer (JASCO Corporation, Tokyo, Japan) was used for circular dichroism (CD) spectra measurements. The analytical parameters were set as follows: wavelength range, 190–250 nm for far-UV and 250–340 nm for near-UV; scanning speed, 50 nm/min for far-UV and 20 nm/min for near-UV; response, 4 s; band width, 1 nm; accumulations, three times; optical path length, 1 mm. The representative lot of each product was used as a sample. For far-UV, the samples were diluted to 0.1 mg/mL (RTX), 0.15 mg/mL (TRA, BEV, and ETN), or 0.16 mg/mL (IFX) using purified water. For near-UV, the samples were diluted to 5 mg/mL in a buffer containing the respective excipients. Each sample was measured in 3 cycles, and a spectrum of blank solution was subtracted. After subtraction, each spectrum was offset.

Statistical Analysis

The quality range of the REF product was used to assess similarity [20]. The quality range was set based on the range of the measurement values of the REF product and expressed as x times the REF product standard deviation (± xSD). As a multiplier (x), 3 was used for all attributes measured in this study.

Results

Commercially available REF products and biosimilars of IFX, TRA, RTX, BEV, and ETN were examined. Although ETN is not an mAb but an Fc-fusion protein having a similar mode of action (MOA) to an anti-TNFα mAb, it was included in this study as a biopharmaceutical of medical importance. The information for the REF products and their biosimilars are listed with respect to the cell substrate used for manufacturing, excipients, and lot number (Tables S1–5; see the electronic supplementary material). The quality ranges calculated from the data of the REF products for each key quality attribute were used to evaluate the similarity to the REF product.

Glycosylation

Protein glycosylation impacts the stability, pharmacokinetics, pharmacological activity, and immunogenicity of therapeutic proteins [21–27]. An IgG molecule contains two N-linked glycosylation sites in its Fc region. The major glycans on IgG include high mannose, hybrid, and biantennary complex types. They exhibit variety, such as the presence or absence of fucose as well as the number of mannose, terminal galactose, and sialic acid moieties [21, 22]. This heterogeneity may be affected by various factors, including the type of cell substrate (i.e., production cell) and manufacturing conditions [21, 28]. To characterize the glycosylation profile, we conducted an oligosaccharide analysis using 2-AB labeling and HILIC analysis, which is a typical analytical method for released N-glycans. The relative content of each glycan species in the products is shown in Fig. 1. The integrated contents of afucosylated, high mannose, galactosylated, and sialylated glycans are plotted in Fig. 2. The glycan terminology abbreviations used are shown in Fig. S1 (see the electronic supplementary material). One dot corresponds to the result of one lot. The representative chromatogram and glycan profile for each lot are shown in Figs. S2–5. For ETN, as the glycan profile of the vial product of ETN biosimilar 1 (ETN_BS1) was confirmed to be the same as that of the syringe type (Fig. S5), only the profile of the syringe product of ETN_BS1 is shown in Figs. 1 and 2.

Fig. 1.

Glycosylation profile of REF and biosimilar products of IFX, TRA, RTX, BEV, and ETN. Mean % of total peak area and SD are plotted for each product. The abbreviations of each glycan are listed in Fig. S1. BEV bevacizumab, ETN etanercept, IFX infliximab, REF reference, RTX rituximab, TRA trastuzumab

Fig. 2.

Comparison of the N-glycan profiles for IFX, TRA, RTX, BEV, and ETN. Scatter plots of afucosylated glycans % (G0 + G1 + G2), high mannose glycans % (Man 5), galactosylated glycans % (G1 + G1F + G2 + G2F), and sialylated glycans % are shown with the mean and SD. Dashed lines represent the quality ranges of the REF products. The abbreviations of each glycan are listed in Fig. S1. BEV bevacizumab, BS biosimilar, ETN etanercept, IFX infliximab, REF reference, RTX rituximab, TRA trastuzumab

The overall glycan profiles were significantly different between the REF and biosimilar products for IFX. IFX products are produced with two different cell substrates derived from different species. The cell substrate for the REF and BS1 is mouse myeloma cell Sp2/0, whereas that for BS2 and BS3 is CHO cells. Depending on the production cells, the glycans eluted after G2F in the chromatogram were different (Figs. 1, 2, Fig. S2, and Fig. S3). The glycans unique to Sp2/0 were considered α-galactosylated or sialylated glycans (mainly N-glycolylneuraminic acid) [28].

ETN is an Fc-fusion dimeric protein that contains the extracellular ligand binding domain of the human TNF receptor 2 (TNF RII) linked to the Fc region of human IgG1. Each chain contains one N-glycosylation site (Asn317) on its Fc region and has two additional N-glycosylation sites (Asn147 and Asn179) and 13 O-linked glycosylation sites on the extracellular domain of TNF RII [29]. Most of the N-linked glycans located at the TNF RII region are mainly sialylated, whereas the N-linked glycans at the Fc region are primarily asialylated [29, 30]. The glycan analysis results of ETN treated with exoglycosidase suggested that the peaks eluted after G2F were primarily sialylated glycans (Fig. S6). The differences observed in the profiles of asialylated and sialylated glycans indicate the differences in the glycans derived from the Fc and TNF RII regions, respectively, among REF, BS1, and BS2 (Figs. 1, 2).

Comparing the content ratio of the afucosylated, high mannose, galactosylated, and sialylated glycans among the products in Fig. 2, the measurement results of IFX showed significant differences in all glycans. The measurement values for the afucosylated glycans, high mannose glycans, and peaks after G2F of IFX_BS3 were not within the quality range of the REF product. The galactosylation level of IFX_BS2 was beyond the quality range. For TRA, BS1 exhibited higher sialylated glycan content. For RTX, a higher content of afucosylated glycans was observed in the BS2 results. The biosimilars of BEV had a higher content of galactosylated glycans compared with the upper limit of the quality range of the REF products. For ETN, BS1, and BS2 tended to have higher levels of afucosylated and galactosylated glycans and exhibited a lower level of high mannose. For the five types of products examined in this study, lot-to-lot variation was observed to some degree for each glycan. Based on the glycan profiles of all the products in Figs. S3–5, no trend was observed in which the specific glycans decreased or increased between lots.

Charge Heterogeneity

The charge heterogeneities of therapeutic mAbs and Fc-fusion proteins may result from deamidation, oxidation, and variations in C-terminal lysine, formation of N-terminal pyroglutamate, aggregation, isomerization, charged (sialylated) glycans, fragmentation, glycation at lysine residues, and succinimide formation [31, 32]. Acidic and basic variants may be formed, in which the acidic fraction typically contains a variety of protein species, such as deamidated, sialylated, and glycated forms, whereas the basic fraction contains N-terminal glutamine, succinimide formation, isomerization of aspartic acid, and oxidated and C-terminal lysine variants [31–33]. The charge heterogeneity of the products in the present study was measured by imaged capillary isoelectric focusing (icIEF). The peak ratio is compared between REF and biosimilar products in Fig. 3. The representative electropherogram of each product is shown in Fig. S7 (see the electronic supplementary material). The content of the main peak and basic peaks of IFX differed greatly between the REF and biosimilar products (Fig. 3, Fig. S7). In particular, IFX_BS2 had lower basic variants compared with the lower limit of the quality range. The electropherogram of samples treated with CPB (blue line in Fig. S7) indicated that basic peaks 1 and 2 of mAbs were considered as variants with one and two C-terminal lysine residue(s), respectively. TRA_BS1, TRA_BS3, RTX_BS1, RTX_BS2, and BEV_BS1 had higher basic peaks compared with each quality range of REF (Fig. 3). The acidic variant content tended to be lower in biosimilar products compared with the REF products of RTX and BEV. In terms of lot-to-lot variation, the IFX_REF product showed a relatively large variation in charge profile. On the other hand, TRA showed relatively small differences between products and between lots compared with other product types. For other IFX, RTX, and BEV products, no lot-to-lot trends were observed, such as an increase in acidic peaks and basic peaks.

Fig. 3.

Charge profiles of IFX, TRA, RTX, and BEV by icIEF analysis. Box plot shows the interquartile range (box), median (band inside of box), maximum, and minimum values. Scatter plots of acidic peaks (%) and basic peaks (%) are shown with the mean and SD. Dashed lines represent the quality ranges of the REF products. BEV bevacizumab, BS biosimilar, icIEF imaged capillary isoelectric focusing, IFX infliximab, REF reference, RTX rituximab, TRA trastuzumab

The representative electropherograms of ETN products are shown in Fig. S7. The effects of isoelectric precipitation made it difficult to perform reproducible analyses of ETN, and comparisons between samples and lots were difficult.

Biological Activities

Biological activities were determined by antigen binding affinities and Fcγ receptor binding affinities based on the mechanism of action of each antibody and Fc-fusion protein. While a cell-based assay is commonly used to evaluate the biological activities of biotherapeutic proteins, it may show high variability in the measurement results. Considering the purpose of evaluating the similarity between products and variation among lots, we determined the K_D using SPR, which shows less variation in measurement. The kinetic data are useful because they provide not only the equilibrium binding constant, but also the rate constants, which characterize the dynamics of the interaction.

We evaluated the binding affinity of IFX and ETN to TNFα, TRA to human epidermal growth factor receptor 2 (HER2), RTX to CD20, and BEV to vascular endothelial growth factor (VEGF). There are two methods for evaluating binding affinity using SPR: (1) immobilizing the antigen on a sensor chip or (2) applying the antigen as an analyte. We used these two approaches for each combination of interactions and selected the appropriate method by comprehensively judging the quality of the sensorgram and the degree of model fit. Only for ETN, approach (1) was applied to satisfy the criteria to ensure the quality of sensorgram and model fitness. The K_D measured for each lot is plotted in Fig. 4, and the kinetic parameters, association rate constant (k_on) and dissociation rate constant (k_off), are listed in Tables S7–S11 (see the electronic supplementary material).

Fig. 4.

Comparison of binding activities against antigen and FcγRIIIa 158V and 157F. Scatter plots for the dissociation constant (K_D) are shown with the mean and SD. Dashed lines represent the quality ranges of the REF products. For ETN, solid lines represent the quality ranges for the vial type of the REF product, whereas dashed lines represent the quality ranges of the syringe type of the REF product. BEV bevacizumab, BS biosimilar, ETN etanercept, IFX infliximab, REF reference, RTX rituximab, s syringe product, TRA trastuzumab, v vial product

No biosimilar products exhibiting higher or lower activity were observed against the quality ranges of the REF products. Some biosimilar products tended to have slightly weaker binding affinities for the REF product of IFX, TRA, RTX, and ETN (Fig. 4). For BEV, BS1 and BS2 tended to have stronger binding affinity compared with the REF. In addition, there was a tendency for TRA_BS1 and BEV_BS1 to have larger lot-to-lot variations compared with each REF and other biosimilar products. In terms of binding kinetic parameters, the interaction behavior was considered similar for all combinations of REF and biosimilar products (Tables S7–S11). Overall, there were no major differences between the products or lots that would significantly affect quality and efficacy.

Among the Fcγ receptors, FcγRIIIa primarily contributes to Fc-mediated antibody-dependent cellular cytotoxicity (ADCC) activity mediated by natural killer (NK) cells. The binding affinity to FcγRIIIa is an important attribute when comparing the quality of a REF and biosimilar products for the mAbs with ADCC in their MOAs, such as TRA and RTX. FcγRIIIa contains a genetic polymorphism at amino acid position 158 (158F, 158V), which may be a factor in individual differences in therapeutic efficacy [34]. Furthermore, genetic polymorphisms affect the ability of ADCC bioassays to discriminate afucosylated glycan content [35, 36]. Thus, the binding affinities to both FcγRIIIa 158V and 158F were evaluated. We observed differences in binding affinities to FcγRIIIa among the products, which may be derived from the variations in glycosylation profiles (Fig. 4). The interaction with FcγRIIIa 158V had a smaller K_D value compared with that of 158F. It exhibited stronger affinity, which was consistent with that of previous reports [36]. Although similar results were obtained overall for FcγRIIIa 158V and 158F, the results measured with 158F showed clearer differences between the products. IFX_BS2 and RTX_BS1 had a lower affinity to FcγRIIIa 158F, and ETN_BS1 and ETN_BS2 had higher affinity compared with each quality range of REF.

Aggregation

Protein molecules can inherently self-associate to form aggregates ranging in size from nanometer-sized dimers to those detected as insoluble particles because of various physicochemical stresses [37]. Multiple studies have suggested that protein aggregates including insoluble particles can elicit immune response [38, 39]. Protein aggregates and insoluble particles are considered impurities to be adequately characterized and controlled to ensure the efficacy and safety of therapeutic protein injections [40, 41]. For the purity test of drug substances and drug products, HMW, including dimers and multimers, are evaluated by size-exclusion chromatography followed by UV detection (SEC-UV). The measurement results of HMW % are shown in Fig. 5. The representative chromatogram of each product is shown in Fig. S8 (see the electronic supplementary material). IFX_BS1 and TRA_BS2 had a high HMW %. For the other products, the HMW % of the biosimilar products tended to be lower or equal compared with that of each REF product. For lot-to-lot variations, the REF products of RTX and ETN exhibited larger variations in HMW % compared with the other products. Overall, no products were found that were likely to significantly affect quality.

Fig. 5.

Comparison of HMW for a IFX, b TRA, c RTX, d BEV, and e ETN. Scatter plots of the HMW % are shown with the mean and SD. Dashed lines represent the quality ranges for the REF products. For ETN, solid lines represent the quality ranges for the vial type of the REF product, whereas dashed lines represent the quality ranges of the syringe type of the REF product. BEV bevacizumab, BS biosimilar, ETN etanercept, HMW high-molecular-weight aggregates, IFX infliximab, REF reference, RTX rituximab, s syringe product, TRA trastuzumab, v vial product

The particulate matter test using LO is generally performed for drug products to evaluate the number of subvisible particles greater than 10 μm and 25 μm according to the general test of the Japanese Pharmacopoeia <6.07> Insoluble Particulate Matter Test for Injections [18]. In the test, the acceptance criteria for the number of particles with a size greater than 10 μm and also those greater than 25 μm are defined. However, several studies indicate the limitation of LO to detect particles with low optical contrast or refractive index, such as protein particles, compared with polystyrene latex beads that are used for the calibration of LO instruments [42–45]. Therefore, FI has emerged as a useful tool to characterize protein particles and to complement LO, as it is more sensitive than LO for highly transparent particles with a low refractive index. Therefore, both LO and FI analyses were used to measure subvisible particles in innovator and biosimilar products.

The total particle concentration of ≥ 2-μm particles is shown in Fig. 6. Detailed data for each particle size range is shown in Fig. S9 and Tables S12–S16. Comparing the results of LO with FI, FI exhibited a higher particle concentration, which was consistent with our previous report [46]. Approximately ten- to 50-fold differences in particle counts between the methods were observed. Except for some results, the trends in differences between the products were generally similar for the LO and FI results. Overall, the number of TRA, RTX, and BEV particles tended to be lower compared with that of the IFX and ETN products. Considering the large variation caused by counting particles non-uniformly dispersed in solutions, no significant differences were observed between the REF and the biosimilars for TRA, RTX, and BEV. For IFX, there was a tendency for the number of particles in BS1 to be higher compared with that in the other products, except for the FI measurement results for some lots of BS2 and BS3, which suddenly showed a high value. For ETN, syringe products, particularly the REF and BS2, exhibited higher particle counts compared with the vial products. When checking the particle images obtained with FI, many silicone oil droplets were detected in the syringe formulation. Although the time until the expiration date of each lot varied, there was a tendency for newer lots to contain lower numbers of particles overall (Fig. S9). On the other hand, for the syringe formulation of ETN_BS2, a tendency was observed for both LO and FI that the number of particles increased with newer lots (Fig. S9).

Fig. 6.

Subvisible particle analysis using light obscuration (left) and flow imaging (right). Scatter plots of the total particle concentration of ≥ 2 μm (#/mL) are shown with the median. BEV bevacizumab, BS biosimilar, ETN etanercept, IFX infliximab, REF reference, RTX rituximab, s syringe product, TRA trastuzumab, v vial product

The LO measurements confirmed that no products exceeded the acceptance criteria for particulate matters as defined in the internationally harmonized general test of the pharmacopoeia [18]: 6000 particles per container or less with a particle size of 10 µm or more, and 600 particles per container with a particle size of 25 µm or more (Tables S12–S16).

Higher Order Structures

To evaluate the structural similarity between the REF and biosimilar products, we conducted secondary and tertiary structural analysis using far- and near-UV CD. The CD spectra of biosimilar products and REF products were shown in Fig. 7. While the biosimilar products had different glycan and charge profiles from that of the REF product, all of the biosimilar products showed similar spectra to the corresponding REF product, indicating the similarity in higher order structure between the REF and biosimilar products.

Fig. 7.

Comparison of far-UV and near-UV CD spectra for IFX, TRA, RTX, BEV, and ETN. The averaged spectra of triplicate measurements were shown for each product. BEV bevacizumab, BS biosimilar, CD circular dichroism, ETN etanercept, IFX infliximab, REF reference, RTX rituximab, TRA trastuzumab

Discussion

Many studies have evaluated similarities between a REF product and its biosimilar; however, reports evaluating differences in quality attributes among several biosimilars are limited. Our data indicated that the range of the individual attributes for a REF product and its biosimilar can be captured by analyzing multiple lots of the same product. It was demonstrated that the extent of similarities or differences between a REF and its biosimilar product depends on each biosimilar product. We confirmed the differences between REF and biosimilar products reported in previous articles or review reports from the Japanese regulatory agency Pharmaceuticals and Medical Devices Agency (PMDA), which were issued at the time of marketing authorization approval of each product.

In the glycan profile for IFX, the review report from PMDA indicated that afucosylated glycans were lower in BS1, whereas afucosylated glycans and high mannose were lower, and galactosylated glycans were higher in BS2 compared with the REF [47, 48]. Our findings in the present study were consistent with the results in the review report. Regarding IFX_BS3, no data on BS3 that quantitatively compared glycans with REF have been published. Our analysis showed that the afucosylated glycans were significantly higher compared with those of the REF and other BS products. Despite the marked difference in glycan profiles, Lerch et al. reported that the structure of BS3 was highly similar to that of REF based on the results of X-ray crystallographic analyses [49]. For the charge variants, the relative ratios of the major peaks were different between the REF and BS1, and the proportion of mono-C-terminal lysine and di-C-terminal lysine residues were lower in BS3, consistent with the review reports [47, 48, 50]. The IFX_REF also showed lot-to-lot variations in charge variants derived from the C-terminal lysine. Because the C-terminal lysine variants are rapidly removed by carboxypeptidase in blood following injection, those variations do not affect the pharmacological activity of therapeutic antibodies [31, 51]. In our biological assays, BS1 and BS2 exhibited a lower affinity for FcγRIIIa compared with the REF, which is consistent with the review reports and other documents [47, 48, 52]. With respect to impurities, a difference in the relative amount of HMW between REF and BS1 was reported in a review report (FDA CT-P13) [53], and our results indicated that HMW were significantly higher in BS1.

The difference in glycan structures between the REF and the biosimilar for ETN was reported in articles and review reports [54–56]. An article on BS1 indicated that BS1 had the highest abundance ratio of G1F, whereas REF had the highest abundance of G0F. BS1 had higher G2 and lower disialoglycan content (G2FS2 and G2S2) compared with the REF [56]. In the present study, no significant differences in disialoglycan content were observed; however, the highest amount of G0F was observed in the REF. Although not mentioned in the article, our results indicated that high mannose was significantly lower in BS1. Although there are no published analytical results for BS2, the review report described the differences in high mannose content between the REF and the BS2, and a higher content of afucosylated glycans and galactosylated glycans in BS2 [54], which is consistent with our results. As reported, we confirmed that BS1 has a higher affinity for FcγRIIIa compared with the REF, and the HMW of the REF was higher compared with both BS1 and BS2.

Consistent with our results for the glycan profiles of the RTX biosimilars, it was shown in the review reports that the major glycans are similar to those of the REF; however, BS1 has fewer high mannose glycans, and BS2 tends to have a higher total afucosylated glycan content compared with the REF [57, 58]. Our results indicated that BS2 also has fewer high mannose glycans than the REF. We confirmed that our RTX charge variant results of lower acidic variant contents and higher basic variant contents in BS1 and BS2 were also consistent with those of previous reports [57–60]. No difference in biological activity was reported in BS1, whereas the affinity for FcγRIIIa was higher in BS2 compared with that in the REF [58]. Indeed, the affinity of BS2 for FcγRIIIa was higher than that of the REF; however, our results indicated that the affinity of BS1 tended to be weaker compared with that of the REF.

There were relatively few quality attributes in which certain differences were observed between the REF and the biosimilar for TRA and BEV in the review reports and other documents [61–70]. The differences between TRA biosimilars and the REF described in the review reports were the high sialic acid content in BS1, the different percentage of acidic peaks in the charge variants of BS2, and the high percentage of basic peaks in BS3, all of which were confirmed by our data. The only differences reported for BEV were slight differences in N-linked glycan profiles and charge heterogeneity for BS1 and BS2. No significant differences in the glycan profiles or charge heterogeneity were observed in our measurements; however, a lower tendency was found for HMW content in biosimilars compared with the REF.

The differences between the REF and biosimilar described in the review report as well as the previous studies were not considered to affect the efficacy and safety of the products based on the results of in vitro bioactivity studies related to the MOA and the results of clinical pharmacokinetics and pharmacodynamics or efficacy studies. Therefore, all of the biosimilar products measured in the present study have been deemed comparable to their REF products with respect to the quality, safety, and efficacy at the time of approval review. In addition, no significant events have been reported on these products under the post-marketing pharmacovigilance activities until now. Therefore, the observed ranges in this study can retrospectively be considered as clinically acceptable ranges. Because of the high structural complexity compared with small molecules, several variations of manufacturing conditions (e.g., pH, temperature in production culture, and manufacturing site) may reduce product consistency and cause a drift in quality attributes [16]. The drifts in N-linked glycans or charge profiles were reported for the REF products of TRA, RTX, and ETN [12–15]. In the present study, although lot-to-lot variations and trends at some degrees were also observed for some products, no clear upward or downward drifts among lots were observed as reported in previous studies. To ensure the safety and efficacy of biosimilars, it would be important to monitor the trend of each quality attribute through the product lifecycle.

The N-linked glycan structure attached to the Fc and Fc-mediated effector function is important in the comparability study of biosimilars for therapeutic antibodies. In the present study, SPR was used as a more robust method than the cell-based assay, and binding affinity to FcγRIIIa was also evaluated, including the effect of the genetic polymorphism 158V/F. In general, the more afucosylated and high mannose glycans lacking fucose, the higher the binding affinity for FcγRIIIa [21, 23]. Galactose addition has also been reported to increase binding affinity to FcγRIIIa [21, 25, 27]. The correlation between afucosylated glycan content, high mannose content, terminal galactosylated glycan content, and binding affinity in this study is shown in Figs. 8, 9 and 10. Regression analysis was performed on the graphs, and coefficients of determination and p values are shown. A coefficient of determination (R²) close to 1 indicates a good fit for the model, whereas a p value below 0.05 indicates that the content of each glycan may have a significant effect on binding affinity. Overall, the R² value of the plots for the affinity for 158F was higher compared with those for 158V, and the affinity for 158F correlated better with the contents of the three glycan types. This result is consistent with our previous report and suggests that 158F has better discriminative performance as an evaluation method compared with 158V [36]. Furthermore, the extent of the contribution of each glycan to the binding affinity of FcγRIIIa may differ depending on the type of antibody and Fc-fusion protein. As for RTX, although there was no significant effect on high mannose, there was a correlation between the binding affinity for FcγRIIIa and the higher content of afucosylated glycan and galactosylated glycan, consistent with the theory mentioned above. On the other hand, no clear correlation between the binding affinity and the content of each glycan was observed for TRA as well as BEV. RTX_BS1 has approximately 1% lower afucosylated glycan content type and BS2 has approximately 3% higher content compared with REF, which may affect biological and pharmacological activity. In the previous study, the ADCC activities of mixtures of the RTX_REF and afucosylated mAbs at different ratios were measured in peripheral blood mononuclear cells (PBMCs) [36]. This resulted in a detectable percentage of afucosylated mAbs varying depending on the PBMC donor. In one donor, a 1% difference of afucosylated glycan resulted in the difference of ADCC activity, whereas in another donor, a difference of 4% or more was required for obtaining significant difference of ADCC activity. The difference in afucosylated glycan content observed between RTX_REF and biosimilars was small enough to be obscured by individual differences and was considered to have little impact on efficacy.

Fig. 8.

Correlation of binding affinity (K_D) to FcγRIIIa with afucosylated glycan (%). Solid line = linear regression line; dashed line = 95% confidence interval. BEV bevacizumab, IFX infliximab, K_D dissociation constant, RTX rituximab, TRA trastuzumab

Fig. 9.

Correlation of the binding affinity (K_D) to FcγRIIIa with high mannose (%). Solid line = linear regression line; dashed line = 95% confidence interval. BEV bevacizumab, IFX infliximab, K_D dissociation constant, RTX rituximab, TRA trastuzumab

Fig. 10.

Correlation of the binding affinity (K_D) to FcγRIIIa with galactosylated glycan (%). Solid line = linear regression line; dashed line = 95% confidence interval. BEV bevacizumab, IFX infliximab, K_D dissociation constant, RTX rituximab, TRA trastuzumab

In this study, we analyzed the quality attributes related to efficacy and safety of the products. Binding affinities to antigens/Fcγ receptors and N-glycan profiles are related to in vivo pharmacological activity and pharmacokinetics of mAbs and Fc fusion protein. Charge variants may substantially affect these properties of mAbs, depending on the locations and types of modification. Some of these quality attributes in some biosimilar products were out of the quality ranges of REF products. The amount of afucosylated and galactosylated glycans may affect the efficacy of products whose MOA has effector activity, through changing in binding affinities to Fcγ receptor IIIa. Indeed, the content of these glycans corelated to the Fcγ receptor IIIa binding affinity. However, as discussed above, the observed differences are likely obscured by individual differences based on our previous report [36]. Regarding charge variants, biosimilar products outside the quality range exhibited lower content of acidic variants such as deamidated forms, indicating lower potential impact on biological activities. While the higher content of basic variants, which would be due to the levels of C-terminal lysine, was observed in some biosimilars, it was confirmed that the difference does not make a significant impact on antigen binding affinities in our study. Among the quality attributes evaluated in this study, aggregates may impact on the safety and immunogenicity of mAbs. One biosimilar product showed higher content of HMW species than a REF product, but the content itself was within the range of HMW species normally found in mAb products. Therefore, it is highly unlikely that this difference would affect the efficacy or immunogenicity.

The accumulating evidence of biosimilars has brought the new trends in the strategy of biosimilar development. As shown in recent reviews [71, 72], biosimilar products approved so far did not indicate significant concerns regarding safety or efficacy, and therefore, the importance of switching studies or clinical efficacy studies is questioned. This means that comparative analysis of quality attributes of biosimilars and REF products and risk assessment for their impact on safety and efficacy becomes more important in the biosimilarity assessment. Our data can be useful in providing knowledge to support efficient development of biosimilars of mAb and related products.

Conclusion

This study revealed the characteristics of biosimilar products approved and distributed in Japan for the first time at post-approval phase. By analyzing multiple lots of each product, we were able to capture the extent of the differences between a REF product and its biosimilar products as well as to confirm the differences in the regulatory review reports and previous studies. In quality attributes that may affect efficacy and safety, such as afucosylated and galactosylated glycans content and HMW species, lot-to-lot variations and deviation from the quality range of the REF product were observed. Considering the review reports and safety information, the observed differences are unlikely to affect efficacy or safety. This type of lot analysis will enable us to accumulate comprehensive information on the quality of biosimilar products distributed in Japan and to understand the trends. This will lead to the maintenance of a stable supply system of high-quality REF biopharmaceuticals and biosimilars. The information provided by this study is useful for understanding the quality of biosimilars and has significant value as prior knowledge for the future development of biosimilars.

Supplementary Information

Below is the link to the electronic supplementary material.

Declarations

Funding

This research was supported in part by a research grant from the Japan Agency for Medical Research Development (AMED) under grant numbers JP19mk0101152 and JP 22mk0101238.

Conflict of interest

All authors declare no competing financial interests.

Availability of data and material

The authors confirm that the data supporting the findings of this study are available within the article and its supplementary materials.

Ethics approval

Not applicable.

Patient consent to participate/publish

Not applicable.

Author contributions

HS and AI designed all of the experiments. AH conducted glycan, HMW, and charge heterogeneity analysis. MK conducted SPR for FcγR binding affinity. HS analyzed the other quality attributes. HS, YS, and AI obtained the grant for this study. All the authors contributed to the preparation and discussion of the manuscript. All the authors read and approved the final manuscript.