FormalPara Key Points

The PD-1 blocking antibody pembrolizumab is currently marketed as Keytruda® for the treatment of various types of cancer.

FYB206 is being developed as a proposed biosimilar to Keytruda® to increase access to pembrolizumab.

For the analytical similarity study of FYB206 and Keytruda® presented in this manuscript, a wide battery of state-of-the-art analytical technologies has been used to show that the proposed biosimilar is ready for testing in clinical studies.

1 Introduction

FYB206 is being developed by Formycon AG (Martinsried, Bavaria, Germany) as a proposed biosimilar to Keytruda® [international nonproprietary name (INN): pembrolizumab], a recombinant mAb targeting programmed cell death protein 1 (PD-1) [1]. Keytruda® was generated by cloning the variable sequence of a high-affinity mouse antihuman PD-1 antibody onto a human IgG4-kappa isotype antibody containing a stabilizing S228P mutation in the fragment crystallizable (Fc) region. It is produced in Chinese hamster ovary (CHO) cells using heterologous expression technology [2]. Keytruda® was the first so-called checkpoint inhibitor directed against PD-1 approved in the United States of America (USA), developed by Merck & Co. for the treatment of metastatic malignant melanoma [3] and is currently available as a solution for infusion. Following the initial approval, the list of indications for Keytruda® has been extended to a wide variety of cancers, such as non-small cell lung cancer, head and neck squamous cell cancer, classical Hodgkin lymphoma, and gastric cancer.

The immune inhibitory surface receptor PD-1, which is expressed on activated T cells and B cells, plays a critical role in regulating immune responses to tumor antigens and autoantigens. Interaction with its ligands PD-L1 or PD-L2 on an adjacent cell inhibits T cell receptor (TCR) signaling and TCR-mediated proliferation, transcriptional activation, and cytokine production. PD-L1 and PD-L2 are usually expressed on immune cells to avoid autoreactivity, which is exploited by tumor cells to evade the immune system [4,5,6,7,8]. Inhibition of the PD-1/PD-L1 interaction via designed therapeutic antibodies (INNs: pembrolizumab, nivolumab, atezolizumab and avelumab) showed promising results in clinical trials and the molecules are successfully used for the treatment of a large variety of cancers [9, 10]. Through its targeted mode of action pembrolizumab blocks the receptor–ligand interaction and as it belongs to the IgG4 subclass, binding of the molecule to PD1 does not engage Fc receptors or activate the complement system. Consequently, pembrolizumab is devoid of antibody-dependent cell mediated cytotoxicity (ADCC) and complement-dependent cytotoxicity (CDC) [1].

Since biologics are large and complex molecules, many of their characteristics can affect the quality, safety, and efficacy of the products. Thus, a comprehensive analytical assessment of physicochemical, biophysical, and functional parameters is required to prove similarity of a proposed biosimilar with the RMP. These analytical similarity studies, together with clinical data, are then used to assess biosimilarity based on the totality of evidence approach [11]. Development of biosimilar mAbs starts with the extensive characterization of the RMP, which results in the establishment of quality ranges for product quality attributes. Based on their criticality, the attributes are ranked in a critical quality attribute (CQA) assessment to guide the development of the biosimilar candidate. Side-by-side analysis of the proposed biosimilar and RMP using qualified, state-of-the-art analytical technologies is then continuously performed during the development program.

Analogous to Keytruda®, FYB206 is expressed in a CHO cell line system. Within the current study, the analytical similarity of Formycon’s proposed biosimilar FYB206 and Keytruda® is assessed using RMP batches marketed in the USA as well as in the European Union (EU) [12]. Structural/physicochemical and functional similarity are thoroughly evaluated, and any differences in quality attributes between the products are discussed regarding their potential impact on safety and efficacy.

2 Materials and Methods

2.1 Data Analysis and Batches

Regarding the requirements set forth by regulatory agencies to assess whether a proposed biosimilar is indeed similar to the RMP, the European Medicines Agency (EMA) published a reflection paper [13] and the Food and Drug Administration (FDA) a draft guidance [14]. Of note, both publications mostly address the requirements for filing and not requirements for entering the clinical phase, when the number of available batches is necessarily much lower than during later stages of product development. In any case, EMA and FDA mention the possibility of calculating ranges based on the mean and a multiple of the standard deviation (SD) of RMP batches as a possibility for a statistical approach to analyze analytical biosimilarity (commonly referred to as "mean ± x × SD", with x being the multiplier for the SD). Various multipliers have been used successfully in biosimilar applications, usually varying from two to three, sometimes depending on the nature of the attribute and other factors. In this manuscript, a multiplier of three is used for analysis as it is a common factor in process control and provides reliable results even for non-normal distributions of stable processes. Regarding the number of batches necessary to assess analytical similarity, FDA states that for the final submission, a minimum of six independent batches of the proposed biosimilar should be compared against at least ten batches of the RMP. The number of batches that can be analyzed before the start of a clinical trial is typically much lower. In this manuscript, one small scale drug substance (DS) batch representing the optimized early-stage development and one large scale drug product (DP) FYB206 batch manufactured under good manufacturing practice (GMP) were compared with up to eight EU and six US batches of Keytruda®. The results can be used to assess the suitability of FYB206 to enter the clinical phase, while the number of batches is foreseen to be increased for the final assessment of similarity that is needed for a biosimilar application.

Keytruda® batches were acquired and analyzed to develop and qualify methods and to define quality ranges for the assessment of FYB206. All batches included in the comparative analytical assessment (FYB206 and Keytruda®) are listed in Supplementary Table 1.

2.2 Sequence Analysis by LC–ESI–MS and LC–MS/MS

For sequence confirmation, proteins were digested with trypsin and LysC/GluC and were reduced by dithiothreitol. In addition, the digested samples were deglycosylated enzymatically with PNGase F. For analysis of post-translational modifications (PTMs), proteins were digested in parallel with trypsin, trypsin/chymotrypsin, chymotrypsin, GluC/AspN, trypsin/AspN, and trypsin/GluC followed by reduction with dithiothreitol and carbamidomethylated using iodoacetamide. Peptides were separated with reversed phase (RP) chromatography on an Accucore RP – MS LC column (2.1 × 100 mm, 2.6 µm, Thermo Fisher Scientific) with a segmented gradient in water/acetonitrile with 0.1% formic acid. Liquid chromatography coupled to electrospray ionization mass spectrometry (LC–ESI–MS) and tandem mass spectrometry (MS/MS) spectra were obtained using the UltiMate® 3000 system (Thermo Fisher Scientific) coupled to a Q Exactive Orbitrap Plus mass spectrometer (Thermo Fisher Scientific). MS and MS/MS spectra were recorded in positive ion mode with internal mass calibration and data sets were imported into the ProteinScape 2 (Bruker Daltonics, Protagene AG) and Byos® (Protein Metrics) bioinformatics platforms and searched against the sequence databases.

2.3 Structural Assessment

2.3.1 FTIR

With fourier-transformation infrared spectroscopy (FTIR), the secondary structure of proteins can be evaluated. Samples were measured diluted in formulation buffer in a 25 °C tempered AquaSpec transmission cell with a Bruker Tensor II FTIR spectrophotometer. All spectra were acquired over a range of ṽ = 3100 – 1000 cm−1 at 4 cm−1 and were blank corrected. The second-derivative spectrum was calculated (nine smoothing points) and used for estimation of the relative secondary structure composition by automatic comparison (multievaluation) with a reference library from Bruker.

2.3.2 CD

With circular dichroism (CD), the secondary structure and the tertiary structure of proteins can be evaluated. Samples were measured diluted in phosphate buffer with a cell path length of 1 cm for near ultraviolet (UV) and 0.1 cm for far UV using a Jasco J-1500 spectropolarimeter. Buffer spectra were acquired for blank correction and acquired spectra were evaluated using Jasco Software Spectra Manager CFR. The blank-corrected sample spectra were then converted to molar ellipticity.

2.3.3 Nano-DSC

With nano-differential scanning calorimetry (nano-DSC), the thermal stability of proteins can be evaluated. Samples were diluted and scans were performed using a MicroCal PEAQ-DSC automated system and Malvern software. Scans ran on the reference as well as sample cell from 20 to 110 °C at 1 °C/min using the feedback mode none and thermograms were baseline corrected (instrumental, formulation buffer, and transition baseline) and normalized to the concentration. Melting temperatures (protein unfolding) represent the maxima of detected peaks.

2.4 Reduced and Nonreduced SDS–cGE

Size variants of analytes were separated by capillary gel electrophoresis (cGE) on a Beckman Coulter (PA800 plus) instrument equipped with a photodiode array (PDA) detector. The strong anionic detergent sodium dodecyl sulfate (SDS) was used to mask the charged residues of the protein. Denatured protein was incubated for 10 min at 70 °C under either reducing (β-mercaptoethanol) or nonreducing (N-ethylmaleimide) conditions and analytes were separated by voltage according to their respective apparent molecular weights in a bare-fused silica capillary (ID 50 µm, 30.2 cm total length with 20 cm effective length to detector) containing an SDS gel. Peaks were detected at 220 nm and cGE analysis was performed using 32 Karat software.

2.5 SE-HPLC

Size variants of the molecule were separated according to their hydrodynamic volume by size exclusion high performance liquid chromatography (SE-HPLC). Measurements were performed on a Waters system with a stationary phase (TSKGel G3000SWXL, 5 µm, 7.8 mm × 300 mm) from TOSOH using an isocratic gradient with slightly modified phosphate buffered saline (PBS) at pH 7.2. Two wavelengths [214 nm and 280 nm (monitoring only)] were used for detection and peaks were integrated with the Empower software.

2.6 icIEF without and with CPB Treatment

Charged variants were separated and the isoelectric point (pI) was determined with imaged capillary isoelectric focusing (icIEF) using the Maurice C instrument of ProteinSimple. Samples were analyzed either without or with carboxypeptidase B (CPB) treatment (sample:CPB ratio of 40:1, 60 min, 25 °C). After sample preparation, molecules were focused in a linear ampholytic pH gradient (pharmalyte 3–10 and servalyte 6–9, urea, methyl cellulose) through an applied electric field until they reached a pH equal to their pI. For detection by UV absorbance (280 nm) upon focusing, the entire capillary was scanned continuously. Linear regression was used to determine the peak pIs of the samples based on two pI markers using the software Compass for iCE. Integration of sample peaks was performed using the Empower software from Waters.

2.7 Product-Related Variants (e.g. Oxidation) by HIC

Product-related variants were determined by hydrophobic interaction chromatography (HIC) with fluorescence detection using a Waters H-Class Bio UPLC system. After enzymatic digest with CPB, proteins bound to the column (TSKgel Butyl-NPR, 4.6 × 35 mm, 2.5 µm, Tosoh Bioscience), which was equilibrated with a high salt concentration (2 M ammonium sulfate). The molecules were separated and eluted with a decreasing salinity caused by the gradient. Emission was recorded at 350 nm with an excitation at 278 nm. Generated data were evaluated using the Empower software from Waters.

2.8 Glycan Pattern with IPC Labeling by HILIC

Following sample denaturation, N-glycan structures were cleaved from the protein with N-glycanase and released glycans were labeled with Instant Procaine (IPC) dye. After removal of proteins and salts, glycan samples were separated by hydrophilic interaction liquid chromatography (HILIC) according to their structure using a Waters Glycan BEH column (130 Å, 1.7 µm, 2.1 mm × 150 mm). N-glycans were eluted via increasing the polarity of the mobile phase (aqueous ammonium formate buffer in combination with acetonitrile) and fluorescence was used for detection. Hyphenation to a mass spectrometer allowed the determination of compositional masses of the N-glycans for structural elucidation and initial peak annotation.

2.9 Protein Concentration

The protein concentration was determined by measuring the absorption at 280 nm on a Shimadzu UV/VIS spectrophotometer. For calculation, the theoretically determined extinction coefficient was used and moreover, to consider light scattering, the absorption at 320 nm was subtracted from the absorption at 280 nm.

2.10 Cell-Based PD-1/PD-L1 Blockade Potency

For demonstration of potency, a reporter gene assay was applied, which is based on the interaction of two genetically engineered cell lines expressing PD-1 and PD-L1 using the Promega PD-1/PD-L1 blockade bioassay kit (catalogue numbers: J1250, J1255). The PD-L1 receptor of PD-L1 aAPC cells binds to the PD-1 receptor of PD-1 effector cells (engineered Jurkat T cells, Promega catalogue number: J115A), which disables the expression of luciferase in the PD-1 effector cells induced by the TCR and the nuclear factor of activated T cell—response element (NFAT-RE). In case binding of PD-1 and PD-L1 is disrupted, e.g., by an antibody such as pembrolizumab, TCR induces the expression of luciferase mediated by NFAT-RE. Addition of Bio-Glo reagents enable the detection of luminescence generated by the substrate turnover of the luciferase with a plate reader in relative luminescence units (RLU).

PD-L1 aAPC/CHO-K1 cells were added to 96-well plates and incubated at 37 °C, 5% CO2. Afterwards, a serially diluted sample and a fixed amount of effector cells were added and the mixture was further incubated at 37 °C, 5% CO2. Bio-Glo reagent was added and after incubation at room temperature, the luminescence was read on a plate reader. The relative potency was determined by comparison of the 10-point dose–response curve of the sample and the reference standard, using the 4-parameter logistic (4-PL) curve fit function of PLA software. Reported are the homogeneously weighted means of three independent experiments.

2.11 PD-1 Potency ELISA

Relative binding potency was evaluated with an indirect enzyme-linked immunosorbent assay (ELISA) in which PD-1 was coated on a 96-well plate. After blocking, serially diluted samples were added and bound to PD-1. Signals were detected using a horseradish peroxidase (HRP)-labeled antibody directed against human kappa light chains in combination with 3,3′,5,5′-tetramethylbenzidin (TMB) as colorimetric substrate. The development was stopped by addition of sulfuric acid and the plate was read on an ELISA plate reader at 450 nm. Obtained signal is proportional to the concentration of the sample. The relative potency was determined by comparison of the 8-point dose response curve of the sample and the reference standard, using the 4-PL curve fit function of PLA 3.0 software. Reported are the means of two independent experiments.

2.12 PD-1 Binding and Fc Receptor Binding Analyzed with SPR

PD-1, neonatal Fc receptor (FcRn) as well as Fc gamma receptor I and III (FcyRI and III, CD64 and CD16) binding was analyzed using surface plasmon resonance (SPR) on a Biacore 8K (Cytivia) instrument. Samples were captured to a streptavidin coupled, CaptureSelect human anti-Fab-kappa conjugate loaded on a CM5 sensor chip. In individual cycles, an increasing concentration series (6-point dilution series in triplicate) of PD-1 or recombinant human Fc receptor was injected across the captured sample with surface regeneration performed between each injection, which reflects an inverted design with sample as capture solution (ligand) and PD-1 or Fc receptor as analyte. Data were analyzed and reviewed with the insight evaluation software version 5.0.18.

For PD-1 and CD64 binding, the 1:1 Langmuir model was applied. The association and dissociation rate constants (ka and kd) were calculated to determine the global dissociation constants (KD). For CD16, the normalized binding response was calculated as the mean steady state response in resonance units (RU) achieved, normalized against the mean capture level and adjusted to response per 500 RU of sample captured. For FcRn binding, the steady state affinity model was applied to calculate the KD.

2.13 C1q Binding ELISA

Complement binding was analyzed using an ELISA with a VersaMax microplate reader and SoftMax Pro software. The sample dilution series was coated onto a 96-well plate and incubated with the complex protein C1q. An HRP labeled anti C1q antibody was used for detection. HRP activity was analyzed using the chromogenic substrate TMB and the enzymatic reaction was stopped by acidification. The optical density (OD) was measured at 450–655 nm and C1q binding was reported in per cent relative to the IgG1 isotype reference at its estimated EC50.

2.14 CDC and ADCC Assay

CDC and ADCC were measured using KILR Jurkat PD-1 cells (DiscoverX catalogue number: 97-1037C019) as target cell line, which was engineered to stably express a housekeeping protein that is tagged with enhanced ProLabel (ePL), a β-gal enzyme fragment, using the KILR Retroparticles. In addition, normal human serum (NHS) complement (for CDC) or CD16 effector cells (for ADCC) were applied using the DiscoverX KILR cytotoxicity assay kit (catalogue number: 97-0001L). Activity was assessed by incubating KILR Jurkat PD-1 cells with high, medium, and low concentrations of test sample together with NHS complement (for CDC) or CD16 effector cells (for ADCC, with an effector to target cell ratio of 5:1). Moreover, a positive control sample consisting of an IgG1 variant of pembrolizumab, a total lysis control and a spontaneous release control were included. Binding of the complement system (for CDC) or CD16 effector cells (for ADCC) to the Fc part of the sample antibody, leads to cell lysis and thereby to a release of the ePL-tagged protein. Addition of detection reagent, which contains an enzyme acceptor fragment of the β-gal enzyme, leads to the formation of the active β-gal enzyme, which hydrolyzes the substrate to give a chemiluminescent output, detected on a luminescence reader. CDC activity was calculated as percentage of the mean response levels of triplicate analyses of each sample, reference and control of the total lysis control. The ADCC activity was calculated in the same way as the CDC activity, except that the spontaneous release control was considered as well.

3 Results and Discussion

Biologic products are typically very complex molecules and difficult to characterize. Applying the totality of evidence approach, which consists of assessing results obtained by analytical as well as clinical studies, the objective of a biosimilar development program is to demonstrate high similarity to the licensed RMP. The analytical panel to evaluate similarity in the current study was designed on the basis of available information on pembrolizumab and knowledge gained during initial RMP analyses. In Table 1, an overview of quality attributes, evaluation criteria, methods, and results are shown. A criticality score is assigned to each quality attribute of pembrolizumab, and quality ranges are provided where applicable. The wide battery of qualified and fit for purpose analytical and functional tests was assigned to categories such as structural characterization, product-related variants, glycosylation, general properties, and biological function. FYB206 was assessed against ranges corresponding to the mean ± three SD of all analyzed RMP batches for quantitative assays. Results from nonquantitative assays were evaluated qualitatively. The results of the extensive data package of the outlined analytical study, which are discussed in detail in the following, show that FYB206 is analytically highly similar to the approved RMP Keytruda®. Thus, FYB206 is well suited to enter the clinical phase.

Table 1 Analytical panel with different categories, quality attributes, and results of Keytruda® and FYB206
Full size table

3.1 Structural Characterization

Primary structure (amino acid sequence) and all higher-order structure attributes are highly similar between FYB206 and Keytruda®. Intact molecular mass spectrometry and LC–ESI–MS/MS peptide map analysis were chosen as methods for amino acid sequence confirmation (Tables 1, 2). Secondary as well as the tertiary structure were evaluated using FTIR, far UV, and near UV CD spectroscopy. Moreover, disulfide linkages and thermodynamic stability of the higher-order structure were analyzed with LC–ESI–MS/MS and nano-DSC, respectively (Table 1). In addition, integrity and a high similarity of the three-dimensional (3D) structure can be deduced indirectly with the help of binding assays, which are described within the section on biological function.

Table 2 Results of LC–ESI–MS/MS sequence analysis
Full size table

3.1.1 Primary Structure

The amino acid sequence of a biosimilar is expected to be identical to the primary structure of the corresponding RMP. With tandem mass spectrometry (LC–ESI–MS/MS) sequencing applied to peptides of two digests (trypsin and LysC/GluC, both deglycosylated), full sequence coverage of heavy chain (HC) and light chain (LC) of all products was achieved as presented in Table 2. In addition, the molecular masses of the unmodified LC and HC as well as the expected major variants (HC with one glutamine to pyroglutamate conversion, HC with nonclipped C-terminal lysine) were confirmed by intact mass spectrometry. The results confirm that FYB206 and Keytruda® have identical primary structures.

3.1.2 Higher-Order Structure

The higher-order structure is crucial for the bioactivity of the molecule and misfolded proteins may induce an immune response. FTIR spectroscopic profiles are shown in Fig. 1a, b. The mean absorption band maximum at 1640 cm−1 indicates a strong beta-sheet content, which is typical for antibodies. The content of alpha-helices was of low abundance and minor amplitude differences can be attributed to method variability. Spectra are qualitatively comparable, demonstrating that the secondary structure of FYB206 and Keytruda® is highly similar.

Fig. 1
figure 1

Secondary structure analysis. Overlay of FTIR spectra (a) and the related second derivative (b) of one FYB206 batch (red) and Keytruda® EU (green) as well as USA (blue). EU European Union, FTIR fourier-transformation infrared spectroscopy, FYB206 project-specific identifier of the proposed biosimilar, USA (US) United States of America

Full size image

Overlays of far UV and near UV CD spectra are depicted in Fig. 2a, b, respectively. Head-to-head secondary structure analysis with far UV CD showed a high content of beta-sheets, a low content of alpha-helices and intermediate amounts of turns and random coils as expected for pembrolizumab. With qualitatively comparable patterns, the analysis confirmed the high similarity in secondary structure between FYB206 and Keytruda®.

Fig. 2
figure 2

Higher-order structure analysis. Overlay of far-UV (a) and near-UV (b) CD spectra as well as overlay of melting curves (c) of one FYB206 batch (red) and Keytruda® EU (green) as well as USA (blue). CD circular dichroism, EU European Union, FYB206 project-specific identifier of the proposed biosimilar, USA (US) United States of America, UV ultraviolet

Full size image

CD effects in the near UV range reflect the native tertiary structure and are considered as a fingerprint region of the protein. Profiles contain signals for aromatic amino acids and disulfides (250–310 nm) in the typically expected regions, which indicates proper folding of the protein. Visual inspection of spectra obtained by near UV CD shows qualitatively comparable spectra for FYB206 and Keytruda®, thus confirming a high similarity between the products also for the tertiary structure.

Unbridged cysteines in mAbs may result in decreased stability of the molecule and corresponding lower melting temperatures [15]. Therefore, disulfide linkages, which are important structural elements and contribute to the integrity of the molecule, were evaluated by LC–ESI–MS/MS. All expected disulfides were detected for Keytruda® and FYB206 (data not shown). Moreover, nano-DSC was employed to analyze thermodynamic stability and melting curves are presented in Fig. 2c. Visual comparison of profiles reveals high similarity between FYB206 and Keytruda®.

3.2 Product-Related Variants

Product-related variants such as low molecular weight (LMW), high molecular weight (HMW), and acidic and basic species, as well as oxidized forms are important attributes for the quality of the product. They may impact potency, pharmacokinetics (PK), and immunogenicity of biological products. To assess size and charge variants as well as oxidized states of FYB206 and Keytruda®, SDS–cGE, SE-HPLC, icIEF, HIC, and LC–ESI–MS/MS were applied. Overall, high similarity was shown between FYB206 and Keytruda®, with only minor differences that are not considered to be clinically meaningful.

3.2.1 LMW Species

LMW species were investigated with nonreduced SDS–cGE as well as reduced SDS–cGE, and the results of a side-by-side analysis of FYB206 and Keytruda® are illustrated in Fig. 3a, b. In electropherograms of the nonreduced condition SDS–cGE analysis of FYB206 and Keytruda®, the same peaks are resolved, i.e., one main peak and several minor LMW peaks. No additional or missing LMW species were observed for FYB206, and the profile is visually similar to Keytruda®. The overall sum of LMW species in FYB206 is 0.2 percentage points lower (small scale batch) or within the mean ± 3 SD range of Keytruda® (large scale batch). Correspondingly, the percentage of the main peak of FYB206 is slightly higher (small scale batch) or within the mean ± 3 SD range of Keytruda® (large scale batch). As the differences are minute, only observed for the small-scale batch and as all biological function attributes show high similarity (see later sections), the minor differences between the sum of LMW peaks are not considered to be clinically meaningful.

Fig. 3
figure 3

Size variant analysis. Overlay of electropherograms from nonreduced SDS-cGE (a) as well as reduced SDS–cGE (b) analysis and zoomed overlay of SE-HPLC chromatograms (c) of one FYB206 batch (red) and Keytruda® EU (green) as well as USA (blue). EU European Union, FYB206 project-specific identifier of the proposed biosimilar, HC heavy chain, HMW high molecular weight, LC light chain, LMW low molecular weight, ngHC non-glycosylated HC, non-red./ red. SDS-cGE non reduced/ reduced sodium dodecyl sulfate capillary gel electrophoresis, SE-HPLC size exclusion high performance liquid chromatography, USA (US) United States of America

Full size image

Regarding reduced SDS–cGE analysis both batches are well within the mean ± 3 SD range of Keytruda® for the main peak as well as for the sum of LMW species. Moreover, electropherograms of FYB206 and Keytruda® show a similar overall profile.

3.2.2 HMW Species

HMW species were investigated with SE-HPLC. Chromatograms of a head-to-head analysis of FYB206 and Keytruda® are depicted in Fig. 3c and visual comparison reveals a similar profile with no new or missing peak for FYB206 compared with Keytruda®. For all batches, one main peak and one HMW peak was resolved with similar peak shapes. HMW levels of FYB206 are very low and were found to be within the mean ± 3 SD range of Keytruda®.

3.2.3 Acidic and Basic Species

Charge variants such as acidic and basic species were analyzed with icIEF under native conditions as well as after CPB (carboxypeptidase B, an enzyme that cleaves off C-terminal lysines) treatment. Without prior CPB digest, basic species of FYB206 are above the mean ± 3 SD range of Keytruda®. However, the sum of acidic species as well as the main peak is within the Keytruda® range for the batch produced at large scale. Electropherograms of representative FYB206 and Keytruda® batches are shown in Fig. 4a, and charge distribution profiles were overall similar with no additional species observed. For both products, seven charge isoforms were determined, with isoelectric points between ~7.0 and ~7.8. The main peak exhibited a pI of ~7.4. As common for IgG molecules, basic species might be further divided into subpopulations containing nonclipped C-terminal lysine on only one of the two C-termini or on both termini leading to different basic peaks in the electropherogram. Data from mass spectrometry (data not shown) as well as icIEF analysis following prior CPB digest demonstrate that the higher sum of basic species is dominated by the higher degree of residual, unprocessed C-terminal lysines of FYB206 when compared with Keytruda®, as can be seen in Fig. 4b. C-terminal lysine variants are known to not affect efficacy or safety [16]. Moreover, analysis with orthogonal cation exchange chromatography (data not shown) further confirmed the observations from icIEF. The slightly different charge profile caused by the variation of C-terminal lysines is not considered to be relevant in patients due to the rapid cleavage of these lysines in serum [17, 18]. In addition, potency of FYB206 within binding and neutralization assays is not influenced by this difference. Thus, the difference in C-terminal lysine content does not preclude overall similarity.

Fig. 4
figure 4

Charged and oxidized species analysis. Overlay of icIEF electropherograms of native (a) as well as CPB treated (b) samples and overlay of HIC chromatograms (c) of one FYB206 batch (red) and Keytruda® EU (green), as well as USA (blue). CPB Carboxypeptidase B, EU European Union, FYB206 project-specific identifier of the proposed biosimilar, HIC hydrophobic interaction chromatography, icIEF imaged capillary isoelectric focusing, USA (US) United States of America

Full size image

3.2.4 Protein Modifications

Protein modifications such as oxidation, deamidation, aspartate/isoaspartate isomerization, pyroglutamate formation, C-terminal lysine clipping, and glycation were analyzed by LC–ESI–MS/MS. Except for lysine clipping, FYB206 data were comparable to the RMP (data not shown).

In the following, results of protein oxidation analysis are outlined in more detail. According to structural considerations (PDB-ID: 5DK3), oxidation of a specific amino acid (Met105, HC) in the Fab domain could presumably directly reduce target binding and therefore potency. In addition to LC–ESI–MS/MS evaluations, oxidation was assessed using HIC with prior CPB digest to remove complexity of the peak profile due to different retention times of lysine-carrying variants; the respective HIC chromatograms are shown in Fig. 4c. Characterization analyses identified peak 1 to contain predominantly species with both Met105 residues oxidized. Peak 2 contains the singly oxidized precursor, and peak 3 corresponds to the main peak with only minor Met105 oxidation. For the most recently tested large scale FYB206 batch, peak areas representing Met105 oxidation are within the mean ± 3 SD range of Keytruda®, and hence, similarity is given. LC–ESI–MS/MS data confirmed similar oxidation levels for the Fab domain for both products while a slightly higher oxidation of FYB206 was detected in one amino acid residue of the Fc domain when compared with Keytruda®. More specifically, for Met252 in the Fc domain the level of oxidation was slightly (by 0.1–0.2%) outside of the mean ± 3 SD range (Table 1), and the modification of Met252 may potentially have an impact on FcRn binding [19]. However, according to literature, much larger differences in oxidation at the FcRn binding site are necessary to affect PK significantly and the minor differences between FYB206 and Keytruda® are expected to have no detectable influence [20]. Indeed, SPR analysis confirmed similarity in FcRn binding between FYB206 and Keytruda® (see further below) and the minute difference in one oxidation site is not considered to pose any clinical safety or efficacy concerns.

3.3 Glycosylation

Keytruda® and FYB206 both contain a single N-glycan at asparagine (N)297 of each heavy chain and glycan site occupancy was assessed by LC–ESI–MS/MS (data not shown) and reduced SDS–cGE. The glycopattern was analyzed with HILIC following IPC labeling of cleaved glycans. While the glycan site occupancy and the qualitative glycan pattern were highly similar between FYB206 and Keytruda®, some minor quantitative differences were detected for afucosylated and galactosylated structures. Quantitative differences in glycans on the Fc part of a monoclonal antibody could theoretically affect the effector functions ADCC and CDC. However, it is known for Keytruda® and was shown for FYB206 that pembrolizumab, which is an IgG4, does not exhibit ADCC or CDC (see further below). This matches the widely spread assertion that the glycosylation state of the IgG4 Fc region is not expected to significantly affect its activity [21, 22]. Therefore, these minor differences are justified as not being clinically meaningful.

3.3.1 N-Glycan Occupancy

Under reducing conditions, disulfide bonds are cleaved and two main peaks corresponding to the LC as well as the HC portion of the molecule are detected with SDS–cGE. In addition, a minor peak, which is assigned to the nonglycosylated (ng)HC, can be observed. Reduced SDS–cGE electropherograms shown above (see Fig. 3b), revealed very low ngHC levels and all tested FYB206 batches are within the mean ± 3 SD range of Keytruda®.

3.3.2 Glycopattern

Side-by-side glycan map analysis comparing FYB206 and Keytruda® is shown in Fig. 5. Peak annotation was possible for all relevant peaks, with the total sum of unknown peaks representing < 1.0% of the total N-glycan pool. Mainly, complex biantennary and high mannose structures (M4 to M8) were detected, with FA2 being the main structure of the pool. The second largest group comprised the galactosylated complex biantennary N-glycans (e.g., FA2G1/FA2G1′). Low contents of sialylated [e.g., FA2G1S1 (Neu5Ac)] and bisected structures (e.g., FA2B) were also found. Overall, the glycan profiles are visually similar with no relevant species newly detected or missing for FYB206 compared with Keytruda®. Further, classification of the identified glycans into four groups (high mannose, sialylated, afucosylated, and galactosylated glycans) was conducted and the structures were quantified. For FYB206, the content of high-mannose type as well as sialylated glycans is within the mean ± 3 SD range of Keytruda®. Afucosylated glycans of FYB206 are slightly outside of the narrow quality range of Keytruda® and the most pronounced difference is in terminal galactosylation. Concerning afucosylation, slightly lower levels can potentially impact and reduce FcyRIIIa binding of IgG1 antibodies, which could in turn lead to differences in ADCC. However, no decrease in binding was detected for FYB206 with SPR analysis (Table 1). Also, Pembrolizumab is an IgG4 antibody and, as expected for this class of molecules, no ADCC activity was observed for FYB206 and Keytruda®. Therefore, the slight difference in afucosylation is not of functional relevance. The higher level of terminal galactosylation of FYB206 similarly also has no functional relevance. Analysis of C1q binding and CDC activity, which could be affected by the galactosylation level [23, 24], showed comparable and expected responses for Keytruda® and FYB206 (Table 1, IgG4 antibodies are not expected to show any relevant C1q binding and CDC activity). Therefore, the identified differences in glycosylation are not considered to be clinically meaningful.

Fig. 5
figure 5

Glycostructure analysis. Stack plot of glycan map chromatograms of one FYB206 batch (red) and Keytruda® EU (green) as well as USA (blue). EU European Union, FYB206 project-specific identifier of the proposed biosimilar, USA (US) United States of America

Full size image

3.4 General Properties

An important attribute is the protein concentration because biosimilars must demonstrate a comparable dosage strength with the RMP to achieve highly similar efficacy. The protein concentration was measured using UV absorption at 280 nm. While the small scale development batch was slightly below the similarity range, the large scale batch was well within the mean ± 3 SD range of Keytruda®, demonstrating the capability of the scaled-up manufacturing process to accurately achieve the required protein concentration.

3.5 Biological Function

All functional investigations confirmed the high similarity of FYB206 to the RMP. SPR was applied to analyze FcRn binding as well as interactions with FcyRI and FcyRIII, which are not part of the mechanism of action (MoA) of Keytruda®. Results demonstrated high similarity between Keytruda® and FYB206 regarding FcRn, FcyRI, and FcyRIII binding (Table 1). When C1q binding was tested with an ELISA format and the effector function mechanisms ADCC and CDC were investigated with cell-based systems, the results showed that neither Keytruda® nor FYB206 exhibited relevant C1q binding, and both were negative for ADCC and CDC (Table 1).

Pembrolizumab competitively blocks the interaction between PD-1 and PD-L1 [25]. A cell-based competitive assay and an SPR-based PD-1 binding set-up were selected to precisely examine the potency and drug target binding [22]. In addition, a binding ELISA was applied to substantiate the results. With all methods, FYB206 was shown to be highly similar to Keytruda® regarding PD-1 binding and bioactivity (Table 1). Consequentially, with highly similar protein–protein and complex interaction analysis results, the correct folding and quaternary 3D structure of FYB206 was further confirmed by comparison with Keytruda® using the biological assays.

3.5.1 FcRn Binding and Fc Related Interactions

FcRn binds to the Fc domain of IgG, ultimately protecting it from degradation and is, thus, one of the key factors for the clinical efficacy of antibody biologics to ensure a long in vivo half-life. Evaluation with SPR identified FYB206 to be within the mean ± 3 SD range of Keytruda® as can be seen in Table 1; a similarity plot is shown in Fig. 6.

Fig. 6
figure 6

FcRn binding evaluation. Similarity plot of the equilibrium dissociation constant from Biacore analysis of the binding of FcRn to FYB206 (red) and Keytruda® EU (green) as well as USA (blue). "Mean Total" refers to the combined range of Keytruda EU and USA. EU European Union, FcRn neonatal fragment crystallizable receptor, FYB206 project-specific identifier of the proposed biosimilar, KD equilibrium dissociation constant, USA (US) United States of America

Full size image

FcyR interactions are not part of the MoA of Keytruda® and IgG4 type antibodies bind to FcyRI but are expected to exhibit only low affinity toward FcyRIIIa. Due to possible effector functions, the ability of FYB206 to bind to these FcyRs was investigated by SPR. The results revealed similarity in binding to FcyRI as well as FcyRIIIaV between FYB206 and Keytruda® (Table 1). Although binding to FcyRI was shown, this receptor has only limited availability to partake in effector functions in vivo since it is heavily occupied by serum IgG1 or IgG3 due to the higher affinity when compared with IgG4 [26,27,28,29]. For the very low affinity FcyRIIIa binding, similarity of FYB206 was demonstrated despite differences detected in afucosylation (Table 1). This lack of correlation between afucosylation and FcyRIIIa binding is most likely since the afucosylation differences between the two IgG4 products are small. Binding of the first subcomponent of the C1q complex of the classical pathway of complement activation was tested for characterization purposes only, as it has no role in efficacy or safety of Keytruda®. Both, FYB206 and Keytruda® did not exhibit any relevant C1q binding activity. As expected, the assessment of effector functions with cell-based assays for ADCC and CDC demonstrated the absence of activity for FYB206 and Keytruda®.

3.5.2 PD-1 Binding and Neutralization of PD-1

The MoA of FYB206 is inhibition of the interaction between PD-1 and PD-L1/PD-L2. The binding to PD-1 was investigated by SPR and parameters of the FYB206 batches (equilibrium dissociation constant, dissociation rate and association rate) are within the mean ± 3 SD range of Keytruda®, which is illustrated in Fig. 7.

Fig. 7
figure 7

PD-1 binding evaluation. Similarity plot of the equilibrium dissociation constant (a), the dissociation rate (b) and the association rate (c) from Biacore analysis of binding of PD-1 to FYB206 (red) and Keytruda® EU (green) as well as USA (blue), respectively. "Mean Total" refers to the combined range of Keytruda EU and USA. EU European Union, FYB206 project-specific identifier of the proposed biosimilar, KD equilibrium dissociation constant, koff dissociation rate, kon association rate, PD-1 programmed cell death protein 1, USA (US) United States of America

Full size image

Target binding analysis with an ELISA format confirmed the results obtained with SPR and a concentration-response curve is shown in Fig. 8a. Further, a cell-based bioluminescent PD-1/ PD-L1 blockade method was applied to closely mimic the in vivo activity of pembrolizumab. When PD-1 effector cells and PD-L1 aAPC/CHO-K1 cells interact, TCR signaling and NFAT-RE-mediated luminescence is inhibited. The assay reflects the neutralization of the PD-1/PD-L1 signaling by pembrolizumab, with TCR activation and subsequent luminescence being triggered by suppression of the interaction. Dose–response curves of the batches were compared with the reference standard to determine relative potencies. A representative evaluation is shown in Fig. 8b. The FYB206 batches were found to be within the mean ± 3 SD range of Keytruda® (Table 1).

Fig. 8
figure 8

Potency analysis. Typical concentration-response curve of the PD-1 binding ELISA (a) and of the cell-based neutralization assay (b) comparing FYB206 (red) and Keytruda® EU (green) as well as USA (blue). Bars indicate the standard deviation. ELISA enzyme-linked immunosorbent assay, EU European Union, FYB206 project-specific identifier of the proposed biosimilar, PD-1 programmed cell death protein 1, USA (US) United States of America

Full size image

Overall, the orthogonal bioassays addressing the MoA led to highly similar results for FYB206 and Keytruda®. Therefore, the biological function due to Fab-mediated binding, which is essential for the clinical efficacy of FYB206, is highly similar to Keytruda® as shown in Fig. 9.

Fig. 9
figure 9

Potency evaluation. Similarity plot of the PD-1 binding ELISA (a) and of the cell-based neutralization assay (b) comparing FYB206 (red) and Keytruda® EU (green) as well as USA (blue)."Mean Total" refers to the combined range of Keytruda EU and USA.  ELISA enzyme-linked immunosorbent assay, EU European Union, FYB206 project-specific identifier of the proposed biosimilar, PD-1 programmed cell death protein 1, USA (US) United States of America

Full size image

4 Conclusions

The purpose of the comparative assessment was to evaluate the analytical and functional similarity of the proposed biosimilar FYB206 in comparison with Keytruda® and to, thus, provide a solid foundation for the clinical phase. In summary, the results from all assessments covering structure, purity, protein variants, binding and potency confirm a high similarity between the two products. The minor differences in physicochemical attributes between FYB206 and Keytruda® are not clinically relevant and FYB206 has the characteristics to be a safe and efficacious alternative to Keytruda®. Thus, the results warrant testing of the proposed pembrolizumab biosimilar FYB206 in clinical studies to confirm similarity between both products. After successful execution of clinical studies and regulatory approval of the biosimilar version of pembrolizumab, patient access to this essential medicine will be improved. In the context of this study, it is worthwhile to point out that the sensitivity that comes with these types of analytical similarity exercises typically far outperforms the sensitivity of a clinical efficacy study. In fact, the outcome of standalone clinical efficacy studies has never been a decisive factor for biosimilar development [30]. While a comparative PK study adds considerable value to the analysis of a proposed biosimilar, the need for a separate efficacy study is less clear, especially when the target and MoA is as well defined as in the case of pembrolizumab (exerted solely via binding to PD-1 receptor), the biosimilar candidate shows high similarity for all clinically relevant functional assays, and safety and immunogenicity are sufficiently addressed. This topic is currently under much discussion in the biosimilar space, where industry and regulators exchange their views on the subject [31]. In the current study, all functional attributes with a potential impact on clinical performance (neutralization of PD-1, PD-1 binding, FcRn binding) show a high degree of similarity, and minor physicochemical differences are justified to not have an impact on clinical performance. While analytical assays may be added during FYB206 development to further broaden and deepen the knowledgebase, one can envision a future scenario in which, for biosimilar candidates such as Pembrolizumab, the requirement for a separate efficacy study is waived on the basis of comprehensive analytical and PK results, always taking into consideration safety and immunogenicity aspects. Waiving the need for a separate study would help to reduce costs and further expand patient access to high quality biologics.