RGB-19 product development was carried out over the course of several years using an extensive panel of state-of-the-art molecular and functional methods to demonstrate analytical similarity to the RoActemra® reference product. All methods used in the similarity study were in their final validated or qualified state. Altogether, 44 analytical methods and functional assays (supplementary Table 1; see the electronic supplementary material) were selected based on their sensitivity and capability to detect differences in the CQAs and used during the similarity assessment. The descriptions of the major and critical methods and the obtained results are given in this publication, while the remaining descriptions and data are provided in the electronic supplementary material, as indicated in supplementary Table 1. Based on the literature of biosimilarity analytical and functional studies [12,13,14,15,16,17,18], a higher-than-average number of methods was applied in the current study. Each method was carefully selected following an extensive review of scientific and regulatory sources, informed by the company’s prior experience in IgG1 characterisation strategies. The method package included a large number of structural characterisation methods providing deep insights into minor molecular features and an understanding of differences experienced in physico-chemical measurements.

2.1 Data Evaluation Strategy for Similarity Assessment

For data analysis, two approaches were used to statistically evaluate the experimental results. For the quantitative results of the physico-chemical and bioassay measurements, the range approach statistical procedure was used. The quality range calculation was based on the RoActemra®(Mean) ± X*SD formula, determined from the results of the analysis of available EU-sourced RoActemra® batches measured continuously during the development and validation phase of the project (mean value of the RoActemra® QTPP data). The “SD” is the standard deviation of the RoActemra® QTPP data, and the “X” value is the multiplier, which was determined to be 3 for the majority of physico-chemical and bioassay methods and determined to be 2.5 for the most critical release bioassay methods (cell-based anti-proliferation assay and sIL-6R binding [enzyme-linked immunosorbent assay ]), which are based on the clinically relevant mechanism of action of the product. Because of the different formulations and drug product manufacturing processes, separate quality ranges were defined for the RGB-19 SC and IV drug products. For some quality attributes such as N-glycan forms and sialic acid content, which are not dependent on the formulation and the drug product manufacturing process, a quality range combined from SC and IV RoActemra® data was used.

The results of structural characterisation identity methods, including intact and subunit analyses, were evaluated by parallel analyses and compared to theoretical values. No multiplier was defined for negative functional assays, some biophysical and sub-visible particle methods.

Overall, 47 EU-sourced RoActemra® IV and 33 EU-sourced RoActemra® SC batches were analysed during the product development and validation phase. The RoActemra® batches were measured over several years to determine the target product profile and quality ranges. In all cases, the measurements were performed within the expiry date of the RoActemra® RMP batches. The final similarity study for marketing authorisation application (MAA) submission was performed with six RGB-19 IV and five RGB-19 SC drug product batches, which were produced from independent drug substance batches. In line with the EMA guidelines [9, 10], all batches of RGB-19 drug product that were used for the analytical similarity assessment study were manufactured using the final commercial process and scale.

2.2 Materials

The biosimilar candidate RGB-19 IV (20 mg/mL concentrate for solution for infusion) and SC (162 mg/0.9 mL solution for injection) drug products were produced by Gedeon Richter Plc. The RoActemra® RMPs were purchased from Roche Registration GmbH and stored and handled according to the EMA summary of product characteristics document [3]. For the comparative analyses, the drug products were taken from their original containers; the active substance was not extracted prior to analysis.

2.3 Analytical Methods2.3.1 Intact and Subunit Molecular Mass Analysis and Intact Glycation Analysis by Liquid Chromatography–Mass Spectrometry (LC–MS)

Intact mass analysis was performed after a dilution step on a Shimadzu Nexera X2 ultra-high-performance liquid chromatography (UHPLC) instrument coupled to a Bruker maXis II Q-TOF mass spectrometer. The on-line desalting of intact samples was performed on a Waters Acquity BEH C4 column (2.1 × 100 mm, 1.7 µm, 300 Å) at 80 ºC. For the separation, gradient elution was applied (eluent A%: H2O = 100 [V/V%]; eluent B%: H2O:acetonitrile [ACN] = 10:90 [V/V%]; eluent C%: ACN = 100 [V/V%]; eluent D%: H2O:concentrated [cc.] trifluoroacetic acid [TFA] = 100:1 [V/V%]) and the flow rate was 0.4 mL/min. The same chromatographic conditions were applied for the intact glycation analysis where the samples were diluted and additionally treated with PNGaseF (New England BioLabs) and CPB (Sigma) enzymes. For the subunit analysis, the samples were digested by IdeS enzyme (Genovis) and were reduced using dithiothreitol (DTT) under denaturing conditions. The separation of the subunits (Fc/2, light chain [LC] and Fd’) was performed on the same high-performance liquid chromatography (HPLC) column at 68 ºC, with a flow rate of 0.33 mL/min, using gradient elution (eluent A%: H2O = 100 [V/V%]; eluent B%: H2O:ACN = 10:90 [V/V%]; eluent C%: H2O:cc. formic acid [FA] = 100:1 [V/V%]; eluent D%: H2O:cc. TFA = 100:1 [V/V%]). The spectra were averaged and deconvoluted using the BioPharma Compass software (Bruker). The deconvoluted mass spectra, the measured average mass (intact) or monoisotopic mass (subunit) values were compared. For glycation analysis, the relative intensity of the mono-glycated peak was determined.

2.3.2 Reduced and Non-reduced Peptide Mapping Analysis by LC-MS (with Lys-C and Trypsin Enzymes)

For reduced peptide mapping, the samples were first denatured at neutral pH, then digested with Lys-C enzyme (Fujifilm Wako Pure Chemical Corporation). Finally, the disulfide bridges were reduced using tris(2-carboxyethyl)phosphine (TCEP). As a first step of the non-reduced peptide mapping, the samples were denatured at neutral pH. The free thiol groups were labelled with N-ethylmaleimide (NEM). The samples were digested by Lys-C (Fujifilm Wako Pure Chemical Corporation) and Trypsin (Promega) enzymes. Finally, the digestion was stopped by the addition of acetic acid. The peptides were separated on a Waters Acquity BEH Phenyl column (1.7 μm, 2.1 × 150 mm, 130 Å) at 73 ºC using gradient elution (eluent A%: H2O:cc. TFA = 100:0.1 [V/V%]; eluent B%: H2O:ACN:methanol [MeOH]:cc. TFA = 20:50:30:0.1 [V/V%]). The flow rate was 0.31 mL/min. The peptide mapping analysis was performed on a Shimadzu Nexera X2 ultra-performance liquid chromatography (UPLC) instrument coupled to an Orbitrap Fusion Tribrid instrument (Thermo Fisher Scientific). Data evaluation of the MS and tandem MS (MS/MS) spectra was performed using the BioPharma Finder software (Thermo Fisher Scientific). A semi-quantitative analysis of post-translational modification (PTM) levels was also carried out.

2.3.3 Reduced Peptide Mapping Analysis by LC-MS (with Chymotrypsin Enzyme)

In order to achieve 100% sequence coverage of the protein at the peptide level, a digestion with Chymotrypsin (Promega) enzyme was also performed. After the denaturation of the protein at neutral pH, the samples were digested with Chymotrypsin enzyme and the disulfide bridges were reduced using TCEP. The peptides were separated on a Waters Acquity UHPLC Peptide CSH C18 column (1.7 μm, 2.1 × 150 mm, 130 Å) at 55 °C column temperature using gradient elution (eluent A%: H2O:cc. FA = 100:0.1 [V/V%]; eluent B%: H2O:ACN:cc. FA = 50:50:0.1 [V/V%]) with a flow rate of 0.3 mL/min. A Shimadzu Nexera X2 UPLC coupled to an Orbitrap Fusion Tribrid instrument (Thermo Fisher Scientific) was used for the measurements.

2.3.4 Free Thiol Content by Ellman’s Assay

The number of free thiol groups (sulfhydryl, -SH) per protein molecule was determined quantitatively using 5,5’-dithiobis-[2-nitrobenzoic acid] (DTNB, Ellman’s reagent). In the rapid and stoichiometric reaction, 1 mole of 2-nitro-5-thiobenzoate (TNB2−) is released per 1 mole of thiol. The amount of the released TNB2− was quantified using an absorbance plate reader (MTP Reader Infinite Series, Tecan). Free thiol concentrations of test samples were calculated from the N-acetyl-L-cysteine standard calibration curve.

2.3.5 Hotspot Peptide Mapping Analysis by Reversed Phase (RP)–Ultra-High-Performance Liquid Chromatography (UHPLC) with UV Detection (with Lys-C Enzyme)

The extent of heavy chain (HC) M254 oxidation was measured with reversed-phase chromatography. HC M254 was the most sensitive amino acid residue to oxidation during the stress studies (H2O2, UV light). After dilution of the sample with methionine solution, DTT was added, and Lys-C enzyme (Wako) was used for the digestion. The oxidised and native HC 251–276 peptides were separated using a Waters Acquity UPLC Peptide BEH C18 reversed phase column (150 × 2.1 mm, 1.7 µm) on a Waters Acquity H-Class UPLC. The applied chromatographic parameters were as follows: column temperature 70 °C; flow rate 0.4 mL/min; gradient elution (eluent A%: H2O:ACN:cc. TFA = 90:10:0.1 [V/V%]; eluent B%: H2O:ACN:cc. TFA = 50:50:0.1 [V/V%]); UV detection 215 nm. The relative amount of the peptides containing the single oxidised and the non-oxidised (native) forms of the M254 were compared using area% evaluation.

2.3.6 N-Glycosylation Profile by Hydrophilic Interaction Liquid Chromatography (HILIC) High-Performance Liquid Chromatography (HPLC) with Fluorescence (FL) Detection

The measurement was performed in two steps. Firstly, N-glycans were enzymatically released (PNGase, GlycoPrep® Kit, Agilent Technologies) from the denatured proteins, and derivatised using an Instant Procaine (InstantPCTM, GlycoPrep® Kit, Agilent Technologies) dye, followed by the removal of the excess dye. Secondly, the purified N-glycan mixture was analysed using an Acquity UPLC BEH Glycan (1.7 μm, 2.1 × 150 mm) column (Waters) on a UPLC/UHPLC (Shimadzu or Waters) with fluorescent (FLR) detection (fluorescence excitation wavelength [λEX] = 285 nm, fluorescence emission wavelength [λEM] = 345 nm). Chromatographic parameters were as follows: column temperature 45 °C; flow rate 0.5 mL/min; gradient elution (eluent A%: 50 mM ammonium formate, pH = 4.4; eluent B%: 100% ACN). Data were acquired and processed using Empower™ 3 (Waters) or LabSolutions CS 6.88 SP1 (Shimadzu) software. As a final result, the relative amounts (area%) of the different N-glycan forms were obtained.

2.3.7 Analysis of Bisecting GlcNAc and Gal-α-1,3-Gal Forms by HILIC-UHPLC-FL/Electrospray Ionisation(ESI) Tandem Mass Spectrometry (MS/MS) with Exoglycosidase Digestions

Samples were prepared according to the hydrophilic interaction liquid chromatography (HILIC) HPLC with fluorescence (FL) detection released glycan protocol divided into aliquots and digested with different combination of exoglycosidases. For the identification of bisecting GlcNAc forms, the results of the α(2-3,6,8,9)-Sialidase A–, β(1-4)-Galactosidase– and β-N-acetylhexosaminidase–digested aliquots were compared to the α(2-3,6,8,9)-Sialidase A– and β(1-4)-Galactosidase–digested samples. For the identification of galactose-α(1-3)-galactose linkage–containing forms, the α(2-3,6,8,9)-Sialidase A–, β(1-4)-Galactosidase– and α(1-3,4,6)-Galactosidase–digested aliquots were compared to the α(2-3,6,8,9)-Sialidase A– and β(1-4)-Galactosidase–digested samples. The samples were analysed on a Shimadzu Nexera X2 UHPLC instrument coupled to a Bruker maXis II Q-TOF mass spectrometer. The same chromatographic conditions were applied as in the HILIC-UHPLC-FL analysis. N-glycans were identified based on their measured mass, relative retention time, MS/MS spectra and the results of the exoglycosidase digestions using DataAnalysis and BioPharma Compass software (Bruker).

2.3.8 Sialic Acid Content by RP-HPLC-FL

N-acetylneuraminic acid (NANA) and N-glycolylneuraminic acid (NGNA) contents were determined in two steps. Firstly, the sialic acids (NANA and NGNA) were released from the sugar chains by acidic hydrolysis (HCl), and derivatised using a fluorescent dye (DMB, Takara Bio Inc.). Secondly, sialic acids were separated using a Phenomenex Kinetex C8 column (2.1 × 100 mm, 2.6 µm) on a UPLC/UHPLC (Shimadzu or Waters) instrument with FL detection (λEX = 373 nm, λEM = 448 nm). Chromatographic parameters were as follows: column temperature 35 °C; flow rate 0.35 mL/min; gradient (eluent A%: MeOH:H2O = 15:85 [V/V%]; eluent B%: H2O:cc. TFA = 100:0.1 [V/V%]). Data based on NANA and NGNA standard (Sigma) calibration curves were acquired and processed by Empower™3 (Waters) or LabSolutions CS 6.88 SP1 (Shimadzu) software.

2.3.9 Glycosylation Site Analysis by LC-MS Non-reduced Peptide Mapping

Samples were digested with rapid PNGase F (New England BioLabs) enzyme. Then, the samples were treated the same way as the samples during the non-reduced peptide mapping analysis. The same chromatographic conditions were applied. The glycosylation site was determined based on the analysis of the MS/MS spectra of the deglycosylated HC_E295-R303 peptide.

2.3.10 Non-glycosylated Heavy Chain (NgHC) Fragment by Reducing Capillary Gel Electrophoresis Sodium Dodecyl Sulphate (R-CE-SDS)

Non-glycosylated heavy chain (NgHC) was measured on a Maurice S. (ProteinSimple) CE instrument. The fragments subjected to analysis (NgHC, HC, LC) were generated by reducing the disulfide bridges of the antibody using β-mercaptoethanol together with sodium dodecyl sulfate (Maurice CE-SDS Plus 1x Sample Buffer, ProteinSimple). Compass for iCE software (ProteinSimple v 2.1.0) was used.

2.3.11 Hydrogen–Deuterium Exchange Mass Spectrometry (HDX-MS)

The samples were prepared by a two-step dilution, first with H2O buffer, then with D2O buffer, and were subjected to deuteration for different time periods. After the quenching with a low pH buffer, samples were kept in an ice bath and the samples were injected into the hydrogen–deuterium exchange (HDX)-manager module of the Waters M-Class liquid chromatograph coupled to a Waters Synapt XS® ion mobility Q-TOF mass spectrometer (Waters Corporation, Milford, MA, USA). The protein was digested on-line by a Waters Enzymate® BEH Pepsin column (30 mm × 2.1 mm, 5 µm), and the peptides were first trapped on a Waters Acquity® UPLC BEH C18 VanGuard Pre-column (5 mm × 2.1 mm, 1.7 µm), then separated on a Waters Acquity® UPLC BEH C18 column (100 mm × 2.1 mm, 1.7 µm) and finally eluted into the mass spectrometer. Chromatographic parameters were as follows: digestion column temperature 20 °C; analytical column temperature 0.1 °C; digestion eluent flow rate 100 µL/min; analytical eluent flow rate 40 µL/min; gradient elution (eluent A%: H2O:cc. FA = 100:0.2 [V/V%]; eluent B%: ACN:cc. FA = 100:0.2 [V/V%]). The mass spectrometer was set to positive ion – HDMSE (ion mobility) mode. The acquired HDX-MS data were analysed in two stages: first, tocilizumab-derived peptides were identified using Waters PLGS 3.0.3 and, subsequently, Waters DynamX 3.0 software was used to calculate the absolute and relative deuteration results for the identified peptides.

2.3.12 Two-dimensional (2D) Nuclear Magnetic Resonance (NMR) Spectroscopy

Two-dimensional (2D) nuclear magnetic resonance (NMR) spectroscopy-based structural fingerprinting was applied for the comparative assessment of the higher order structure on the atomic level. 1H-13C heteronuclear single quantum coherence (HSQC) NMR spectra of intact protein samples in uniformised near-formulation conditions were collected at 320 K on a Bruker Avance III HD 800 MHz spectrometer, equipped with a 5-mm cryogenically cooled triple-resonance Z-pulsed field gradient TCI probe (Bruker Corporation, Billerica, MA, USA). Parameters of a gradient-selected sensitivity-enhanced HSQC pulse sequence were optimised for methyl groups. 2D methyl-HSQC spectra were compared visually, as well as by a chemometric approach: sets of combined chemical shift difference (CCSD) values of pairwise comparisons of peak lists were evaluated statistically [19].

2.3.13 High Molecular Weight (HMW) Species by Size Exclusion Chromatography (SEC) (SEC-HPLC)

Chromatographic resolution was achieved with a Tosoh Bioscience TSKgel G3000SWXL column (7.8 × 300 mm, 5 µm, Tosoh) together with a Tosoh Bioscience TSKgel SWXL Guard precolumn (6.0 × 40 mm, 7 µm, Tosoh) on a Shimadzu Nexera HPLC system. Chromatographic parameters were as follows: column temperature 21 °C; flow rate 0.5 mL/min; isocratic flow (eluent: mix of solvent A: 100 mM NaH2PO4, 250 mM Na2SO4; solvent B: 100 mM Na2HPO4, 250 mM Na2SO4, pH 6.8); UV detection 280 nm. Data were acquired and processed by Empower™3 (Waters) or LabSolutions CS 6.88 SP1 (Shimadzu) software.

2.3.14 Low Molecular Weight (LMW) Species by Non-reducing Capillary Electrophoresis Sodium Dodecyl Sulphate (NR-CE-SDS)

The Maurice S. (ProteinSimple) CE instrument was used with Compass for iCE software (ProteinSimple v 2.1.0). Samples were denatured with sodium dodecyl sulfate (Maurice CE-SDS Plus 1x Sample Buffer, ProteinSimple) and acetylated with iodoacetamide. The relative amount of low molecular weight (LMW) fragments was summed as the result.

2.3.15 Charge Variants by Ion Exchange Chromatography (IEX-HPLC) with Native Treatment or CPB Digestion

Charge variants were separated on a strong cation-exchange column (YMC-BioPro SP-F, 100 mm × 4.6 mm, 5 µm) using a multi-step salt gradient on a Shimadzu Nexera HPLC instrument. Chromatographic parameters were as follows: column temperature 25 °C; flow rate 0.85 mL/min; gradient (eluent A: 25 mM MES buffer; eluent B: 25 mM MES buffer + 200 mM NaCl), FL detection (λEX = 280nm, λEM = 350 nm). The identity evaluation was based on the retention time values of the main peak. Using as a purity method, area% data of the main peak and sum of acidic and basic variants were evaluated.

2.4 Functional Assays2.4.1 Binding to Soluble Interleukin-6 Receptor (sIL-6R) by Enzyme-Linked Immunosorbent Assay (ELISA)

Recombinant human sIL-6R was coated onto a 96-well plate. After a blocking step, serial dilutions in three replicates were prepared and transferred onto the coated plates. Binding was detected using horseradish peroxidase (HRP)-conjugated anti-human IgG (Fc-specific) followed by 3,3',5,5'-Tetramethylbenzidine (TMB) reagent. The emerging enzymatic reaction was stopped by adding sulfuric acid. Dose-response data were fit to a 4-parameter logistic (4PL) curve using PLA software (Stegmann System). The relative binding activity of the test sample was calculated from half maximal effective concentration (EC50) values derived from the dose-response curve, expressed as a relative percentage of the reference standard.

2.4.2 Kinetic Analysis of sIL-6R, Neonatal Fc Receptor (FcRn), Fcγ Receptors (FcγRI/CD64, FcγRIIa/CD32a, FcγRIIIa/CD16a) and the Complement Component 1q (C1q) Binding by Biolayer Interferometry (BLI)

The kinetic constants were determined by biolayer interferometry (BLI) using Octet RED96e instrument (Sartorius). As ligands, biotin-labelled human sIL-6R (Acro Biosystems), His-tagged human neonatal Fc receptor (FcRn), CD64 (Sino Biological) and CD32a 131R (R&D Systems) were immobilised on Streptavidin (SA, Sartorius) or HIS1K biosensors (Anti-Penta-HIS, Sartorius), respectively, and varying concentrations of test samples were used as analytes. In the case of FcγRIIIa/CD16a and complement component 1q (C1q) binding, test samples were immobilised on FAB2G biosensors (Anti-Human Fab-CH1 2nd Generation, Sartorius) as ligands, and varying concentrations of human CD16a 158V receptor (Acro Biosystems) and human C1q (Merck) were used as analytes, respectively. During each kinetic analysis, the sensorgrams were collected and processed using Octet Data Analysis HT software (Sartorius). The kinetic parameters of the interactions (equilibrium dissociation constant [KD], association rate constant [ka] and dissociation rate constant [kdis]) were determined by fitting Langmuir 1:1 model to the kinetic curves.

2.4.3 Cell Surface IL-6R (Membrane-Bound IL-6R) Binding by Flow Cytometry

Human B lymphocyte-derived cell line, DS-1 (ATCC) was used, which stably and abundantly expresses IL-6R on its surface. The amount of test sample bound to IL-6R was determined by indirect immunofluorescence labelling. Samples of varying concentrations were added to the cells, and after washing steps, fluorescein isothiocyanate (FITC)-labelled anti-human IgG F(ab’)2 (Thermo Fisher Scientific) was used. The detection was performed by BD FACSVerse flow cytometer (Becton Dickinson). The results were collected by BD FACSuite software (Becton Dickinson) and evaluated using PLA software (Stegmann Systems). The measured median fluorescence intensity (MFI) values were plotted as a function of treatment concentration after log transformation. A 4PL curve was fitted to the points, and the relative biological activity of the sample was calculated from the EC50 values.

2.4.4 Anti-proliferation Assay

TF-1 cells (ATCC® CRL-2003) were cultured in RPMI-1640 medium with fetal bovine serum (FBS) and granulocyte-macrophage colony-stimulating factor (GM-CSF) before being seeded into a 96-well assay plate in the presence of FBS and recombinant human interleukin-6 (rhIL-6). Serial dilutions of the test sample were added to the IL-6–treated cells, and after incubation, viable cells were detected by adding AlamarBlue® reagent. The relative biological activity result of the test sample was calculated from EC50 values derived from the dose-response curve with respect to reference standard, using a 4PL model fit by PLA software (Stegmann System).