Acetonitrile and formic acid for an LC-MS analysis (LC-MS grade) were obtained from Fujifilm Wako Pure Chemical (Japan). The reagents used were of the highest grade obtained from Fujifilm Wako Pure Chemical (Japan), unless otherwise stated. All the aqueous solutions and solvents were prepared using ultrapure water purified with the Milli-Q purification system (Merck Millipore, USA).

The candidate material was expressed using the CHO-derived cell line. After fermentation of the antibody-producing CHO cell line in a serum-free culture medium for 7 days using the Allegro XRS 25 Bioreactor (Pall, USA), the culture supernatant was prepared using the Millistak Pod depth filtration system (EMD Millipore, USA) to remove cells and debris. Thereafter, it was purified further by three-step chromatographic technique using protein A affinity (RTP MabSelect SuRe, Cytiva, USA), anion-exchange (RTP Capto Q, Cytiva, USA), and cation-exchange (RTP Capto S, Cytiva, USA) columns. The purified material was treated using the Planova 20 N virus removal filter (Asahi Kasei Medical, Japan), concentrated through Pellicon ultrafiltration (EMD Millipore, USA) and buffer exchange to 10 mmol/L potassium phosphate buffer (pH 7.0) in a good manufacturing practice (GMP) grade facility. The raw material (1 ml) was dispensed into polypropylene vials sterilely using a Microlab STARlet 8ch liquid handling system (Hamilton, USA) with a FluidX XSD-48Pro automated capper/decapper system (Azenta Life Sciences, USA) and stored at −80°C. These processes were performed at Manufacturing Technology Association of Biologics at Kobe, Japan.

LC-MS for structural analyses was performed using a maXis II electrospray ionization quadrupole time-of-flight mass spectrometer (Bruker, Germany) in a positive ion mode coupled with an LC-30A Nexera HPLC system (Shimadzu, Japan). The data were analyzed using Data Analysis 4.3 software (Bruker, Germany).

Cation exchange chromatography (CEX) was performed using an LC-20A Prominence HPLC system with an ultraviolet (UV) detector (Shimadzu, Japan) and a BioPro IEX SF column (5 μm, 4.6 mm diameter × 100 mm length, YMC, Japan). A volume of 5 μl of the candidate RM was injected into the LC-UV system. The measurement conditions were as follows: mobile phase A: 20 mmol/L 2-(N-morpholino) ethanesulfonic acid (MES) (pH 6.0), mobile phase B: 20 mmol/L MES, 200 mM NaCl (pH 6.0), flow rate: 0.5 ml/min, column temperature: 30°C, and gradient condition: 10–80 %B 30 min. Absorbance was detected at a wavelength of 215 nm. The performance criteria were set to 5.6 (± 1.2) min for α-chymotrypsinogen A from bovine pancreas (Sigma-Aldrich, USA), and 23.2 (± 0.9) min for equine myoglobin (Serva, USA) based on two sigmas of averaged retention time.

Size-exclusion chromatography (SEC) was performed using the LC-UV system used for CEX. The chromatography column used was a TSK gel G3000SW_XL (5 μm, 7.8 mm diameter × 300 mm length, TOSOH, Japan). A volume of 10 μl of the candidate RM was injected into the LC-UV system. The measurement conditions were as follows: 100 mmol/L sodium phosphate buffer containing 100 mmol/L Na₂SO₄ (pH6.8) in isocratic elution, flow rate: 0.4 ml/min, and column temperature: 25°C. Absorbance was detected at a wavelength of 280 nm. The system performance was evaluated using molecular weight marker proteins for SEC from Oriental Yeast (Japan) consisting of five proteins. The performance criteria based on two sigmas of averaged retention time were set to 20.0 (± 0.1) min for glutamate dehydrogenase, 23.1 (± 0.1) min for lactose dehydrogenase, 24.9 (± 0.1) min for enolase, 27.2 (± 0.1) min for myokinase, and 28.9 (± 0.1) min for cytochrome C.

Microchip electrophoresis was performed using the LabChip GXII Touch24 electrophoresis system (PerkinElmer, USA). The sample solution was prepared by using 2 μl of diluted the candidate RM at approximately 2.5 mg/g using the Protein Express Assay Reagent Kit (PerkinElmer, USA) at 70°C for 10 min, according to the manufacturer’s instruction. The results of the three lanes were averaged to obtain the measurement results. Measured molecular masses were calibrated using a molecular weight marker supplied with the Protein Express Kit (PerkinElmer, USA). The standard deviation of the measured molecular masses was confirmed to be less than 5%.

Capillary isoelectric focusing was performed using an iCE3/Alcott720NV capillary isoelectric focusing system with Fc cartridge (100 μm diameter × 50 mm length, ProteinSimple, USA). The sample solution was prepared with 0.4 mg/mL as the final concentration of the candidate RM in 4% pharmalyte (pH 3–10), 0.35% methylcellulose, and 10 mmol/L arginine solution. Electrophoresis conditions were 1500 V for 1 min while prefocusing and 3 kV for 4.5 min when focusing on detection at 280 nm. The measurements were repeated thrice. Samples were measured after confirming that the isoelectric points of human hemoglobin (Sigma-Aldrich, USA) were within 7.1 ± 0.1 and 7.2 ± 0.1, and the variation in triplicate measurements of the peak height of a high pI marker (pI 9.77) supplied with High pI marker (102219, ProteinSimple, USA) was less than 10%.

Sample preparation including hydrolysis by peptide N-glycosidase F (PNGaseF) and 2-aminobezamide (2-AB) derivatization of glycans from the candidate RM was performed using the EZGlyco mAb-N Kit with 2-AB (BS-X4410, Sumitomo Bakelite, Japan). A volume of 2 μl of the resulted glycan mixture solution obtained with 8 μl of candidate RM was injected into an LC-fluorescence detection (LC-FL) system for analysis. The LC-FL system used was a Nexera 30A HPLC system with a fluorescence detector (Shimadzu, Japan) with an AQUITY UPLC Protein BEH Amide column (1.7 μm, 2.1 mm diameter × 150 mm length, Waters, USA). The measurement conditions were as follows: mobile phase A: 100 mmol/L ammonium formate (pH 4.5), B: acetonitrile, flow rate: 0.2 ml/min, column temperature: 45°C, gradient condition: 75–50 %B 50 min, fluorescent detection: excitation at 330 nm, and detection at 420 nm.

The standard solutions of natural amino acids were gravimetrically prepared by dissolving the following NMIJ CRMs in 10 mmol/L HCl: l-aspartic acid (NMIJ CRM 6027a), l-glutamic acid (NMIJ CRM 6026a), l-proline (NMIJ CRM 6016a), l-valine (NMIJ CRM 6015a), l-isoleucine (NMIJ CRM 6013a), l-leucine (NMIJ CRM 6012a), l-phenylalanine (NMIJ CRM 6014a), and l-alanine (NMIJ CRM 6011a). The standard mixture of amino acids was gravimetrically prepared by mixing the standard solution of each amino acid in the same molar ratio of as that of each amino acid composition in the monoclonal antibody molecule. The following isotopically labeled amino acids (Cambridge Isotope Laboratories, USA) were used as the internal standard: l-¹³C₄ ¹⁵N-Asp, l-¹³C₅ ¹⁵N-Glu, l-¹³C₅ ¹⁵N-Pro, l-¹³C₆ ¹⁵N₂-Lys, l-¹³C₅ ¹⁵N-Val, l-¹³C₆ ¹⁵N-Ile, l-¹³C₆ ¹⁵N-Leu, l-¹³C₉ ¹⁵N-Phe, and l-¹³C₃ ¹⁵N-Ala. The candidate RM (0.1 ml) and the mixture of isotopically labeled amino acids (0.1 ml) were gravimetrically dispensed into a glass vial. The dispensed solutions were dried with gentle nitrogen flow. Calibration blends were gravimetrically prepared by mixing the standard mixture of natural amino acids and mixture of isotopically labeled amino acid solutions. The calibration blends (0.2 ml) were dispensed into a glass vial and dried by gentle nitrogen flow.

For liquid-phase hydrolysis, dried sample blend and calibration blend in a glass vial were dissolved by adding 0.2 ml of 6 mol/L HCl and 0.1% phenol. Samples were hydrolyzed at 150°C for 1 and 3 h, 160°C for 1 and 3 h, and 170°C for 1 and 3 h using an ETHOS One microwave digestion system (Milestone SRL, Italy) after purging with nitrogen. The hydrolysate was analyzed by pre-column derivatization with N-butylnicotinic acid N-hydroxysuccinimide ester iodide and measured via LC-MS/MS under the selected ion monitoring (SRM) mode.

For gas-phase hydrolysis, the dried sample in the glass vial was hydrolyzed under the gas phase by 6 mol/L HCl and 2% phenol at 130°C for 18, 24, and 48 h as well as 150°C for 18, 24, and 48 h using a Pico-tag workstation hydrolysis system (Waters, USA) after purging with nitrogen. The hydrolysate was analyzed by hydrophilic interaction chromatography (HILIC)-MS/MS under the SRM mode. The detailed measurement conditions for amino acid analyses are described in the supplementary material.

Homogeneity was assessed by measuring the relative area percentage of the main peak obtained by CEX in triplicate for twelve selected vials.

Stability was evaluated as acceleration (stored at 4, or 25°C), long-term (stored at −80°C), short-term (stored at −20°C), and freeze-thaw cycles (up to five times) tests. The long-term stability test was performed via CEX and UV absorption at 280 nm (see the below section) using a previous lot of the preceding product. The acceleration, short-term, and freeze-thaw cycles tests were performed via CEX, SEC, and UV absorption at 280 nm using the candidate RM.

The density of the candidate RM was measured using a vibration type of the density meter (DMA5000EX, Anton Paar, Austria) in triplicate. Calibration was performed using dried air and pure water density standard (QAT182462, Kyoto Electronics, Japan).

UV absorbance measurement was performed using a UV-2550 spectrophotometer (Shimadzu, Japan) calibrated with potassium dichromate solution and optical filter, according to the Japanese Industrial Standards (JIS) K0115 (JIS K0115, 2004). A quartz cell with nominal 1-mm optical path length (1/Q/1, Starna Scientific, England) was used to measure the absorbance of the candidate RM at 280 nm and at 1 nm of the band path. Absorbance was measured using a double-beam spectrophotometer without a reference cell, using the same procedure as that used for a single-beam type instrument. A solvent blank sample was measured first, then the sample solution was measured using the same optical cell, and their difference was used to determine the absorption of the sample solution.

The desired target structure of IgG1 is a heterotetrameric structure with two heavy chains (HC) and two light chains (LC; full body, 2LC:2HC). However, biopharmaceutical products usually contain free subunit chains, fragments of each subunit, or products with insufficient tetramer formation owing to disulfide bond scrambling as the size variants (Gaza-Bulseco and Liu, 2008; Leblanc et al., 2017; Turner and Schiel, 2018). These small size variants were analyzed via electrophoresis under denatured and non-reducing condition using a microchip electrophoresis system, whose result is presented in Figure 1.

Peaks were assigned according to the molecular size, indicating free LC, HC, partial molecules, 2HC and 2LC:HC, and non-glycosylated. The peak area ratio of the full body was estimated to be 93%, which was considered to be a minimum value. This is because the sample may dissociate into the partial molecules depending on the sample preparation conditions, and the area ratio may vary depending on the measurement conditions. In addition to the evaluation via electrophoresis, the size heterogeneity was assessed via SEC, which enabled to analyze fragmented products to aggregated macro molecule. As shown in Figure 2, the chromatogram obtained via SEC with UV detection exhibited four peaks assigned as oligomers (oligomeric form 1 and 2) and truncated products (truncated form 1 and 2) with the area percentage of (93.7 ± 0.2) % of the monomer peak. The entities regarding oligomeric forms 1 and 2 were attributed to trimer and dimer with area percentage of (0.42 ± 0.01) % and (5.72 ± 0.14) %, respectively. The entities of truncated forms 1 and 2 were (0.16 ± 0.01) %, which was negligibly small compared to the sum of the monomeric and oligomeric forms.

The heterogeneity owing to charge variants was assessed via CEX and capillary isoelectric focusing (cIEF). The net charge of the monoclonal antibody may depend on the PTMs including deamidation and conformational changes, resulting in the charge distribution of the molecule (Du et al., 2012; Fekete et al., 2015; Turner and Schiel, 2018). The chromatogram obtained via CEX is shown in Figure 3A. The chromatographic peaks on the chromatogram were divided into three, and the area percentages of each peak were (35.9 ± 0.1) % for the acidic peak (47.8 ± 0.2) % for the main peak, and (16.4 ± 0.1) % for the basic peaks. The electropherogram obtained via cIEF is shown in Figure 3B, and exhibited area percentages of indicated (54.3 ± 0.9) % for the acidic peak (42.1 ± 0.6) % for the main peak, and (3.6 ± 0.3) % for basic peak. The charge distribution was evaluated using two independent methods with different measurement principles, and results indicated similar main peak contents with area percentages of 47.8 and 42.1%, although not identical. The isoelectric point of the candidate RM obtained via cIEF was 9.03.

Confirmation of the primary structure and identification of the PTM were performed by mass spectrometry–based analyses, including intact mass spectrometry, middle-down structural characterization, and peptide mapping.

N-linked glycans of the candidate RM were analyzed by glycan mapping using LC-FL for the 2-AB–derivatized glycans. Various sample preparations including glycan release with PNGaseF, derivatization with 2-AB and clean-up were conducted using a commercial preparation kit. The chromatogram obtained via LC-FL is shown in Figure 6.

The area percentages of triplicate measurements are summarized in Supplementary Table S4. The major assigned 10 peaks were from typical hybrid and high mannose types of N-linked glycan of monoclonal antibody produced by CHO cells (Hossler et al., 2009). Based on the results of glycan mapping and analysis of scFc in the previous section, we identified 12 major glycoforms. However, 20 peaks with the relative area ratios greater than 0.1% were observed on the chromatogram, and almost half of the structures are currently unknown. Reliable absolute quantification of glycans has been difficult, and it is necessary to conduct more in-depth analysis in the future by accumulating data through inter-method comparisons and collaborative measurements for structural analysis of unidentified glycans and precise quantification (Wada et al., 2007; De Leoz et al., 2020). The use of recently available certified reference material for glycan analysis, NIST SRM 3655, glycans in solution (Lowenthal and Phinney, 2021) enables highly reliable and accurate analysis.

Peptide mapping is a widely used bottom-up technique for identifying conformation of the primary structure and modifications such as PTM, N/C-terminal extension, and truncation (Beck et al., 2013). Peptide mapping typically involves denaturation, reduction, and alkylation of protein prior to enzymatic digestion, followed by separation and analysis of the resulting peptide mixture by reversed phase LC-MS/MS (Ren et al., 2009; Lauber et al., 2013; Mouchahoir and Schiel, 2018). For peptide mapping of the candidate RM, enzymatic digestion was performed using trypsin, Lys-C, and Glu-C with different sequence specificities to achieve the highest possible sequence coverage. The results of peptide mapping using these three proteases under reduced condition are presented in Figure 7A for LC and (B) for HC, and the results with the retention times, calculated masses (monoisotopic masses), and observed masses for the identified peptides are summarized in Supplementary Table S5A–C.

Trypsin is the most commonly used protease because of the strict substrate specificity, high activity, and the favorable size of peptides yielded for mass spectrometry (Olsen et al., 2004). Digested peptide fragments generated using trypsin covered most of the entire sequence, but small peptides consisting of a few residues were undetectable. However, peptide mapping with Lys-C and Glu-C digestion in addition to trypsin digestion complementarily and successfully covered the entire sequence. The peptide mapping with the three proteases confirmed the amino acid sequence of the candidate RM to be exactly as designed.

Furthermore, peptide mapping by trypsin digestion revealed the modified structure including N and C terminal structure of HC, deamidation on asparagine residues, oxidation on methionine residues, as well as glycosylation site. The results of peptide mapping regarding these modifications are summarized in Table 1.

Peptide mapping by trypsin digestion showed partial N-terminal pyroglutamylation of the HC and addition of a lysine residue at the C-terminus of the HC. For deamidation of asparagine residues, the isotopic patterns of Asn and deamidated Asn overlapped when high-resolution mass spectrometry was used. Therefore, peptides with variation in both mass and retention time were determined to be deamidated. The results showed that N30 at the LC, and N55, N84, N318, N387, N392, or N393 at the HC were deamidated. A tryptic peptide, H374-395 carried three asparagine residues (N387, N392, and N393) in the peptide chain. Although we were unable to identify the deamidation site from these data, N387 was a part of the -Asn-Gly- sequence, which is known to be susceptible to deamidation, and those deamidation has been reported (Zhang et al., 2014; Huang et al., 2016). Therefore, N387 could be deamidated among the three asparagine residues. Oxidized methionine residues were found at M107 and M255 of HC.

It is known that IgG1 forms 16 pairs of disulfide bonds to form a heterotetrameric structure with two light chains and two heavy chains (Liu and May 2012). For the analyses of disulfide bond formation, trypsin digestion was performed under non-reducing condition by skipping the reduction and alkylation steps, and the results are summarized in Table 2.

The peptide mapping under non-reducing condition successfully revealed disulfide bond sites within the LC and HC, between the LC and HC, as well as between the heavy chains. Based on the above results, the structure determined by mass spectrometry–based analyses is shown in Figure 8.

Assigning the indicative value (antibody concentration) of candidate RM was performed according to the NMIJ quality system in accordance with ISO 17025 and 17034. Assessing the homogeneity and stability is essential to ensure consistency of the indicative value, and it is mandatory to state their conformity to the international standards.

Homogeneity was assessed through three repeat measurements of the relative area percentage of the main peak obtained by CEX. 12 vials taken by stratified random sampling from whole batch were used for this test. The homogeneity data obtained by analysis of variance (ANOVA) showed that the mean square between-vial variance could not be obtained because of larger value of the variance of within-vial. Therefore, the uncertainty associated with the inhomogeneity was estimated using Eq. 1, considering the measurement variability described in ISO Guide 35 (ISO Guide 35, 2017). u b b = M S w i t h i n n × 2 ν M S w i t h i n 4 (1) where MS _within represents the mean square within-vial variance, n represents number of measurement replicates per vial, ν represents degree of freedom of MS _within.

The calculated u _bb according to Eq. 1 was 0.103%, and this value was used as the relative standard uncertainty.

The stability was evaluated as acceleration, long-term, short-term stabilities, and free-thaw cycle tests via CEX. The acceleration test was performed at 4 and 25°C by CEX, SEC, and UV absorption at 280 nm. Time course of the relative area percentage of the main peak obtained by CEX is shown in Supplementary Figure S2A at 4°C and (B) at 25°C. The area ratio of the main peak obtained by CEX decreased after 10 days at 4°C and 3 days at 25°C. Monomer peak ratio on SEC analysis decreased only by 0.8% in 40 days at 25°C (data not shown). No change was observed by UV absorption for 122 days at 4°C and for 40 days at 25°C (data not shown). These results indicated that CEX was the most sensitive method for monitoring the stability.

The long-term stability test was performed using an antibody solution from the previous production batch produced using same production process, and it was monitored for up to 675 days at −80°C via CEX. The area ratio of the main peak by CEX over the monitoring period is shown in Figure 9.

A trend analysis was performed according to ISO Guide 35 to determine whether a linear approximation to instability would show a trend in the data over the monitoring period. The result indicated that the significance of the linear regression was not found. By multiplying the uncertainty derived from the slope of the linear regression by 3 years, the uncertainty associated with long-term stability was determined to be 0.923%. The short-term stability was monitored for up to 304 days at −20°C via CEX (data not shown). The results of the trend analysis over the monitoring period according to ISO Guide 35 indicated that the significance of the linear regression was not found. The uncertainty associated with the short-term stability was determined to be 0.955% based on the uncertainty derived from the slope of the linear regression.

In addition, freeze-thaw cycle tests were examined as follows: The candidate RM stored at −80°C was moved from a freezer and placed in an incubator at 25°C for 1 h and then returned to the freezer at −80°C. The freeze-thaw cycle test was conducted by repeating this cycle up to five times using CEX, SEC, and UV absorption at 280 nm. The results of the test showed no significant difference over five-times as summarized in Supplementary Table S6, and it can be concluded that freeze-thaw up to five times does not affect the results of any of the methods examined, as well as the indicative value.

Antibody concentration, which is as an indicative value of the RM, was determined via amino acid analyses. To ensure the reliability of the amino acid analysis, hydrolyses with hydrochloric acid by microwave-assisted liquid phase (liquid phase) and gas phase were performed independently, and isotope dilution mass spectrometry was used to determine the hydrolyzed amino acids. Amino acid analysis requires complete hydrolysis of the sample into its constituent amino acids. Because the optimal conditions for hydrolysis vary depending on the sample protein, we examined the optimal conditions for the candidate RM (Burkitt et al., 2008; Feng et al., 2020). The hydrolysis conditions were examined at 150, 160, and 170°C for 1 and 3 h for the liquid phase hydrolysis and 130 and 150°C for 18, 24, and 48 h for the gas phase hydrolysis. The optimum conditions were 160°C for 3 h for microwave-assisted liquid phase hydrolysis and 150°C for 48 h for gas phase hydrolysis based on the highest recovery of the eight measured amino acids and the smallest difference between the measured amino acids.

Hydrolyses of the candidate RM were performed under the optimized condition on three separate days and three repeated measurements using LC-MS/MS. The measured concentrations for each amino acid were divided by the number of the amino acid residues in the candidate RM. The quantitative results of each amino acid along with the associate uncertainties are summarized in Table 3 for the liquid phase hydrolysis and for the gas phase hydrolysis. The detailed uncertainty budget is shown in Supplementary Table S7.

The protein concentration was obtained by calculating the weighted mean of the quantitative results of the eight amino acids. From the calculation, the concentrations of the candidate RM obtained via liquid phase and gas phase hydrolyses were (33.68 ± 0.42) nmol/g and (33.94 ± 0.98) nmol/g, respectively (the number following ± represents the standard uncertainty.).

The validity of the analyses was confirmed using certified reference material of human serum albumin (NMIJ CRM 6202a) with a certified value of the concentration (74.3 ± 3.2) g/L (expanded uncertainty with coverage factor, k = 2) (Kinumi et al., 2017). The quantitative results by liquid phase and gas phase hydrolyses were (74.4 ± 0.9) g/L and (74.1 ± 0.8) g/L, respectively (the number following ± represents the standard deviation.). This indicates that the amino acid analysis was sufficiently accurate and suitable for the value assignment.

In addition, we assessed protein, peptide, and amino acid impurities in the candidate RM which may affect the result of the amino acid analyses. The raw material was highly purified by three-step-chromatography as equivalent to that used in the production of biopharmaceutical. Possible impurities should be considered are HCP and protein A, which could be contaminated during the purification process (Hogwood et al., 2014). The quantitative results of HCP and protein A by ELISA were 3.2 ng/mg protein and 0.2 ng/mg protein respectively, which were significantly low, and did not affect the results of amino acid analyses of the material. To assess peptide and free amino acid contaminants, amino acid analysis was performed for hydrolysate by hydrochloric acid hydrolysis of filtrates with a 10 kDa ultrafiltration device (Amicon ultra, Millipore, USA) for the candidate RM and phosphate buffer (as the blank sample). The quantitative results indicated that contents of the eight measured amino acid in the filtrate of the candidate RM were equivalent or lower compared to those of the blank. These results indicate no impurities that could affect the results of amino acid analysis, and the values obtained can be corresponded to the concentration of antibody.

Furthermore, quantitative analysis was performed for host cell–derived DNA and endotoxin which are often considered process-related impurities, and their quantification results were below the lower limit of quantification (Host cell–derived DNA, 10.0 pg/mg protein; endotoxin, 0.01 EU/mg protein).

As the measurand is a heterotetrameric structure including oligomeric forms, the indicative value can be determined by using the results of amino acid analysis. The concentration of the candidate RM was calculated from the weighted mean of the results of the amino acid analyses by liquid phase and gas phase hydrolyses, which was to be 33.75 nmol/g. Finally, the indicative value of the candidate RM was determined to be 5.00 g/L by converting the unit to mass concentration using the molecular weight (148,056) and the density (1.0008 g/cm³ at 20°C) measured by a vibration-type density meter. The molecular weight was calculated based on the amino acid sequence with 16 disulfide linkages and the most abundant glycan structure (G0F/G0F) described in previous section.

The uncertainty associated with the concentration from two amino acid analyses was calculated according Eq. 2 (Kinumi et al., 2017): u 2 ( x ) = ∑ w i 2 u 2 ( x i ) + D i f 2 , (2) where w _i represents the weight of the quantitative results by liquid phase and gas phase hydrolyses, u (x_i) represents the uncertainty of the quantitative results by the two hydrolysis methods, and Dif represents the difference between the two quantitative methods calculated as a rectangular distribution of the two quantitative results. From the equation, the standard uncertainty associated with amino acid analysis was calculated to be 0.19 g/L.

In conclusion, the uncertainty budget is summarized in Table 4 according to ISO Guide 35, considering the following uncertainty components: amino acid analysis for value assignment, inhomogeneity of the material, long-term instability, and short-term instability. The uncertainty estimation of the indicative value of the candidate RM was calculated as the combined standard uncertainty by summing the squares of the uncertainty components listed in Table 4.

Overall, the indicative value of the candidate RM was determined to be (5.00 ± 0.19) g/L, and the number following ± represents the expanded uncertainty (coverage factor k = 2).

Because the absorbance of the protein solution can be measured using a low cost instrument with high reproducibility and simple measurement principle based on the Lambert-beer’s law, the absorbance at 280 nm is frequently used to determine antibody concentration (Miranda-Hernandez et al., 2016; Cole et al., 2018). Antibody concentration is essential in the quality control of an antibody drug, and the accurate determination of the extinction coefficient at 280 nm of the RM can lead to various application of the RM with high accuracy.

The extinction coefficient of the candidate RM was obtained according to the Lambert-Beer equation: the absorbance measured by a spectrophotometer at 280 nm using an optical cell with an assessed optical path length; the antibody concentration determined as the indicative value by amino acid analysis. The absorbance of the candidate RM was measured using a quartz cell with an optical path length of 1 mm. Because the absorbance at 280 nm is approximately 0.7 when measured with an optical path length of 1 mm, which is appropriate for accurate measurement. Because the only available optical cell whose optical path length was evaluated was a 10 mm cell supplied by Hitachi Ltd determined to be (10.00 ± 0.05) mm, the extinction coefficient at 280 nm of potassium hydrogen phthalate solution was determined using this 10 mm cell, and the optical path length of the 1 mm cell was obtained using the determined extinction coefficient of the potassium hydrogen phthalate solution. Finally, the optical path length of the 1 mm optical cell was determined to be (1.02 ± 0.01) mm (the number following ± represents the standard uncertainty.). To measure the absorbance of protein solutions containing aggregates, it is necessary to consider the effect of light scattering. The correction of the light scattering at 280 nm was made using the results of absorbance measurements at 320–350 nm according to literature (Maity et al., 2009). Using the indicative value (measured antibody concentration), optical path length, and absorbance at 280 nm using the 1 mm optical cell, the extinction coefficient was found to be (1.41 ± 0.03) L/(g cm) with the correction of light scattering (the number following ± represents the standard uncertainty.). The extinction coefficient of proteins at 280 nm is often predicted using the equation by Pace et al., based on the numbers of tryptophan, tyrosine, and cystine residues in the protein molecule (Pace et al., 1995). The calculated absorption coefficient of the candidate RM is 1.42 L/(g cm), which is consistent with the experimental result. We performed an inter-laboratory comparison on UV measurement at 280 nm for monoclonal antibody solution, and we found that the standard deviation of the measurement results of 34 laboratories was approximately 10%, indicating a large variation among laboratories (Kinumi et al., 2021). It has been shown that this difference in absorbance can be reduced by conducting system suitability test. The use of this material enables validation of absorbance measurement and it can realize more accurate measurement.

The physicochemical properties of the candidate RM regarding its particle size and higher-order structure were also analyzed by SEC-MALS (Supplementary Figure S3), DLS (Supplementary Figure S4), NTA (Supplementary Figure S5), FI (Supplementary Figure S6), CD (Supplementary Figure S7), and TSA (Supplementary Figure S8). The molecular weight calculated from SEC-MALS was 1.35 × 10⁵, indicating that the main peak corresponded to the IgG monomer. The relative UV peak area (94.6%) demonstrated that most of the candidate RM was dispersed as monomeric form. The DLS measurements presented a typical value (11.4 nm) of hydrodynamic diameter for the IgG monomer and no significant peaks for aggregates, which was consistent with the SEC-MALS results. The presence of detectable amounts of submicrometer-sized particles was clarified via NTA analysis. The mode diameter, mean diameter, and particle concentration were 194 nm, 206 nm, and 5.81 × 10⁸ particles/mL, respectively. The FI measurements demonstrated that the candidate RM contained considerable number of micrometer-sized particles (e.g., ∼7,000 particles/mL for a particle of ≥5 nm diameter) compared with trastuzumab-US (Hutterer et al., 2019). The CD spectrum of the candidate RM exhibited a typical shape of a globular protein rich in β-sheet structures. The thermal denaturation temperatures revealed by TSA analysis were 69.7 and 80.4°C, which were comparable to the values reported for trastuzumab-US (Hutterer et al., 2019). These results indicate that the candidate RM maintains the higher-order structure of the native IgG.

NISTmAb (RM 8671) from NIST is the humanized IgG1κ solution produced using mouse-derived NS0 cells, and assigned the antibody concentration, size heterogeneity, and charge heterogeneity as the reference values. This material also provides with the case study of various properties.

AIST-MAB was produced using CHO cell line, which is used in the production of over 70% of biopharmaceuticals of which almost all of the monoclonal antibodies (Lalonde and Durocher, 2017). The difference in the host cells for the production, can lead to a large difference in properties, especially in the glycan structure. Even both materials are recombinant humanized IgG1κ, the amino acid sequence, buffer formulation, and method used for value assignment in addition to host cells were between them differed. The development of antibody reference material with different properties, specifically produced by the CHO cells, presents considerable advantage in antibody characterization. Conversely, to evaluate the various properties of monoclonal antibodies, orthogonal approaches of analyses using multiple reference materials can provide complementary information and improve the quality of characterization. The reference materials are not exclusive to each other, and the availability of multiple reference materials for antibody characterization could expand the scope of the utilization of each reference material.