NIST Interlaboratory Study on Glycosylation Analysis of Monoclonal Antibodies: Comparison of Results from Diverse Analytical Methods

A broad-based interlaboratory study of glycosylation profiles of a reference and modified IgG antibody involving 103 reports from 76 laboratories.


Supplementary Discussion 1. Study Rationale and Design
Prior to the full-scale study, several laboratories with broad experience in glycosylation analysis performed a preliminary "Stage 1" series of analyses. This step enabled the refinement of reports, verified that no problems existed in sample preparation and shipping, and generated helpful suggestions on how to improve the study. Data and method reporting templates 24 were modified based on feedback from these experienced laboratories. For example, sixteen glycan compositions were initially listed in the data reporting template but after Stage 1, 68 N-glycan entries consisting of 54 unique compositions were listed. Also, both Consortium for Functional Glycomics (CFG) and Oxford (UOXF) names and structures were listed in the final template.
Because of differences in method details, several reporting options were allowed. To accommodate laboratories that could only identify compositions and those that could separate isomers, glycan compositions were listed in the data reporting template as main glycans then isomers under them.
There is no current standard to denote monosaccharide compositions in N-glycans. As described in Methods, a new notation was devised. This notation has been adapted for use in the NIST glycan tandem MS library (www.chemdata.nist.gov/glycan).
Participants were asked to use their method of choice to determine the relative glycan content in the two samples. Some participants provided multiple reports describing results of different methods. Common methods were pre-filled in drop down menus of the method reporting template. Participants could write comments or add other methods. Laboratories were requested to create separate reports for each method of analysis.
Data were analyzed using a variety of robust data analysis techniques to assess measurement reproducibility and to characterize glycan distributions. Results were compiled and evaluated for the community's consensus medians, within-laboratory precision, and concordance within the laboratories. A technical summary 24 of reported and derived values from all laboratories, a table of all identified glycans, and an individualized graphical analysis of their performance for the exercise were sent to participants. Appropriate modifications were made as necessary.

Supplementary Discussion 2. Rationale for the Samples
Samples sent to participants were intact mAbs rather than cleaved glycans to permit all varieties of glycosylation determination, including protein digestion and intact protein analysis. Our goal was to assess net measurement variability starting with a given protein sample, as done when comparing innovator and biosimilar drugs.
As outlined in Materials and Methods, two samples were sent to laboratories, namely NISTmAb and a modified-NISTmAb. Measurements on two related samples helped to assess within-lab and between-lab errors using Youden two-sample plots. 25,26 This scheme is a simple technique that plots, for a given glycan, measurement results for one sample against those of the second. The spread among the left-to-right diagonal relates to systematic between-participant bias; the off-axis distance from this diagonal relates to within-participant variability.
NISTmAb was chosen as the primary sample because it is a well-characterized reference material 31 readily available and now widely distributed to researchers. Around 1500 vials of NISTmAb were sold as of November 2019. Choosing a second sample was a challenge. An ideal second sample would be a monoclonal antibody that is not greatly different than the NISTmAb. Techniques such as chemoenzymatic glycosylation remodeling and modification by sialyltransferases (α2-6), galactosyltransferases, fucosidases (α1-2,4,6), neuraminidases, and galactosidases (β1-4, β1-4,6, and α1-3,6) were explored to produce a second sample. Some of the techniques produced complex mixtures that were difficult to purify or produced too little of the product.
Use of β1,4-galactosidase partly addressed the challenge. This enzyme breaks the glycosidic bond of the terminal β1,4-galactose residue in a glycan. However, a complete reaction using this exoglycosidase entirely eliminates certain glycans such as G1F and G2F, leading to drastic differences from the NISTmAb. To circumvent this issue, β1,4-galactosidase-treated NISTmAb was added back to the original NISTmAb sample in a 30:70 ratio. The resulting solution was used as the second sample for the study, referred to as mod-NISTmAb.  Sum of results reported for "unidentified" glycan-like signals a Index: Each different glycan composition is represented by an integer. Individual glycans are represented by a digit following a decimal point -isomers of these are represented by an additional digit. Indices in bold are glycans with complete structural assignments.

Supplementary
b Code: These correspond to numbers in the data reporting template. Entries with code Other are glycans reported by participants but not in the data reporting template.
c Measurands: text in square brackets correspond to monosaccharide compositions (see Composition). Common names are listed when available.
d Oxford: Oxford naming convention: All N-glycans have two core GlcNAcs; F at the start of the abbreviation indicates a core fucose, (6) after the F indicates that the fucose is 1-6 linked to the inner GlcNAc; Mx, number (x) of mannose on core GlcNAcs; Ax, number of antenna (GlcNAc) on trimannosyl core; A2, biantennary with both GlcNAcs as 1-2 linked; A3, triantennary with a GlcNAc linked 1-2 to both mannose and the third GlcNAc linked 1-4 to the 1-3 linked mannose; A3', triantennary with a GlcNAc linked 1-2 to both mannose and the third GlcNAc linked 1-6 to the 1-6 linked mannose; A4, GlcNAcs linked as A3 with additional GlcNAc 1-6 linked to 1-6 mannose; B, bisecting GlcNAc linked 1-4 to 1-3 mannose; Gx, number (x) of linked galactose on antenna, (4) or (3) after the G indicates that the Gal is 1-4 or 1-3 linked; [3]G1 and [6]G1 indicates that the galactose is on the antenna of the 1-3 or 1-6 mannose; Gax, number (x) of linked alpha galactose on antenna; Sx, number (x) of sialic acids linked to galactose; Sa is used when the sialic acid is known to be N-acetylneuraminic acid, or Sg is used when it is known to be N-glycolyl neuraminic acid; the numbers 3 or 6 in parentheses after S indicate whether the sialic acid is in an 2-3 or 2-6 linkage.

Headings and Subheadings Definition Measurand
The glycan (or identified combination of glycans) for which measurement results were reported. Index: Decimal index of identified glycan compositions assigned by decreasing number of NISTmAb results.
Composition: Composition of the glycan in the De Leoz-Stein notation (see Table of Identified Glycans for details.) The result listed for a given composition is the sum of the reported results for all glycoforms of that composition, where the glycoforms are indicated by "[composition] Gly". The result listed for a given glycoform is the sum of the reported results for all isomers of that glycoform where the isomers are indicated by "[composition] GlyIso".
Common Name: When available, a common name of glycoforms and isomers. See Table of Identified Glycans for Oxford names. Sample A, % Summary statistics for results reported for sample A, a modified version of the NISTmAb material #: The number of participants reporting this measurand in this sample. 25%: The 25th percentile (1st quartile) of the distribution of the reported results. Median: The consensus median (50th percentile or 2nd quartile) of the distribution of the reported results. 75%: The75th percentile (3rd quartile) of the distribution of the reported results. Sample B, % Summary statistics for results reported for sample B, the NISTmAb material. #: The number of participants reporting this measurand in this sample. 25%: The 25th percentile (1st quartile) of the distribution of the reported results. Median: The consensus median (50th percentile or 2nd quartile) of the distribution of the reported results. 75%: The75th percentile (3rd quartile) of the distribution of the reported results. A/B Ratio Summary statistics for the ratio A/B when results were reported for both samples A and B. #: The number of A/B ratios. 25%: The 25th percentile (1st quartile) of the distribution of the calculated ratios. Median: The consensus median (50th percentile or 2nd quartile) of the distribution of the calculated ratios. 75%: The75th percentile (3rd quartile) of the distribution of the calculated ratios.