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

Abstract

Glycosylation is a topic of intense current interest in the development of biopharmaceuticals because it is related to drug safety and efficacy. This work describes results of an interlaboratory study on the glycosylation of the Primary Sample (PS) of NISTmAb, a monoclonal antibody reference material. Seventy-six laboratories from industry, university, research, government, and hospital sectors in Europe, North America, Asia, and Australia submitted a total of 103 reports on glycan distributions. The principal objective of this study was to report and compare results for the full range of analytical methods presently used in the glycosylation analysis of mAbs. Therefore, participation was unrestricted, with laboratories choosing their own measurement techniques. Protein glycosylation was determined in various ways, including at the level of intact mAb, protein fragments, glycopeptides, or released glycans, using a wide variety of methods for derivatization, separation, identification, and quantification. Consequently, the diversity of results was enormous, with the number of glycan compositions identified by each laboratory ranging from 4 to 48. In total, one hundred sixteen glycan compositions were reported, of which 57 compositions could be assigned consensus abundance values. These consensus medians provide community-derived values for NISTmAb PS. Agreement with the consensus medians did not depend on the specific method or laboratory type. The study provides a view of the current state-of-the-art for biologic glycosylation measurement and suggests a clear need for harmonization of glycosylation analysis methods.

Keywords: Glycomics; NISTmAb; fluorescence; glycan; glycopeptide; glycoproteins; glycosylation; interlaboratory study; mass spectrometry; reference antibody.

Graphical abstract

Fig. 1.

Number of unique glycan compositions reported, grouped by method, analyte, and sector. The boxes span the central 50% of reported values, 25% to 75%; the whiskers span the central 90%, 5% to 95%; the central line marks the median, 50%. Box widths are proportional to the number of reports. Groups within each category are presented in order of decreasing number of reports. Solid circles represent individual results within categories of fewer than six reports. The dotted line marks the median number of compositions reported in the 103 reports provided by 76 laboratories.

Fig. 2.

Proportion of glycan compositions reported as isomers. The boxes span the central 50% of reported values, 25% to 75%; the whiskers span the central 90%, 5% to 95%; the central line marks the median, 50%. Box widths are proportional to the number of reports. Categories are presented in order of increasing median proportion. The dotted line marks the median proportion of compositions reported as isomers.

Fig. 3.

Summary results for the 57 most frequently reported unique glycan compositions. Box plots for A) mod-NISTmAb, B) mod-NISTmAb/NISTmAb ratio, and C) NISTmAb PS 8670. Glycan compositions in red have terminal β1,4-gal as their dominant structure. Each box represents the distribution of the central 50% of the mean of the reported replicate values for one glycan. The horizontal middle line in each box represents the consensus median. The width of each box is proportional to the square root of the number of values defining the distribution. The dashed red line in the display of the Fig. 3B denotes the expected ratio, 1.0, when a glycan result is the same in mod-NISTmAb as it is in NISTmAb. Glycans are sorted in order of increasing mod-NISTmAb/NISTmAb ratio. D) Targetplot summary of mod-NISTmAb/NISTmAb ratios relative to the consensus medians. Each dot represents one set of results. Dot diameter is proportional to number of mod-NISTmAb/NISTmAb ratios reported. The dots are color-coded by distance from the (0, 0) origin: dots within two comparability units are colored green, between two and three units are colored yellow, and greater than three units are colored red. The “Z-score Mean” axis displays the average bias estimated as the mean of the “Z-score” values of the ratios. The “Z-Score SD.” axis displays the variability of individual bias estimates, estimated as the standard deviation of the Z-scores.

Fig. 4.

Targetplot summary of mod-NISTmAb/NISTmAb ratios relative to the consensus medians. Each dot represents one set of results. Dot diameter is proportional to number of ratios reported. The “Average Bias” axis displays values estimated as the mean of the “Z-scores” of the ratios. The “Bias Variability axis displays values estimated as the standard deviation of the Z-scores. The subplots are colored by: A) analytical technique, B) analyte, C) laboratory type, and D) number of replicates.

Fig. 5.

Youden two-sample plots for the four most abundant glycan compositions in NISTmAb. Each panel displays the bivariate distribution for one composition. A) [h4n4f1], B) [h3n4f1], C) [h5n4f1], and D) [h3n3f1] (see Materials and Methods for key). The panels are centered on the univariate medians and scaled to display all values from 0 to twice the median. Values that are greater than twice the median are assigned a value of twice the median. The ellipse includes about 68% of the pairs (≈ 1 SD). The diagonal line represents the expected relationship when measurement systems have the same bias for both samples. Each symbol represents the (mod-NISTmAb, NISTmAb) pair for one data set. Symbols are coded and labeled by separation technique.

Fig. 6.

Scatterplot of the relationship between measurement repeatability, estimated as the CV, and glycan amount, estimated as the mean of the replicates. The black line represents a consensus power curve fit to all available (mean of replicates, relative standard deviation) pairs, denoted by the light gray dots: CV = 5.0 × Mean^−0.35 (or SD = 0.050 × Mean^0.65). The red line is the power curve fit to the pairs, denoted by the blue diamonds, reported in one data set. The measurement repeatability or this data set is somewhat better than average.

Fig. 7.

Scatterplot of the closeness to consensus of the reported values as a function of measurement repeatability estimated as CV. The symbols are coded by the user-stated nature of the reported replicates. The plot shows the data point, in blue diamond, of one data set.

Fig. 8.

Comparison Consensus Medians to Published Peak Areas²². Each symbol represents this interlaboratory study's consensus median % proportion as a function of the published peak areas for one composition or defined group of compositions. The bars span the central 50% of the distribution of reported values. The solid red circles denote compositions where the central 50% of the values does not include the published peak area. The dashed line represents equality between the two estimates.