Evaluating Similarity

Biologics are biopharmaceuticals, e.g. proteins, vaccines, mAbs (monoclonal Antibodies). The HOS (Higher Order Structure) and dynamics of proteins are directly linked to their function. NMR is a key analytical technology that provides information with atomic resolution about these large molecules’ conformation, aggrega-tion, stability and modifications such as glycosylation - critical factors in the biological efficacy. NMR is a uniquely valuable tool in the study of HOS because of its intrinsically high information content and ease of use. Recently, new developments in acqui-sition and data analysis have emerged for HOS characterization of the intact molecule at natural abundance, using both 1D and 2D NMR methods. Together with Mestrelab, newly developed “BiologicsHOS” software provides an efficient workflow for eval-uating HOS.

Advances in AcquisitionAtomic resolution HOS evaluation with NMR involves measuring 2D methyl fingerprint spectra of large biologics molecules. For this purpose, we use a 2D 1H-13C ALSOFAST-HMQC with 50% NUS to shorten spectral acquisition times. If required, selective removal of excipient signals by the new SIERRA filter [1] added to the AF-HMQC is used. Examples at 600MHz, 5mm CryoProbe, 2D 1H-13C AF-HMQC and 50% NUS are:�� 170 kDa intact mAb, 0.3 mmol, 7.5 hours (with SIERRA)�� 60 kDa intact protein, 0.4 mmol, 90 minutes (no SIERRA)

Acquisition time at 800MHz: reduction by ~30%.

Peak Position ChangesThe signal positions are sensitive to the HOS of the sample. To track changes in a supervised mode, a peak list from a reference sample is transferred to the test spectrum, and peak position changes are measured. These peak position changes are combined into one number, CCS. The CCSD (Combined Chemical Shift Difference)[2] between peak positions in reference and test spectrum tracks changes in peak positions which report changes in the HOS.

Pointwise Comparison of SpectraECHOS (Easy Comparability of HOS)[3] is a simple method for comparing two spectra pointwise. All points in the region of interest are compared if the intensity is above the noise threshold in at least one spectrum. A linear regression is performed. For easy comparison, the residuals are displayed so that differences are highlighted.

Batch to Batch VariationAllowed variations from batch to batch requires statistical analysis. PCA (Principal Component Analysis) is a robust, commonly used non-supervised dimension reduction method. It searches for variations in the dataset. Either the peak list (CCS), bins or the entire spectrum can be used as input for the analysis. While the CCS based statistics are often well separated, using the entire spectrum may show additional variation.

Biologics and Biosimilars

Fig. 1: The chemical shift differences track changes in peak positions. They are shown in two different plots: one with shifts only and in combination with amplitude ratios.

Fig. 2: ECHOS plot comparting NIST mAb with modified NIST mAb. From left to right: overlaid spectra to compare; linear regression result; differences are highlighted on top of the overlaid spectra. Typical pattern are visible in the regression plot, e.g. in this example a missing peak.

Fig. 3: PCA analysis of HSQC spectra of the NIST mAb shows a clear grouping in the scores plot according to the temperature. The loadings plot allows identification of the responsible signals. Data taken from NIST coordinated interlaboratory study. [4]

Summary�� Routine assessment of HOS of intact biopharmaceuticals

at natural abundance.�� Fingerprints of 2D spectra show differences at atomic

resolution.�� Newly developed Mestrelab / Bruker SW offers conve-

nient evaluation of HOS.

For further information and beta version of the software please contact us at biologicsHOS@bruker.com

[1] Arbogast, Luke W, et.al. Journal of Biomolecular NMR (2018) 72:149–161 [2] Amezcua, Carlos A., et.al.. Journal of Pharmaceutical Sciences 102.6 (2013): 1724-1733. [3] Arbogast, Luke W, et.al.. Analytical Chemistry 87.7 (2015): 3556-3561 [4] Brinson, Robert G., et. al., mAbs, 11:1 (2019), 94-105