Characterization of lysozyme PEGylation products using polarized excitation-emission matrix spectroscopy

The growing use of therapeutic proteins requires accurate analytical techniques for measuring biophysical and structural changes during manufacturing. This is particularly true for the PEGylation of proteins, because characterization of PEGylation reactions and products can often be difficult due to the relatively small impact on protein structure, the lack of an accessible polyethylene glycol (PEG) chromophore, and the heterogeneous final product mixtures. Intrinsic fluorescence spectroscopy is one potential solution due to its relatively high sensitivity to small changes in protein structure and its suitability for online or atline measurements. In this study, we use the PEGylation of lysozyme as a model system to determine the efficacy of polarized excitation-emission matrix (pEEM) spectroscopy as a rapid tool for characterizing the structural variability of the lysozyme (LZ) starting materials and PEGylated products with varying PEG-to-protein ratios (PPR). Dynamic light scattering showed that as PPR increased from 0 to 2.8, the hydrodynamic radius increased from ∼2.2 to 4.8 nm. pEEM measurements provided several sources of information: Rayleigh scattering to identify size changes and aggregate/particle formation, and fluorescence emission to assess chemical and structural changes. PEGylation induced sufficient physicochemical changes in LZ, which produced changes in the pEEM spectra, largely due to variations in the hydrophobic environments of tryptophan residues close to a PEG attachment site. These significant spectral changes when modeled using conventional multivariate analysis methods were able to easily discriminate the raw product solutions according to the degree of PEGylation and were also able to predict PPR with reasonable accuracy (root mean square error for calibration ∼10%, relative error of prediction < 20%), considering the reference size exclusion chromatography method error of ∼7.2%. The variable selection of the pEEM data suggests that equivalent predictions could be obtained with faster and simpler two-dimensional spectra, making the method a more viable online measurement method.