Selective protein quantification for preparative chromatography using variable pathlength UV/Vis spectroscopy and partial least squares regression

In preparative protein chromatography, broad dynamic ranges of protein concentrations as well as coelution of product and impurities are common. Despite being the standard in biopharmaceutical production, monitoring of preparative chromatography is generally limited to surrogate signals, e.g. UV absorbance at 280 nm. To address this problem, variable pathlength (VP) spectroscopy in conjunction with Partial Least Squares regression (PLS) was used to monitor preparative chromatography. While VP spectroscopy enabled the acquisition of absorbance data for a broad concentration range, PLS modelling allowed for the differentiation between the protein species. The approach was first implemented for monitoring the separation of lysozyme from cytochrome c at an overall loading density of 92 g/l. The same method was then applied to the polishing step of a monoclonal antibody (mAb) at 40 g/l loading density. For PLS model prediction of the mAb monomer and the high molecular weight variants (HMWs), the root mean square error (RMSE) was 1.07 g/l and 0.42 g/l respectively. To demonstrate the usability of the approach for in-line control, pooling decisions for both separation problems were subsequently taken based on the computed concentrations or thereof derived purities. In summary, VP spectroscopy in conjunction with PLS modelling is a promising option for in-line monitoring and control of future chromatography steps at large scale. (C) 2017 The Authors. Published by Elsevier Ltd.