Multivariate curve resolution -alternating least squares augmented with partial least squares baseline correction applied to mid-IR laser spectra resolves protein denaturation by reducing rotational ambiguity

Abstract

High spectral power density provided by advances in external cavity quantum cascade lasers (EC-QCL) have enabled increased transmission path lengths in mid-infrared (mid-IR) spectroscopy for more sensitive measurement of proteins in aqueous solutions. These extended path lengths also facilitate flow through measurements by avoiding congestion of the flow cell by protein aggregates. Despite the advantages presented by laser-based mid-IR spectroscopy of proteins, extraction of secondary structure information from spectra, especially in the presence of complex multi-component matrices with overlapping spectral features, remains an impediment that requires fine tuning of evaluation algorithms (e.g., band fitting, interpretation of second derivative spectra etc.). In this work, the use of multivariate curve resolution alternating least squares (MCR-ALS) for the analysis of a chemical de- and renaturation experiment has been demonstrated, since this technique offers the second-order advantage of extracting spectral signatures and concentration profiles even in the presence of unknown, uncalibrated constituents. Furthermore, we exhibit a partial least squares regression (PLSR) based subtraction of matrix component spectra prior to MCR-ALS as a method to obtain secondary structure information even in the absence of reference spectra. These approaches are showcased using the online reaction monitoring of the titration of β-lactoglobulin (β-LG) in water against the surfactants sodium dodecyl sulfate (SDS) and octaethylene glyol monododecyl ether (C12E8), using a commercially available laser-based IR spectrometer. Results for the automated PLSR correction plus MCR-ALS approach compare favorably to an MCR-ALS standalone approach using initial estimates as well as analysis of secondary structure using data processed with a manual baseline correction. The herein described chemometric approach suggests a way to simplify the challenge of handling complex matrices in protein structure analysis by isolating the background from the protein contributions, prior to analysis via other soft-modelling techniques. Consequently, the findings of this study indicate the suitability of online reaction monitoring through mid-IR spectroscopy combined with chemometric techniques as a potential tool in downstream quality control and process automation.