Bioprocess monitoring using Polarised Multidimensional Fluorescence (pMDF) spectroscopy

Bioprocess monitoring is essential for biopharmaceutical manufacturing as it helps to ensure the required quality of the therapeutic proteins and maximise the productivity of the bioprocess as per Quality-by-Design (QbD) principles. The monitoring of critical process parameters (CPPs) such as viable cell density and protein quantification is complicated by the complex chemical composition of the bioreactor broth. This thesis presents the systematic development and application of polarised multidimensional fluorescence (pMDF) spectroscopy as an analytical technique for bioprocess monitoring.First, we show the early-stage development and optimization of a polarised Total Synchronous Fluorescence Spectroscopy (pTSFS) method for protein quantification in a hydrolysate-protein model (mimics clarified bioreactor broth samples) using a standard benchtop laboratory fluorometer. We used UV transmitting polarizers to provide wider range pTSFS spectra for screening of the four different TSFS spectra generated by the measurement. We identified TSFS||(parallel polarised) measurements were the best for protein quantification compared to standard unpolarized measurements and the Bradford assay. This was because TSFS||spectra had a better analyte signal to noise ratio (SNR), due to the anisotropy of protein emission making protein signals better resolved from the background emission of small molecule fluorophoresin the cell culture media. SNR of > 5000 was achieved for concentrations of BSA/YST 1.2/10 g L–1with TSFS||. Optimisation using genetic algorithm and interval Partial Least Squares based variable selection enabled reduction of spectral resolution and number of excitation wavelengths required without degrading performance. This enables fast (<3.5 minutes) online/at-line measurements, and the method had an LOD of 0.18 g L–1and high accuracy with predictive error of <9%. Next, we evaluate the relative efficacy of polarised MDF methods such Excitation Emission Matrix (EEM), TSFS, and Resonant/90° Light Scattering (RLS) spectroscopy (90LSS) measurements for the quantitative analysis of a more complex, multi-protein bioreactor broth model using chemometricmethods including Principal Component Analysis (PCA) and Partial Least Squares (PLS) regression. TSFS||measurements yielded predictive accuracies of % REP = 11% for total protein and % REP = 12% for IgG quantification.For this system we also realised that EEMǁmeasurements, being more sensitive to light scatter contamination were less robust compared toEEM⊥. Normalised RLS||was more accurate in predicting the total protein concentration (%REP = 14%) compared to specific IgG quantification (%REP = 20%) due to the difference in size between BSA and IgG.In our final study we evaluate the application of pMDF in biopharmaceutical process development which often involves the use of small-scale bioreactors (SSBR) for optimising media formulations and process conditions during scale up to commercial scale production. Specifically, we explore the efficacy of polarised total synchronous fluorescence spectroscopy (pTSFS) and Resonant Light Scattering (RLS) to qualitatively monitor and quantitatively predict titre and VCD respectively in a large-scale media optimisation SSBR study. The study involved 71 different media formulations, with over 50 components of varying concentrations, and the bioprocess was run for 13 days or more. Samples were extracted at set times (Day 0, 3, 6, 9, and 13), clarified by centrifugation, and then measured using pTSFS and RLS. These showed significant decrease in fluorescence intensity, changes in spectral profile and increased light scatter over the bioprocess. However, simple mean and standard deviation analysis and Principal component analysis (PCA) of the pTSFS data showed that the degree of variation was greater between media formulations than due to the evolving bioprocess. Spectral clustering methods were used to generate sample subsets for more accurate PLS regression modelling for real time monitoring of CPPs. Classification methods were used to predict the product titer at day 9 from the fluorescence spectra measured at the start of the bioprocess (day 0) with acceptable accuracy. RLS|| measurements correlated well with DLS and showed the potential in predicting VCD in future samples with less variation in media composition such as in actual commercial manufacturing bioprocess. Combining emission and scatter measurements with multivariate data analysis provides a more holistic, multi-attribute bioprocess monitoring method that minimises the need to use different offline analytical methods. We believe this approach will be very valuable for optimising media compositions with potential for high productivity while avoiding the cost of running to completion media formulations that results in poor productivity

Publisher

NUI Galway

Rights

Attribution-NonCommercial-NoDerivs 3.0 Ireland
CC BY-NC-ND 3.0 IE

cbn

Collections

University of Galway Theses (PhD Theses)