Anna Osberghaus

Determination of parameters for the steric mass action model—A comparison between two approaches

The application of mechanistic modeling for the optimization of chromatographic steps increased recently due to time efficiency of algorithms and rising calculation power. In the modeling of ion exchange chromatography steps, the sorption processes occurring on adsorbent particle surfaces can be simulated with the steric mass action (SMA) model introduced by Brooks and Cramer (1992) [14]. In this paper, two approaches for the determination of SMA parameters will be carried out and discussed concerning their specific experimental effort, quality of results, method differences, reasons for uncertainties and consequences for SMA parameter determination: Approach I: estimation of SMA parameters based on gradient and frontal experiments according to instructions in Brooks and Cramer (1992) [14] and Shukla et al. (1998) [16]. Approach II: application of an inverse method for parameter estimation, resulting in SMA parameters that induce a best fit of chromatographic data to a mechanistic model for column chromatography. These approaches for SMA parameter determination were carried out for three proteins (ribonuclease A, cytochrome c and lysozyme) at pH 5 and pH 7. The results were comparable and the order of parameter values and their relations to the chromatographic data similar. Nevertheless, differences in the complexity and effort of methods as well as the parameter values themselves were observed. The comparison of methods demonstrated that discrepancies depend mainly on model sensitivities and additional parameters influencing the calculations. However, the discrepancies do not affect predictivity; predictivity is high in both approaches. The approach based on an inverse method and the mechanistic model has the advantage that not only retention times but also complete elution profiles can be predicted. Thus, the inverse method based on a mechanistic model for column chromatography is the most comfortable way to establish highly predictive SMA parameters lending themselves for the optimization of chromatography steps and process control.

A tool for selective inline quantification of co‐eluting proteins in chromatography using spectral analysis and partial least squares regression

Selective quantification of co‐eluting proteins in chromatography is usually performed by offline analytics. This is time‐consuming and can lead to late detection of irregularities in chromatography processes. To overcome this analytical bottleneck, a methodology for selective protein quantification in multicomponent mixtures by means of spectral data and partial least squares regression was presented in two previous studies. In this paper, a powerful integration of software and chromatography hardware will be introduced that enables the applicability of this methodology for a selective inline quantification of co‐eluting proteins in chromatography. A specific setup consisting of a conventional liquid chromatography system, a diode array detector, and a software interface to Matlab® was developed. The established tool for selective inline quantification was successfully applied for a peak deconvolution of a co‐eluting ternary protein mixture consisting of lysozyme, ribonuclease A, and cytochrome c on SP Sepharose FF. Compared to common offline analytics based on collected fractions, no loss of information regarding the retention volumes and peak flanks was observed. A comparison between the mass balances of both analytical methods showed, that the inline quantification tool can be applied for a rapid determination of pool yields. Finally, the achieved inline peak deconvolution was successfully applied to make product purity‐based real‐time pooling decisions. This makes the established tool for selective inline quantification a valuable approach for inline monitoring and control of chromatographic purification steps and just in time reaction on process irregularities. Biotechnol. Bioeng. 2014;111: 1365–1373.

Optimizing a chromatographic three component separation: A comparison of mechanistic and empiric modeling approaches

The search for a favorable and robust operating point of a separation process represents a complex multi-factor optimization problem. This problem is typically tackled by design of experiments (DoE) in the factor space and empiric response surface modeling (RSM); however, separation optimizations based on mechanistic modeling are on the rise. In this paper, a DoE-RSM-approach and a mechanistic modeling approach are compared with respect to their performance and predictive power by means of a case study - the optimization of a multicomponent separation of proteins in an ion exchange chromatography step with a nonlinear gradient (ribonuclease A, cytochrome c and lysozyme on SP Sepharose FF). The results revealed that at least for complex problems with low robustness, the performance of the DoE-approach is significantly inferior to the performance of the mechanistic model. While some influential factors of the system could be detected with the DoE-RSM-approach, predictions concerning the peak resolutions were mostly inaccurate and the optimization failed. The predictions of the mechanistic model for separation results were very accurate. Influences of the experimental factors could be quantified and the separation was optimized with respect to several objectives. However, the discussion of advantages and disadvantages of empiric and mechanistic modeling generates synergies of both methods and leads to a new optimization concept, which is promising with respect to an efficient employment of high throughput screening data.

Adjoint-based estimation and optimization for column liquid chromatography models

Abstract Simulation and optimization of chromatographic processes are continuously gaining practical importance, as they allow for faster and cheaper process development. Although a lot of effort has been put into developing numerical schemes for simulation, fast optimization and estimation algorithms also are of importance. To determine parameters for an a priori defined model, a suited approach is the inverse method that fits the measurement data to the model response. This paper presents an adjoint method to compute model parameter derivatives for a wide range of differentiable liquid chromatography models and provides practical information for the implementation in a generic simulation framework by the example of ion-exchange chromatography. The example shows that the approach is effective for parameter estimation of model proteins and superior to forward sensitivities in terms of computational effort. An optimization of peak separation in salt step elution demonstrates that the method is not restricted to inverse parameter estimation.

Examination of a genetic algorithm for the application in high-throughput downstream process development

Compared to traditional strategies, application of high-throughput experiments combined with optimization methods can potentially speed up downstream process development and increase our understanding of processes. In contrast to the method of Design of Experiments in combination with response surface analysis (RSA), optimization approaches like genetic algorithms (GAs) can be applied to identify optimal parameter settings in multidimensional optimizations tasks. In this article the performance of a GA was investigated applying parameters applicable in high-throughput downstream process development. The influence of population size, the design of the initial generation and selection pressure on the optimization results was studied. To mimic typical experimental data, four mathematical functions were used for an in silico evaluation. The influence of GA parameters was minor on landscapes with only one optimum. On landscapes with several optima, parameters had a significant impact on GA performance and success in finding the global optimum. Premature convergence increased as the number of parameters and noise increased. RSA was shown to be comparable or superior for simple systems and low to moderate noise. For complex systems or high noise levels, RSA failed, while GA optimization represented a robust tool for process optimization. Finally, the effect of different objective functions is shown exemplarily for a refolding optimization of lysozyme.

Integrated development of up- and downstream processes supported by the Cherry-Tag™ for real-time tracking of stability and solubility of proteins

Product analytics is the bottleneck of most processes in bioprocess engineering, as it is rather time-consuming. Real-time and in-line product tracing without sample pre-treatment is only possible for few products. The Cherry-Tag™ (Delphi Genetics, Belgium) which can be fused to any target protein allows for straightforward product analytics by VIS absorption measurements. When the fused protein becomes unstable or insoluble, the chromophore function of the group is lost, which makes this technology an ideal screening tool for solubility and stability in up- and downstream process development. The Cherry-Tag™ technology will be presented for the tagged enzyme glutathione-S-transferase (GST) from Escherichia coli in a combined up- and downstream process development study. High-throughput cultivations were carried out in a 48-well format in a BioLector system (m2p-Labs, Germany). The best cultivation setup of highest product titer was scaled up to a 2.5L shake flask culture, followed by a selective affinity chromatography product capturing step. In upstream applications the tag was capable of identifying conditions where insoluble and non-native inclusion bodies were formed. In downstream applications the red-colored product was found to be bound effectively to a GST affinity column. Thus, it was identified to be a native and active protein, as the binding mechanism relies on catalytic activity of the enzyme. The Cherry-Tag™ was found to be a reliable and quantitative tool for real-time tracking of stable and soluble proteins in up- and downstream processing applications. Denaturation and aggregation of the product can be detected in-line at any stage of the process. Critical stages can be identified and subsequently changed or replaced.

A comprehensive molecular dynamics approach to protein retention modeling in ion exchange chromatography.

In downstream processing, the underlying adsorption mechanism of biomolecules to adsorbent material are still subject of extensive research. One approach to more mechanistic understanding is simulating this adsorption process and hereby the possibility to identify the parameters with strongest impact. So far this method was applied with all-atom molecular dynamics simulations of two model proteins on one cation exchanger. In this work we developed a molecular dynamics tool to simulate protein-adsorber interaction for various proteins on an anion exchanger and ran gradient elution experiments to relate the simulation results to experimental data. We were able to show that simulation results yield similar results as experimental data regarding retention behavior as well as binding orientation. We could identify arginines in case of cation exchangers and aspartic acids in case of anion exchangers as major contributors to binding.

Calibration-free inverse modeling of ion-exchange chromatography in industrial antibody purification

The identification of optimal process parameters for the isolation of a target component from multicomponent mixtures is especially challenging in industrial applications. With constantly increasing time‐to‐market pressure, screening a large parameter space is not feasible and design‐of‐experiment approaches with few experiments might fail due to dynamic and nonlinear reactions to small parameter changes. Model‐based optimization can determine optimal operating conditions, once the model has been calibrated to the specific process step. In this work, parameters for the steric mass action model were estimated for the target protein and three impurities of an industrial antibody cation‐exchange purification step using only chromatograms at different wavelengths and additional fraction analyses with size exclusion chromatography. Information on the molar or mass concentrations in the feed are not available. The model‐based optimization results coincide with conventional chromatogram‐based optimization.

High-throughput cell quantification assays for use in cell purification development - enabling technologies for cell production.

High-throughput screening (HTS) technology is gaining increasing importance in downstream process development of cell-based products. The development of such HTS-technologies, however, is highly dependent on the availability of robust, accurate, and sensitive high-throughput cell quantification methods. In this article, we compare state-of-the-art cell quantification methods with focus on their applicability in HTS-platforms for downstream processing of cell-based products. Sensitivity, dynamic range, and precision were evaluated for four methods that differ in their respective mechanism. In addition, we evaluated the performance of these methods over a range of buffer compositions, medium densities, and viscosities, representing conditions found in many downstream processing methods. We found that CellTiter-Glo™ and flow cytometry are excellent tools for high-throughput cell quantification. Both methods have broad working ranges (3-4 log) and performed well over a wide range of buffer compositions. In comparison, CyQuant® Direct and CellTracker™ had smaller working ranges and were more sensitive to changes in buffer composition. For fast and sensitive quantification of a single cell type, CellTiter-Glo™ performed best, while for more complex cell mixtures flow cytometry is the method of choice. Our analysis will facilitate the selection of the most suitable method for a specific application and provides a benchmark for future HTS development in downstream processing of cell-based products.

Deconvolution of high‐throughput multicomponent isotherms using multivariate data analysis of protein spectra

Gaining a more profound understanding of biopharmaceutical downstream processes is a key demand of the Quality by Design (QbD) guidelines. One of the most dominant approaches to gain process understanding is the extensive use of experimental high‐throughput formats, such as batch chromatography on robotic liquid handling stations. Using these high‐throughput experimental formats, the generation of numerous samples poses an enormous problem to subsequent analytical techniques. Here, a high‐throughput case study for batch chromatographic multicomponent isotherms is presented. To debottleneck the subsequent analytics, a noninvasive technique using UV spectra and multivariate statistics was adapted to a batch chromatographic format. Using this approach, it was possible to integrate the entire analytical setup into the robotic workflow. As a case study, batch isotherms for sulfopropyl sepharose fast flow and the model proteins cytochrome c and lysozyme at various pH values and ionic strengths were recorded. A successful examination of the quality of the analytical procedure compared to classical single wavelength photometry was carried out. To address the growing demand for a more profound process understanding, the experimental data were fitted to the steric mass action isotherm, getting a more detailed insight into the competitive binding behavior at various pH values and ionic strengths.