Development of a scale down cell culture model using multivariate analysis as a qualification tool

In characterizing a cell culture process to support regulatory activities such as process validation and Quality by Design, developing a representative scale down model for design space definition is of great importance. The manufacturing bioreactor should ideally reproduce bench scale performance with respect to all measurable parameters. However, due to intrinsic geometric differences between scales, process performance at manufacturing scale often varies from bench scale performance, typically exhibiting differences in parameters such as cell growth, protein productivity, and/or dissolved carbon dioxide concentration. Here, we describe a case study in which a bench scale cell culture process model is developed to mimic historical manufacturing scale performance for a late stage CHO‐based monoclonal antibody program. Using multivariate analysis (MVA) as primary data analysis tool in addition to traditional univariate analysis techniques to identify gaps between scales, process adjustments were implemented at bench scale resulting in an improved scale down cell culture process model. Finally we propose an approach for small scale model qualification including three main aspects: MVA, comparison of key physiological rates, and comparison of product quality attributes. © 2013 American Institute of Chemical Engineers Biotechnol. Prog., 30:152–160, 2014     

Characterization of a Saccharomyces cerevisiae fermentation process for production of a therapeutic recombinant protein using a multivariate Bayesian approach

The principle of quality by design (QbD) has been widely applied to biopharmaceutical manufacturing processes. Process characterization is an essential step to implement the QbD concept to establish the design space and to define the proven acceptable ranges (PAR) for critical process parameters (CPPs). In this study, we present characterization of a Saccharomyces cerevisiae fermentation process using risk assessment analysis, statistical design of experiments (DoE), and the multivariate Bayesian predictive approach. The critical quality attributes (CQAs) and CPPs were identified with a risk assessment. The statistical model for each attribute was established using the results from the DoE study with consideration given to interactions between CPPs. Both the conventional overlapping contour plot and the multivariate Bayesian predictive approaches were used to establish the region of process operating conditions where all attributes met their specifications simultaneously. The quantitative Bayesian predictive approach was chosen to define the PARs for the CPPs, which apply to the manufacturing control strategy. Experience from the 10,000 L manufacturing scale process validation, including 64 continued process verification batches, indicates that the CPPs remain under a state of control and within the established PARs. The end product quality attributes were within their drug substance specifications. The probability generated with the Bayesian approach was also used as a tool to assess CPP deviations. This approach can be extended to develop other production process characterization and quantify a reliable operating region. © 2016 American Institute of Chemical Engineers Biotechnol. Prog., 32:799–812, 2016     

Optimization of a Saccharomyces cerevisiae fermentation process for production of a therapeutic recombinant protein using a multivariate bayesian approach

Various approaches have been applied to optimize biological product fermentation processes and define design space. In this article, we present a stepwise approach to optimize a Saccharomyces cerevisiae fermentation process through risk assessment analysis, statistical design of experiments (DoE), and multivariate Bayesian predictive approach. The critical process parameters (CPPs) were first identified through a risk assessment. The response surface for each attribute was modeled using the results from the DoE study with consideration given to interactions between CPPs. A multivariate Bayesian predictive approach was then used to identify the region of process operating conditions where all attributes met their specifications simultaneously. The model prediction was verified by twelve consistency runs where all batches achieved broth titer more than 1.53 g/L of broth and quality attributes within the expected ranges. The calculated probability was used to define the reliable operating region. To our knowledge, this is the first case study to implement the multivariate Bayesian predictive approach to the process optimization for the industrial application and its corresponding verification at two different production scales. This approach can be extended to other fermentation process optimizations and reliable operating region quantitation. © 2012 American Institute of Chemical Engineers Biotechnol. Prog., 28: 1095–1105, 2012     

Fed-batch bioreactor performance and cell line stability evaluation of the artificial chromosome expression technology expressing an IgG1 in Chinese hamster ovary cells

The artificial chromosome expression (ACE) technology system uses an engineered artificial chromosome containing multiple site-specific recombination acceptor sites for the rapid and efficient construction of stable cell lines. The construction of Chinese hamster ovary(CHO) cell lines expressing an IgG1 monoclonal antibody (MAb) using the ACE system has been previously described (Kennard et al., Biotechnol Bioeng. 2009;104:540-553). To further demonstrate the manufacturing feasibility of the ACE system, four CHO cell lines expressing the human IgG1 MAb 4A1 were evaluated in batch and fed-batch shake flasks and in a 2-L fed-batch bioreactor. The batch shake flasks achieved titers between 0.7 and 1.1 g/L, whereas the fed-batch shake flask process improved titers to 2.5–3.0 g/L. The lead 4A1 ACE cell line achieved titers of 4.0 g/L with an average specific productivity of 40 pg/(cell day) when cultured in a non optimized 2-L fed-batch bioreactor using a completely chemically defined process. Generational stability characterization of the lead 4A1-expressing cell line demonstrated that the cell line was stable for up to 75 days in culture. Product quality attributes of the 4A1 MAb produced by the ACE system during the stability evaluation period were unchanged and also comparable to existing expression technologies such as the CHO-dhfr system. The results of this evaluation demonstrate that a clonal, stable MAb-expressing CHO cell line can be produced using ACE technology that performs competitively using a chemically defined fed-batch bioreactor process with comparable product quality attributes to cell lines generated by existing technologies.     

Forecasting and Uncertainty Quantification Using a Hybrid of Mechanistic and Non-mechanistic Models for an Age-Structured Population Model

In this paper, we present a new method for the prediction and uncertainty quantification of data-driven multivariate systems. Traditionally, either mechanistic or non-mechanistic modeling methodologies have been used for prediction; however, it is uncommon for the two to be incorporated together. We compare the forecast accuracy of mechanistic modeling, using Bayesian inference, a non-mechanistic modeling approach based on state space reconstruction, and a novel hybrid methodology composed of the two for an age-structured population data set. The data come from cannibalistic flour beetles, in which it is observed that the adults preying on the eggs and pupae result in non-equilibrium population dynamics. Uncertainty quantification methods for the hybrid models are outlined and illustrated for these data. We perform an analysis of the results from Bayesian inference for the mechanistic model and hybrid models to suggest reasons why hybrid modeling methodology may enable more accurate forecasts of multivariate systems than traditional approaches.     

Maintaining optimal state probabilities in biological systems

A biological problem is usually studied experimentally by reducing it into a number of modules. In contrast, the systems biology approach seeks to address the collective behavior of interacting molecules vis-a-vis the corresponding higher level behavior. Various attributes of a biological system are conditionally dependent on each other, and these conditionalities are usually represented through Bayesian networks for computing easily the joint probability for a state of an attribute. In this article, a genetic algorithm is investigated to a biological system, by representing it through a Bayesian network, for evaluating the optimum state probabilities of different attributes, in order to obtain a desired joint probability for a given state of an attribute. We believe that such a study would be helpful in achieving a desired health condition by maintaining various attributes of a system to their estimated optimum levels.     

A statistical approach to study the interactive effects of process parameters on succinic acid production from Bacteroides fragilis

A statistical approach response surface methodology (RSM) was used to study the production of succinic acid from Bacteroides fragilis. The most influential parameters for succinic acid production obtained through one-at-a-time method were glucose, tryptone, sodium carbonate, inoculum size and incubation period. These resulted in the production of 5.4gL(-1) of succinic acid in 48h from B. fragilis under anaerobic conditions. Based on these results, a statistical method, face-centered central composite design (FCCCD) falling under RSM was employed for further enhancing the succinic acid production and to monitor the interactive effect of these parameters, which resulted in a more than 2-fold increase in yield (12.5gL(-1) in 24h). The analysis of variance (ANOVA) showed the adequacy of the model and the verification experiments confirmed its validity. On subsequent scale-up in a 10-L bioreactor using conditions optimized through RSM, 20.0gL(-1) of succinic acid was obtained in 24h. This clearly indicated that the model stood valid even on large scale. Thus, the statistical optimization strategy led to an approximately 4-fold increase in the yield of succinic acid. This is the first report on the use of FCCCD to improve succinic acid production from B. fragilis. The present study provides useful information about the regulation of succinic acid synthesis through manipulation of various physiochemical parameters.     

Visualizing medium and biodistribution in complex cell culture bioreactors using in vivo imaging

There is a dearth of technology and methods to aid process characterization, control and scale‐up of complex culture platforms that provide niche micro‐environments for some stem cell‐based products. We have demonstrated a novel use of 3d in vivo imaging systems to visualize medium flow and cell distribution within a complex culture platform (hollow fiber bioreactor) to aid characterization of potential spatial heterogeneity and identify potential routes of bioreactor failure or sources of variability. This can then aid process characterization and control of such systems with a view to scale‐up. Two potential sources of variation were observed with multiple bioreactors repeatedly imaged using two different imaging systems: shortcutting of medium between adjacent inlet and outlet ports with the potential to create medium gradients within the bioreactor, and localization of bioluminescent murine 4T1‐luc2 cells upon inoculation with the potential to create variable seeding densities at different points within the cell growth chamber. The ability of the imaging technique to identify these key operational bioreactor characteristics demonstrates an emerging technique in troubleshooting and engineering optimization of bioreactor performance. © 2013 American Institute of Chemical Engineers Biotechnol. Prog., 30:256–260, 2014     

Development of a scale-up strategy for Chinese hamster ovary cell culture processes using the kLa ratio as a direct indicator of gas stripping conditions

Herein, we described a scale-up strategy focused on the dissolved carbon dioxide concentration (dCO₂) during fed-batch cultivation of Chinese hamster ovary cells. A fed-batch culture process for a 2000-L scale stainless steel (SS) bioreactor was scaled-up from similarly shaped 200-L scale bioreactors based on power input per unit volume (P/V). However, during the 2000-L fed-batch culture, the dCO₂ was higher compared with the 200-L scale bioreactor. Therefore, we developed an alternative approach by evaluating the k_La values of O₂ (k_La[O₂]) and CO₂ [k_La(CO₂)] in the SS bioreactors as a scale-up factor for dCO₂ reduction. The k_La ratios [k_La(CO₂)/k_La(O₂)] were different between the 200-L and 2000-L bioreactors under the same P/V condition. When the agitation conditions were changed, the k_La ratio of the 2000-L scale bioreactor became similar and the P/V value become smaller compared with those of the 200-L SS bioreactor. The dCO₂ trends in fed-batch cultures performed in 2000-L scale bioreactors under the modified agitation conditions were similar to the control. This k_La ratio method was used for process development in single-use bioreactors (SUBs) with shapes different from those of the SS bioreactor. The k_La ratios for the SUBs were evaluated and conditions that provided k_La ratios similar to the 200-L scale SS bioreactors were determined. The cell culture performance and product quality at the end of the cultivation process were comparable for all tested SUBs. Therefore, we concluded that the k_La ratio is a powerful scale-up factor useful to control dCO₂ during fed-batch cultures.     

Fed-batch bioreactor process scale-up from 3-L to 2,500-L scale for monoclonal antibody production from cell culture

This case study focuses on the scale-up of a Sp2/0 mouse myeloma cell line based fed-batch bioreactor process, from the initial 3-L bench scale to the 2,500-L scale. A stepwise scale-up strategy that involved several intermediate steps in increasing the bioreactor volume was adopted to minimize the risks associated with scale-up processes. Careful selection of several available mixing models from literature, and appropriately applying the calculated results to our settings, resulted in successful scale-up of agitation speed for the large bioreactors. Consideration was also given to scale-up of the nutrient feeding, inoculation, and the set-points of operational parameters such as temperature, pH, dissolved oxygen, dissolved carbon dioxide, and aeration in an integrated manner. It has been demonstrated through the qualitative and the quantitative side-by-side comparison of bioreactor performance as well as through a panel of biochemical characterization tests that the comparability of the process and the product was well controlled and maintained during the process scale-up. The 2,500-L process is currently in use for the routine clinical production of Epratuzumab in support of two global Phase III clinical trials in patients with lupus. Today, the 2,500 L, fed-batch production process for Epratuzumab has met all scheduled batch releases, and the quality of the antibody is consistent and reproducible, meeting all specifications, thus confirming the robustness of the process.     

Design and scale up of a process for manganese peroxidase production using the hypersecretory strain Phanerochaete chrysosporium I-1512.

Manganese peroxidase (MnP) production was performed in a airlift bioreactor in which Phanerochaete chrysosporium I-1512, an MnP hypersecretory strain, was immobilized on a stainless steel mesh. Production was scaled up from a 2.5-L bench scale to a 100-L bioreactor. The yield of MnP was increased 2-fold and reached 6600 U L(-1). These results indicate the feasibility of MnP production on a medium scale, which promises sufficient MnP availability for its use in pulp bleaching at industrial scale.     

Computational fluid dynamics modeling,a novel,and effective approach for developing scalable cell therapy manufacturing processes

Induced pluripotent stem cells (iPSCs) hold great potential to generate novel, curative cell therapy products. However, current methods to generate these novel therapies lack scalability, are labor-intensive, require a large footprint, and are not suited to meet clinical and commercial demands. Therefore, it is necessary to develop scalable manufacturing processes to accommodate the generation of high-quality iPSC derivatives under controlled conditions. The current scale-up methods used in cell therapy processes are based on empirical, geometry-dependent methods that do not accurately represent the hydrodynamics of 3D bioreactors. These methods require multiple iterations of scale-up studies, resulting in increased development cost and time. Here we show a novel approach using computational fluid dynamics modeling to effectively scale-up cell therapy manufacturing processes in 3D bioreactors. Using a GMP-compatible iPSC line, we translated and scaled-up a small-scale cardiomyocyte differentiation process to a 3-L computer-controlled bioreactor in an efficient manner, showing comparability in both systems.     

Effective temperature shift strategy development and scale confirmation for simultaneous optimization of protein productivity and quality in Chinese hamster ovary cells

Temperature shifts to lower culture temperatures are frequently employed in the manufacturing of protein therapeutics in mammalian cells to improve productivity, viability, or quality attributes. The direction and extent to which a temperature shift affects productivity and quality may vary depending on the expression host and characteristics of the expressed protein. We demonstrated here that two Chinese hamster ovary (CHO) clones expressing different human monoclonal antibodies responded differently to a temperature shift despite sharing a common parental CHO cell line. Within a single CHO line, we observed a nonlinear response to temperature shift. A moderate shift to 35°C significantly decreased final titer relative to the unshifted control while a larger shift to 32°C significantly increased final titer by 25%. Therefore, we proposed a systematic empirical approach to assess the utility of a temperature shift for faster implementation during process development. By testing multiple shift parameters, we identified optimum shift conditions in shake flasks and successfully translated findings to benchtop bioreactors and 1,000-L bioreactor scale. Significant differences in final antibody titer and charge variants were observed with temperature shift increments as small as Δ1.5°C. Acidic charge variants decreased monotonically with decreasing shift temperature in both cell lines; however, final antibody titer required simultaneous optimization of shift day and temperature. Overall, we were able to show that a systematic approach to identify temperature shift parameters at small scales is useful to optimize protein production and quality for efficient and confident translation to large-scale production.     

Joint regression analysis of correlated data using Gaussian copulas

Summary .  This article concerns a new joint modeling approach for correlated data analysis. Utilizing Gaussian copulas, we present a unified and flexible machinery to integrate separate one-dimensional generalized linear models (GLMs) into a joint regression analysis of continuous, discrete, and mixed correlated outcomes. This essentially leads to a multivariate analogue of the univariate GLM theory and hence an efficiency gain in the estimation of regression coefficients. The availability of joint probability models enables us to develop a full maximum likelihood inference. Numerical illustrations are focused on regression models for discrete correlated data, including multidimensional logistic regression models and a joint model for mixed normal and binary outcomes. In the simulation studies, the proposed copula-based joint model is compared to the popular generalized estimating equations, which is a moment-based estimating equation method to join univariate GLMs. Two real-world data examples are used in the illustration.     

Joint prediction of multiple quantitative traits using a Bayesian multivariate antedependence model

Predicting organismal phenotypes from genotype data is important for preventive and personalized medicine as well as plant and animal breeding. Although genome-wide association studies (GWAS) for complex traits have discovered a large number of trait- and disease-associated variants, phenotype prediction based on associated variants is usually in low accuracy even for a high-heritability trait because these variants can typically account for a limited fraction of total genetic variance. In comparison with GWAS, the whole-genome prediction (WGP) methods can increase prediction accuracy by making use of a huge number of variants simultaneously. Among various statistical methods for WGP, multiple-trait model and antedependence model show their respective advantages. To take advantage of both strategies within a unified framework, we proposed a novel multivariate antedependence-based method for joint prediction of multiple quantitative traits using a Bayesian algorithm via modeling a linear relationship of effect vector between each pair of adjacent markers. Through both simulation and real-data analyses, our studies demonstrated that the proposed antedependence-based multiple-trait WGP method is more accurate and robust than corresponding traditional counterparts (Bayes A and multi-trait Bayes A) under various scenarios. Our method can be readily extended to deal with missing phenotypes and resequence data with rare variants, offering a feasible way to jointly predict phenotypes for multiple complex traits in human genetic epidemiology as well as plant and livestock breeding.     

Scale-up of naringinase production process based on the constant oxygen transfer rate for a novel strain of Bacillus methylotrophicus

Naringinase bioprocess based on Bacillus methylotrophicus was successfully scaled up based on constant oxygen transfer rate (OTR) as the scale-up criterion from 5-L bioreactor to 20-L bioreactor. OTR was measured in 5 and 20-L bioreactor under various operating conditions using dynamic method. The operating conditions, where complete dispersion was observed were identified. The highest OTR of 0.035 and 0.04?mMol/L/s was observed in 5 and 20-L bioreactor, respectively. Critical dissolved oxygen concentration of novel isolated strain B. methylotrophicus was found to be 20% of oxygen saturation in optimized medium. The B. methylotrophicus cells grown on sucrose had maximum oxygen uptake rate of 0.14?mMol/L/s in optimized growth medium. The cells produced the maximum naringinase activity of 751 and 778?U/L at 34?hr in 5 and 20-L bioreactors, respectively. The maximum specific growth rate of about 0.178/hr was observed at both the scales of operations. The maximum naringinase yield of 160 and 164?U/g biomass was observed in 5 and 20-L bioreactors, respectively. The growth and production profiles at both scales were similar indicating successful scale-up strategy for B. methylotrophicus culture.     

Characterization and feasibility of a miniaturized stirred tank bioreactor to perform E. coli high cell density fed-batch fermentations

The use of small scale bioreactors that are mechanically and functionally similar to large scale reactors is highly desirable to accelerate bioprocess development because they enable well-defined scale translations. In this study, a 25-mL miniaturized stirred tank bioreactor (MSBR) has been characterized in terms of its power input, hydrodynamics, and volumetric oxygen transfer coefficient (k(L)a) to assess its potential to grow high cell density (HCD) cultures using adequate scale-down criteria. Engineering characterization results show scale down, based on matched specific power input (P(G)/V), is feasible from a 20-L pilot scale stirred tank bioreactor. Results from fed-batch fermentations performed using Fab' producing E. coli W3110 at matched (P(G)/V) in the MSBR and 20-L STR demonstrated that the MSBR can accurately scale down the 20-L fermentation performance in terms of growth and Fab' production. Successful implementation of a fed-batch strategy in the MSBR resulted in maximum optical density of ca. 114 and total Fab' concentration of 940 μg/mL compared with ca. 118 and 990 μg/mL in 20-L STR. Furthermore, the use of the MSBR in conjunction with primary recovery scale-down tools to assess the harvest material of both reactors showed comparable shear sensitivity and centrifugation performance. The conjoint use of the MSBR with ultra scale-down (USD) centrifugation mimics can provide a cost-efficient manner in which to design and develop bioprocesses that account for good upstream performance as well as their manufacturability downstream.     

Reduction of product-related species during the fermentation and purification of a recombinant IL-1 receptor antagonist at the laboratory and pilot scale

Through a parallel approach of tracking product quality through fermentation and purification development, a robust process was designed to reduce the levels of product-related species. Three biochemically similar product-related species were identified as byproducts of host-cell enzymatic activity. To modulate intracellular proteolytic activity, key fermentation parameters (temperature, pH, trace metals, EDTA levels, and carbon source) were evaluated through bioreactor optimization, while balancing negative effects on growth, productivity, and oxygen demand. The purification process was based on three non-affinity steps and resolved product-related species by exploiting small charge differences. Using statistical design of experiments for elution conditions, a high-resolution cation exchange capture column was optimized for resolution and recovery. Further reduction of product-related species was achieved by evaluating a matrix of conditions for a ceramic hydroxyapatite column. The optimized fermentation process was transferred from the 2-L laboratory scale to the 100-L pilot scale and the purification process was scaled accordingly to process the fermentation harvest. The laboratory- and pilot-scale processes resulted in similar process recoveries of 60 and 65%, respectively, and in a product that was of equal quality and purity to that of small-scale development preparations. The parallel approach for up- and downstream development was paramount in achieving a robust and scalable clinical process.