An industrial perspective on bioreactor scale-down: what we can learn from combined large-scale bioprocess and model fluid studies

For industrial bioreactor design, operation, control and optimization, the scale-down approach is often advocated to efficiently generate data on a small scale, and effectively apply suggested improvements to the industrial scale. In all cases it is important to ensure that the scale-down conditions are representative of the real large-scale bioprocess. Progress is hampered by limited detailed and local information from large-scale bioprocesses. Complementary to real fermentation studies, physical aspects of model fluids such as air-water in large bioreactors provide useful information with limited effort and cost. Still, in industrial practice, investments of time, capital and resources often prohibit systematic work, although, in the end, savings obtained in this way are trivial compared to the expenses that result from real process disturbances, batch failures, and non-flyers with loss of business opportunity. Here we try to highlight what can be learned from real large-scale bioprocess in combination with model fluid studies, and to provide suitable computation tools to overcome data restrictions. Focus is on a specific well-documented case for a 30-m(3) bioreactor. Areas for further research from an industrial perspective are also indicated.     

Study and modeling of fluctuating dissolved oxygen concentration impact on Corynebacterium glutamicum growth in a scale-down bioreactor

In this study, the impact of dissolved oxygen concentrations oscillations on Corynebacterium glutamicum 2262 ΔldhA growth was studied experimentally and modeled. Aiming at this, a dedicated two-compartment scale down set-up composed of two interconnected aerobic/anaerobic stirred tank bioreactors was used. The mean residence time of bacteria in each compartment was modified by adapting circulation rates and culture volumes in each bioreactor and the resulting temporal ratio of aeration was calculated. The five growth kinetics were then modeled using an original kinetic model coupling Monod growth modeling and the Residence Time Distributions. Our study showed that the microbial growth rate and macroscopic yields were clearly linked to the temporal ratio of aeration, allowing the definition of simple but robust law for process scale-up purpose. It was also revealed that the model proposed precisely agreed with the experimental growth data, whatever the fractions of aeration time imposed experimentally.     

A two-compartment bioreactor system made of commercial parts for bioprocess scale-down studies: impact of oscillations on Bacillus subtilis fed-batch cultivations

This study describes an advanced version of a two-compartment scale-down bioreactor that simulates inhomogeneities present in large-scale industrial bioreactors on the laboratory scale. The system is made of commercially available parts and is suitable for sterilization with steam. The scale-down bioreactor consists of a usual stirred tank bioreactor (STR) and a plug flow reactor (PFR) equipped with static mixer modules. The PFR module with a working volume of 1.2 L is equipped with five sample ports, and pH and dissolved oxygen (DO) sensors. The concept was applied using the non-sporulating Bacillus subtilis mutant strain AS3, characterized by a SpoIIGA gene knockout. In a fed-batch process with a constant feed rate, it is found that oscillating substrate and DO concentration led to diminished glucose uptake, ethanol formation and an altered amino acid synthesis. Sampling at the PFR module allowed the detection of dynamics at different concentrations of intermediates, such as pyruvic acid, lactic acid and amino acids. Results indicate that the carbon flux at excess glucose and low DO concentrations is shifted towards ethanol formation. As a result, the reduced carbon flux entering the tricarboxylic acid cycle is not sufficient to support amino acid synthesis following the oxaloacetic acid branch point.     

An integrated scale-down plant for optimal recombinant enzyme production by Pichia pastoris

An integrated bioprocess was created in a scale-down production plant by developing a two-stage enzyme production process with Pichia pastoris, containing a cell-breeding reactor and a production reactor in combination with a three-stage downstream process. To harvest the secreted enzymes, a disc separator and a cross-flow microfiltration clear the broth from the cells. Purification with hydrophobic interaction chromatography removes other proteins, concentrates the product, and prepares the enzyme solution for lyophilization. Fully automated and broad observable multi-stage parallel process courses have been developed using industrial process control systems and at-line measurements for enzyme concentration and enzyme activity. Optimal process conditions were found by application of Design of Experiments (DoE) for the production process.     

Advanced microscale bioreactor system: a representative scale-down model for bench-top bioreactors

In recent years, several automated scale-down bioreactor systems have been developed to increase efficiency in cell culture process development. ambr™ is an automated workstation that provides individual monitoring and control of culture dissolved oxygen and pH in single-use, stirred-tank bioreactors at a working volume of 10–15 mL. To evaluate the ambr™ system, we compared the performance of four recombinant Chinese hamster ovary cell lines in a fed-batch process in parallel ambr™, 2-L bench-top bioreactors, and shake flasks. Cultures in ambr™ matched 2-L bioreactors in controlling the environment (temperature, dissolved oxygen, and pH) and in culture performance (growth, viability, glucose, lactate, Na⁺, osmolality, titer, and product quality). However, cultures in shake flasks did not show comparable performance to the ambr™ and 2-L bioreactors.     

Modeling industrial centrifugation of mammalian cell culture using a capillary based scale-down system

Continuous-flow centrifugation is widely utilized as the primary clarification step in the recovery of biopharmaceuticals from cell culture. However, it is a challenging operation to develop and characterize due to the lack of easy to use, small-scale, systems that can be used to model industrial processes. As a result, pilot-scale continuous centrifugation is typically employed to model large-scale systems requiring a significant amount of resources. In an effort to reduce resource requirements and create a system which is easy to construct and utilize, a capillary shear device, capable of producing energy dissipation rates equivalent to those present in the feed zones of industrial disk stack centrifuges, was developed and evaluated. When coupled to a bench-top, batch centrifuge, the capillary device reduced centrate turbidity prediction error from 37% to 4% compared to using a bench-top centrifuge alone. Laboratory-scale parameters that are analogous to those routinely varied during industrial-scale continuous centrifugation were identified and evaluated for their utility in emulating disk stack centrifuge performance. The resulting relationships enable bench-scale process modeling of continuous disk stack centrifuges using an easily constructed, scalable, capillary shear device coupled to a typical bench-top centrifuge.     

Green fluorescent protein (GFP) leakage from microbial biosensors provides useful information for the evaluation of the scale-down effect

Mixing deficiencies can be potentially detected by the use of a dedicated whole cell microbial biosensor. In this work, a csiE promoter induced under carbon-limited conditions was involved in the elaboration of such biosensor. The cisE biosensor exhibited interesting response after up and down-shift of the dilution rate in chemostat mode. Glucose limitation was accompanied by green fluorescent protein (GFP) leakage to the extracellular medium. In order to test the responsiveness of microbial biosensors to substrate fluctuations in large-scale, a scale-down reactor (SDR) experiment was performed. The glucose fluctuations were characterized at the single cell level and tend to decrease the induction of GFP. Simulations run on the basis of a stochastic hydrodynamic model have shown the variability and the frequencies at which biosensors are exposed to glucose gradient in the SDR. GFP leakage was observed to a great extent in the case of a culture operated in well-mixed fed-batch mode, by comparison with those operated in SDR. GFP leakage seems to be correlated to a higher membrane permeability, confirming previous studies highlighting a better cell viability in cultures operated in a fluctuating environment. Our results suggest that GFP leakage could be used in parallel to the normal GFP biosensor function in order to assess microbial viability in process conditions.     

The scale-down of an industrial disc stack centrifuge

The means are described whereby a disc stack centrifuge may be scaled-down by up to 10-fold of its separation capacity. The centrifuge separation characteristics so measured are suitable for direct scale-up predictions of centrifuge performance where only small volumes of particle suspension are available for study. Such an ability to scale-down is especially important in the processing of biological particles where for example, in the early stage of process development, there is often insufficient fermentation broth for fullscale studies. Scale-down is achieved by the reduction of the number of discs available for separation purposes and by the careful positioning of these discs in the overall disc stack. A combination of dye tracer and particle separation studies are used to optimise the disc stack configuration. The resulting grade efficiency curve is an accurate reflection of the curve for the full-scale centrifuge especially in the critical design region specifying centrifuge throughput for near complete particle recovery.     

Host load prediction using linear models

This paper evaluates linear models for predicting the Digital Unix fivesecond host load average from 1 to 30 seconds into the future. A detailed statistical study of a large number of long, fine grain load traces from a variety of real machines leads to consideration of the Box–Jenkins models (AR, MA, ARMA, ARIMA), and the ARFIMA models (due to selfsimilarity.) We also consider a simple windowedmean model. The computational requirements of these models span a wide range, making some more practical than others for incorporation into an online prediction system. We rigorously evaluate the predictive power of the models by running a large number of randomized testcases on the load traces and then datamining their results. The main conclusions are that load is consistently predictable to a very useful degree, and that the simple, practical models such as AR are sufficient for host load prediction. We recommend AR(16) models or better for host load prediction. We implement an online host load prediction system around the AR(16) model and evaluate its overhead, finding that it uses miniscule amounts of CPU time and network bandwidth.     

Anthraquinones production in Rubia tinctorum cell suspension cultures: Down scale of shear effects

The effect of turbulence on suspended cells is one of the most complex problems in the scale-up of cell cultures. In the present paper, a direct comparison of the effects of turbulence on suspension cultures of Rubia tinctorum in a standard bioreactor and in shake flask cultures was done. A procedure derived from the well known global method proposed by Nishikawa et al. (1977) [39] was applied. Standard flasks and four-baffled shake flasks were used. The effect of turbulence and light irradiation on cell viability, biomass, and anthraquinones (AQs) production was evaluated. The biomass concentration and AQs production obtained using baffled shake flasks agitated at 360 rpm were similar to that achieved in R. tinctorum suspension cultures growing in a stirred tank bioreactor operating at 450 rpm, previously published (Busto et al., 2008 [17]). The effect of light on AQs production was found to be very significant, and a difference of up to 48% was found in cells with and without illumination after 7 days of culture. It is concluded that this down-scaled and simple flask culture system is a suitable and valid small scale instrument for the study of intracellular mechanisms of turbulence-induced AQs production in R. tinctorum suspension cultures.     

Performance of reclassification statistics in comparing risk prediction models

Concerns have been raised about the use of traditional measures of model fit in evaluating risk prediction models for clinical use, and reclassification tables have been suggested as an alternative means of assessing the clinical utility of a model. Several measures based on the table have been proposed, including the reclassification calibration (RC) statistic, the net reclassification improvement (NRI), and the integrated discrimination improvement (IDI), but the performance of these in practical settings has not been fully examined. We used simulations to estimate the type I error and power for these statistics in a number of scenarios, as well as the impact of the number and type of categories, when adding a new marker to an established or reference model. The type I error was found to be reasonable in most settings, and power was highest for the IDI, which was similar to the test of association. The relative power of the RC statistic, a test of calibration, and the NRI, a test of discrimination, varied depending on the model assumptions. These tools provide unique but complementary information.     

Local protein structure prediction using discriminative models

Background  

In recent years protein structure prediction methods using local structure information have shown promising improvements. The quality of new fold predictions has risen significantly and in fold recognition incorporation of local structure predictions led to improvements in the accuracy of results.     

An ultra scale-down approach to assess the impact of the choice of recombinant P. pastoris strain on dewatering performance in centrifuges

Pichia pastoris is becoming a desirable host in the biopharmaceutical industry for therapeutics production. It grows on methanol to high cell densities ≥100 g DCW/L and secretes foreign proteins at high titers. However, the culture conditions to reach high cell densities pose a challenge to the processability by primary recovery operations, in particular centrifugation, used for cell removal. This work aims to assess the impact of recombinant P. pastoris strain selection on centrifugal dewatering. Normally, the choice of P. pastoris recombinant strain is based on best target protein expression levels; however, it is unknown whether the choice of strain will have an impact on performance of centrifugation operation. To achieve this aim, a previously developed laboratory ultra‐scale down (USD) methodology that successfully predicted centrifugal dewatering of pilot‐scale disk‐type machines, was used in this work. Two recombinant P. pastoris strains, namely a X‐33 and a glycoengineered Pichia strain, were used to perform fermentations secreting different products. The resulting harvested fermentation culture properties were analyzed and the dewatering performances of a pilot‐ and a large‐scale disk‐type centrifuge were evaluated using the USD methodology. The choice of P. pastoris strain was found to have a considerable impact on dewatering performance, with P. pastoris X‐33 strain reaching better dewatering levels than the glycoengineered strain. The USD method proved to be a useful tool to determine optimal conditions under which the large scale centrifuge needed to be operated, reducing the need for repeated pilot‐scale runs during early stages of process development for therapeutic products. © 2012 American Institute of Chemical Engineers Biotechnol. Prog., 28: 1029–1036, 2012     

Ab initio protein structure prediction using pathway models

Ab initio prediction is the challenging attempt to predict protein structures based only on sequence information and without using templates. It is often divided into two distinct sub-problems: (a) the scoring function that can distinguish native, or native-like structures, from non-native ones; and (b) the method of searching the conformational space. Currently, there is no reliable scoring function that can always drive a search to the native fold, and there is no general search method that can guarantee a significant sampling of near-natives. Pathway models combine the scoring function and the search. In this short review, we explore some of the ways pathway models are used in folding, in published works since 2001, and present a new pathway model, HMMSTR-CM, that uses a fragment library and a set of nucleation/propagation-based rules. The new method was used for ab initio predictions as part of CASP5. This work was presented at the Winter School in Bioinformatics, Bologna, Italy, 10-14 February 2003.     

Semi-continuous scale-down models for clone and operating parameter screening in perfusion bioreactors

Perfusion cell culture, confined traditionally to the production of fragile molecules, is currently gaining broader attention in the biomanufacturing of therapeutic proteins. The development of these processes is made difficult by the limited availability of appropriate scale-down models. This is due to the continuous operation that requires complex control and cell retention capacity. For example, the determination of an optimal perfusion and bleed rate for continuous cell culture is often performed in scale-down bioreactors and requires a substantial amount of time and effort. To increase the experimental throughput and decrease the required workload, a semi-continuous procedure, referred to as the VCD_max (viable cell density) approach, has been developed on the basis of shake tubes (ST) and deepwell plates (96-DWP). Its effectiveness has been demonstrated for 12 different CHO-K1-SV cell lines expressing an IgG1. Further, its reliability has been investigated through proper comparisons with perfusion runs in lab-scale bioreactors. It was found that the volumetric productivity and the CSPR_min (cell specific perfusion rate) determined using the ST and 96-DWP models were successfully (mostly within the experimental error) confirmed in lab-scale bioreactors, which then covered a significant scale-up from the half milliliter to the liter scale. These scale-down models are very useful to design and scale-up optimal bioreactor operating conditions as well as screening for different media and cell lines.     

Characteristics of mechanically disrupted bakers' yeast in relation to its separation in industrial centrifuges

The viscosity, density, and sedimentation characteristics of suspensions of whole and mechanically disrupted yeast cells were measured. Mechanical disruption increases the suspension viscosity and its non-Newtonian behavior. Experiments showed a good correlation between laboratory- and industrial-scale centrifugation results.     

Disulfide connectivity prediction with 70% accuracy using two-level models

Disulfide bridges stabilize protein structures covalently and play an important role in protein folding. Predicting disulfide connectivity precisely helps towards the solution of protein structure prediction. Previous methods for disulfide connectivity prediction either infer the bonding potential of cysteine pairs or rank alternative disulfide bonding patterns. As a result, these methods encode data according to cysteine pairs (pair-wise) or disulfide bonding patterns (pattern-wise). However, using either encoding scheme alone cannot fully utilize the local and global information of proteins, so the accuracies of previous methods are limited. In this work, we propose a novel two-level framework to predict disulfide connectivity. With this framework, both the pair-wise and pattern-wise encoding schemes are considered. Our models were validated on the datasets derived from SWISS-PROT 39 and 43, and the results demonstrate that our models can combine both local and global information. Compared to previous methods, significant improvements were obtained by our models. Our work may also provide insights to further improvements of disulfide connectivity prediction and increase its applicability in protein structure analysis and prediction.