May 28, 2021
revised December 16, 2022
Using dynamic feeds to facilitate successful bioprocess development and cell line screening
Many bioprocess feed strategies are either fixed to a predetermined time series of setpoints or adjusted manually throughout the process in response to laborious offline measurements of residual substrate. However, neither fixed feed regimes nor intermittent adjustments can deliver feed optimally due to the kinetic nature of feed requirements over the course of a production run. The consequence of overfeeding or underfeeding, even by a small magnitude, is starvation or overflow metabolism, respectively, and both can result in subpar performance.
In the example below (Figure 1), a fed-batch bioprocess was developed using a time profile-based feed scheme developed for an individual cell line with defined setpoints in the process. Manual adjustments to the feed rate were made based on the buildup of residual glucose in the cell culture fluid, which is indicative of overfeeding. Fixed feed schemes rely on consistent performance of a cell, which is unlikely given changing setpoints in the context of process development or genetics when cell line screening. Any changes to the feed rates are reactive, based on overfeeding that has already happened in the period since the last sample, which can be many hours, and there is rarely an indication of slight underfeeding.
Fixed feed scheme fed-batch culture
In contrast, a dynamic fed-batch process is one where substrate feeds are delivered in response to physiological signals received from the cells in response to local substrate availability. When variables that impact substrate demand are modified during bioprocess development, dynamic feed strategies can help ensure that substrates are delivered in an optimal manner. While dynamic feed strategies can be extremely valuable in facilitating successful bioprocess development and cell line screening, they must be developed and optimized with the individual organism’s metabolism in mind.
A common requirement for all dynamic processes is an online signal with sufficient sensitivity and reliability to detect the physiological cues of increased substrate demand, excess feeding, and overflow metabolism. Thus, the design of dynamic processes is limited to signals that can be measured, such as pH, dissolved oxygen, and off-gas measurements. The measurement frequency must be high enough for feeds to be adjusted in a timely manner in response to signals. Given these options, many different dynamic feeding strategies can be supported by Culture Biosciences’ 250mL reactors. Dynamic schemes can then be further customized based on client requirements for each cultivation process.
pH and DO stats
Hardware for continuous online measurement of pH and dissolved oxygen (DO) are standard on bioreactors and therefore allow the development of dynamic processes using readily integrated equipment. In the absence of the production of a basic product, cell metabolism of glucose is known to drive down the pH of the cell culture fluid, necessitating the addition of base in order to maintain the optimal pH. This phenomenon can be leveraged in a pH-stat culture, where the exhaustion of substrate results in an increase in pH, which is used as a signal for feeding. Similarly, active metabolism in aerobic processes results in the consumption of oxygen, which can be measured by a dissolved oxygen probe. Upon exhaustion of a substrate, the decrease in metabolic activity can be detected by an increase in dissolved oxygen, indicating a decrease in oxygen demand. In both cases, the increase in pH or dissolved oxygen can be used to indicate substrate depletion and initiate the addition of more substrate.
OUR, CER and RQ-stats
Oxygen Uptake Rate (OUR), Carbon Dioxide Evolution Rate (CER) and Respiratory Quotient (RQ) stats leverage off-gas measurements to determine when to deliver feeds. These have the advantage of being tuned to quantitative metabolic metrics. For example, the timing of feeds could be tuned to maintain a constant oxygen uptake rate or a defined respiratory quotient to minimize overflow metabolism. However, these methods require continuous monitoring of off-gas, which requires hardware that is not standard with many basic bioprocess systems.
Pulse feed strategies
Pulse feed strategies can be used for various dynamic feeding schemes. These schemes rely on the exhaustion of the primary carbon source in the medium resulting in a sharp signal such as an increase in pH, increase in DO, or a decrease in OUR or CER. This signal is then used to initiate a bolus of feed which, when consumed, results in another spike in signal and the sequence is repeated continuously until the conclusion of the bioprocess. These strategies have an additional advantage of being able to leverage both a robust signal and simulate periods of substrate excess and limitation similar to what may be observed in a large-scale reactor.
In the example below (Figure 2), a DO-stat feed strategy was implemented for a fed-batch process. Here (Figure 2a) an increase in the DO over a 15% threshold was used to trigger a bolus of feed. Over the course of the run (Figure 2b), the frequency of the boluses decreases later in the cultivation, as the cells become less metabolically active over time. This strategy ensures that the feed is delivered at the correct rate throughout the bioprocess, based on demand from the cells.
DO-stat pulse feed fed-batch bioprocess
Feed rate adjustment strategies
Setpoint maintenance strategies use similar measurements, but avoid the periods of excess and starvation that are generated in the pulse feed strategies described above. With this feed architecture, a baseline feed rate is programmed in the recipe, and is then adjusted up or down (or cycled on and off) in order to maintain the specified setpoint using control loops.
While this strategy avoids the oscillations in substrate concentration observed in pulse feeding schemes, there are multiple organisms where overflow metabolism will not be detected from pH and DO signals generated from this scheme. For this reason, this strategy may be best suited for organisms without robust overflow metabolism or after thoroughly characterizing the process to identify setpoints that do not result in overfeeding. Adjusting the feed rate based on RQ, which can be tuned to target the desired metabolic state, is well suited for this mechanism of feed control.
Below is an example of a feed-rate adjustment OUR-stat fed-batch process (Figure 3). For this strategy, a desired oxygen uptake rate was defined. At the conclusion of the batch phase, which can be observed by a dip in OUR at 16.75 hours EFT, the substrate feed is initiated (purple lines) and used to maintain the OUR at a constant setpoint (blue lines).
Strategies optimized for specific organism metabolic profiles - Saccharomyces cerevisiae case study
Many microbes, including those widely used in industrial biotechnology, are subject to overflow metabolism, whereby they produce various metabolites in response to high exogenous levels of substrate. One of the most extensively characterized types of overflow metabolism is the Crabtree effect, first described by Herbert Crabtree in the yeast Saccharomyces cerevisiae, where excess glucose is converted into ethanol under aerobic conditions (Crabtree, 1929). Similarly, E.coli will produce the inhibitory metabolite acetate in the presence of excess glucose under aerobic conditions (Luli and Strohl, 1990). Consequently, various dynamic feeding strategies have been developed by different groups to account for the features of each microbe’s metabolism. These strategies can facilitate optimal substrate delivery in the context of other changes such as routine week-to-week variability, different cell line genotypes or different process setpoints as a part of process development.
There are a number of different published strategies available for a variety of organisms. The selection of a dynamic feed strategy should be based not only on the basis of the organism, but also the desired metabolic state and carbon fluxes to be targeted during the bioprocess.
For example, different strategies to modulate carbon fluxes throughout the bioprocess have been used to develop dynamic processes for Saccharomyces cerevisiae based not only on the physiology of the organism, but also on the type of product.
Henes and Sonnleitner (2007) employed a dynamic strategy based on the response of the dissolved oxygen (DO) signal upon a brief disruption to the feed. In cultures where substrate limitation is maintained and there is no excess substrate available (glucose or ethanol), the DO will rise sharply in response to cessation of the feed. If the culture is largely overfed, there will be no immediate DO response when the feed stops. In the case of slight overfeeding, the peak observed in response to the interruption of the feed will be smaller in magnitude compared to the previous peak. These features can be used to maintain an optimal feed rate by adjusting the feed rate up or down based on the DO response in reaction to regular interruptions to the feed.
Many in the field (Tippman et al, Xiong et al, Zigova) have leveraged the fact that the state of yeast metabolism can be measured using RQ. These groups maintained an RQ setpoint using traditional proportional integral derivative (PID) control loops to adjust the feed rates. RQ setpoints ranging from 0.6 - 1.0 were selected to maintain the desired control over the metabolism throughout the fermentation and prevent the buildup of overflow metabolites.
Other microbes require different dynamic feeding strategies based on the unique physiology of each organism. The cases described above are just a few examples of dynamic feeding strategies for Saccharomyces cerevisiae published in the literature. A host of other strategies have been developed for other organisms, allowing bioprocess substrate delivery to be optimally controlled with a variety of cell line genotypes or process setpoints. This can simplify experimental conclusions and increase confidence that the optimal performance of the culture is measured in each run.
Rapidly Develop Dynamic Processes with Culture Biosciences
Culture Biosciences’ cloud-based bioreactors are able to support many different dynamic feeding strategies, including pH and DO-stats, OUR, CER and RQ-stats and custom schemes developed for specific organisms and bioprocesses. In addition, Culture’s bioreactor capacity allows for the rapid development, optimization, and validation of dynamic processes, accelerating customers' timelines toward measuring optimal cell line performance. Conversely, failure to invest in the development of these dynamic strategies can lead to increased complexity in data interpretation and potentially inaccurate conclusions about cell line performance or process modifications, both of which will delay progress on a project.
For more on feed strategies for process development, including key challenges to keep in mind and best practices, check out this resource.
1. Crabtree, HG (1929). "Observations on the carbohydrate metabolism of tumours". The Biochemical Journal. 23 (3): 536–45. doi:10.1042/bj0230536. PMC 1254097. PMID 16744238.
2. Luli GW, Strohl WR. 1990. Comparison of growth, acetate production, and acetate inhibition of Escherichia coli strains in batch and fed-batch fermentations. Appl Environ Microbiol 56:1004–1011. Majewski RA, Domach MM. 1990.
3. Henes B, Sonnleitner B. Controlled fed-batch by tracking the maximal culture capacity. J Biotechnol. 2007 Oct 31;132(2):118-26. doi: 10.1016/j.jbiotec.2007.04.021. Epub 2007 May 3. PMID: 17573139.
4. Tippmann, S., Scalcinati, G., Siewers, V. and Nielsen, J. (2016), Production of farnesene and santalene by Saccharomyces cerevisiae using fed‐batch cultivations with RQ‐controlled feed. Biotechnol. Bioeng., 113: 72-81. https://doi.org/10.1002/bit.25683
5. Zhi-Qiang Xionga,b, Mei-Jin Guoa,∗, Yuan-Xin Guoa, Ju Chua, Ying-Ping Zhuanga, Nam Sun Wangb, Si-Liang Zhanga RQ feedback control for simultaneous improvement of GSH yield and GSH content in Saccharomyces cerevisiae T65. Enzyme and Microbial Technology 46 (2010) 598–602. doi:10.1016/j.enzmictec.2010.03.003
6. Effect of RQ and pre-seed conditions on biomass and galactosyl transferase production during fed-batch culture of S. cerevisiae BT150. Jana Zigova. Institute of Biotechnology, Research Centre Juelich, D-52425 Juelich, Germany. Journal of Biotechnology 80 (2000) 55–62 DOI: 10.1016/s0168-1656(00)00231-5
An interview with Hardware Engineering Manager and first employee Collin Edington about the new capabilities you can expect from Culture's bioreactors.
Receive occasional updates about all the interesting and exciting things bubbling up in the world of fermentation!