Software sensors as a tool for optimization of animal-cell cultures

R.C. Dorresteijn

    Research output: Thesisinternal PhD, WU


    <p>In this thesis software sensors are introduced that predict the biomass activity and the concentrations of glucose, glutamine, lactic acid, and ammonium on line, The software sensors for biomass activity, glucose and lactic acid can be applied for any type of animal cell that is grown in a bioreactor system. The glutamine and ammonium software sensors are determined experimentally by correlating them to the acid-production rate. Therefore, they can only be used for Vero cells.<p>In the development of these software sensors, much attention is paid to the verification of their basic elements, i.e. the oxygen uptake rate (OUR) and the acid production rate (APR). For the estimation of the OUR, information about the course of the oxygen solubility and the oxygen transfer coefficient during cultivation is considered very important, since fluctuations in one of these parameters will introduce a systematic error in the OUR. Experiments have shown that the oxygen solubility remains approximately constant throughout the cultivation, However, the oxygen transfer coefficient can vary significantly during cultivation. Because of this, a method for on-line determination of the oxygen transfer coefficient is developed. Application of this method results in an estimation of the OUR that is more reliable and contains less noise than the determination of the OUR using an off-gas mass spectrometer.<p>The lactic-acid production rate (LPR) is estimated from the APR, which in turn is determined from the amount of base utilized to control the pH at a fixed level. The estimated LPR provides reliable results, until nutrient limitation occurs. After that, the predicted LPR is higher than the measured LPR. Since at the end of the cultivation the lactic-acid concentration starts to drop, the cells are likely to take up lactic acid as an energy source. This uptake is supposed to be at least partially responsible for the discrepancy between estimated and measured LPR. However, since the estimation of the LPR is accurate up to the moment that nutrient limitation occurs and since the aim of control is to prevent nutrient limitation, the estimated LPR can still be used for control purposes.<p>After evaluation of the two basic elements of the software sensors, the sensors themselves are tested. Two software sensors use both the OUR and the LPR. These sensors, the glucose sensor and the biomass activity sensor, defined as the total energy requirement of the cultivated cells, both provide reliable signals. It is shown that the maximum error in the biomass activity signal, caused by the error in the LPR, is 4.6%. This error is small enough to be neglected. Also the estimation of the glucose concentration is accurate throughout the culture. The overall error in the estimation of the glucose uptake rate is found to be 6%.<p>The software sensors for lactic acid, glutamine and ammonium are solely based on the APR. Since the APR only provides reliable values until nutrient limitation occurs, these software sensors can only be used up to that point too. However, in the range at which the software sensors are valid, they are accurate.<p>Applying the software sensors, the effect of controlled addition of glucose and glutamine is studied. For hybridoma cells a significant reduction of the lacticacid production rate is described in literature when the glucose concentration is controlled at a fixed low level. Contradictory to this, the specific lactic-acid production rate for Vero cells is found not to change when glucose is controlled at different levels from 10 down to 0.25 mol.m <sup>-3</SUP>. In fact, no significant changes in the cultivation conditions are observed at all. Reducing the glutamine concentration from 4 down to 0.5 mol.m <sup>-3</SUP>yields in a significant reduction (38%) of the specific ammonium production rate. This reduction is corrected for the ammonium that is formed during the spontaneous decomposition of glutamine. Lowering the glutamine concentration in the culture also results in a reduced decomposition of glutamine. The total reduction in ammonium formation caused by these two effects is 62%. Although a significant reduction of the ammonium formation is achieved, no effect is observed on the growth rate and the maximum cell density of Vero cells.<p>Analysis of the cultivation medium using off-line HPLC has shown that after approximately 100 hours of cultivation, two amino acids are depleted. To find out whether or not these amino acids, methionine and serine, are responsible for the limitations in growth rate and cell density, they are added sequentially at the end of a culture. At that point the biomass activity is already dropping. Adding first methionine results in a stabilization of the biomass activity, whereas first addition of serine does not seem to have any effect. Further addition of the second amino acid instantaneously results in an increase of the biomass activity, the growth rate and the cell density in both cases. This indicates that even serine, which is a nonessential amino acid, is essential to maintain cell growth. At the moment serine is depleted, the cells have to make serine themselves, which causes the growth rate to drop. When methionine, an essential amino acid, is depleted the cell growth stops completely.<p>In an attempt to increase the maximum cell density, additional amounts of both amino acids are added to the cultivation medium prior to the start of the culture. Unexpectedly, serine again is depleted after 100 hours of cultivation, in spite of an initial serine concentration of 0.4 instead of 0.1 mol-m <sup>-3</SUP>. The biomass activity drops instantaneously at the moment serine is depleted. After the serine concentration in the cultivation medium is restored, the biomass activity only partially recovers. Despite the short serine limitation, the cell density doubles compared to the reference culture that is executed without glucose and glutamine control and without supplementing limiting amino acids.<p>If this kind of control systems are to be implemented into production systems, some kind of plant automation is required. This can vary from computer-controlled measurement- and control systems up to fully automized manufacturing execution systems (MES) that control all parts of the production process. These processes can vary from stock management and generation of reports, to manmachine interfaces that operate the measurement- and control systems. An MES becomes important when the production facility has to produce according to governmental regulations such as current good manufacturing practice (cGMP). The system ensures that all parts of the production process are executed according to these rules. Additional features are process- dependent control, the use of software sensors and predictive models, the use of historical data and on-line statistical techniques for trend analysis and detection of instrumentation failures.
    Original languageEnglish
    QualificationDoctor of Philosophy
    Awarding Institution
    • Tramper, J., Promotor, External person
    • Beuvery, E.C., Promotor, External person
    Award date14 Jan 1997
    Place of PublicationS.l.
    Print ISBNs9789054856375
    Publication statusPublished - 1997


    • experiments
    • tissue culture
    • cell culture
    • meristems

    Fingerprint Dive into the research topics of 'Software sensors as a tool for optimization of animal-cell cultures'. Together they form a unique fingerprint.

    Cite this