Plasmid stability analysis in a non‐homogeneous bioreactor for a recombinant fed‐batch fermentation

Abstract

Fed-batch fermentation for tryptophan synthetase by a recombinant Escherichia coli strain has been analyzed through a model incorporating segregational loss of the plasmid pPLc23rrpAl and imperfect macromixing of the broth. These features become significant in large fermentation vessels, where fluid circulation in the bioreactor influences the rates of cell growth and productivity. A simple model consisting of two interacting reactors describes the degree of macromixing, which is characterized by the respective (internal) dilution rates. Controlled imperfect macromixing is superior to other methods in arresting, and even reversing, the normal decline in the concentration of recombinant cells, thereby providing a method to exploit the non-homogeneity of the broth in large bioreactors. Some physical implications are discussed.On a analyst la fermentation en alimentation discontinue pour la synthttase de tryptophan par une souche de Escherichia coli recombinante au moyen d'un modkle incorporant la perte stgrkgationnelle du plasmide pPLc23ttpA I et le macromklange imparfait du bouillon. Ces caracttristiques deviennent significatives dans les rtservoirs de fermentation de grande taille, oh la circulation des fluides dans le biorkacteur influence la vitesse de croissance et la productivitt des cellules. Un modble simple composk de deux rtacteurs qui interagissent dtcrit le degrt de macromtlange, qui est caracttrisk par les vitesses de dilution respectives (internes). Le macromklange imparfait contr6lk est suptrieur ? I celui d'autres mtthodes pour stopper, et m&me inverser, le dkclin normal dans la concentration de cellules recombinantes, ce qui fournit une mtthode pour exploiter la non-homogkntitt du bouillon dans les grands biortacteurs. Certaines implications physiques sont analyskes.

Neural and Hybrid Neural Modeling and Control of Fed‐Batch Fermentation for Streptokinase: Comparative Evaluation under Nonideal Conditions

“…It is required now to decide how many and which of the other fi ve variables should be estimated by neural networks. In the absence of any rules to determine the number of variables to be estimated by neural networks, previous experience with recombinant fermentations (Patnaik, 1997(Patnaik, , 1999a(Patnaik, , 2001b and similarities between S. equisimilis cultivation and that of recombinant strains were used to decide upon two neural networks for on-line estimations of two of the fi ve concentrations. Measuring all by instrumental methods forfeits the advantages of a hybrid neural network (Schubert et al, 1994;van Can et al, 1997), besides the practical diffi culties involved (Locher et al, 1992).…”

“…Measuring all by instrumental methods forfeits the advantages of a hybrid neural network (Schubert et al, 1994;van Can et al, 1997), besides the practical diffi culties involved (Locher et al, 1992). In the absence of any rules to determine the number of variables to be estimated by neural networks, previous experience with recombinant fermentations (Patnaik, 1997(Patnaik, , 1999a(Patnaik, , 2001b and similarities between S. equisimilis cultivation and that of recombinant strains were used to decide upon two neural networks for on-line estimations of two of the fi ve concentrations. Based on these studies, Elman neural networks were selected.…”

“…Since the expression of a recombinant protein in a heterogeneous broth is improved by optimizing the starting distribution of the plasmid copy number (Patnaik, 2000), the degree of mixing or dispersion (Patnaik, 1999a), and the variance of the fi ltered noise entering along with the feed stream (Patnaik, 1999b), it is useful to develop methods to optimize these variables on-line during a fermentation. Since the expression of a recombinant protein in a heterogeneous broth is improved by optimizing the starting distribution of the plasmid copy number (Patnaik, 2000), the degree of mixing or dispersion (Patnaik, 1999a), and the variance of the fi ltered noise entering along with the feed stream (Patnaik, 1999b), it is useful to develop methods to optimize these variables on-line during a fermentation.…”

“…While allowing the inherent non-homogeneity of the broth in large bioreactors to be profi tably exploited, a variation of the copy number, with time and among the cells at a given time, also complicates on-line monitoring, modelling and control. Since the expression of a recombinant protein in a heterogeneous broth is improved by optimizing the starting distribution of the plasmid copy number (Patnaik, 2000), the degree of mixing or dispersion (Patnaik, 1999a), and the variance of the fi ltered noise entering along with the feed stream (Patnaik, 1999b), it is useful to develop methods to optimize these variables on-line during a fermentation.…”

Neural Optimization of Fed‐batch Streptokinase Fermentation in a Non‐ideal Bioreactor

“…Most of these studies have analyzed recombinant fermentations, where it has been shown that optimizing the degree of mixing or dispersion (Patnaik, 1999a(Patnaik, , 2001a and the variance of the filtered noise (Patnaik, 1999b) generates greater activities of the cloned-gene protein than is possible either with unfiltered noise or the complete removal of noise. The latter result may be surprising in that it invalidates the classical belief that a noise-free fermentation in a fully dispersed medium should generate the highest productivity.…”